# IMPROVING THE NUTRITIONAL CONTENT AND QUALITY OF CROPS: PROMISES, ACHIEVEMENTS, AND FUTURE CHALLENGES

EDITED BY : Felipe Klein Ricachenevsky, Marta Wilton Vasconcelos, Huixia Shou, Alexander Arthur Theodore Johnson and Raul Antonio Sperotto PUBLISHED IN : Frontiers in Plant Science

#### Frontiers Copyright Statement

© Copyright 2007-2019 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.

The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.

Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.

Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.

As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.

All copyright, and all rights therein, are protected by national and international copyright laws.

The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use. ISSN 1664-8714 ISBN 978-2-88963-018-9 DOI 10.3389/978-2-88963-018-9

### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

## IMPROVING THE NUTRITIONAL CONTENT AND QUALITY OF CROPS: PROMISES, ACHIEVEMENTS, AND FUTURE CHALLENGES

Topic Editors:

Felipe Klein Ricachenevsky, Federal University of Santa Maria, Brazil Marta Wilton Vasconcelos, Universidade Católica Portuguesa, Portugal Huixia Shou, Zhejiang University, China Alexander Arthur Theodore Johnson, The University of Melbourne, Australia Raul Antonio Sperotto, University of Taquari Valley - Univates, Brazil

Citation: Ricachenevsky, F. K., Vasconcelos, M. W., Shou, H., Johnson, A. A. T., Sperotto, R. A., eds. (2019). Improving the Nutritional Content and Quality of Crops: Promises, Achievements, and Future Challenges. Lausanne: Frontiers Media. doi: 10.3389/978-2-88963-018-9

# Table of Contents

*06 Editorial: Improving the Nutritional Content and Quality of Crops: Promises, Achievements, and Future Challenges*

Felipe Klein Ricachenevsky, Marta Wilton Vasconcelos, Huixia Shou, Alexander Arthur Theodore Johnson and Raul Antonio Sperotto

*10 From in* planta *Function to Vitamin-Rich Food Crops: The ACE of Biofortification*

Simon Strobbe, Jolien De Lepeleire and Dominique Van Der Straeten

*37 Natural Variation in Physiological Responses of Tunisian* Hedysarum carnosum *Under Iron Deficiency*

Heithem Ben Abdallah, Hans Jörg Mai, Tarek Slatni, Claudia Fink-Straube, Chedly Abdelly and Petra Bauer


Bingbing Luo, Jingguang Chen, Longlong Zhu, Shuhua Liu, Bin Li, Hong Lu, Guoyou Ye, Guohua Xu and Xiaorong Fan

*77 Elemental Profiling of Rice FOX Lines Leads to Characterization of a New Zn Plasma Membrane Transporter, OsZIP7*

Felipe K. Ricachenevsky, Tracy Punshon, Sichul Lee, Ben Hur N. Oliveira, Thomaz S. Trenz, Felipe dos Santos Maraschin, Maria N. Hindt, John Danku, David E. Salt, Janette P. Fett and Mary Lou Guerinot

*89 Genetic Basis and Breeding Perspectives of Grain Iron and Zinc Enrichment in Cereals*

Ana Luisa Garcia-Oliveira, Subhash Chander, Rodomiro Ortiz, Abebe Menkir and Melaku Gedil


Grace Z. H. Tan, Sudipta S. Das Bhowmik, Thi M. L. Hoang, Mohammad R. Karbaschi, Hao Long, Alam Cheng, Julien P. Bonneau, Jesse T. Beasley, Alexander A. T. Johnson, Brett Williams and Sagadevan G. Mundree

*132 Biofortification of Cereals With Foliar Selenium and Iodine Could Reduce Hypothyroidism*

Graham Lyons

### *140 Should Heavy Metals be Monitored in Foods Derived From Soils Fertilized With Animal Waste?*

Rafael da Rosa Couto, Jucinei J. Comin, Monique Souza, Felipe K. Ricachenevsky, Marcos A. Lana, Luciano C. Gatiboni, Carlos A. Ceretta and Gustavo Brunetto

*145 Dynamic Modeling of Silicon Bioavailability, Uptake, Transport, and Accumulation: Applicability in Improving the Nutritional Quality of Tomato*

Mari C. López-Pérez, Fabián Pérez-Labrada, Lino J. Ramírez-Pérez, Antonio Juárez-Maldonado, América B. Morales-Díaz, Susana González-Morales, Luis R. García-Dávila, Jesús García-Mata and Adalberto Benavides-Mendoza

*155 Genetic Basis of Variation in Rice Seed Storage Protein (Albumin, Globulin, Prolamin, and Glutelin) Content Revealed by Genome-Wide Association Analysis*

Pingli Chen, Zhikang Shen, Luchang Ming, Yibo Li, Wenhan Dan, Guangming Lou, Bo Peng, Bian Wu, Yanhua Li, Da Zhao, Guanjun Gao, Qinglu Zhang, Jinghua Xiao, Xianghua Li, Gongwei Wang and Yuqing He


Yu-Min Shi, Chun-Chao Wang, Li-Yan Zhang, Jun-Tao Ma, Ling-Wei Deng, Wan Li, Tian-Tian Xu, Cheng-Zhi Liang, Jian-Long Xu and Zhi-Kang Li

*210 Toward Eradication of B-Vitamin Deficiencies: Considerations for Crop Biofortification*

Simon Strobbe and Dominique Van Der Straeten

*229 You Shall not Pass: Root Vacuoles as a Symplastic Checkpoint for Metal Translocation to Shoots and Possible Application to Grain Nutritional Quality*

Felipe K. Ricachenevsky, Artur T. de Araújo Junior, Janette P. Fett and Raul A. Sperotto


Bal R. Singh, Yadu N. Timsina, Ole C. Lind, Simone Cagno and Koen Janssens

*254 Analysis of Yellow Striped Mutants of* Zea mays *Reveals Novel Loci Contributing to Iron Deficiency Chlorosis*

David Chan-Rodriguez and Elsbeth L. Walker

*264 Contribution of NtZIP1-Like to the Regulation of Zn Homeostasis* Anna Papierniak, Katarzyna Kozak, Maria Kendziorek, Anna Barabasz, Małgorzata Palusińska, Jerzy Tiuryn, Bohdan Paterczyk, Lorraine E. Williams and Danuta M. Antosiewicz


Swati Puranik, Jason Kam, Pranav P. Sahu, Rama Yadav, Rakesh K. Srivastava, Henry Ojulong and Rattan Yadav

*385 Genome-wide Identification, Characterization, and Expression Analysis of PHT1 Phosphate Transporters in Wheat*

Wan Teng, Yan-Yan Zhao, Xue-Qiang Zhao, Xue He, Wen-Ying Ma, Yan Deng, Xin-Ping Chen and Yi-Ping Tong

# Editorial: Improving the Nutritional Content and Quality of Crops: Promises, Achievements, and Future Challenges

Felipe Klein Ricachenevsky <sup>1</sup> \*, Marta Wilton Vasconcelos <sup>2</sup> \*, Huixia Shou<sup>3</sup> \*, Alexander Arthur Theodore Johnson<sup>4</sup> \* and Raul Antonio Sperotto<sup>5</sup> \*

<sup>1</sup> Biology Department, Federal University of Santa Maria, Santa Maria, Brazil, <sup>2</sup> Universidade Católica Portuguesa, CBQF - Centro de Biotecnologia e Química Fina - Laboratório Associado, Escola Superior de Biotecnologia, Porto, Portugal, <sup>3</sup> The State Key Lab of Plant Physiology and Biochemistry, College of Life Sciences, Zhejiang University, Hangzhou, China, <sup>4</sup> School of BioSciences, The University of Melbourne, Melbourne, VIC, Australia, <sup>5</sup> Graduate Program in Biotechnology, University of Taquari Valley - Univates, Lajeado, Brazil

Keywords: nutritional quality, biofortification, plant nutrition, iron, zinc, transporter

### **Editorial on the Research Topic**

### Edited and reviewed by:

Jan Kofod Schjoerring, University of Copenhagen, Denmark

#### \*Correspondence:

Felipe Klein Ricachenevsky felipecruzalta@gmail.com Marta Wilton Vasconcelos mvasconcelos@porto.ucp.pt Huixia Shou huixia@zju.edu.cn Alexander Arthur Theodore Johnson johnsa@unimelb.edu.au Raul Antonio Sperotto rasperotto@univates.br

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 08 May 2019 Accepted: 17 May 2019 Published: 06 June 2019

#### Citation:

Ricachenevsky FK, Vasconcelos MW, Shou H, Johnson AAT and Sperotto RA (2019) Editorial: Improving the Nutritional Content and Quality of Crops: Promises, Achievements, and Future Challenges. Front. Plant Sci. 10:738. doi: 10.3389/fpls.2019.00738 **Improving the Nutritional Content and Quality of Crops: Promises, Achievements, and Future Challenges**

Plants are the ultimate source of nutrients for humans and livestock. To date, we use only a handful of species for our subsistence and focus mostly on carbohydrate-rich grains such as wheat, corn and rice. However, these starch-rich cereals are typically nutrient-poor, leaving populations that base their diets on starchy grains with low intakes of many essential nutrients. At the same time, seeds, leaves and roots can accumulate toxic elements depending on species, genotype and growth conditions, representing potential hazards to human health. Improving the nutritional quality of plants, which includes both nutrient enrichment of edible tissues and minimizing the likelihood of contamination, has been a focus of research for years. In this Research Topic a diverse collection of reviews, opinions and original research articles highlight the achievements and future directions in this field.

### BIOFORTIFICATION/NUTRITIONAL QUALITY: AN OVERVIEW

Garg et al. review a range of biofortification studies in different crops involving transgenic, conventional, and agronomic approaches and the use of biotechnology, crop breeding, and fertilization, respectively. The authors discuss the limitations of each approach and the challenges that transgenic biofortified crops face regarding consumer acceptance. Taking a different perspective, Capstaff and Miller review the current knowledge and future direction for yield and nutritional quality improvement of forage crops, which have received relatively limited attention compared to cereals, fruits, and vegetables. The authors also discuss the applicability of knowledge obtained from model plants and grain crops coupled with the availability of genomics and bioinformatics to generate improved forage crops for food security. The review by Hameed et al. presents an overview of nutritional improvement efforts in potato (Solanum tuberosum) using both traditional breeding and genetic engineering techniques, including nutrient concentration increase and anti-nutrient decrease. The authors also discuss constraints to potato production, such as biotic and abiotic stresses, post-harvest quality, unfavorable soil and climatic factors. Moreover, genetically modified potato risk assessment/regulation and

**6**

future breeding techniques using TALENs and CRISPR/Cas9 are discussed.

Paul et al. review progress in the Banana21 biofortification program, which aims to increase pro-vitamin A content (and iron in the future) in cooking banana, which is a staple food in Uganda. Vitamin A deficiency is one of the most widespread nutritional problems worldwide, affecting millions of people, and is especially prevalent in children. Importantly, the program includes technology transfer and aims to produce and deregulate transgenic local varieties of banana. Authors provide an overview of the strategies used, to achieve the goal of delivering at least 50% of the estimated average requirement of pro-vitamin A in 300 g of banana per day.

Other reviews discuss biofortification for specific nutrients. Strobbe and Van Der Straeten focus on the role of thiamin (B1), pyridoxine (B6), and folates (B9) in plant physiology. Biofortification strategies to enhance B-vitamin in crops using metabolic engineering or breeding are presented. The authors also discuss the concept of multi-biofortification, the simultaneous biofortification of multiple vitamins and minerals, and the possible synergistic or adverse effects of such combinations. Lyons reviews the approaches used for agronomic biofortification of crops with iodine (I) and selenium (Se) in major crops, with a special focus on wheat. Although I is provided by salt iodization, this approach has been shown to be insufficient. The author shows problems derived from combined deficiency in both nutrients, and how current approaches to provide enough I and Se to humans are not sufficient. The review discusses approaches to I and Se agronomic biofortification, comparing results of soil and foliar application of I and Se fertilizers. Finally, Puranik et al. assess recent advances and challenges for Ca biofortification in finger millet (Eleusine coracana), a crop with inherently high grain Ca content, but which needs to be in a bioavailable form to impact human nutrition. As such, integration of a naturally Ca-rich crop like finger millet into global biofortification programs could be a good starting point to alleviate human Ca malnutrition. According to the authors, large-scale protein profiling to identify the complete set of proteins involved in finger millet Ca homeostasis is still unavailable.

López-Pérez et al. present a dynamic model to describe absorption, transport, and deposition rates of silicon (Si) in tomato. The model consists of six state equations, using as inputs key environmental factors related to Si absorption and mobilization, such as temperature, pH, CO<sup>2</sup> concentration, and soil organic matter. Use of the model can increase understanding of the agronomic management of Si in plants. On a similar line, da Rosa Couto et al. explore the plants grown in soils with long-term animal waste applications as fertilizers and the potential grain contamination with heavy metals. Although important for both traditional and organic farming, the recurrent use of animal waste can result in changes to soil chemistry that make plants more prone to accumulation of heavy metals, which can harm consumers. The manuscript estimates the potential contamination risks associated with soil concentration of heavy metals after organic waste usage reported in the literature, and calls attention to the need for more careful monitoring to understand how different plant species and their edible parts can become a source of dietary contamination for humans.

### REVIEWS ON FE AND ZN BIOFORTIFICATION

Nutritional quality includes many topics, but certainly iron (Fe) and zinc (Zn) biofortification has been one of the most prolific areas in the field. Here we feature updated reviews and opinions focused on these two essential micronutrients that are commonly lacking in human diets. Garcia-Oliveira et al. review the genetic basis of Fe and Zn accumulation in cereals and describe how modern breeding technologies are helping to promote essential element accumulation and bioavailability, while minimizing the accumulation of anti-nutrients such as phytate and hazardous heavy metals that often are transported along with Fe and Zn. The authors discuss how existing genetic variability can contribute to breeding of biofortified cereals, and some of the methodological difficulties in reliably measuring Fe and Zn. Sperotto and Ricachenevsky discuss common bean (Phaseolus vulgaris) biofortification efforts, and how model species lessons can provide shortcuts in finding pathways and candidate genes for Fe, Zn and anti-nutrient manipulation in beans. In particular, the possibility of tissue-specific biofortification of common bean seed coat and cotyledons, as well as decreasing phytate and polyphenols anti-nutrients in the same tissues, is discussed.

The mini-review by Ricachenevsky et al. describes the role of root vacuoles in controlling symplastic concentrations of nutrients and toxic trace elements, which in turn affects shoot accumulation. Examples from the literature regarding natural variation in vacuolar transporters and loss-of-function mutants show that the more abundantly an element is stored in root vacuoles, the less it is loaded into the xylem and translocated to shoots and seeds. Thus, manipulation of root storage capacity should be considered in biofortification approaches. In another mini-review, Nozoye highlights the usefulness of increasing nicotianamine (NA) levels in plants, which is achieved by overexpressing nicotianamine synthase (NAS) genes. NA is a metal chelator involved in metal translocation. Plants that accumulate more NA increase Fe and Zn concentrations in edible tissues and can also become more tolerant to Fe deficiency.

### EXPERIMENTAL APPROACHES TO IMPROVE NUTRITIONAL QUALITY

Manuscripts in this section describe experimental methodologies important for the improvement of nutritional quality in plants. Due to the high bioavailability of rice seed storage protein (SSP) for human and animal nutrition, increased SSP content is one of the main breeding objectives for improving the nutritional quality of rice. Chen et al. investigate albumin, globulin, prolamin, glutelin, and total SSP contents in milled rice of 527 rice accessions grown in two environments. By associating these nutrient traits with genome sequencing data, they identify novel SNPs and candidate genes related to rice seed protein content and composition which may be useful for future rice molecular breeding strategies aiming at quality improvement.

It is well-known that daily consumption of fruits and vegetables lowers the risk of several chronic illnesses, including cardiovascular disease, diabetes, and cancer. The health benefits of vegetable and fruit crops are often attributed to their high content of specific health promoting compounds such as fibers, polyphenols, and vitamins. Glycine is the most abundant free amino acid in horticultural soils and Yang et al. show that exogenous glycine supplementation can increase the accumulation of health-promoting compounds and enhance antioxidant activity in hydroponically grown lettuce.

Starch is the main carbohydrate form in wheat grain. Other than the major components of amylose and amylopectin, starch can also interact with minor components such as lipids, proteins, and phosphorus. It is understandable that phosphorus supplies must affect the yield and quality of wheat grain. Zhang et al. employ three levels of phosphorus fertilizer application to wheat fields and find that different levels of phosphorus significantly influence the expression of starch biosynthesis genes, starch synthesis, degradation, and microstructure in wheat grains. The results provide knowledge about the importance of applying appropriate amounts of phosphorus fertilizers for the improvement of wheat yield and starch quality.

The high affinity nitrate transporter OsNRT2.1 plays a role in nitrogen uptake and translocation in rice. Luo et al. show that OsNRT2.1 overexpression increases not only nitrate uptake, but also Mn accumulation in rice grains. The reason for high accumulation of Mn by OsNRT2.1 overexpression is probably due to elevated expression of Mn transporter genes, including OsNRAMP3, 5, and 6. This work provides an alternative approach for increasing Mn uptake in plants and could have implications for grain quality.

### EXPERIMENTAL APPROACHES TO FE AND ZN BIOFORTIFICATION

We highlight a series of original research papers focused on increasing Fe and Zn accumulation in crops. Díaz-Benito investigate wild-type rice and six transgenic rice lines overexpressing nicotianamine synthase (OsNAS1) and/or barley nicotianamine amino transferase (HvNAATb) in order to elucidate the role of 2-deoxymugineic acid (DMA) and nicotianamine (NA) on metal distribution in the rice embryo and endosperm. Using a series of approaches the authors conclude that when there are increases in DMA concentration alone or in combination with NA, the prevalent mechanism of seed Fe loading is via Fe(III)-DMA. The study also highlights that a better understanding of transgenic plant phenotypes, using in-depth localized quantification of targeted nutrients, will improve the efficacy of future biofortification strategies. In a similar topic, Tan et al.reports on a strategy for Fe biofortification of chickpea (Cicer arietinum) using genetic engineering. The authors successfully transformed cultivar HatTrick with chickpea nicotianamine synthase 2 (CaNAS2) and soybean (Glycine max) ferritin (GmFER) constitutive expression cassettes. Analysis of NA and Fe levels in the transformed seeds revealed that NA levels were twice those of control lines. The authors suggest that this may have important ramifications in terms of increasing Fe bioavailability in chickpea grains.

Zhang et al. examine two sets of backcrossed inbred lines derived from the same donor, and two recipient elite varieties from Southwestern China, to determine the effect of genetic background and environment on grain mineral concentration by QTL mapping. This study allowed confirmation of the results of a genome-wide association study (GWAS) using a set of 698 sequenced accessions, and favorable haplotypes of Fe, Zn, Cd, Mn, Cu, and Se candidate genes were identified. In particular, 37 genes (19.3%) were found to be significantly associated between the QTL targeting traits and the haplotype variations by pairwise comparison, and these genes may be useful for future rice biofortification strategies.

Singh et al. tackle the challenge of increasing Zn, Fe and protein concentration in wheat grain by applying exogenous nitrogen (N), Fe, and Zn at different rates and application times. Apart from increased grain yield, relatively higher protein content and Fe/Zn concentration were recorded in wheat grain when a split N application was applied. Furthermore, soil and foliar Fe/Zn supplies combined with a single application of N at sowing increased Zn and Fe concentrations by 46% and 35%, respectively, relative to controls. These results indicate that proper management of N, Fe, and Zn application may could enhance grain protein content and Fe/Zn concentration in wheat.

### FE AND ZN PHYSIOLOGY STUDIES TO PROVIDE TOOLS FOR BIOFORTIFICATION

Aiming to develop a new method to quickly characterize rice genes related to metal homeostasis (in particular Fe and Zn), Ricachenevsky et al. investigate the ionome of Arabidopsis Full Length Over-eXpressor (FOX) lines with heterologous expression of rice cDNAs driven by the 35S promoter. The authors identified two lines overexpressing OsZIP7, which had 25% increase in shoot Zn concentrations compared to control. Ricachenevsky et al. found that the gene was able to complement a Zn transport defective yeast mutant, the protein is localized to the plasma membrane, and the seeds of the overexpressing lines had significantly higher Zn concentrations. This technique shows promise toward identifying candidate genes for mineral enhancement.

Phytoremediation has frequently been defined as a sustainable bioremediation process that uses various types of plants to remove, transfer, stabilize, and/or destroy heavy metals from the environment in which they grow. In this context, it is important to choose the best "crop for the job". Papiernak et al. suggest tobacco as a possible crop to fulfill this goal. However, despite the interest in the use of tobacco to remove metals from contaminated soil, knowledge of the processes by which tobacco exerts this role is still limited. The authors strived to identify new Zn transport genes in tobacco using an in silico approach, gene expression data, gene cloning and functional analysis in yeast (heterologous expression). The authors conclude that NtZIP1 like is localized at the plasma membrane, in the roots and shoots, and is involved in Zn transport. NtZIP1-like appears responsible for Zn uptake by root cells of the mature basal zone and may be involved in a mechanism to protect the root and leaf cells from accumulating excess Zn.

Chan-Rodriguez and Walker perform a genetic dissection of yellow-stripe mutants available in the Maize Genetic Cooperation Stock Center (MGCSC). Grasses are known to rely on Fe(III) chelation for Fe uptake, which involves phytosiderophore secretion and Fe(III)-phytosiderophore complex transport into root cells. Yellow-stripe phenotypes are indicative of Fe deficiency in leaves, and one of the first mutants characterized in maize (ys1) contained mutations in the Yellow-Stripe 1 (YS1) transporter, which is the Fe(III)-phytosiderophore transporter (Curie et al., 2001), while another (ys3) is suggested to have mutations in the phytosiderophore secretion transporter TOM1 (Nozoye et al., 2013). Authors screen 31 yellow-stripe mutants from maize, identifying some allelic to ys1 and ys3, as well as three new, non-allelic mutants which have low Fe levels in shoots and may represent new players in Fe uptake mechanisms of grasses.

Abdallah et al. describe a non-model plant species (Hedysarum carnosum), from semi-arid areas in Tunisia, where soils are saline-alkaline and present low Fe availability. Three isolates from H. carnosum were characterized for morphological and physiological responses to low Fe conditions, showing rhizosphere acidification and IRT1-ortholog upregulation, but no Fe reductase activity in roots. Interestingly, one ecotype showed increased Fe deficiency tolerance, which may be linked to more pronounced IRT1 expression under control conditions. This study lays the foundation for using H. carnosum to study adaptation to extreme conditions.

### GENE FAMILY CHARACTERIZATION

Two studies provide basic knowledge on important transporter gene families which could be important for further biofortification and/or plant nutrition studies. Qin

### REFERENCES


et al. identify 13 NRAMP genes in the soybean genome and document variable expression profiles of GmNRAMP genes among tissues and in response to nutrient stress, suggesting that GmNRAMP proteins perform a range of functions in specific tissues throughout development. Subcellular localization analysis in Arabidopsis protoplasts confirm the tonoplast or plasma membrane localization of these proteins. Teng et al. conduct a genome-wide sequence analysis of PHT1 (phosphate (Pi) transporters) genes in wheat and clone 21 of the genes. The cloned PHT1 genes show Pi-transport activity in yeast cells grown under both low and high Pi conditions. Expression of some TaPHT1 genes at flowering stage positively correlated with P uptake after stem elongation across three P application rates and two wheat varieties, suggesting that modification in PHT1 gene expression may improve P use efficiency under a wide range of P concentrations.

### FINAL COMMENT

This Research Topic provides readers with a wide range of manuscripts in plant nutrition and describes novel results relevant to food security and global health.

### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

### FUNDING

This work was supported by National Funds from FCT - Fundação para a Ciência e a Tecnologia through project UID/Multi/50016/2019.

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Ricachenevsky, Vasconcelos, Shou, Johnson and Sperotto. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# From in planta Function to Vitamin-Rich Food Crops: The ACE of Biofortification

#### Simon Strobbe† , Jolien De Lepeleire† and Dominique Van Der Straeten\*

Laboratory of Functional Plant Biology, Department of Biology, Ghent University, Ghent, Belgium

Humans are highly dependent on plants to reach their dietary requirements, as plant products contribute both to energy and essential nutrients. For many decades, plant breeders have been able to gradually increase yields of several staple crops, thereby alleviating nutritional needs with varying degrees of success. However, many staple crops such as rice, wheat and corn, although delivering sufficient calories, fail to satisfy micronutrient demands, causing the so called 'hidden hunger.' Biofortification, the process of augmenting nutritional quality of food through the use of agricultural methodologies, is a pivotal asset in the fight against micronutrient malnutrition, mainly due to vitamin and mineral deficiencies. Several technical advances have led to recent breakthroughs. Nutritional genomics has come to fruition based on marker-assisted breeding enabling rapid identification of micronutrient related quantitative trait loci (QTL) in the germplasm of interest. As a complement to these breeding techniques, metabolic engineering approaches, relying on a continuously growing fundamental knowledge of plant metabolism, are able to overcome some of the inevitable pitfalls of breeding. Alteration of micronutrient levels does also require fundamental knowledge about their role and influence on plant growth and development. This review focuses on our knowledge about provitamin A (beta-carotene), vitamin C (ascorbate) and the vitamin E group (tocochromanols). We begin by providing an overview of the functions of these vitamins in planta, followed by highlighting some of the achievements in the nutritional enhancement of food crops via conventional breeding and genetic modification, concluding with an evaluation of the need for such biofortification interventions. The review further elaborates on the vast potential of creating nutritionally enhanced crops through multi-pathway engineering and the synergistic potential of conventional breeding in combination with genetic engineering, including the impact of novel genome editing technologies.

Keywords: vitamin metabolism, crop improvement, hidden hunger, malnutrition, plant development, carotenoids, ascorbate, tocochromanols

## INTRODUCTION

Ensuring food security to all populations is considered a top priority for global societal progress. Undernourishment has dropped severely in the last decades, from roughly 20% of the world population in 1990 to little above 10% in 2016 (Food and Agriculture Organization [FAOSTAT], 2017). It stands undisputed that continuing efforts should be undertaken to further reduce the

### Edited by:

Raul Antonio Sperotto, University of Taquari Valley, Brazil

### Reviewed by:

Jorge E. Mayer, Ag RD&IP Consult P/L, Australia Khurram Bashir, RIKEN, Japan

### \*Correspondence:

Dominique Van Der Straeten Dominique.VanDerStraeten@UGent.be

> †These authors have contributed equally to this work

### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 28 September 2018 Accepted: 03 December 2018 Published: 18 December 2018

### Citation:

Strobbe S, De Lepeleire J and Van Der Straeten D (2018) From in planta Function to Vitamin-Rich Food Crops: The ACE of Biofortification. Front. Plant Sci. 9:1862. doi: 10.3389/fpls.2018.01862

**10**

number of undernourished people in the world, which is still close to 800 million. The successful reduction of malnourishment can partly be attributed to the increase in staple crop yield witnessed over the last decades. Indeed, in the last 25 years, the production per hectare of rice, wheat and potato has risen by 30% (Food and Agriculture Organization [FAOSTAT], 2017). However, these crops often fail to supply adequate amounts of micronutrients, thereby augmenting the prevalence of micronutrient malnutrition (MNM, 'hidden hunger'). These micronutrients include minerals such as iron, zinc, selenium, and manganese, as well as a wide range of vitamins (Miller and Welch, 2013). Hidden hunger affects an alarming two billion people (Bailey et al., 2015; Rautiainen et al., 2016), mostly in the form of anemia, occurring in one-fourth of the human population (McLean et al., 2009). The case of anemia clearly demonstrates the physiological impact of MNM, as its onset has been linked to deficiencies in different micronutrients such as iron, vitamin B1, B9, and B12 (Green, 2003; Imdad and Bhutta, 2012; Stabler, 2013). The importance of MNM is further highlighted by the large, calculated economic benefit a reduction of child malnutrition would have on development. Among 19 prioritized investment-for-development targets listed in the Post-2015 Consensus, the Copenhagen Consensus Center think-tank has ranked the reduction of child malnutrition as the human development investment with the highest potential economic returns (Copenhagen Consensus, 2012).

Vitamin deficiencies can be combatted by supplementation, industrial fortification, biofortification, and educational interventions encouraging dietary diversification. It should be noted that choice of the intervention strategy to be implemented depends on regional dietary and cultural differences (Bailey et al., 2015). However, some universally valid remarks can be made. Supplementation, whether by administration of (multi-)vitamin pills or by fortification of cereal products (mandatory in many countries), has shown to be a fast and powerful means to reduce vitamin deficiencies (Sandjaja et al., 2015; Atta et al., 2016; Wang et al., 2016). Unfortunately, this intervention is not easily applicable to poor rural populations in need (Blancquaert et al., 2014). Furthermore, supplementation could exhibit adverse effects, as demonstrated by the observation of increased mortality and higher risk of colorectal cancer in males upon vitamin A and B9 supplementation, respectively (Benn et al., 2015; Cho et al., 2015). Educational efforts, aimed to change the diet and/or processing of food by populations suffering from vitamin deficiencies, are an excellent way to fight MNM, tackling the root causes of the problem. However, these interventions are expensive and imply cultural and agronomical changes, the feasibility of which cannot be guaranteed (Low et al., 2007; Faber and Laurie, 2011). Biofortification, which consists of enhancing the natural vitamin level of food crops, is advocated as a powerful complementary method to fight vitamin malnutrition, circumventing the aforementioned obstructions (Fitzpatrick et al., 2012; Blancquaert et al., 2017; Saltzman et al., 2017).

Biofortification of local crops can be considered a sustainable and cost-effective means to reduce vitamin shortage (Meenakshi et al., 2010; De Steur et al., 2015). Two methods of biofortification, apart from agronomical interventions (Cakmak and Kutman, 2017; Watanabe et al., 2017), can be distinguished. First, biofortified crops can be obtained by conventional breeding or using molecular techniques, to obtain novel high-vitamin lines (Ortiz-Monasterio et al., 2007; Bouis and Saltzman, 2017). Unfortunately, this approach relies on the presence of sufficient variation of vitamin levels in sexually compatible germplasm collections (Shimelis and Laing, 2012; Strobbe and Van Der Straeten, 2017). Furthermore, introgression of a certain trait of interest into various region-specific crops demands time-consuming selection over several generations. Novel breeding techniques, however, enable more rapid retrieval of the desired trait via genome wide association mapping (GWAS) or accelerated selection of the introgression lines using marker-assisted breeding (MAB) (Borrill et al., 2014; Esuma et al., 2016). Second, metabolic engineering via GM-technology allows introduction of one or multiple genes of interest, influencing plant metabolism toward increased accumulation of the particular vitamin. As it is not dependent on sexual compatibility of gene source, genetic elements from a very diverse pool could be utilized, including the vast genetic diversity of prokaryotes. Moreover, metabolic engineering can be implemented in a time and tissue-specific manner via selection of promoters with the desired temporal and spatial characteristics. This method, however, demands prior knowledge about specific vitamin metabolism as well as availability of adequate promoters. In principle, it allows the creation of a model vitamin engineering strategy, which can be implemented in a variety of cultivars and crops. However, this cannot be generalized, due to differences in vitamin regulation and metabolism in different crops and tissues (Strobbe and Van Der Straeten, 2017). Interestingly, novel genome editing techniques such as the CRISPR/Cas system allow directed mutagenesis and editing of targeted genomic regions (Cong et al., 2013; Luo et al., 2016), enabling targeted metabolic engineering approaches, though still constrained by the limitation of genetic diversity of the engineered species. A combination of the aforementioned techniques, could offer powerful solutions to alleviate vitamin deficiencies.

Biofortification should be carried out with due consideration to its effects on the plant's physiology and not only with the consumers' vitamin needs in mind. The health impact of a biofortified crop could be region specific, due to genetic, environmental and dietary factors. Massive consumption of staple crops with low content of one or more micronutrients appears to be a major factor aggravating the incidence of the deficiency. Therefore, biofortification of these crops is advised. Biofortification endeavors should, however, not solely focus on vitamin content, but take all factors influencing vitamin-specific nutritional value of the particular crop into considerations, such as storage and processing stability, as well as bioavailability (Fitzpatrick et al., 2012; Blancquaert et al., 2015; Diaz-Gomez et al., 2017b).

The three vitamins covered in this review–namely vitamin A, C and E–have been the subject of various biofortification approaches due to their impact on human health and very low content in the six major staple crops consumed worldwide (**Table 1**). But because of their roles in key


(g/capita.day)Dataportionconsumptionon the food matrix, can easily amount to 50% each. 1Highest recommended daily allowance (RDA) is depicted (µg/day). 2Retinol activity equivalent. 3Milled equivalent. 4in 2013.

enzymatic and stress-related stress response roles, there is a need to bundle the existing knowledge of in planta vitamin metabolism, taking possible detrimental effects on crop growth into consideration. Consequently, proper design of metabolic engineering approaches for vitamin biofortification requires a profound understanding of in planta vitamin biosynthesis as well as its metabolism.

In the past decades, major advances have been accomplished in biofortification of different food crops. Fortunately, some of these are already being used to combat MNM. However, the use of metabolically engineered, biofortified crops has not been implemented to date. Interestingly, the imminent commercialization of provitamin A-rich 'Golden Rice' might open doors toward application of other engineered biofortified crops. In this review, the incidence and pathophysiology of the different vitamin deficiencies are discussed, alongside with the status of knowledge on plant vitamin biosynthesis and physiology and the advances made in crop biofortification with these vitamins.

### PROVITAMIN A – CAROTENOIDS

Vitamin A is a collective term for different fat-soluble retinoid molecules (Bai et al., 2011), defined as every chemical structure able to fulfill the biological activity of all-trans-retinol (**Figure 1C**) upon human consumption (Eitenmiller et al., 2016). Carotenoids, comprise over 600 different compounds, only three of which can be metabolically converted to active vitamin-A substances such as retinol (**Figure 1**) and its oxidized equivalents retinal and retinoic acid (Asson-Batres and Rochette-Egly, 2016). Carotenoids represent the major source of provitamin A in the diet and are present throughout the plant kingdom. The general backbone is formed by head-to-tail linking of eight isoprene units, resulting in a C40-unsaturated chain, lycopene (**Figure 1A**), a carotenoid precursor (Eitenmiller et al., 2016). The most important carotenoid, β-carotene (**Figure 1B**), harbors cyclized β-ionone rings on both ends of the C40-chain (**Figure 1**). Because these molecules consist of long-chain conjugated polyene units, they are sensitive to oxidation, light, heat and acids (Asson-Batres and Rochette-Egly, 2016). Their sensitivity to oxidation, however, enables them to serve as antioxidants in plants and animals, as the radical resulting from interaction with reactive oxygen species (ROS), is much less hazardous by stabilization of the polyene groups. Vitamin A function, however, greatly exceeds its antioxidant properties, as it plays multiple roles in plant and animal physiology.

### Vitamin A Biosynthesis

The principal provitamin A for humans is β-carotene, which is composed of two symmetrical retinyl groups. One such retinyl group consists of a retinyl isoprenoid chain and a β-ionone ring which is important for vitamin A action (**Figure 1**) (Send and Sundholm, 2007). Hence, as α-carotene, γ-carotene and β-cryptoxanthin also carry 1 β-ionone ring, they possess 50% vitamin A activity. Provitamin A is synthesized in plastids in all photosynthetic organisms by enzymes associated

fpls-09-01862 December 15, 2018 Time: 15:9 # 3

with the thylakoid membrane, namely phytoene desaturase (PDS), ζ-carotene desaturase (ZDS), lycopene-β-cyclase (β-LCY) and lycopene-ε-cyclase (ε-LCY); or associated in multienzyme complexes (Cunningham and Gantt, 1998).

The direct precursor for provitamin A is geranylgeranyl diphosphate (GGPP) (see also vitamin E biosynthesis, 4.1), which is formed by the condensation of the building blocks isopentenyl diphosphate (IPP) and 3 dimethylallyl diphosphate (DMAPP) molecules, by GGPP synthase (GGPPS) (Ruiz-Sola et al., 2016) (**Figure 2**). IPP is produced in the plastidlocalized 2-C-methyl-D-erythritol 4-phosphate (MEP) pathway and DMAPP is its isomerisation product catalyzed by isopentenyl diphosphate isomerase (IDI). GGPP is also the precursor for chlorophylls, ubiquinones, tocopherols, gibberellins and terpenoids (Saini et al., 2015). The first step of the actual provitamin A biosynthetic pathway is the condensation of two GGPP molecules by phytoene synthase (PSY) forming 15 cis-phytoene, assumed to be a rate-limiting step (Fray and Grierson, 1993; Li F.Q. et al., 2008; Cazzonelli and Pogson, 2010). In most plant species multiple redundant PSY genes are present which are differentially regulated. Salt and drought, are environmental factors which induce PSY expression, thereby enhancing carotene levels (Ruiz-Sola et al., 2014; Nisar et al., 2015). Moreover, ethylene is known to have a positive influence on accumulation of carotenoids, inducing PSY expression (Zhang et al., 2018). This aspect is particularly important in fruit ripening and has therefore been studied in mango (Mangifera indica) (Ma et al., 2018), durian (Durio zibethinus) (Wisutiamonkul et al., 2017) and tomato (Solanum lycopersicum) (Su et al., 2015; Cruz et al., 2018). A recent study identified the tomato transcription factor SlCMB1 as a regulator of both ethylene production and carotenoid accumulation (via PSY and PDS) (Zhang et al., 2018). PSY can therefore, in most plants, be considered a master regulator of carotenoid accumulation, given that it is also stimulated by light, directly controlled by transcription factors PHYTOCHROME INTERACTING FACTOR 1 (PIF1) and LONG HYPOCOTYL 5 (HY5) in Arabidopsis photomorphogenesis (Toledo-Ortiz et al., 2010; Llorente et al., 2017). In the subsequent biosynthesis step, directly downstream of PSY, 15-cis-phytoene is transformed into 9,15,9<sup>0</sup> tri-cis-ζ-carotene via a 15,9-di-cis-phytofluene intermediate by two consecutive desaturation reactions catalyzed by phytoene desaturase (PDS) (Pecker et al., 1992; Li et al., 1996; Qin et al., 2007). Subsequently, either a photoisomerization or an isomerization by ζ-carotene isomerase (ZISO) (Pecker et al., 1992; Li et al., 2007) results in 9,9<sup>0</sup> -di-cis-ζ-carotene. Reiteratively, two desaturation reactions are performed by ζ-carotene isomerase (ZDS) producing neurosporene followed by 7,9,7<sup>0</sup> ,90 -tetra-cis-lycopene (prolycopene) (Wong et al., 2004; Dong et al., 2007). Finally, either light or carotene isomerase (CRTISO) isomerizes the cis bonds into all trans-lycopene. This enzyme is a secondary point of regulation, as it is epigenetically regulated via methylation (Cazzonelli et al., 2009). Several cyclization reactions result in the production of bicyclic carotenoids. Lycopene-β-cyclase (β-LCY) catalyzes the addition of β-ionone rings. One β-ionone ring leads to the formation of γ-carotene; a second one forms β-carotene. Lycopene-ε-cyclase (ε-LCY) catalyzes addition of ε-ionone rings, forming δ-carotene. Addition of one β-ionone ring and an ε-ionone ring on the other side of the linear backbone results in production of α-carotene. Essentially, the pathway bifurcates after lycopene synthesis into β,β- and ε,β-carotenoids, and the relative activities of β-CLY and ε-CLY determine the proportion of lycopene funneled to the two branches (Cazzonelli and Pogson, 2010). Hydroxylation of

FIGURE 2 | Provitamin A biosynthesis. Enzymes involved in its biosynthesis are marked in blue. Connections to other vitamin pathways are indicated in red. Filled yellow boxes indicate the external influences on the biosynthesis, affected enzymes surrounded by a yellow square. The regulatory influences on DXS are derived from studies on Arabidopsis (Estevez et al., 2001), those on PSY from studies in maize, rice and tomato (Li F.Q. et al., 2008; Welsch et al., 2008). Cofactors are encircled in gray. Abbreviations (in order of appearance in the pathway): G3P, glyceraldehyde-3-phosphate; DXS, 1-deoxy-D-xylulose-5-phosphate synthase; DXP, 1-deoxy-D-xylulose-5-phospate; DXR, DXP reductoisomerase; IPP, isopentenyl diphosphate isomerase; IDI, isopentenyl diphosphate isomerase; DMAPP, dimethylallyl diphosphate; GGPPS, geranylgeranyl diphosphate synthase; GGPP, geranylgeranyl diphosphate; PSY, phytoene synthase; PDS, phytoene desaturase; ZISO, ζ-carotene isomerase; ZDS, ζ-carotene desaturase; CRTISO, carotene isomerase; β-LCY, lycopene-β-cyclase; ε-LCY, lycopene-ε-cyclase; BCH1, β-carotene hydroxylase; ZEP1, zeaxanthin epoxidase; VDE, violaxanthin de-epoxidase; AsA, ascorbate.

α-carotene gives rise to lutein, while hydroxylation of β-carotene leads to formation of zeaxanthin.

### Provitamin A Functions in planta

Oxygenated carotenoid derivatives are termed xanthophylls, whereas the non-oxygenated analogs are designated as carotenes. Distinct functions are attributed to these two classes of carotenoids.

### Enhancing Light Harvesting and Photoprotection

Lipid soluble carotenoids play a major role in photoprotection. The conjugated double bonds in the carbon skeleton function as chromophore, allowing light absorption in the range of 450–570 nm, covering the absorption gap of chlorophyll. Consequently, they function as accessory pigments in photosynthesis, enhancing light harvesting in the blue– green spectral domain (Cogdell and Frank, 1987; Havaux et al., 2004), while also being required for the correct assembly of photosystems (Formaggio et al., 2001).

Xanthophylls are crucial in non-photochemical quenching (NPQ) of excess photon energy by thermal dissipation through molecular vibrations (Demmig-Adams and Adams, 1996). The xanthophyll cycle encompasses two antagonistic enzymes, violaxanthin de-epoxidase (VDE) which converts violaxanthin via antheraxanthin into zeaxanthin, and zeaxanthin epoxidase (ZEP) which performs the reversed reactions. This protective mechanism prevents the over-reduction of photosystem II (PSII) and the generation of ROS (Briantais, 1994). When the level of absorbed light exceeds the photochemical capacity of PSII, the acidification of the thylakoid lumen activates VDE. Additionally, ethylene was found to be a negative regulator of the cycle as it influences the activity and activation of VDE (Chen and Gallie, 2015). Overexpression of β-carotene hydroxylase (BCH1), causing a simultaneous increase in zeaxanthin and xanthophyll levels, enhances tolerance to high light and heat stress (Davison et al., 2002). The extra xanthophyll was shown to be associated with the PSII light-harvesting complexes (LHCII), and the plants exhibited reduced leaf necrosis and lipid peroxidation.

Carotenes are important to mitigate the generation of ROS during photosynthesis. Carotenoids can quench both triplet chlorophyll (3Chl<sup>∗</sup> ) and singlet oxygen (1O2), protecting PSI and PSII from photoinhibition (Edge et al., 1997; Triantaphylides and Havaux, 2009). On the other hand xanthophylls like zeaxanthin are involved in the protection of the photosynthetic membranes against lipid peroxidation (Havaux and Niyogi, 1999; Davison et al., 2002).

### Stress Signaling

Besides their role in photosynthesis, carotenoids perform a function in stress signaling, as stress-imposed singlet oxygen production can lead to a variety of oxidative cleavage carotenoid derivatives, several of which are reactive electrophile species (RES). One example of RES is the volatile β-cyclocitral (β-CC), which is capable of altering <sup>1</sup>O<sup>2</sup> responsive gene expression in relation to stress acclimation (Havaux, 2014). This RESinduced <sup>1</sup>O<sup>2</sup> response could interact with jasmonic acid (JA) signaling and thus compromise the JA-mediated responses to pathogens and herbivores in high light acclimated plants (Ramel et al., 2012). Another example in which carotenoidderived signals are implicated in retrograde signaling resides in the control of chloroplast and leaf development. The albino Arabidopsis (Arabidopsis thaliana) null mutant of ZDS, Arabidopsis zds/chloroplast biogenesis5 (clb5), exhibits abnormal leaf development and cell differentiation with weakened auxin responses. Introduction of the pds3 mutation, compromising ζ-carotene synthesis, rescued the clb5 mutant gene expression and leaf development phenotypes. This suggests that ζ-carotene isomers are implicated in regulating chloroplast biogenesis and leaf development (Avendano-Vazquez et al., 2014).

### Shoot and Root Development

Inhibition of carotenoid production disturbs the rhythmic oscillation of the lateral root (LR) clock, necessary for establishment of pre-branch sites (Van Norman et al., 2014). The same decrease in LR capacity was observed when using an inhibitor of carotenoid cleavage dioxygenases (CCDs), but the carotenoid-derived signaling molecule responsible for the influence on root branching remains to be identified (Van Norman et al., 2014). Earlier mutant analysis has revealed the necessity for other carotenoid derived signals in normal development. The bypass1 (bps1) mutant has short roots, a malfunctioning shoot apical meristem and leaf vasculature with an increasing severity in lower temperatures. Grafting experiments suggested the constitutive presence of a mobile root derived 'bypass' signal, which required β-carotene synthesis, but no CCDs. (Van Norman et al., 2004, 2014; Van Norman and Sieburth, 2007). CAROTENOID CHLOROPLAST REGULATORY1 (CCR1) which encodes a histone methyltransferase Set Domain Group8 (SDG8), defines yet another link of carotenoids to shoot development. SDG8 is important for expression of CRTISO (**Figure 2**). Besides enhanced rosette growth and cauline branching, altered carotenoid content was observed in ccr1 (Cazzonelli et al., 2009).

### Vitamin A in Human Health

### Function and Pathophysiology of Vitamin A Deficiency

During the last decades, knowledge on vitamin A functioning in humans has greatly increased, emphasizing its tremendous clinical importance (Wiseman et al., 2017). Retinol and retinal vitamer forms of vitamin A play a pivotal role in proper function of vision and dark adaptation. Human vision depends on the regeneration of the vitamin A derivative 11-cis-retinal, necessary for the formation of rhodopsin (Tang et al., 2013; Hanneken et al., 2017; Tian et al., 2017). Rhodopsin in turn is required as pigment in the retinal receptor responsible for dark adaptation (Sommer, 2008; Wiseman et al., 2017). This explains why vitamin A deficiency (VAD) can lead to xerophthalmia, a pathophysiological condition of impaired vision, starting with night blindness, and ultimately leading to complete blindness due to corneal damage (Sommer, 2008; Chiu et al., 2016). Furthermore, vitamin A is known to have a beneficial impact on innate and adaptive immunity (Lima et al., 2010; Wiseman et al., 2017). Consequently, VAD induces increased

susceptibility toward a variety of infections, particularly gastrointestinal conditions (Brown and Noelle, 2015). Anemia, the most prevalent of all micronutrient deficiency-induced disorders, has also been linked to VAD, as vitamin A is able to influence iron metabolism (Semba et al., 1992; West et al., 2007). Human reproduction also depends on vitamin A, more particularly retinoic acid, as it is shown to be necessary in spermatogenesis as well as for proper embryo growth (Hogarth and Griswold, 2010; Clagett-Dame and Knutson, 2011; Wiseman et al., 2017). The above is, however, but a selection of the vast impact of vitamin A in all its vitamer entities on basic human physiology. It also illustrates the urgency of cutting back VAD incidence on a global scale.

### Global Burden of Vitamin A Deficiency

Occurrence of VAD can be explained by poor dietary diversification, likely caused by high consumption of staples with low vitamin content (**Table 1**). An estimated 250 million children of preschool age suffer from VAD (Wiseman et al., 2017; World Health Organization [WHO], 2018). Moreover, 250–500 thousand children develop VAD-induced full blindness each year, half of these cases resulting in death within a year (World Health Organization [WHO], 2018). As VAD is known to have a negative impact on the human immune system (Brown and Noelle, 2015), many infection-related deaths could also, at least partially, be attributed to low vitamin A status, indicating that the incidence of VAD-induced mortality is potentially underestimated. UNICEF reported that vitamin A supplementation programs are able to save 350 thousand children lives annually (Dalmiya and Palmer, 2007; Sommer and Vyas, 2012). Despite these efforts, coverage of the supplementation programs remains poor in many regions, explaining the persistent occurrence of VAD in these populations. Though VAD is much more prevalent in low-income countries (Bailey et al., 2015; Wiseman et al., 2017), there is also a great need to enhance (pro)vitamin A uptake on a global scale, given the existence of VAD-induced disorders in the developed world (Chiu et al., 2016).

### Sources of Vitamin A

Provitamin A is present in animal as well as plant derived foods (Bai et al., 2011; Mody, 2017). Meat and dairy products are typically rich in retinyl esters, which can be metabolized to retinol in the human body (Bai et al., 2011). In plant-derived food sources on the other hand, the provitamin A content is represented by carotenoids, β-carotene being the most prevalent (Grune et al., 2010). β-carotene can be converted to retinal by the human β-carotene 15,150–monooxygenase (Lindqvist and Andersson, 2002), which is typically absent in strictly carnivorous mammals. Richly colored fruit and vegetables are good sources of provitamin A. Examples of high provitamin A carotenoid containing crops are carrots, sweet potatoes, pumpkin, kale and spinach (Harrison, 2005). The food matrix within which the vitamin is delivered is also of great importance, as it determines bioavailability, demonstrated by the increasing portion of bioavailable provitamin A in orange juice upon pasteurization (Aschoff et al., 2015). As the provitamin A content of a food source can be the result of a whole array of provitamin A (mostly carotenoids) compounds, the vitamin content is often described as retinol activity equivalent (RAE). The RAE measures the amount of provitamin A expressed as having the same bioactive power of a certain amount of retinal, taking bioavailability into account. The highest recommended daily allowance is 1.3 mg for lactating females. As all major staples, with the exception of plantain (cooking banana, Musa sp.), are poor sources of provitamin A (**Table 1**), there is a strong case for raising its level in those crops (De Moura et al., 2016).

### Biofortification: Toward Higher Provitamin A Levels in Crops Metabolic Engineering

Over the last decades tremendous efforts has been invested in the augmentation of provitamin A levels in different crops (Giuliano, 2017). PSY, responsible for the first committed step of carotenoid biosynthesis, has been pinpointed as rate-limiting step, thereby serving as an ideal candidate gene in biofortification strategies (Fitzpatrick et al., 2012). A well-known example is the genetically engineered Golden Rice (Oryza sativa) (Ye et al., 2000; Beyer et al., 2002; Paine et al., 2005), which has a yellow color, due to its high carotenoid nature. In Golden Rice (Ye et al., 2000), metabolic engineering was achieved via endospermspecific induction of the daffodil (Narcissus pseudonarcissus) PSY and bacterial (Erwinia uredovora) carotene desaturase (CRT), representing the steps in carotenoid synthesis which are naturally not expressed in rice endosperm. The Golden Rice engineering strategy was later improved by replacing the daffodil-derived PSY by a maize ortholog showing a stronger enzymatic activity in rice than the originally used daffodil enzyme and thus leading to higher beta-carotene levels in the so called Golden Rice 2 (GR2) (Paine et al., 2005). The latter rice lines contain up to 3.7 mg/100 g dry weight (DW) carotenoids in the endosperm. GR2 delivers 50% of a child's RDA of provitamin A in 72 g dry rice. On top of its ability to be deployed to minimize VAD, Golden Rice can be considered a solid proof-of-concept, enabling implementation of this metabolic engineering strategy in a range of crops. Indeed, adopting this strategy into Zea mays yielded maize kernels with 6 mg/100 g DW β-carotene (Naqvi et al., 2009), corresponding to a 112-fold increase in total carotenoid content over the WT corn variety used in this study. Also in wheat (Triticum aestivum), this strategy has led to a 10-fold increase in endosperm carotenoid levels, reaching almost 500 µg/100 g dry weight (Cong et al., 2009).

Interestingly, a one-gene metabolic engineering approach, overexpressing only PSY, has also led to several successfully biofortified crops. In canola (rapeseed, Brassica napus), PSY introduction yielded 50-fold increase in seed carotenoid content (Shewmaker et al., 1999). Similarly, carotenoid content of potato (Solanum tuberosum) was elevated (up to 3.5 mg/100 g DW) mostly caused by strongly enhanced β-carotene levels (up to 1.1 mg/100 g DW) (Ducreux et al., 2005). In cassava (Manihot esculenta), root specific ectopic expression of PSY resulted in carotenoid levels to be elevated 20-fold, reaching 2.5 mg/100 g DW (Sayre et al., 2011). Finally, a recent cis-genic PSYoverexpression engineering approach resulted in banana lines

reaching up to 5.5 mg/100g DW β-carotene equivalent content of fruits (Paul et al., 2017).

A different one-gene approach has been applied in tomato fruit engineering (Rosati et al., 2000; Ralley et al., 2016), as this tissue harbors high expression of genes controlling biosynthesis of lycopene, such as the aforementioned PSY. Therefore, a carotenoid biosynthesis gene, downstream of lycopene was a more appropriate choice for biofortification of carotenoid content in tomato fruit (Rosati et al., 2000). The lycopene β-cyclase gene (β-LCY), catalyzing the cyclization of the lycopene molecule by introduction of the β-ionone rings yielding β-carotene (Cunningham et al., 1996) (**Figures 1A,B**), was engineered in tomato fruit, resulting in high β-carotene tomatoes (Rosati et al., 2000; D'Ambrosio et al., 2004; Ralley et al., 2016).

Single gene approaches, despite reaching satisfying levels of provitamin A, could be strengthened by introduction of additional genes, further increasing flux through the biosynthetic pathway. Indeed, further research in canola resulted in seeds with over 1,000-fold increase in β-carotene, reaching over 20 mg/100 g fresh weight (FW) (Fujisawa et al., 2009). This was accomplished by introduction of seven bacterial genes, highlighting the power of multiple gene engineering as well as the applicability of prokaryotic genes (Fujisawa et al., 2009; Bai et al., 2011). Similarly, in potato, combined tuber-specific boosting of PSY, PDS and β-LCY (**Figure 2**) generated 'golden potato' tubers having 11 mg/100 g DW of carotenoids of which 4.7 mg/100 g DW is represented by β-carotene (Diretto et al., 2007).

Another interesting gene in carotenoid biofortification is the gene encoding 1-deoxyxylulose-5-phosphate synthase (DXS). The DXS enzyme acts in the MEP pathway, upstream of IPP formation, in the plastid isoprenoid pathway (Estevez et al., 2001; Sayre et al., 2011; Ruiz-Sola and Rodríguez-Concepción, 2012), thereby acting also upstream of biosynthesis of a whole range of metabolites depending on this pathway, including tocochromanols (see vitamin E). This approach has been adopted in cassava, tomato and Arabidopsis (Estevez et al., 2001; Enfissi et al., 2005; Sayre et al., 2011). The idea of changing carotenoid content via engineering of a further upstream component proves to be applicable, as shown in tomato, as fruit-specific down-regulation of DE-ETIOLATED1 (DET1) [a light signaling pathway controlling gene (Schafer and Bowler, 2002)], leads to enhancement of both carotenoid and flavonoid levels (Davuluri et al., 2005). Fruit-specific RNAi suppression of an epoxycarotenoid deoxygenase (NCED), a key enzyme in abscisic acid (ABA) biosynthesis, resulted in enhanced lycopene and β-carotene levels (Sun et al., 2012). Strikingly, metabolism of different vitamins could be intertwined, potentially positively influencing their accumulation and stability, as was the case with the combined biofortification of vitamin E and carotenoids in 'Golden Sorghum' (Che et al., 2016). This further emphasizes the importance of considering vitamin stability, especially upon long-time storage. In this respect, down-regulation of a lipoxygenase gene (r9-LOX1), known to cause carotenoid oxidation (Wu et al., 1999; Blancquaert et al., 2017) in rice endosperm yielded enhanced provitamin A stability in Golden Rice upon storage (Gayen et al., 2015). Suppressing enzymes involved in vitamin breakdown has also been implemented as a metabolic engineering strategy and successfully demonstrated in wheat. Endosperm-specific stimulation of carotenoid biosynthesis by bacterial phytoene synthase was combined with silencing of carotenoid hydroxylase, leading to kernels accumulating up to 500 µg/100 g DW of β-carotene (Zeng et al., 2015).

These strategies are, however, species and likely tissue-specific, as different crops require adjusted engineering approaches. Assessment of their implementation in different agronomically important crops would be a great leap forward (Kang et al., 2017). In this respect, the ability of processing habits to lower vitamin bioavailability should be taken into consideration (Diaz-Gomez et al., 2017a). Interestingly, interventions in provitamin A metabolism resulted in remarkable alterations in crop properties. This has been reported for provitamin A biofortified cassava, achieved by DXS and CTRb (bacterial phytoene synthase) introduction, resulting in prolonged shelf-life upon storage as well as aberrant carbon partitioning causing a significant reduction in dry matter content (Beyene et al., 2018). This further emphasizes the importance of taking all aspects of plant physiology into consideration, not only upon designing but also upon evaluation of biofortified crops.

### Breeding

Enhancement of provitamin A content in food crops has not been limited to transgenic metabolic engineering approaches, as different breeding projects have also led to successes (Giuliano, 2017; Haskell et al., 2017). Interestingly, studies implementing genome-wide association (GWAS), association analysis and quantitative trait locus (QTL) mapping, pinpoint the factors strongly influencing carotenoid accumulation. Indeed, as maize exhibits a strong natural variation in carotenoid content, germplasm analysis indicated a lycopene cyclase to be the major determinant of the vitamin level (Harjes et al., 2008). QTL analysis of different crops mostly revealed the same genes to be major effectors in carotenoid accumulation, corresponding to those genes also implemented in successful metabolic engineering approaches such as PSY, LCY, and DXS genes (Giuliano, 2017). Analysis of carotenoid variation could also highlight negative regulators, as was the case for the gene encoding BCH1 (Yan et al., 2010). Molecular techniques have enabled breeding of high vitamin yielding crops. Exemplary cases include biofortified corn (up to 1.5 mg/100 g DW of β-carotene) (Muzhingi et al., 2011; Palmer et al., 2018; Zunjare et al., 2018), cassava (800 µg/100 g DW of β-carotene) (Welsch et al., 2010; Ilona et al., 2017) and sweet potato (400 µg/100 g FW of RAE of provitamin A) (Low et al., 2017). The latter is already reaching almost three million households in Sub-Saharan Africa, thanks to the Sweet Potato for Profit and Health Initiative (SPHI), which aims to provide this orange-fleshed sweet potato (OFSP) to 10 million households (Laurie et al., 2018). Unfortunately, satisfactory variation in rice germplasm to support adequate breeding for enhanced provitamin A content of the endosperm, has not been found (De Moura et al., 2016). A nice overview of achievements in provitamin

A biofortified crops is given in a recent review of Giuliano (2017).

## Provitamin A: Major Problems and Future Prospects

The successful creation of provitamin A rich rice, coined Golden Rice, is a good example of a product with great potential, the introduction of which is hampered by regulatory obstructions (Potrykus, 2010, 2017). Indeed, though the potential humanitarian benefit as well as adequate cost-effectiveness of Golden Rice are well known (Stein et al., 2006), current societal perception, strongly following the precautionary principle, has blocked the implementation of Golden Rice for almost two decades. Ingo Potrykus, one of the creators of Golden Rice, has referred to this impediment as 'a crime against humanity' (Potrykus, 2010). The rationale behind this, is the calculated amount of Disability-Adjusted Life Years (DALY) (over a million) as well as deaths (over 40 thousand) that could be saved annually by Golden Rice implementation (Potrykus, 2010; De Steur et al., 2017; Wesseler and Zilberman, 2017). The case of Golden Rice holds an important lesson to minimize regulatory obstructions for products of genetic engineering. Satisfactory proof-of-concepts are often difficult to commercialize due to intellectual and tangible property right (Kowalski et al., 2002). When the ultimate goal of a biofortification endeavor goes beyond the academic proof-of-concept, one must thoroughly examine every patent or intellectual property right attached to it. In the case of the Golden Rice project, all licenses -for the technologies involved- have been acquired, enabling free distribution to farmers, provided that the transgenic event is approved (Potrykus, 2017). This was possible, as it is considered a humanitarian project, allowing to be deployed in developing countries by a Humanitarian Use Technology Transfer (HUTT) license. More strikingly, the Golden Rice event GR2-R1 was found to disrupt the native OsAUX1 (encoding an auxin influx transporter) expression, yielding detrimental consequences for plant growth and development (Bollinedi et al., 2017). This further emphasizes the importance of characterizing the genomic place of insertion and potential influences on growth and development.

Provitamin A is an example of a micronutrient for which major progress has been achieved in biofortification over the last decades (Giuliano, 2017). A substantial part of these accomplishments has been realized via breeding endeavors (Bouis and Saltzman, 2017; Ilona et al., 2017), without the use of genetic engineering and therefore more readily accepted for commercial release (Potrykus, 2017). Focus should now be directed toward proper information of the public on allowing provitamin A rich crops created via GM-technology, so that these can be deployed to decrease VAD in populations which are in need. Moreover, the case of tomato fruit, which naturally contains sufficient lycopene, thus requiring a downstream metabolic engineering intervention to redirect the biosynthetic pathway, is a nice example on how general knowledge of a food crop steers biofortification approaches. Therefore, acquiring a general metabolic engineering strategy is difficult and future research should first be directed to understanding provitamin A biosynthesis within the target crop tissue as well as natural variation in the germplasm thereof. The latter could put breeding strategies forward as a valuable solution to fight VAD. Finally, given the success of breeding strategies in provitamin A biofortification and the natural variation of sexually compatible germplasm they depend on, expanding the available germplasm of a certain crop could have very beneficial impacts.

## VITAMIN C – ASCORBATE

Ascorbate or L-ascorbic acid (AsA), referred to as vitamin C, is a potent water-soluble antioxidant (Iqbal et al., 2004; Macknight et al., 2017). This molecule is, however, unstable, as it easily deteriorates, being sensitive to heat, alkaline environments and oxygen (Iqbal et al., 2004). Vitamin C sensu lato includes all molecules (vitamers) which can be metabolized to form ascorbic acid in human metabolism, including dehydroascorbic acid (Wilson, 2002). Ascorbic acid is a weak sugar acid, related to, and derived from, hexoses (Pohanka et al., 2012).

### Vitamin C Biosynthesis

The sole physiologically significant source of AsA is provided via the Smirnoff-Wheeler pathway, following a route via Dmannose (D-Man) and L-galactose (L-Gal), essentially taking place in the cytosol, with the exception of the final mitochondrial step generating L-AsA (**Figure 3**) (Ishikawa et al., 2008). Therefore, hexoses need to be directed into D-Man metabolism by phosphomannose isomerase (PMI), followed by the conversion of D-Man-6-P into D-Man-1-P by phosphomannomutase (PMM) (Qian et al., 2007; Maruta et al., 2008). The reversible phosphorylation of D-mannose-1-phosphate (D-Man-1-P) by GDP-D-mannose pyrophosphorylase (GMP/VTC1) results in GDP-D-Man, which is subsequently equilibrated with its epimer, GDP-L-galactose (GDP-L-Gal), through GDP-D-mannose-3,5 epimerase (GME) (Wolucka and Van Montagu, 2007). However, this enzyme can also produce GDP-L-gulose, which occurs in 25% of the epimerization events. This leads to the alternative biosynthesis route, named the L-gulose pathway, which might be species or tissue specific. GDP-Gal is converted to L-galactose-1-Phosphate (L-Gal-1-P) by GDP-L-Gal phosphorylase/L-Gal guanylyltransferase (GGP/VTC2), the first committed and ratelimiting step in the vitamin C biosynthesis pathway (Linster and Clarke, 2008). Both transcription and activity of GGP appear light-regulated, explaining the increase in ascorbate levels in high light conditions. Furthermore, as their diurnal pattern of expression was also observed in constant darkness, GGP is assumed to be under circadian clock control (Dowdle et al., 2007) (Ishikawa et al., 2018). Additionally, VTC2 is suggested to be controlled by a cis-acting upstream open reading frame in high ascorbate conditions (Laing et al., 2015). Several other enzymes are also feedback-inhibited by AsA, including PMI (Maruta et al., 2008), GME (Wolucka and Van Montagu, 2003) and LGalDH (Mieda et al., 2004). In the subsequent step in ascorbate biosynthesis, L-Gal-1-P is hydrolyzed to L-galactose (L-Gal) by L-Gal-phosphate phosphatase (GPP/VTC4) (Conklin et al., 2006),

followed by an NAD-dependent oxidation into L-galactono-1,4 lactone (GalL) by L-galactose dehydrogenase (L-GalDH). The last step is yet another oxidation, exerted in mitochondria by the flavin containing L-galactono-1,4-lactone dehydrogenase (L-GalLDH), forming AsA which uses cytochrome c as an electron acceptor (Wheeler et al., 1998; Leferink et al., 2008). This enzyme also shows a diurnal expression pattern (Tamaoki et al., 2003). In the case of the alternative gulose pathway, L-gulono-1,4-lactone is formed, and further converted into AsA by L-GulL dehydrogenase (Wolucka and Van Montagu, 2003). Both GalL and AsA, being low molecular-weight solutes, might cross the outer membrane without the need of a carrier.

### Vitamin C Functions in planta

The physiologically active form of vitamin C is its anionic form, ascorbate. The water soluble ascorbate anion (AH−) is a universal player in both enzymatic and non-enzymatic antioxidant defense systems and therefore implicated in a range of processes in plants. Its efficiency as an antioxidant most probably relies on the (relative) stability of its primary oxidation product, the monodehydroascorbate radical (MDA) and moreover, on its capacity to terminate radical chain reactions by spontaneously disproportionating into the non-toxic, nonradical product AsA and dehydroascorbate (DHA) (Noshi et al., 2016).

### Antioxidant

AsA is of great importance during photosynthesis, firstly because it is capable to donate electrons to PSI and PSII in both normal and stress conditions (Mano et al., 2004; Ivanov, 2014). Moreover it eliminates directly superoxide (O<sup>−</sup> 2), hydroxyl radicals (•OH) and singlet oxygen (1O2) coming from photoreduction and photorespiration and aids in the scavenging of hydrogen peroxide being a cofactor of ascorbate peroxidase (APX) in the Asada-Halliwel pathway or Mehler-peroxidase pathway (Foyer and Halliwell, 1976; Shigeoka et al., 2002). The latter pathway, also known as the ascorbate-glutathion cycle (ASC-GSH cycle), involves APX, monodehydroascorbate reductase (MDHAR), dehydroascorbate reductase (DHAR) and glutathione reductase (GR) and is of uttermost importance in the antioxidant defense of plants (Foyer and Halliwell, 1976). Despite the multiple scavenging processes present in plants, lipids still receive the burden of oxidative stress leading to the generation of lipid peroxyl radicals. Clearing thereof is accomplished by α-tocopherols (see vitamin E), which in turn are recycled through the oxidative action of AsA (Davey et al., 2000). In addition, ascorbate, being the cofactor of violaxanthin de-epoxidase (VDE), plays a role in the xanthophyll cycle, as mentioned above in the section of vitamin A, protecting PSII from photoinhibition (Eskling et al., 1997).

### Development

A wide range of hormone-AsA interactions influence plant physiology. First, AsA is involved as a cofactor of GA3-oxidase and ACC-oxidase in the biosynthesis of gibberellin (GA) and ethylene, respectively (Arrigoni and De Tullio, 2000; Van de Poel and Van Der Straeten, 2014). Second, hormones can also control AsA biosynthesis. In seed tissue, enhanced levels of ABA suppresses activity of NADPH oxidases, the main producers of ROS in seeds (Ishibashi et al., 2017). The resulting decrease of ROS in the aleuron layers inhibits AsA and concomitant GA biosynthesis (Ye et al., 2012). On the other hand, ROS, and more specifically exogenous H2O2, were shown to enhance expression in imbibed seeds of biosynthesis genes of GA, an essential hormone in seed germination (GA20ox1, GA20ox2, GA20ox3, GA3ox1, and GA3ox2) (Liu et al., 2010; Ye et al., 2012).

Furthermore, AsA was shown to be implicated in sustaining seedling growth. Simultaneous loss of function of two homologs (vtc2-1 and vtc5-1 or vtc5-2) encoding the biosynthesis enzyme GGP, results in growth inhibition after cotyledon expansion, followed by bleaching. In later stages of development, ascorbate is required for growth, as the older leaves of the rescued double mutants started to bleach again when transferred back to L-Gal-free medium, the immediate downstream product of these isoforms. Moreover, growth reduction was already observed in the vtc2 null mutant, in accordance with its low ascorbate level (20%) as compared to wild-type (Dowdle et al., 2007). AsA is also linked with cell expansion and division. Culture experiments showed an increase in ascorbate levels during cell elongation in tobacco, while addition of an ascorbate biosynthesis inhibitor (lycorin) induced cell cycle arrest in G1 in onion root cells (Liso et al., 1984; Kato and Esaka, 1999). This link could partially be attributed to its function as a cofactor of prolyl hydroxylase which converts proline residues in hydroxyprolinerich glycoproteins such as extensins in the cell wall (Fry, 1986; Kerk and Feldman, 1995; De Tullio et al., 1999; Joo et al., 2001; Sanmartin et al., 2007). Moreover, the observation of a depleted level of ascorbate together with an increased activity of ascorbate oxidase (AOX) in the quiescent center (QC) in maize roots are suggestive for a role of ascorbate in the maintenance of QC identity. The concomitant augmented auxin level revealed a regulatory role of the latter on AOX expression (Kerk and Feldman, 1995). Moreover, shoot apical dominance is stimulated by ascorbate (Barth et al., 2006; Kotchoni et al., 2009; Zhang C.J. et al., 2011). Through control of GA and ABA, AsA is also involved in flowering, programmed cell death and senescence (Barth et al., 2006; Kotchoni et al., 2009). AsA and ABA were also shown to influence the expression of senescence associated genes (SAGs) in an antagonistic way (Barth et al., 2006).

Finally, fruit ripening is also related to AsA (Sanmartin et al., 2007). Ascorbate aids in fruit ripening by its counterintuitive site-specific pro-oxidant function. This involves the apoplastic conversion of O<sup>2</sup> and Cu2<sup>+</sup> into H2O<sup>2</sup> and Cu+, which thereupon combine to generate OH radicals. The presence of the latter results in polysaccharide degradation causing fruit softening (Fry, 1998). In addition, AsA is involved in ethylene biosynthesis (see above), which is essential to induce ripening, and in turn, induces AsA biosynthesis via upregulation of VTC4 expression (**Figure 3**) (Ioannidi et al., 2009).

FIGURE 3 | Biosynthesis of vitamin C. The enzymes committed to vitamin C (VTC) biosynthesis are marked in blue. Feedback regulations are illustrated in purple. Filled yellow boxes indicate the external influences on the biosynthesis, regulating enzymes surrounded by a yellow square. Biosynthesis and salvage links to vitamins A and E, in dark red, are indicated with a double and a dashed arrow, respectively. The dashed oval arrow represents the recycling of α-tocopherol, within which ascorbate aids in the detoxification of tocopheroxyl radicals. Abbreviations (in order of appearance in the pathway): PMI, phosphomannose isomerase; PMM, phosphomannose mutase; VTC1/GMP, GDP-D-mannose pyrophosphorylase; GDP-D-mannose, guanosine diphosphate mannose; GME, GDP-mannose-3<sup>0</sup> ,50 -epimerase; VTC2/GPP, GDP-L-galactose-phosphorylase/L-galactose guanylyltransferase; VTC4/GPP, L-galactose 1-phosphate phosphatase; L-GalDH, L-galactose dehydrogenase; L-GalLDH, L-galactono-1,4-lactone dehydrogenase; L-GulLDH, L-gulono-1,4-lactone dehydrogenase.

### Vitamin C in Human Health

fpls-09-01862 December 15, 2018 Time: 15:9 # 12

### Functions and Pathophysiology of the Deficiency

In human physiology, ascorbate functions as an important scavenger of ROS, such as hydrogen peroxide (Lobo et al., 2010). Importantly, ascorbate is also required as a reducing agent in the conversion of iron from ferric (Fe3+) to ferrous (Fe2+) oxidation state, thereby aiding in sufficient iron uptake and thus indirectly linked to anemia in case of deficiency (Iqbal et al., 2004; Macknight et al., 2017). Furthermore, vitamin C assists in the metabolism of tryptophan, tyrosine and folate (Iqbal et al., 2004). Moreover, AsA aids in lowering excess cholesterol levels, thereby reducing atherosclerosis (Das et al., 2006; Chambial et al., 2013). This vitamin is also known to function as a cofactor in several reactions such as hydroxylation of muscle carnitine, amidation of several hormones, and the conversion of the neurotransmitter dopamine into norepinephrine (Chambial et al., 2013). Hence, the function of ascorbate is evidently linked to energy metabolism. In collagen biosynthesis, prolyl and lysyl hydroxylases utilize AsA as a enzymatic cofactor (Myllylä et al., 1984; Pimentel, 2003). This explains the pathogenesis of scurvy, a vicious disease, caused by severe vitamin C deficiency, characterized by bleeding gums and eventually leading to edema, jaundice, hemolysis, spontaneous bleeding, neuropathy and death (Leger, 2008). Strikingly, there have been indications that ascorbate supplementation could have a negative impact on tumor development (Cha et al., 2013; Mastrangelo et al., 2018). Moreover, high vitamin C status could prevent or cure several infections (Carr and Maggini, 2017). Evidence indicates that low vitamin C status, though not immediately depicting clinical symptoms, hampers ideal human functioning, as increasing vitamin C uptake is known to be beneficial (Johnston et al., 2006, 2014).

### Prevalence of Vitamin C Deficiency

Incidence of Vitamin C deficiency is difficult to quantify, as clear deficiency-induced disorders only occur upon very severe ascorbate shortage. Furthermore, there is no consensus on ideal vitamin C intake quantities (Frei et al., 2012; Hickey et al., 2014). Indeed, retrieving an ideal recommended daily intake for vitamin C has been a heavily debated issue, even tackled by Nobel Prize winner Linus Pauling (Pauling, 1974). However, it remains undeniable that increasing the vitamin C status would exhibit positive effects on general human health (Macknight et al., 2017). There is, however, no controversy about the presence of vitamin C deficiency in the general public, despite the infrequency of scurvy. Vitamin C status was reported as being deficient in about 20% of the low-income population of the United Kingdom (Mosdol et al., 2008). Comparable results were obtained by analysis of the north-American population, where smoking and low socio-economic status were identified as risk factors for vitamin C deficiency (Cahill et al., 2009; Schleicher et al., 2009).

### Vitamin C Sources

Most animals are capable of de novo ascorbic acid biosynthesis, given its vital role in their metabolism. However, humans (but also guinea pigs and bats) have lost this privilege due to mutation in the L-gulono-γ-lactone oxidase (GLO) gene (cf. L-GulLDH in **Figure 3**) (Nishikimi et al., 1988; Imai et al., 1998), the evolutionary reason of which has been questioned (De Tullio, 2010). This leaves humans dependent on sufficient dietary ascorbate intake to preserve vital functioning. Fresh (citrus) fruits, tomatoes, broccoli and leafy vegetables are considered excellent sources of vitamin C (Iqbal et al., 2004; Chambial et al., 2013). Unfortunately, ascorbate is prone to deteriorate upon storage or processing, as its content declines upon exposure to heat and oxygen (Lee and Kader, 2000). Vitamin C losses during storage can be decreased via limited exposure to heat and oxygen (Lee and Kader, 2000; Sapei and Hwa, 2014). Though most staples are poor sources of vitamin C, potato and cassava do supply a significant amount of the vitamin to the populations relying on these crop products (**Table 1**). However, elevating the levels of vitamin C in these crops could deliver additional health advantages.

### Vitamin C Biofortification Metabolic Engineering

Metabolic engineering strategies, aimed at elevating ascorbate levels in a specific crop/tissue, have been deployed by increasing either ascorbate biosynthesis, salvage or altered pathway regulation (Macknight et al., 2017). Interestingly, these approaches possess the ability to increase tolerance to abiotic stresses such as drought, salinity, cold, heat and high light. In ascorbate biosynthesis, the conversion of GDP-L-galactose to L-galactose-1-P, the central step in plant ascorbic acid biosynthesis, carried out by the GDP-L-galactose phosphorylase (GGP, VTC2) enzyme (see **Figure 3**), is mainly considered as being rate-limiting, thereby a prime target for metabolic engineering approaches (Bulley and Laing, 2016; Macknight et al., 2017). This has been adequately demonstrated in tomato and potato, where introduction of the kiwi and potato GGP gene, respectively, yielded an ascorbate increase up to sixfold in tomato fruit and threefold in potato tubers (Bulley et al., 2012). Though other steps in ascorbate biosynthesis have been evaluated in metabolic engineering, GGP remains the most successful (Macknight et al., 2017). Ascorbate salvage on the other hand, the retrieval of ascorbic acid from the oxidized dehydroascorbic acid vitamer, has been tackled using dehydroascorbate reductase (DHAR) (Li et al., 2012). Similarly, ascorbate degradation has been engineered via RNAi-mediated downregulation of AOX in tomato fruit, resulting in augmented vitamin C levels (Zhang Y.Y. et al., 2011). Furthermore, the Arabidopsis ethylene response factor AtERF98, positively regulating ascorbate biosynthesis, has been implemented in metabolic engineering attempts, as its overexpression in Arabidopsis resulted in enhanced ascorbate levels concomitant with increased salt tolerance (Zhang et al., 2012). This should, however, be approached with caution, as the impact on other aspects of plant metabolism/physiology requires in-depth knowledge of the affected metabolic pathways (Macknight et al., 2017). Moreover, AsA stability should be considered upon evaluation of metabolic engineering strategies. Indeed, after 8 months storage, a drop of vitamin C levels of almost 90% was demonstrated in pasteurized pink guava nectar juice (Psidium

Strobbe et al. Vitamins A, C, E Enhancement in Crops

guajava L.) (Ordonez-Santos and Vazquez-Riascos, 2010). Thus, a metabolic engineering approach combining multiple aspects of ascorbate metabolism including as biosynthesis, recycling, stability and potentially regulation, might prove to yield higher but also stable vitamin C augmentation.

### Breeding

Given the relatively low increase in ascorbate levels upon metabolic engineering approaches, breeding methods might catch up with these interventions. In pepper (Capsicum annuum), which can be considered a rich source of vitamin C, a 2.5 fold variation was observed within the 7 genotypes examined (Geleta and Labuschagne, 2006). The high heritability of this trait indicates a great potential in breeding programs in vitamin C biofortification of pepper. In tomato, transcriptomic analysis of an introgression line exhibiting 4-fold difference in fruit AsA content, pinpointed pectine degradation (particularly pectinesterases) as an important determinant for vitamin C accumulation (Di Matteo et al., 2010; Ruggieri et al., 2015). By QTL mapping of introgression lines, tomato fruit ascorbate levels were also linked to a single nucleotide polymorphism (SNP) near the MDHAR genomic region (Sauvage et al., 2014; Bulley and Laing, 2016). Subsequently, analysis of a high ascorbate/carotenoid introgression line enabled identification of an L-ASCORBATE OXIDASE allele (AOX) as a determinant for AsA levels, the expression of which negatively correlated with vitamin C content (Calafiore et al., 2016). Interestingly, the same study identified an NCED allele, to indirectly control AsA accumulation. In apple, a sixfold variation in AsA content found over 28 commercial varieties allowed creation of a mapping population, pinpointing GGP alleles as major determinants of fruit vitamin C content (Mellidou et al., 2012). Together, these findings illustrate the vast potential of screening crop germplasms for high vitamin C accumulating varieties, and implement these plants in GWAS and breeding programs.

### Ascorbate: Major Problems and Future Prospects

Given its antioxidant nature and a diversity of potential roles, pathophysiological manifestations are not easily attributable to ascorbate deficiency. This is likely the main cause for the dissent on the ascorbate RDA value, which in turn provokes an underestimation of vitamin C deficiency. Therefore, there is a great need to further underline the tremendous health benefit of improving ascorbate status on a global scale, despite the absence of typical deficiency symptoms. As inherent ascorbate levels in wheat and rice endosperm are negligible (**Table 1**), metabolic engineering strategies in these tissues might be challenging. However, ascorbate metabolic engineering strategies could be fruitful in helping these crops cope with abiotic stresses. Moreover, metabolic engineering has the potential to convert potato into an ideal medium to deliver sufficient quantities of a potent water-soluble antioxidant, ascorbate, to the population. Future biofortification strategies on the other hand, should, based on the available knowledge on ascorbate function in plant physiology, try to exploit ascorbate accumulation to enable creation of

nutritionally enhanced crops with concomitant increased stress tolerance.

## VITAMIN E – TOCOCHROMANOLS

Vitamin E or tocochromanols, which includes tocopherols and tocotrienols, are fat-soluble, amphipathic molecules (Colombo, 2010). These molecules consist of a lipophilic isoprenoid chain carrying a polar chromanol ring, providing their amphipathic nature (**Figure 4**). The molecular structure of these vitamers contains three chiral centers, resulting in 8 stereoisomers of each vitamin E entity (**Figure 4**). Depending on the substituents on the chromanol ring, both tocochromanols groups exist as α-, β-, γ-, and δ-isomers. Vitamin E molecules are known as potent antioxidants, as they are free radical scavengers, of which α-vitamers are most powerful (Niki and Traber, 2012).

### Vitamin E Biosynthesis

Tocochromanols are synthesized only in the plastids of photosynthetic organisms. While tocopherols are present throughout the plant, tocotrienol is found almost exclusively in seeds and fruits. Both groups and their isoforms occur in different tissues and exert different functions. α-tocopherol resides mainly in the leaves of vascular plants, while γ-tocopherol is the predominant form in seeds (Grusak and DellaPenna, 1999; Abbasi et al., 2007). Indeed, as seen in Arabidopsis, seeds typically exhibit a more pronounced γ-tocopherol contribution to the total tocopherol pool (Gilliland et al., 2006). The precursors of tocochromanols are derived from two different pathways, the shikimate and the MEP pathway, which are also delivering the precursors of the plastidial biosynthesis of folate (B9) and

carotenoids (provitamin A), respectively (Mène-Saffrané and Pellaud, 2017).

The polar phenolic p-hydroxyphenylpyruvic acid (HPP), synthesized from tyrosine by tyrosine aminotransferase (TAT) and therefore the shikimate pathway (**Figure 5**), is used to produce the aromatic ring of the tocochromanols (**Figure 4**). HPP dioxygenase (HPPD) catalyzes the onset of the actual tocochromanol biosynthesis by converting HPP into homogentisic acid (HGA) after which the pathway bifurcates toward the production of tocopherols and tocotrienols through condensation of two different metabolites bearing the polyprenyl chains (**Figure 5**).

The MEP pathway delivers the precursors for the biosynthesis of prenyl side chains of tocochromanols, as described for provitamin A biosynthesis (see section "Provitamin A Functions in planta"). This branch of tocochromanol biosynthesis utilizes GGPP (geranylgeranyl diphosphate). Interestingly, this product serves as a substrate of multiple enzymes in biosynthesis of different metabolites, including carotenoids, gibberellins, and plastoquinones (Ruiz-Sola et al., 2016). Reduction of GGPP by geranylgeranyldiphosphate reductase (GGDR) yields phytyl diphosphate (PPP) (Gramegna et al., 2018). In the absence of light, PIF3 physically interacts with the promoter of GGDR, down-regulating its expression. Light activation of phytochromes prevents that interaction, leading to transcriptional derepression of the GGDR promotor. The resulting product of GGDR activity, PPP, can be utilized for both tocopherol and chlorophyll biosynthesis (Tanaka et al., 1999). Moreover, PPP is recycled from chlorophyll breakdown, by phytol kinase (VTE5) and phytolphosphate kinase (VTE6) (Vom Dorp et al., 2015). This was revealed by feeding studies in Arabidopsis which demonstrated the incorporation of labeled phytol in tocopherols in seedlings (Ischebeck et al., 2006). Notably, in ripening fruit tissues, often an important source of tocochromanols, recycling of phytol from chlorophyll breakdown is witnessed to be the predominant PPP source (Gramegna et al., 2018).

Condensation of PPP and HGA by HGA phytyl transferase (HPT/VTE2) leads to the formation of 2-methyl-6-phytylbenzoquinol (MPBQ), a step toward creation of tocopherols (Sattler et al., 2004). On the other hand, HGA geranylgeranyl transferase (HGGT) catalyzes the condensation of GGPP with HGA, yielding 6-geranylgeranyl-benzoquinol (MGGBQ), leading toward the formation of tocotrienols (Cahoon et al., 2003; Mène-Saffrané and Pellaud, 2017). These two benzoquinol products, MPBQ and MGGBQ, resulting from HGGT and MGGBQ action, giving rise to tocopherols and tocotrienols, respectively, mark the branch point of tocopherol/tocotrienol biosynthesis. This is illustrated by higher accumulation of tocotrienols in HGGT-overexpressing barley (Hordeum vulgare) lines, depicting decreased tocopherol levels and therefore relatively unaltered total tocochromanol levels (Chen et al., 2017). Downstream reactions follow a similar pattern for both tocopherols and tocotrienols, as the catalysis is performed by shared enzymes. Cyclization of MPBQ and MGGBQ results in δ-tocochromanols (δ-tocopherol and δ-tocotrienol, respectively), a reaction which is executed by tocopherol cyclase (TC, VTE1) (Porfirova et al., 2002; Semchuk et al., 2009). However, MPBQ and MGGBQ can take a different route by methyltransferase reactions (MPBQMT, VTE3), resulting in the formation of 2,3-dimethyl-6-phytyl-1,4-benzoquinone (DMPBQ) and 2,3-dimethyl-6-geranylgeranyl-1,4-benzoquinone (DMGGBQ) (Cheng et al., 2003). Cyclization of these products by the aforementioned TC results in the formation of γ-tocochromanols. These γ-tocochromanols and δ-tocochromanols can thereafter be methylated by γ-tocopherol methyltransferase (γ-TMT,VTE4) to α-tocochromanols and β-tocochromanols, respectively (Bergmuller et al., 2003).

### Vitamin E Functions in planta Scavenger of Lipid Peroxyl Radicals

The most important role of vitamin E in vivo is the termination of a chain reaction of polyunsaturated fatty acid (PUFA) free radicals generated by lipid oxidation. Hence, they play a vital role in scavenging lipid peroxyl radicals during germination and early seedling growth. The detrimental decrease in germination potential of TC mutants (vte1-1) show they are indispensable to preserve the viability of seeds during seed quiescence, which might explain the elevated level of γ-tocopherol in seeds (Sattler et al., 2004). The upstream biosynthesis mutant vte2, which lacks the intermediary DMPBQ, displays difficulties in early seedling development attributable to a decrease in both synthesis and catabolism of lipids as well as an increase in lipid oxidation (Sattler et al., 2004). The few vte2 plants that survive up to the adult stage display no phenotypical differences from wild type which is explained by a predominant need for tocopherols during early development when essential carbon is recruited from lipid catabolism and gluconeogenesis. At later stages, other antioxidants can mitigate the deficiency of tocopherol-mediated ROS scavenging. Hence, tocopherols and its precursors are important to attenuate lipid peroxidation at specific developmental or stress-related periods (Sattler et al., 2006).

### Antioxidant, Photoprotectant, and Stress Signaling

The antioxidant function of tocopherols is supported by the ascorbate-glutathione cycle which recycles tocopheroxyl radicals produced during the reaction of tocopherols with lipid peroxyl radicals. Moreover, tocochromanols are, albeit with a lower rate constant than carotenoids, quenchers of singlet oxygen (1O2) (Kaiser et al., 1990). Up to 120 molecules of <sup>1</sup>O<sup>2</sup> can be neutralized by one molecule of α-tocopherol through resonance energy transfer (Fahrenholtz et al., 1974). Related to their scavenging capability, tocochromanols have strong photoprotective properties. When exposing the alga Chlamydomonas to high light, the inhibition of HPP-dioxygenase led to decreased levels of α-tocopherol and concomitantly, to the inactivation of PSII (Trebst et al., 2002). Addition of synthetic, cell-wall permeable, short-chain tocopherol derivatives could partly restore photosynthesis, hence tocopherols are implicated in the maintenance of PSII function, supplemental to the photoprotective function of NPQ (Trebst et al., 2002; Havaux et al., 2005; Kruk et al., 2005). Thus, tocochromanols together with carotenoids and zeaxanthin are the major protectors of PSII

vitamins, in dark red, are indicated with a double and a dashed arrow, respectively. Filled yellow boxes indicate the inducers, regulating enzymes surrounded by a yellow square. Tyr, tyrosine; TAT, tyrosine aminotransferase; HPP, p-hydroxyphenylpyruvic acid; HPPD, HPP dioxygenase; PDS1, PHYTOENE DESATURATION1; HGA, homogentisic acid; GGDR, geranylgeranyl diphosphate reductase; HGGT, geranylgeranyl transferase; HPT, homogentisate phytyltransferase; VTE, VITAMIN E PATHWAY gene (1–6); GGPP, geranylgeranyl pyrophosphate; PPP, phytyl pyrophosphate; MGGBQ, 6-geranylgeranyl-benzoquinol; MPBQ, 2-methyl-6-phytyl-1,4-benzoquinone; MPBQMT, MPBQ methyltransferase; DMGGBQ, 2,3-dimethyl-6-geranylgeranyl-1,4-benzoquinone; DMPBQ, 2,3-dimethyl-6-phytyl-1,4-benzoquinone; TC, tocopherol cyclase; γ-TMT, γ-tocopherol methyltransferase; α-,β-,γ-,δ-toc, α-,β-,γ-,δ-tocopherol; α-,β-,γ-,δ-T3, α-,β-,γ-,δ-tocotrienol; SA, salicylic acid; ABA, abscisic acid.

against photoinhibition, as they control D1 protein degradation by scavenging singlet oxygen molecules in PSII, and they also protect the whole thylakoid membrane against photooxidative stress, by controlling lipid peroxidation (Trebst et al., 2004). In young leaves of a carotenoid mutant devoid of zeaxanthin, high light stress induced accumulation of tocopherols, conferring tolerance to the mutant, suggesting overlapping functions for these antioxidants (Havaux et al., 2000; Golan et al., 2006). Recently, it was found that an oxidation product, tocopherol quinone, can function as an indicator of oxidative stress, transforming into a signal for programmed cell death upon severe stress. Herewith, the plant protects itself from propagation of stress from the infection point (Li Y. et al., 2008). Moreover, defense-related genes were expressed at higher levels in vte2 plants in response to an increase in peroxidized lipids, suggesting that tocopherol plays a role in gene regulation and modulation of defense responses (Sattler et al., 2006). In this respect, α-tocopherol was found to be important in the mitigation of

salt and heavy metal stresses (Jin and Daniell, 2014). In rice, expression of the VTE1 gene was induced by high salt, H2O2, drought and cold, while overexpression led to increased tolerance to salt stress (Ouyang et al., 2011). Conversely, tocopherol deficient Arabidopsis mutants displayed similar phenotypes as wild types under most stress conditions (high light, salinity and drought) applied (Maeda et al., 2006). Hence, in case of tocopherol shortage, other antioxidants can take over its role in stress, yet, vitamin E is an additive value in harsh conditions.

### Membrane Fluidity and Phloem Transport

Besides their role as lipophilic antioxidant, tocochromanols also act as important structure-stabilizing agents of membranes (Wang and Quinn, 1999). Their concentration in the chloroplast is most probably tightly regulated as a low concentration of α-tocopherol, comparable with the physiological plastidial concentration, seemed to have an important effect on membrane stability during freezing (Hincha, 2008). On that account tocopherols help, together with other components, to maintain the fluidity and thus the function of photosynthetic membranes.

Furthermore, tocopherols have been suggested to play a role in the regulation of photoassimilate export and thus be involved in carbohydrate metabolism, source-sink relationships and growth (Sattler et al., 2003; Hofius et al., 2004). In that respect, a tocopherol cyclase mutant of maize sucrose export defective1 (sxd1) suggested the link between the tocopherol pathway and carbohydrate metabolism as it accumulated carbohydrates in leaves (Russin et al., 1996). The same was observed in StSXD1 RNA interference knockdown lines in potato, but surprisingly not in the vte1 mutant in Arabidopsis, suggesting species-specific differences to tocopherol reduction or a possible additional role of tocopherol in signal transduction (Sattler et al., 2003; Hofius et al., 2004; Li Y. et al., 2008). The biosynthesis mutants vte2 and, to a lesser extent, vte1 revealed inhibition of photoassimilated carbon transport at low temperatures and thus indicated a crucial role of tocopherol in low-temperature adaptation. Cold, non-freezing conditions resulted in a dramatic growth reduction and seed production in these mutants due to structural changes in the phloem parenchyma transfer cells induced by callose deposition and thus leading to reduced photoassimilate export. Lipid peroxidation and photoinhibition were not intensified in vte2, leading to the conclusion that vitamin E function in phloem transport might be more important than its photoprotective role. Apparently the intermediate redoxactive DMPBQ can compensate for the absence of tocopherols as the phenotype of vte1 is not as pronounced as of vte2 (Maeda et al., 2006).

### Vitamin E in Human Health Function and Onset of Deficiency

As antioxidants, the different E-vitamers play an important role in neutralizing ROS and inhibiting membrane peroxidation, very much like they do in plants. Due to their amphipathic character, they reside in the membranes, where they perform their peroxyl scavenging function (Brigelius-Flohe, 2009). The main role of these vitamers is to maintain the integrity of long-chain polyunsaturated fatty acids, thereby ensuring their bioactivity (Traber and Atkinson, 2007). Vitamin E deficiency can induce changes in phospholipid composition of membranes, possibly leading to reduced fertility (Infante, 1999). Indeed, vitamin E, together with the micronutrient selenium, has been suggested to serve as a supplement to treat male infertility (Keskes-Ammar et al., 2003). Though tocopherols, predominantly α-tocopherols, are present at higher levels in the human body, significance of tocotrienols should not be neglected (Sen et al., 2006; Colombo, 2010). Indeed, tocotrienols have shown to be effective in inhibiting proliferation of cancers (Aggarwal et al., 2010; Kannappan et al., 2012), albeit that the ability to impede tumorigenesis also has been documented for tocopherols (Li et al., 2011). Vitamin E is also known to have a positive effect on human health by negatively influencing the occurrence of atherosclerosis and cardiovascular diseases (Mathur et al., 2015). Furthermore, vitamin E, α-tocopherol in this case, was shown to delay the development of Alzheimer's disease in patients (Dysken et al., 2014; La Fata et al., 2014). Indeed, vitamin E deficiency aggravates or even induces neurodegenerative disorders (Berman and Brodaty, 2004; Wysota et al., 2017). Hence, vitamin E has been proposed as a therapeutic agent for Alzheimer's disease (Ibrahim et al., 2017). Vitamin E deficiency can impair cognitive functioning, particularly in elderly people (Ortega et al., 2002), which could be explained by aberrant brain energy metabolism, also known to be associated with thiamin deficiency (Sang et al., 2018; Strobbe and Van Der Straeten, 2018) and phospholipid composition (McDougall et al., 2017).

### Global Vitamin E Status

Vitamin E deficiency, though not often identified as the causative agent of pathophysiological disorders, is known to be highly prevalent in different populations. Strikingly, a vast majority of the US population is characterized by insufficient intake of dietary α-tocopherol (Maras et al., 2004), the predominant dietary source of vitamin E (Chun et al., 2006). Assessment of vitamin E intake in the French and Italian population, indicated a significant prevalence of suboptimal vitamin E levels (Polito et al., 2005). Interestingly, vitamin E status of the Italian population appeared superior compared to the French, which could be attributed to the typical dietary habits in the Italian culture (see below). More recently, approximately onefourth of the Korean population (in the Seoul metropolitan area) was found to be vitamin E deficient, based on plasma α-tocopherol levels (Kim and Cho, 2015). Furthermore, analysis of blood α-tocopherol levels, confirmed the presence of vitamin E deficiency in many developing countries (Dror and Allen, 2011).

### Sources of Vitamin E

Good plant-based sources of dietary (bioactive) vitamin E, in some cases interpreted as supply of α-tocopherol, are fat and oily products such as dried nuts, seeds and almonds (Maras et al., 2004). Tomatoes, avocadoes, spinach, and olives deliver a significant portion of vitamin E (Chun et al., 2006). Though vegetables are generally not a good source of vitamin E (α-tocopherol), soybean and dark leafy greens do exhibit relatively high tocochromanol content. This could explain the rather high vitamin E status of the Italian population (Polito et al., 2005), given the consumption of vitamin E-rich vegetable oil in this region (Huang and Sumpio, 2008). Indeed, the traditional Mediterranean diet has been associated with health benefits, similar to vitamin E, such as reduced incidence of cardiovascular diseases and decreased lipid oxidation (Fito et al., 2007). Starchy, energy-rich staples on the other hand, can be considered rather poor contributors to dietary the vitamin E supply (**Table 1**).

### Vitamin E Biofortification

fpls-09-01862 December 15, 2018 Time: 15:9 # 17

### Metabolic Engineering

Biofortification to enhance vitamin E content in different crops has been successfully deployed over the last decades (Mène-Saffrané and Pellaud, 2017). To understand the rationale behind these strategies, one must first consider the different biological activities of the E-vitamers. As mentioned above, in many cases, α-tocopherol is considered the most potent, bioactive E-vitamer, as confirmed in a rat fetal resorption assay (Bunyan et al., 1961; Mène-Saffrané and Pellaud, 2017). Interestingly, important vitamin E sources such as vegetable oils (soybean, corn, canola and palm) contain a high ratio (up to 10:1) of γ-tocopherol over α-tocopherol (Eitenmiller, 1997). As α-tocopherol was determined to be ten times more bioactive as compared to γ-tocopherol, the idea arose to design metabolic engineering approaches shifting this ratio toward an enhanced relative α-tocopherol content (Shintani and DellaPenna, 1998). However, this objective needs to be justified by assessing the bioavailability as well as storage stability of these vitamers. Indeed, no compelling differences in bioavailability of these E-vitamers were found (Reboul et al., 2008; Reboul, 2017). Unfortunately, α-tocopherol appears less stable in storage, as it reacts faster with peroxy radicals, confirmed by the higher instability of α-tocopherol compared to γ-tocopherol in storage of camelina (Camelina sativa) oil (Abramovic et al., 2007). Although this issue should not be neglected, the higher bioactivity of the α-tocopherol vitamer could outweigh this disadvantage. Introduction of a γ-tocopherol methyltransferase (γ-TMT) (**Figure 5**) (Tewari et al., 2017), catalyzing the addition of the required methyl group to form α-tocopherol from γ-tocopherol (**Figures 4**, **5**), was therefore conducted. This strategy was proven successful in Arabidopsis, where the α/γ-tocopherol ratio was completely reversed in favor of α-tocopherol accumulation in seeds overexpressing the γ-TMT gene (Shintani and DellaPenna, 1998; Mène-Saffrané and Pellaud, 2017). This strategy has been implemented in several crops, including corn (Zhang L. et al., 2013), soybean (Glycine max) (Arun et al., 2014) and lettuce (Lactuca sativa L.) (Cho et al., 2005). Theoretically, the biological activity of the crop vitamin E pool can be increased up to 10-fold by this strategy (Mène-Saffrané and Pellaud, 2017). In rice endosperm, ectopic γ-TMT expression yielded no significant change in α-tocopherol content, explained by low γ-tocopherol levels, yet significantly altered tocotrienol levels, in favor of α-tocotrienol (Zhang G.Y. et al., 2013). Interestingly, implementation of this metabolic engineering approach, yielding higher α-tocopherol content in

alfalfa leaves (Medicago sativa), coincided with a delayed leaf senescence phenotype as well as enhanced tolerance to osmotic stress (Jiang et al., 2016). As this strategy does not greatly influence accumulation of the absolute tocochromanol content, applicability is confined to crops accumulating higher levels of E-vitamers with lowered bioactivity, such as γ-tocopherols and δ-tocopherols. Furthermore, generalization of E-vitamers into absolute values of 'bioactivity' could prove to be difficult. Indeed, different vitamers could exhibit different potencies in a whole range of biological functions, but without a single vitamer being omnipotent. This is indicated by the observed higher ability of γ-tocopherol to reduce 8-isoprostane [oxidative stress marker (Elfsmark et al., 2018)] (Jiang et al., 2002; Jiang and Ames, 2003). Assigning a universal (vitamin E) bioactivity to a specific vitamer could miss identifying its full biological potential. Moreover, the typical accumulation of γ-tocopherols witnessed in seeds (Sattler et al., 2004; Gilliland et al., 2006) (and therefore contributing to the vitamin E content of oils), might hint at its physiological importance in planta. Fortunately, no aberrant growth and fertility have been reported in the γ-TMT-engineered biofortified crops, indicating that the altered tocopherol ratio has marginal effects on plant growth and development (Mène-Saffrané and Pellaud, 2017).

Besides redirection of tocopherol homeostasis toward a more satisfactory vitamer composition, increase of (absolute) vitamer content has been tackled in metabolic engineering approaches (Cahoon et al., 2003). Engineering the HGGT gene, catalyzing the committed step in tocotrienol biosynthesis (**Figure 5**), resulted in an increase in total tocochromanol content of maize kernels and up to 18-fold enhancement in tocotrienol accumulation (Dolde and Wang, 2011). Furthermore, engineering HPPD, a key enzyme in the biosynthesis of the tocochromanol precursor HGA (**Figure 5**), generated a massive accumulation of tocotrienols, provided that prephenate dehydrogenase (shikimate pathway) was also engineered to ensure sufficient flux toward tyrosine (Rippert et al., 2004). Building further on this approach, high tocochromanol accumulating soybean was created via additional introduction of HPT and GGDR (see **Figure 5**) (Karunanandaa et al., 2005). However, biofortification approaches should not neglect tocochromanol stability, as vitamin E levels were shown to halve in freeze-dried fortified apple upon 6 months storage (Cortes et al., 2009). Further details on the different strategies employed in biofortification of crops toward higher vitamin E content have been elaborated by Mène-Saffrané and Pellaud (2017).

### Breeding

From the perspective of plant breeders, an interesting amount of variation in vitamin E content has been observed in different agronomical important crops (Mène-Saffrané and Pellaud, 2017). In rice, total kernel vitamin E content was found to vary up to threefold in different Malaysia-grown varieties (Shammugasamy et al., 2015). Similarly, a study in canola, which is important for oil production and therefore tocochromanol delivery, identified VTE3 and PDS as important

determinants of tocopherol content, based on screening of 229 accessions (Fritsche et al., 2012). Moreover, a measured variation of almost sixfold in maize kernel α-tocopherol content enabled conducting a GWAS wherein a HGGT gene, a prephenate dehydratase paralog [participating in tyrosine biosynthesis (El-Azaz et al., 2016)] and a tocopherol cyclase were recognized to contribute to tocotrienol content (Lipka et al., 2013). The same study further confirmed the link between γ-TMT alleles and α-tocopherol content. Interestingly, more recent GWAS in maize revealed many significant QTL loci, attributed to genes harboring novel activities as well as participating outside the tocopherol pathway (Diepenbrock et al., 2017; Wang et al., 2018). In conclusion, this is a nice example of GWAS and assignment of candidate genes to the identified QTLs to pinpoint potential factors for novel metabolic engineering approaches.

### Tocochromanols: Major Problems and Future Perspectives

The case of tocochromanols, comprising tocopherols and tocotrienols, is a good example on how simplifying these distinct groups of molecules to their collective term 'vitamin E' can be misleading. As previously mentioned, the bioactivity of E-vitamers is diverse. However, bioactivity alters depending on which tocochromanol-related process is utilized to assess it. Moreover, there is no one-to-one relationship between a certain vitamer and a given function. One could therefore argue that grouping tocochromanols into one group of 'vitamin E' is incorrect. This notion becomes more important given the existence of different metabolic engineering approaches aimed at altering E-vitamer ratios (e.g., increasing α-tocopherol/γ-tocopherol ratio) while keeping total tocochromanol levels intact (γ-TMT-engineering). Similarly, bioavailability as well as (storage) stability should not be neglected. Moreover, whether engineering approaches are based on altering tocochromanol ratio (e.g., via γ-TMT-engineering) or enhancing total tocochromanol content (e.g., HGGTengineering), the impact on plant growth and development should be closely monitored. Finally, seeds, being an important target for metabolic engineering approaches, often depict a typical tocochromanol signature (Sattler et al., 2004), related to their function therein, which could be disrupted upon engineering approaches. Future research should therefore further unravel the in planta role of the different vitamin E entities. Similarly, the pathophysiological significance of the different vitamers in humans should be thoroughly examined.

### INTERTWINING OF VITAMIN METABOLISM AND ITS SIGNIFICANCE IN MULTI-BIOFORTIFICATION

A simultaneous increase of several micronutrients in a particular crop/tissue, referred to as multi-biofortification, is a powerful means to tackle MNM. This strategy aims at obtaining adequate levels of multiple micronutrients in a single staple crop, which is massively consumed by the local population in need. Such endeavor might encounter synergistic but also potentially detrimental effects, due to micronutrient interactions. Taking the example of the antioxidant ascorbate, protection of components sensitive to oxidative damage (e.g., carotenoids) is expected, thereby contributing to their accumulation as well as stability upon storage, an advantage which could also be expected from the combination with vitamin E. Furthermore, the ascorbate-glutathione pathway is needed in the 'detoxification' of tocopheroxyl radicals in vitamin E salvage (Szarka et al., 2012). In addition, ascorbate is known to ameliorate iron uptake in humans (Iqbal et al., 2004). Consequently, ascorbate biofortified crops could also aid in combatting iron deficiency indirectly. Similarly, provitamin A and vitamin E biofortification have shown to be positively affect one another (Che et al., 2016; Muzhingi et al., 2017). In the example of biofortified sorghum, the raised level of vitamin E, obtained by genetic engineering, enhanced provitamin A stability (Che et al., 2016). Interestingly, a synergistic interrelationship between ascorbate and vitamin B9 (folates) has been proposed, justified by their coextensive increase during germination (Liu et al., 2017). This study also proposes that folates (vitamin B9) biosynthesis counteracts vitamin E biosynthesis by its competition for the precursor GTP. Competition for precursors could prove to have a substantial influence on vitamin metabolism, considering the fact that vitamin E biosynthesis requires precursors from shikimate and MEP pathways, which are also required in the folate and provitamin A pathways, respectively. Conversely, folates are proposed to aid in maintaining high ascorbate content, as they contribute in supplying NADPH to the cell, which could support adequate ascorbate salvage (Gorelova et al., 2017; Liu et al., 2017). Moreover, DXS activity, which has been enhanced in different metabolic engineering approaches aimed at augmenting plant provitamin A content, requires active B1 vitamer cofactor (thiamin pyrophosphate) for its functioning (White et al., 2016) and is also required in the biosynthesis of tocochromanols. This nicely illustrates how different vitamins are part of a potentially strong network of interactions in plant as well as in human metabolism. This aspect certainly deserves proper consideration upon evaluation of novel biofortification strategies (Strobbe and Van Der Straeten, 2018). Furthermore, certain environmental influences could alter the accumulation of multiple vitamins, illustrated by the light-dependent accumulation of both provitamin A and tocochromanols (Cruz et al., 2018; Gramegna et al., 2018). This aspect can therefore be considered upon setting light conditions in vertical farming projects (Bantis et al., 2018).

Last but not least, biofortification could have the beneficial 'side-effect' of enhancing tolerance to abiotic stresses, as reported in metabolic engineering approaches enhancing plant ascorbate content (Macknight et al., 2017). This is particularly important given the increased exposure to abiotic stresses, but also to biotic stresses crops will have to face as a result of climate change (Cheeseman, 2016).

## CONCLUSION

fpls-09-01862 December 15, 2018 Time: 15:9 # 19

Vitamin biofortification of food crops holds the potential to alleviate the global burden of vitamin deficiencies (Blancquaert et al., 2017; Garcia-Casal et al., 2017; Jiang L. et al., 2017; Martin and Li, 2017; Van Der Straeten et al., 2017; Garg et al., 2018). In doing so, staple crops will play a predominant role, as they hold the impressive capability to deliver cheap calories to populations in need and have the potential to be nutritionally enhanced via metabolic engineering or breeding approaches. Both conventional breeding and metabolic engineering should coexist in the battle against vitamin deficiencies, thereby reciprocally strengthening their potential. Molecular breeding techniques such as GWAS promise to facilitate enhancement of crop vitamin content whilst uncovering potential new determinants in vitamin accumulation in the particular crop/tissue, subsequently applicable in new engineering approaches. In some cases, downregulation of genes impeding vitamin accumulation is advised (see provitamin A biofortification). Here, metabolic engineering strategies utilizing genome-editing techniques such as the CRISPR/Cas system are promising, especially considering they might suffer less from regulatory issues blocking their commercialization (Potrykus, 2017), in cases where no transgenes are introduced. However, this technology still faces crop-specific limitations toward the maximal vitamin enhancement possible. Therefore, a combination with metabolic engineering strategies employing transgenes, is advisable, in which CRISPR/Cas technology could still be utilized to allow specific T-DNA insertion the genome position of interest.

When using a biofortification approach, several aspects should be considered, including bioavailability, bioactivity, stability and impact on crop yield and/or physiology. Bioavailability, bioactivity and stability can be addressed by

### REFERENCES


examination of these properties on the specific biofortified crop, and targeted by specific strategies to confer these properties to the crop product [e.g., engineering toward more stable folates in rice (Blancquaert et al., 2015), or engineering toward more potent α-tocopherol (Shintani and DellaPenna, 1998)]. Assessment of biofortification interventions influencing plant physiology (and thereby yield) requires indepth analysis and knowledge of micronutrient metabolism as well as post hoc examination of plant physiology in field conditions.

Given their potential to provide sufficient micronutrients, (multi-)biofortified crops are a crucial piece of the puzzle in eradicating micronutrient deficiencies on a global scale. Moreover, biofortified crops are already contributing to sustainable food security in a time of increasing global demographic pressure and climate change. Last but not least, they hold great potential to contribute even more to maintaining a healthy world population into the future, provided that novel approaches to biofortification are embraced.

### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

### FUNDING

SS was indebted to the Agency for Innovation by Science and Technology in Flanders (IWT) for a predoctoral fellowship. DV acknowledges support from Ghent University (Bijzonder Onderzoeksfonds, BOF2009/G0A/004, BOF2018/GOA/042), and the Research Foundation—Flanders (FWO, project 3G012609).



a plant ascorbic acid biosynthetic enzyme. J. Biol. Chem. 281, 15662–15670. doi: 10.1074/jbc.M601409200




enhanced seed tocopherol. Metab. Eng. 7, 384–400. doi: 10.1016/j.ymben.2005. 05.005




fruit metabolic traits. Plant Physiol. 165, 1120–1132. doi: 10.1104/pp.114.24 1521


dehydrogenase which affects ascorbate pool size in Arabidopsis thaliana. Plant Sci. 164, 1111–1117. doi: 10.1016/S0168-9452(03)00122-5



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Strobbe, De Lepeleire and Van Der Straeten. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Natural Variation in Physiological Responses of Tunisian Hedysarum carnosum Under Iron Deficiency

Heithem Ben Abdallah<sup>1</sup> , Hans Jörg Mai<sup>1</sup> , Tarek Slatni<sup>2</sup> , Claudia Fink-Straube<sup>3</sup> , Chedly Abdelly<sup>2</sup> and Petra Bauer1,4 \*

1 Institute of Botany, Heinrich Heine University Düsseldorf, Düsseldorf, Germany, <sup>2</sup> Laboratory of Extremophile Plant, Center of Biotechnology of Borj Cedria, Hammam-Lif, Tunisia, <sup>3</sup> INM-Leibniz Institute for New Materials, Saarbrücken, Germany, <sup>4</sup> Cluster of Excellence on Plant Sciences, Heinrich Heine University, Düsseldorf, Germany

Iron (Fe) is an essential element for plant growth and development. The cultivation of leguminous plants has generated strong interest because of their growth even on poor soils. Calcareous and saline soils with poor mineral availability are wide-spread in Tunisia. In an attempt to select better forage crops adapted to Tunisian soils, we characterized Fe deficiency responses of three different isolates of Hedysarum carnosum, an endemic Tunisian extremophile species growing in native stands in salt and calcareous soil conditions. H. carnosum is a non-model crop. The three isolates, named according to their habitats Karkar, Thelja, and Douiret, differed in the expression of Fe deficiency symptoms like morphology, leaf chlorosis with compromised leaf chlorophyll content and photosynthetic capacity and leaf metal contents. Across these parameters Thelja was found to be tolerant, while Karkar and Douiret were susceptible to Fe deficiency stress. The three physiological and molecular indicators of the iron deficiency response in roots, Fe reductase activity, growth medium acidification and induction of the IRON-REGULATED TRANSPORTER1 homolog, indicated that all lines responded to −Fe, however, varied in the strength of the different responses. We conclude that the individual lines have distinct adaptation capacities to react to iron deficiency, presumably involving mechanisms of whole-plant iron homeostasis and internal metal distribution. The Fe deficiency tolerance of Thelja might be linked with adaptation to its natural habitat on calcareous soil.

Keywords: legume, natural diversity, iron deficiency, chlorophyll, acidification, Fe reductase activity, IRT1

### INTRODUCTION

Iron (Fe) is an essential micronutrient with numerous cellular functions, e.g., in photosynthesis, respiration, DNA synthesis, and N<sup>2</sup> fixation. Plants are frequently challenged by Fe deficiency, especially on alkaline and calcareous soils due to poor Fe solubility under these conditions. In Tunisia, the exploration of such kinds of natural habitats and saline environments revealed that they are colonized by a native leguminous vegetation which might have specific adaptations to both, salinity and nutrient deficiencies, especially Fe (Ben Abdallah et al., 2017). Leguminous plants take up reduced Fe using mainly the so-called Strategy

Edited by:

Huixia Shou, Zhejiang University, China

#### Reviewed by:

Ping Lan, Institute of Soil Science (CAS), China Marta Dell'Orto, Università degli Studi di Milano, Italy

> \*Correspondence: Petra Bauer petra.bauer@uni-duesseldorf.de;

#### Specialty section:

Petra.Bauer@hhu.de

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 08 December 2017 Accepted: 30 August 2018 Published: 02 October 2018

#### Citation:

Ben Abdallah H, Mai HJ, Slatni T, Fink-Straube C, Abdelly C and Bauer P (2018) Natural Variation in Physiological Responses of Tunisian Hedysarum carnosum Under Iron Deficiency. Front. Plant Sci. 9:1383. doi: 10.3389/fpls.2018.01383

I. The main feature of Strategy I plants, e.g., in Arabidopsis thaliana and leguminous plants, is that they acidify the soil via proton extrusion through an ATPase, reduce ferric to ferrous Fe by a ferric chelate reductase and take up the divalent Fe via divalent metal IRON-REGULATED TRANSPORTER1 (Brumbarova et al., 2015), being a member of the ancient ZIP ( = ZRT/IRT1) protein family (Eng et al., 1998). IRT1 homologs were found Fe-regulated in roots of multiple legumes like Pisum sativum (Cohen et al., 2004), Medicago truncatula (Lopez-Millan et al., 2004), Arachis hypogaea (Ding et al., 2010), Glycine max (Brear et al., 2013), and Vigna radiata (Muneer et al., 2014).

Natural variation studies make use of the existing natural allelic diversity in plant populations as a source to pinpoint the adaptive alleles for relevant traits. Natural variation was successfully applied in model plants to identify causal alleles by genome-wide association studies for such different traits as environmental adaptation in Arabidopsis (Li et al., 2010), or nutritional quality and agronomic traits in maize (Diepenbrock et al., 2017) and rice (Si et al., 2016). The model legumes M. truncatula and G. max are particularly suited for natural biodiversity studies (Gentzbittel et al., 2015). Prerequisites for association studies are genome sequence variation reflected by a broad collection of ecotypes demonstrating phenotypic diversity for given traits. However, alternative procedures are available for studying natural diversity of small population collections in the absence of large genome sequence data, e.g., by making use of recombinant or near-isogenic inbred lines suitable for mapping and gene identification (Yan et al., 2017), by isolating candidate genomic regions and genes based on comparative genomics (Friesen et al., 2010, 2014; Turner et al., 2010) and transcriptomics or proteomics of genetically divergent lines (Voelckel et al., 2017).

In Tunisia, a large area is arid to semi-arid with calcareous and alkaline-saline soil conditions, where bioavailability of Fe, Mg, and other minerals is low, posing a major problem for crop yield (Rabhi et al., 2007). Such areas are often used for forage crop and cattle production. To improve agricultural land usage and provide better perspectives to farmers, there is a need to select tolerant crops adapted to such poor soils.

Perennial Hedysarum carnosum was proposed as a prospective well-palatable pasture crop being a naturally adapted halophyte to the Mediterranean basin with good potential for rehabilitation strategies (Le Houerou, 1996). The Hedysarum (sweetvetch) genus belongs to the Fabaceae plant family, which contains many of the very important crops. Species of this genus have arisen in a non-monophyletic manner, as recently established based on multiple sequence alignments and phylogenetic tree constructions of 58 accessions accounted to this genus using nuclear and plastid gene sequences (Liu et al., 2017). H. carnosum (also known as Sulla carnosa) that is subject of this study is most related to Hedysarum coronarium (also known as Sulla coronaria) (Liu et al., 2017). H. coronarium is wide-spread around the Mediterranean basin. In contrast, H. carnosum is endemic in Tunisia where it grows in different climates, ranging from semi-arid (Karkar) to arid regions (Thelja and Douiret). This species prefers slightly acid to alkaline soils (pH 5.5–8.5), sandy loams and clays, and good growth is achieved on alkaline soils. As an extremophile, H. carnosum also grows on Tunisian saline sodic soils, represented especially by Chotts and Sebkhas (Dallali et al., 2012). This species plays important roles in animal feed due to its high protein contents and tannins (Aissa et al., 2016). H. carnosum is a non-model crop and molecular investigation is difficult because of the lack of molecular data and gene sequences. H. carnosum responds to salt, potassium and magnesium deficiencies and low Fe availability (Farhat et al., 2016; Elkhouni et al., 2017; Hafsi et al., 2017). However, in all these studies only a single H. carnosum line was studied, rendering it difficult to judge the natural adaptation potential within the species. Moreover, these studies primarily focused on photosynthetic parameters in leaves. The objective of this work was to investigate the effect of Fe deficiency at the physiological level in H. carnosum and to compare the responses in three different isolates collected from different natural sites. Our data show that the Thelja isolate is the most tolerant to Fe deficiency stress and therefore will be the best choice for use in future genetic and RNAseq studies to identify the natural basis for calcareous soil-induced Fe deficiency. Thelja will also represent the most promising ecotype for agricultural purposes.

### MATERIALS AND METHODS

### Plant Material and Growth Condition

Three isolates of H. carnosum were acquired by collecting seeds from Karkar, Thelja, and Douiret in Tunisia (see **Supplementary Figure 1** for species characteristics and geographic repartition of collected isolates). About 300 seeds per isolate were collected from an average of 10 plants distributed in a diameter of 100 m. Seeds of the three isolates were germinated and grown in separate green houses. Due to allogamy plants were multiplied by crossfertilization of plants within each isolate. The F3 generation was used for analyses.

Seeds were mechanically scarified by rubbing in between fine grit sand paper sheets. The seeds were sterilized in 10% sodium hypochlorite for 8 min and then abundantly rinsed with distilled water. After a 10 min imbibition phase, they were germinated for 4 days at 20◦C in Petri dishes on constantly moistened filter paper.

Four day-old seedlings were transferred to a half strength aerated liquid nutrient solution for 2 days. Similarly sized seedlings were then selected and cultured in groups of 4 or 10 plants in 1 or 10 L of full strength aerated nutrient solution [1.5 mM Ca(NO3)2, 1.25 mM KNO3, 0.75 mM MgSO4, 0.5 mM KH2PO<sup>4</sup> and 10 µM H3BO3, 1 µM MnSO4, 0.5 µM ZnSO4, 3 µM MOO4Na2, 0.5 µM CuSO4, and 50 µM Fe-EDTA]. At day 6, the following treatments were conducted for the amount of time indicated in the text and figure legends: +Fe, control (Fe-sufficient medium with 50 µM Fe) and −Fe, Fe deficiency (medium without Fe). The pH was adjusted to 6.0 with NaOH for both, the +Fe (control) and −Fe (iron deficiency) treatments. Aerated hydroponic cultures were maintained in a growth chamber with a day/night regime of 16/8 h light-dark-cycle, a 24/18◦C temperature cycle and a constant relative humidity of 70%. The

FIGURE 1 | Effect of Fe deficiency stress on Karkar, Thelja, and Douiret lines. (A) Root dry weight (RDW), n = 3; (B) shoot dry weight (SDW), n = 3; (C) main root length, n = 4–5; (D) total chlorophyll (Chl) content, n = 3; (E,F) photon yield of PSII calculated by Fv/Fm and Fv/F0, n = 4; (G) chlorosis phenotype symptoms of young leaves calculated according to the indicated scale from 1, green leaves up to 5,white-yellow leaves, n = 4. H. carnosum plants were exposed for 10 days (A–F) or up to 10 days (G) to Fe-sufficient (+Fe) and Fe-deficient (–Fe) hydroponic growth conditions. Data are means ± SD; means with the same letter are not significantly different with P ≤ 0.05 according to ANOVA and Tukey's HSD test.

solution was renewed every 4 days. The standard experiment was conducted using ten-day Fe sufficiency and deficiency treatments.

## Morphological Root and Shoot Phenotypes

Roots and shoots were harvested, dried in an oven at 70◦C for 48 h and the dry weights determined per plant. The main root lengths were measured. The degree of leaf chlorosis was assessed in the youngest expanded leaves. The leaf chlorosis scale was determined as previously described (Schuler et al., 2012), ranging from 1 = green, 2 = light green, partially yellow, 3 = yellow-green, 4 = yellow to 5 = white-yellow, as shown in **Figure 1G**.

### Chlorophyll Measurements

Total chlorophyll was extracted from fresh leaves in 80% acetone and assayed photometrically at 645 nm and 663 nm. The OD values were used to calculate the total chlorophyll content in mg/g fresh weight of the leaves as published (Arnon, 1949).

### Pulse Amplitude Measurements (PAM)

Pulse Amplitude Measurements (PAM) was determined with the FluorCam FC 800-C machine (Photon Systems InstrumentsTM). Plants were adapted to darkness for about 15 min. Then single leaves were measured for F<sup>0</sup> (minimal fluorescence) up to Fm (maximal fluorescence). To analyze photosystem II activity, Fv/Fm values and Fv/F<sup>0</sup> values were calculated (Murchie and Lawson, 2013). F<sup>0</sup> and Fm represent the minimum and maximum values of chlorophyll fluorescence, while F<sup>V</sup> is the variable fluorescence.

### Acidification of the Growth Medium

The acidification capacity was determined after placing plants into 1 L nutrient solution (at pH 6.2), respectively, and measuring the pH of the nutrient solution in the subsequent days as indicated in the text and figure legend.

### Measurements of Root Fe Reductase Activity

Intact root systems were washed with 100 mM Ca(NO3)<sup>2</sup> solution and submerged in the Fe reductase assay solution containing 0.1 mM Fe3+-NaEDTA and 0.3 mM ferrozine at pH 5.0 for 1 h in the dark. Then the absorbance was determined at 562 nm. The concentration of the Fe2+-ferrozine complex was calculated using the molar extinction coefficient of 28.6 mM−<sup>1</sup> cm−<sup>1</sup> . The amount of Fe2<sup>+</sup> was normalized to the root weight in the assay and Fe reductase activity was calculated.

## Mineral Element Analysis

To determine the metal ion content the youngest expanded leaves of H. carnosum were harvested and dried over 72 h at 70◦C. After drying, the harvested leaves were finely powdered with an achat mortar and pestle. Metal contents (Zn, Fe, Cu) were determined using inductively-coupled plasma optical emission spectrometry (ICP-OES) at the Leibniz Institute for New Materials (INM, Saarbrücken).

FIGURE 2 | Metal contents of Karkar, Thelja, and Douiret. (A) Fe; (B) Zn; (C) Cu contents of the youngest expanded leaves. H. carnosum plants were exposed for 10 days to Fe-sufficient (+Fe) and Fe-deficient (–Fe) hydroponic growth conditions. Data are means ± SD; n = 4; means with the same letter are not significantly different with P ≤ 0.05 according to ANOVA and Tukey's HSD test.

### Obtention of H. carnosum cDNA Sequences and Multiple Sequence Alignment of Amino Acid Sequences

Obtention of H. carnosum cDNA sequences is outlined in **Supplementary Figure 2A**. A. thaliana, M. truncatula, G. max, Lotus japonicus sequences of IRT1 and β-ACTIN (ACT) were aligned. Conserved regions near the start and stop codons were identified and primers matching 100% the M. truncatula sequences were designed (Mt primers, **Supplementary Figure 3**). With these Mt primers 1 µL of template root cDNA of H. carnosum was used to amplify the IRT1 and ACT internal coding sequences in a standard PCR. PCR amplicon bands were

purified from agarose gels according to standard procedures and sequenced.

Next, TAIL-PCR (**Supplementary Figure 2B**) was used to identify the unknown upstream 5<sup>0</sup> and downstream 3<sup>0</sup> cDNA sequences adjacent to the determined HcIRT1 and HcACT partial sequences. We used three nested specific primers (S1–S3) that aligned near the edge of the known cDNA (**Supplementary Figure 3**). For extension in the opposite direction AD (arbitrary degenerate) primers were used. The AD primers were 64x–256x degenerate and designed to be relatively short (15–16 nt) with a low melting temperature (ca. 40–50◦C) (**Supplementary Figure 3**). Three consecutive TAIL-PCR reactions were conducted as described (Liu et al., 1995). The third step PCR products were sequenced and HcIRT1 and HcACT sequences assembled and provided to GenBank (accession numbers MH879027 and MH879028).

Multiple sequence alignment and construction of neighborjoining trees using amino acid sequences was performed using the Clustal Omega tool at https://www.ebi.ac.uk.

### RNA Isolation and Quantitative Real-Time PCR

Gene expression was analyzed using three biological replicates. RNA isolation and reverse transcription-quantitative PCR were carried out as described previously (Ben Abdallah and Bauer, 2016). Briefly, total RNA prepared from 100 mg H. carnosum root tissue was used for cDNA synthesis using an oligo-dT primer. qPCR was conducted using the SYBR Green detection method. RT-qPCR primers were used for qPCR. The absolute quantity of initial transcripts was determined for the genes IRT1 and ACT by standard curve analysis using mass standards prepared from H. carnosum cDNA PCR products amplified with Mt primers. Absolute expression data of IRT1 was obtained after normalization to the internal control ACT gene. Each biological cDNA sample was tested in two technical qPCR replicates.

### Statistical Analysis

Morphological, physiological, and molecular data were obtained in at least three biological replicates, as detailed in the figure legends. Data of biological replicates were used to calculate mean values and standard deviations. Statistical significance was determined by applying t-tests (for two sample comparisons) and One-way ANOVA followed by Tukey's HSD test (for more than two sample comparisons) designated as "ANOVA and Tukey's HSD test" in the figure legends.

## RESULTS

### Morphological and Physiological Shoot Responses to Fe Deficiency

Seeds from H. carnosum plants were collected in three different locations in Tunisia characterized by semi-arid, arid and Saharan conditions, named Karkar, Thelja, and Douiret, with salinesodic, calcareous and sandy soil characteristics (**Supplementary Figure 1**). After seeds were multiplied for three generations, morphological and physiological experiments were carried out. We were interested in obtaining an ecotype with high tolerance to prolonged Fe deficiency growth conditions, a trait expected to be beneficial upon growth on calcareous soil. We therefore hypothesized that the three isolates might show different adaptation and respond differently to Fe deficiency conditions. Plant seedlings were grown in controlled hydroponic conditions and exposed to sufficient iron (+Fe) or deficient iron supply (−Fe) for 10 days. At first, we compared the tolerance/sensitivity of the lines to −Fe by measuring different growth parameters. The three isolates did not behave any different from each other in terms of root biomass production under + and −Fe (**Figure 1A**). Also in terms of shoot biomass production, the ecotypes were very similar (**Figure 1B**). Only one comparison resulted in a significant difference in biomass, which was the fourfold higher shoot dry weight of Douiret versus Karkar at −Fe (**Figure 1B**). However, none of the lines showed lower root or shoot biomass when grown at – compared to +Fe (**Figures 1A,B**). Karkar displayed a shorter main root compared to Douiret at +Fe. When comparing the main root length at −Fe versus +Fe, there was a significant decrease only in the case of Douiret but not Karkar and Thelja (**Figure 1C**).

Fe is required in high amounts during plant growth in the leaves to sustain photosynthesis and for chlorophyll synthesis. Fe

n = 4. H. carnosum plants were exposed for 10 days (A) or up to 10 days (B) to Fe-sufficient (+Fe) and Fe-deficient (–Fe) hydroponic growth conditions. Data are means ± SD; means with the same letter are not significantly different with P ≤ 0.05 according to ANOVA and Tukey's HSD test, means with <sup>∗</sup> label in B show a significant difference of + versus –Fe with P ≤ 0.05 according to a t-test.

can also be stored in chloroplasts in the form of ferritin. Lack of Fe results in the typical leaf chlorosis symptoms especially in the expanding leaves. Leaf chlorosis is caused by low chlorophyll contents under Fe deficiency. Karkar and Douiret had higher total chlorophyll contents at + than at −Fe (**Figure 1D**). Thelja, on the other hand, displayed no significant difference at + versus −Fe (**Figure 1D**). No significant differences were detectable between the lines at either + or −Fe (**Figure 1D**). PAM measurements based on chlorophyll fluorescence are an indicator for the photosynthetic performance under stress conditions. Low Fv/Fm and Fv/F<sup>0</sup> ratios are indicative of stress affecting negatively the photosystem activity. We found that Karkar had lower Fv/Fm and Fv/F<sup>0</sup> ratios at – versus +Fe, while no significant differences were found in Thelja and Douiret (**Figures 1E,F**). When comparing the lines with each other, Karkar had a lower Fv/Fm ratio than Thelja and Douiret and Thelja had a higher Fv/F<sup>0</sup> ratio than Karkar and Douiret (**Figures 1E,F**). We were also interested in comparing the development of the leaf chlorosis during the 10 days of exposure to −Fe. Leaf chlorosis started 2 days earlier in Douiret than in Karkar and Thelja, but after 10 days the chlorosis had reached similar levels, as determined above from the chlorophyll measurements (**Figures 1D,G**).

Next, we investigated whether the observed leaf chlorosis and impact on photosynthesis could be related to the amount of Fe taken up. We determined metal contents in the expanding leaves, as these are the organs where Fe deficiency symptoms are noted. Roots were not used since under hydroponic growth Fe is available in the Form of Fe Na EDTA, resulting in an accumulation and high Fe content, e.g., in the apoplasts of roots. In addition to Fe we measured Zn and Cu contents. Arabidopsis IRT1 can take up Zn but not Cu (Vert et al., 2002) and MtZIP6 can also transport Zn (Lopez-Millan et al., 2004). Karkar and Douiret had lower Fe contents upon −Fe than under +Fe, but not Thelja, which had comparable levels under both conditions (**Figure 2A**). Thelja also had higher Fe contents upon −Fe compared to Karkar and Douiret (**Figure 2A**). The Zn content was decreased at – versus +Fe growth conditions only in Douiret (**Figure 2B**). However, Thelja had a higher Zn content than Karkar and Douiret at – but not +Fe (**Figure 2B**). The Cu content was decreased at – versus +Fe in Karkar and Douiret, but again not in Thelja (**Figure 2C**). When comparing the lines with each other, a significant difference of Cu was only found in the comparison of Thelja versus Karkar at −Fe (**Figure 2C**).

To conclude from this physiological and growth analysis of the three lines exposed to + and −Fe, we summarized the

selected legumes and Arabidopsis. The multiple amino acid sequence alignment was produced with H. carnosum IRT1 and all identified ZIP protein sequences from A. thaliana, M. truncatula, G. max, and L. japonicus. The box indicates the closest relatives of AtIRT1. The star indicates HcIRT1.

comparative outcomes for the parameters measured at + versus −Fe for each line and designated a significant decrease at – versus +Fe as "sensitive" and no decrease as "tolerant" behavior (**Figure 3**). Karkar received four sensitivity and five tolerance labels, Douiret six sensitivity and three tolerance labels, and Thelja nine tolerance labels. Root and shoot biomass were not identified as parameters that could be used to discriminate the behavior of the lines at + and −Fe, while leaf chlorosis, photosystem activity and metal contents were well suited to do so. Taken together, it can be deduced that Thelja shows tolerance to −Fe in contrast to the other two lines.

### Physiological and Molecular Root Responses to Fe Deficiency

Roots of strategy I plants show typical Fe deficiency symptoms like enhanced soil acidification, Fe reduction and increased IRT1 gene expression. Quantification of these responses is used to judge the degree of tolerance to Fe deficiency (Brumbarova et al., 2015). Therefore, we tested next for potential differences in the level of Fe deficiency responses in the root. None of the plants subjected to Fe deficiency showed a significant increase in root Fe reductase activity (**Figure 4A**). In a time-course experiment we found that the growth medium was acidified significantly starting as early 2 days after exchange to Fe deficiency and continued until 10 days in all three lines (**Figure 4B**). H. carnosum gene sequences were not deposited in the database. To conduct gene expression analysis by the RT-qPCR method, we selected IRT1 as a target gene to reflect molecular Fe uptake regulation. Gene expression of additional genes were not investigated here as it is more useful to conduct RNAseq studies in the future. First, we identified homologs of IRT1 and of the reference gene β-ACTIN (ACT) from H. carnosum using PCR and TAIL-PCR by exploiting sequence similarities among leguminous plant IRT1 sequences and available microarray-based gene expression data for M. truncatula (see section "Materials and Methods"; outline in **Supplementary Figure 2**). Since we found no differences in the amino acid sequences of IRT1 between the three H. carnosum lines, we compared the sequence to ZIP sequences from Arabidopsis and other legumes. The full-length HcIRT1 amino acid sequence was found most related to MtZIP6 in a neighbor-joining tree derived from a multiple sequence alignment of the entire families of A. thaliana, M. truncatula, G. max, and L. japonicus ZIP protein sequences (**Figure 5**). MtZIP6 was the only M. truncatula ZIP protein with high sequence similarity to HcIRT1 and AtIRT1 (**Figure 5**). MtZIP6 is up-regulated by −Fe in roots (He et al., 2009; Benedito et al., 2010) and it was characterized as Fe transporter (Lopez-Millan et al., 2004). All other M. truncatula ZIP proteins group along with other branches of A. thaliana ZIP proteins (**Figure 5**), indicating that these other MtZIP proteins have different functions in metal homeostasis. Interestingly, this analysis also shows that A. thaliana has a high expansion of four IRT1-like proteins (IRT1, IRT2, ZIP8, ZIP10). In this same branch of IRT1-like sequences, there are two G. max, two L. japonicus and only one M. truncatula ZIP sequence (highlighted by a red box in **Figure 5**). Thus, legumes have fewer IRT1-like proteins than Arabidopsis. One possible explanation could be the different genome duplication histories during evolution. HcIRT1 and MtZIP6 share two important functional sequence features in the predicted variable cytoplasmic loop region with AtIRT1, namely two conserved lysine positions used for ubiquitination in metal-directed IRT1 turnover (Kerkeb et al., 2008) and the histidine-rich stretch for metal-binding relevant

for metal import by ZIPs into the cell (Zhang et al., 2017; **Figure 6A**). Hence, the sequence analysis convincingly suggests that HcIRT1 encodes a functional IRT1 homolog. HcIRT1 gene expression was found significantly induced by −Fe in Karkar, Thelja, and Douiret (**Figure 6B**). Thelja displayed a higher base level of HcIRT1 expression in the +Fe control situation compared to Karkar and Douiret (**Figure 6B**). Thelja and Douiret had a higher HcIRT1 expression level at −Fe versus Karkar (**Figure 6B**).

In summary, the three lines displayed root Fe deficiency response reactions which were most pronounced in case of HcIRT1 induction and medium acidification, while Fe reductase activity increases were not found to be significant. Perhaps the constitutively elevated IRT1 expression level of Thelja is linked with its higher Fe content under −Fe as an adaptation to growth on calcareous soil.

### DISCUSSION

Here, we show that the extremophile H. carnosum shows natural variation and phenotypic plasticity with regard to Fe deficiency responses in 8 out of 12 measured parameters. This species is an endemic growing on Tunisian saline and calcareous soil conditions that are known to affect micronutrient use efficiency. Overall, Thelja is the most tolerant isolate showing tolerance to Fe deficiency perhaps as a consequence of its adaptation to calcareous soils.

All H. carnosum isolates sensed Fe deficiency and responded to this stress, while the outcome of −Fe stress was different among the ecotypes. The common −Fe symptom elicited by all three lines was the development of a leaf chlorosis. All lines acidified the plant medium and had induced expression of HcIRT1 under −Fe versus +Fe. Several other −Fe symptoms were, however, only displayed by Karkar and Douiret, but not by Thelja. Karkar and Douiret exhibited quite drastic leaf chlorosis at −Fe. This was evident from accelerated leaf chlorosis and the low chlorophyll contents at – versus +Fe after 10 days of −Fe. Leaf chlorosis is a frequently occurring stress symptom in plants since under unfavorable conditions plants tend to reduce photosystem activity by removing chlorophyll and degrading chloroplasts to avoid additional stress caused by the light. This phenomenon can be measured by PAM chlorophyll fluorescence, which was lower in Karkar and Douiret and fits to the leaf chlorosis observations. Moreover, several steps in photosynthetic pigment metabolism and chloroplast ultrastructure are dependent on Fe, which explains the leaf chlorosis in the young expanding leaves after transfer of the plants to −Fe conditions. Fe deficiency resulted in a decrease of Fe contents in Karkar and Douiret expanding leaves, and hence the low Fe status can be regarded as reason for the leaf chlorosis. It is surprising that biomass production was not affected by −Fe in our experiments. We explain this partly by the fact that with ANOVA and Tukey HSD we applied the appropriate but comparably conservative statistical test for multiple comparisons that keeps the family-wise error rate (FWER) at 0.05. Hence, an increasing number of

comparisons increases the Type II error (false negative) rate and thus decreases the power for the single comparisons. Perhaps, less conservative tests such as Fisher LSD or two-sample t-tests would have resulted in statistically significant differences. The data indicate that there was a tendency for lower values at – versus +Fe for many parameters even in Thelja, but the differences were not significant according to ANOVA/Tukey HSD.

Zn and Cu contents were affected by −Fe in Karkar and Douiret in addition to Fe contents, but not in Thelja. The reduced Fe contents are explained as primary reaction by the low amount of Fe to which the plants were exposed. However, the reduced Zn and Cu contents must have been a secondary reaction of the plant to −Fe. Normally, it would be expected that Zn contents might increase upon −Fe, because increased IRT1 would take up Zn (Li et al., 2014). Perhaps, Hedysarum plants have different capacities to regulate metal homeostasis, and this capacity differs between Thelja, Douiret, and Karkar. Differences in the regulation of Fe reductase activity, IRT1 gene expression and metal contents between different ecotypes were also found for M. truncatula (Li et al., 2014). Under Fe deficiency, plants can suffer from oxidative stress (Ranieri et al., 2001; Zaharieva and Abadia, 2003; Waters et al., 2012; Ramirez et al., 2013). In A. thaliana, the CuSOD (copper/zinc superoxide dismutase) genes CSD1 and CSD2 are induced under Fe deficiency and have been suggested to replace FeSOD's (iron superoxide dismutases) to cope with oxidative stress under iron deficient conditions (Waters et al., 2012). It can be assumed that a similar mechanism exists in leguminous plants and elevated Cu and Zn contents could contribute to the effectiveness of this mechanism. Our observation of higher Cu and Zn levels in Thelja could be one possible explanation for the increased resistance of Thelja to iron deficiency compared to Karkar and Douiret. Hence, the different efficiencies of Cu and Zn uptake under Fe deficiency in Thelja, Karkar, and Douiret could be an important distinctive factor with respect to Fe deficiency tolerance.

The stronger −Fe leaf symptoms of Karkar and Douiret suggest that these lines should sense −Fe stress stronger than Thelja. But the two lines did not activate their root Fe mobilization in a stronger manner than Thelja. Karkar reduced more Fe in the root than did Douiret and Thelja. On the other hand, medium acidification capacities were similar between the lines. Thelja and Douiret displayed higher HcIRT1 gene expression than did Karkar. An interesting regulatory phenomenon could be seen for Thelja HcIRT1 gene expression, which was higher at +Fe compared to Karkar and Douiret. One possible explanation is that Thelja might take up more Fe at +Fe than Karkar and Douiret, but store this Fe in the root. Upon Fe deficiency the Fe stores could be remobilized and effectively transported to the shoots. Hence, Thelja could survive better upon −Fe conditions and maintain Fe levels. Since Thelja was collected in a region with calcareous soil condition it is tempting to speculate that the constitutive HcIRT1 expression might contribute to adaptation. On the other hand, Karkar might profit from an inefficient Fe usage, perhaps caused by ineffective internal mobilization and transport in its natural habitat with saline-sodic soil. Some abiotic stress factors induced for example by salt stress affect Fe uptake negatively, which can be explained by the toxicity of metals under water loss and the increased risk of oxidative stress (Le et al., 2016). Several proteins relevant for Fe regulation including IRT1 are controlled at post-translational level (Brumbarova et al., 2015) and hence it would be interesting to combine in the future transcriptomic in comparison to protein studies to assess the physiological activities.

The present study focussed on Fe deficiency responses, which are physiologically distinct from responses to calcareous-alkaline medium conditions. Future studies should focus on natural calcareous and saline-calcareous soil conditions. Multiple factors will differ in such experiments, including pH, soil texture, other mineral availabilities and microbial communities. Quite possibly, the ecotypes may differ in their way to mobilize Fe under additional mineral deficiencies and salt stress. Furthermore, the internal iron homeostasis regulation and allocation upon −Fe should gain attention (Schuler et al., 2012). In this respect it is noteworthy that natural variation for Fe efficiency in crops can be manifested at the level of metabolite changes, citrate concentrations for Fe-citrate movement, oxidative stress scavenging and Fe-mobilizing riboflavin patterns (Kabir et al., 2012, 2013, 2015; Jelali et al., 2014; Ben Abdallah et al., 2017). Genome sequence variation of metal homeostasis-relevant genes might account for differences in the gene expression levels or functional SNPs in coding regions. Comparing stress and −Fe responses between young and adult stages as well as under double stress may lead to better understanding of the mechanism of −Fe regulation in this leguminous species. One possibility would be to conduct comparative RNAseq. Gene expression differences between the lines can be used to build novel hypotheses on the physiological mechanisms of tolerance, which could subsequently be validated in biochemical experiments. Our work lies the ground for experiments addressing the mechanistics using the characterized Thelja and Karkar ecotypes as an extreme pair for a detailed natural variation study.

### AUTHOR CONTRIBUTIONS

PB, TS, and CA designed the experiments. HBA carried out the experiments. PB, HBA, and HM analyzed the data. CF-S performed the metal determination. PB wrote the manuscript. HBA and HM commented on the manuscript.

### FUNDING

This work was supported through an internship and STIBET fellowships from the DAAD and the Tunisian Ministry of Higher Education and Scientific Research (LR10CBBC02).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018.01383/ full#supplementary-material

### REFERENCES

fpls-09-01383 September 29, 2018 Time: 16:44 # 10



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Ben Abdallah, Mai, Slatni, Fink-Straube, Abdelly and Bauer. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

## Iron and Zinc in the Embryo and Endosperm of Rice (Oryza sativa L.) Seeds in Contrasting 2 0 -Deoxymugineic Acid/Nicotianamine Scenarios

### Edited by:

Felipe Klein Ricachenevsky, Universidade Federal de Santa Maria, Brazil

#### Reviewed by:

Elsbeth L. Walker, University of Massachusetts Amherst, United States Tracy Punshon, Dartmouth College, United States

#### \*Correspondence:

Ana Álvarez-Fernández ana.alvarez@eaad.csic.es

#### †Present address:

Pablo Díaz-Benito, Centro de Biotecnologia e Química Fina – Laboratório Associado, Escola Superior de Biotecnologia, Universidade Católica Portuguesa, Porto, Portugal Raviraj Banakar, Department of Agronomy, Iowa State University, Ames, IA, United States

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 27 March 2018 Accepted: 25 July 2018 Published: 21 August 2018

#### Citation:

Díaz-Benito P, Banakar R, Rodríguez-Menéndez S, Capell T, Pereiro R, Christou P, Abadía J, Fernández B and Álvarez-Fernández A (2018) Iron and Zinc in the Embryo and Endosperm of Rice (Oryza sativa L.) Seeds in Contrasting 2 0 -Deoxymugineic Acid/Nicotianamine Scenarios. Front. Plant Sci. 9:1190. doi: 10.3389/fpls.2018.01190 Pablo Díaz-Benito<sup>1</sup>† , Raviraj Banakar<sup>2</sup>† , Sara Rodríguez-Menéndez<sup>3</sup> , Teresa Capell<sup>2</sup> , Rosario Pereiro<sup>3</sup> , Paul Christou2,4, Javier Abadía<sup>1</sup> , Beatriz Fernández<sup>3</sup> and Ana Álvarez-Fernández<sup>1</sup> \*

<sup>1</sup> Department of Plant Nutrition, Estación Experimental de Aula Dei, Consejo Superior de Investigaciones Científicas, Zaragoza, Spain, <sup>2</sup> Departament de Producció Vegetal i Ciència Forestal, Universitat de Lleida-Agrotecnio Center, Lleida, Spain, <sup>3</sup> Department of Physical and Analytical Chemistry, Faculty of Chemistry, University of Oviedo, Oviedo, Spain, <sup>4</sup> ICREA, Catalan Institute for Research and Advanced Studies, Barcelona, Spain

Iron and Zn deficiencies are worldwide nutritional disorders that can be alleviated by increasing the metal concentration of rice (Oryza sativa L.) grains via bio-fortification approaches. The overproduction of the metal chelator nicotianamine (NA) is among the most effective ones, but it is still unclear whether this is due to the enrichment in NA itself and/or the concomitant enrichment in the NA derivative 2<sup>0</sup> -deoxymugineic acid (DMA). The endosperm is the most commonly consumed portion of the rice grain and mediates the transfer of nutrients from vegetative tissues to the metal rich embryo. The impact of contrasting levels of DMA and NA on the metal distribution in the embryo and endosperm of rice seeds has been assessed using wild-type rice and six different transgenic lines overexpressing nicotianamine synthase (OsNAS1) and/or barley nicotianamine amino transferase (HvNAATb). These transgenic lines outlined three different DMA/NA scenarios: (i) in a first scenario, an enhanced NA level (via overexpression of OsNAS1) would not be fully depleted because of a limited capacity to use NA for DMA synthesis (lack of -or low- expression of HvNAATb), and results in consistent enrichments in NA, DMA, Fe and Zn in the endosperm and NA, DMA and Fe in the embryo; (ii) in a second scenario, an enhanced NA level (via overexpression of OsNAS1) would be depleted by an enhanced capacity to use NA for DMA synthesis (via expression of HvNAATb), and results in enrichments only for DMA and Fe, both in the endosperm and embryo, and (iii) in a third scenario, the lack of sufficient NA replenishment would limit DMA synthesis, in spite of the enhanced capacity to use NA for this purpose (via expression of HvNAATb), and results in decreases in NA, variable changes in DMA and moderate decreases in Fe in the embryo and endosperm. Also, quantitative LA-ICP-MS metal map images of the embryo structures show that the first and second scenarios altered local distributions of Fe, and to a lesser extent of Zn. The roles of DMA/NA levels in the transport of Fe and Zn within the embryo are thoroughly discussed.

Keywords: metals, laser ablation, ligands, mass spectrometry, rice, seeds

### INTRODUCTION

fpls-09-01190 August 18, 2018 Time: 11:57 # 2

The deficiencies of iron (Fe) and zinc (Zn) are among the most important nutritional disorders in plants and humans. These elements play key roles as cofactors and structural components (e.g., Fe in cytochromes and Zn in Zn-finger proteins, respectively) in many proteins. Near 33% of world human population is affected by ferropenic anemia, a low red blood cell count due to Fe deficiency (McLean et al., 2009), whereas Zn deficiency causes about 1.5% of all deaths and about 20% of the perinatal mortality worldwide (Nriagu, 2007). Furthermore, a potential outcome of both metal deficiencies is neuropsychological impairment (Sandstead, 2000). Many of these cases of malnutrition could be solved with a diet enriched in Fe and Zn (De Benoist et al., 2008; White and Broadley, 2009). Foods rich in micronutrients such as meat and vegetables, unlike staple foods, are expensive and cannot be stored for long periods. Since rice is a staple food in large areas of the world, particularly in underdeveloped regions, biofortification of rice grains with Fe and Zn is a realistic target to alleviate these nutritional disorders. In most parts of the world, rice is traditionally cooked after milling and polishing, reducing the nutritional value because of the removal of the metal-rich bran and embryo, with only the endosperm remaining. Also, rice can be treated hydrothermally (parboiling) prior to milling to reduce breakage, increasing the nutritional value because of micronutrient transport from bran to endosperm, although changes of color, odor and texture, as well as mycotoxin risks, may arise (Rohman et al., 2014). Conventional, agronomic and transgenic approaches have been used for rice biofortification (reviewed by Garg et al., 2018), with approaches based on molecular genetics having the advantage that any gene with a demonstrated utility may be further used for targeting the biofortification of other crops. Furthermore, whereas diet diversification might be an option in principle, in practical terms poor people in developing counties cannot afford a diverse diet. Therefore, a major challenge for biofortification strategies in rice is to increase the concentrations of Fe and Zn in the endosperm.

Rice plants take up Fe from the soil using mechanisms classically ascribed to Strategies I and II (Ishimaru et al., 2006). Strategy I, used by non-graminaceous species, involves the uptake of Fe(II) via a Fe-Regulated Transporter (IRT) (Vert et al., 2002). Rice roots do express OsIRT1 and this transporter is strongly upregulated upon Fe deficiency (Ishimaru et al., 2006). In Strategy II, used by Gramineae, Fe acquisition is mediated by the synthesis and secretion of phytosiderophores (PSs) of the mugineic acid family (MAs) (Kobayashi et al., 2006). The synthesis of MAs starts from the condensation of three S-adenosyl methionine molecules to produce nicotianamine (NA) via nicotianamine synthase (NAS). Then, 2<sup>0</sup> -deoxymugineic acid (DMA) is synthesized from NA via nicotianamine aminotransferase (NAAT) and DMA synthase. In response to Fe deficiency, rice roots synthesize DMA (Takagi, 1993), which is secreted to the rhizosphere via TOM1 (Transporter Of Mugineic acid 1) (Nozoye et al., 2011). The secreted DMA is able to solubilize sparingly soluble Fe(III) by forming Fe(III)-DMA complexes, which are then taken up by root cells via transporters of the YSL (Yellow Stripe 1-Like) family (Curie et al., 2009). Zinc is usually taken up by plants as the free Zn(II) ion by root epidermal cells (Ishimaru et al., 2011; Sinclair and Krämer, 2012). Also, PSs can form Zn complexes that are as stable as Fe(III)-PS (Murakami et al., 1989), and both the secretion of PS and uptake of Zn-PS via YSL transporters have been observed in grasses (von Wirén et al., 1996; Suzuki et al., 2006; Widodo et al., 2010).

The short- and long-distance transport of Fe and Zn in grasses occurs both as free ions and metal complexes. Different PS-metal complexes have been found in plant fluids, including Fe(III)-DMA and Zn(II)-DMA in the xylem sap of wheat (Xuan et al., 2006), and Fe(III)-DMA and Zn(II)-NA in the phloem sap of rice (Nishiyama et al., 2012). Since both plant fluids transport nutrients from maternal to filial tissues at the reproductive stage (Waters and Grusak, 2007), Fe and Zn in grains can originate either from a direct root-to-seed route via xylem or from the remobilization of Fe from old and senescing leaves via phloem (Grillet et al., 2014; Yoneyama et al., 2015). An internal transport of these metals also occurs once they are in the grain, since the developing embryo is a sink for nutrients and the endosperm constitutes a nutrient reservoir. A complex network of transporters belonging to different families mediates the movement of both metals within cells and the whole plant; some proteins (e.g., IRT, P1B-type heavy metal ATPases) are capable of transporting Zn and Fe divalent ions (Takahashi et al., 2012; Kolaj-Robin et al., 2015), whereas others are capable of transporting Zn and Fe complexes (e.g., YSL family transport metal-NA or metal-PSs complexes; Curie et al., 2009).

The fact that mutations in the genes involved in NA and DMA synthesis and those of YSL transporters do not cause substantial decreases in the seed Fe concentrations (e.g., osnas3, Lee et al., 2009b) supports that metal transport systems in plants are redundant. On the other hand, the fact that the overexpression of genes involved in DMA/NA synthesis and those of YSL transporters only lead to limited increases in seed Fe concentrations (Banakar et al., 2017a,b) support the existence of regulatory feedback loops. Once in the rice grain, Fe may be stored in ferritins to avoid toxicity (Briat et al., 2010), sequestered in vacuoles (Kim et al., 2006) or bound to phytate, a P-rich molecule preferentially located in the aleurone layer (Persson et al., 2009). Phytate is also considered to control Zn levels in seeds (Raboy, 2003), and has a low bioavailability during human digestion (Guttieri et al., 2004; Gupta et al., 2015).

A range of transgenic approaches have been used to increase micronutrient concentrations in rice grains (Bashir et al., 2013; Nozoye, 2018), including: (i) increases in the expression of NA, DMA, YSL and ferritin synthesis genes; (ii) increases in absorption promoters, and/or (iii) decreases in inhibitors of absorption in human gut. The highest increase achieved so far in the concentrations of Fe and Zn in polished seeds has been achieved with an OsNAS2-SoyferH-1 construct, leading to 6- and 4-fold increases, respectively, over the WT values (Trijatmiko et al., 2016). Several studies have found a positive effect of increasing NA synthesis in achieving rice biofortification with Fe and Zn (Nozoye, 2018), but it is still unclear whether this is due to

the enrichment in NA itself and/or the concomitant enrichment in the NA derivative DMA. Moreover, most previous studies have focused on the grain endosperm, with the embryo, a part of the seed of outmost importance for seed formation, germination and viability, being studied in less detail.

In this work, the impact of contrasting levels of NA and DMA on the distribution of metals in the embryo and endosperm of rice seeds has been studied, using wild-type (WT) rice and six transgenic lines overexpressing OsNAS1 and/or expressing barley NAAT (HvNAATb). Increasing only DMA led to Fe enrichments in the embryo and endosperm, whereas increasing DMA in combination with NA produced Fe and Zn enrichments in both tissues. Laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) was also used to determine the spatial localization and concentrations of Fe, Zn and other elements within the embryo structures, providing the first quantitative set of data for these metals in the embryo tissues of biofortified rice seeds. We discuss the changes induced in the elemental distribution in contrasting DMA/NA scenarios, providing new insights into the Fe and Zn transport mechanisms within the embryo.

### MATERIALS AND METHODS

### Plant Material

Rice plants (Oryza sativa L. cv EYI 105) were transformed to obtain genotypes overexpressing OsNAS1 and/or HvNAATb genes and therefore with high levels of NA and/or DMA. The details of the cloning, expression and transformation were described in detail in Banakar et al. (2017b). The six lines used are two overexpressing OsNAS1 alone (N1 and N2), two expressing HvNAATb alone (D1 and D2) and two expressing OsNAS1 and HvNAATb together (ND1 and ND2). Lines ND1 and ND2 were also used, although in different growth conditions, in Banakar et al. (2017b).

### Gene Cloning

The cDNAs of OsNAS1 (GenBank ID: AB021746.2, 999 bp) and/or HvNAATb (GenBank ID: AB005788.1, 1,656 bp) were cloned from roots of rice (O. sativa cv EYI 105) and barley (Hordeum vulgare L. cv Ordalie) grown in vitro on MS medium without Fe (Murashige and Skoog, 1962) for 2 weeks. Total RNA was extracted with RNeasy Plant Mini kit (Qiagen, Hilden, Germany) and 1 µg of RNA was used for reverse transcription using Omniscript RT Kit (Qiagen) by RT-PCR. The full-size cDNAs for OsNAS1 (999 bp) and HvNAATb (1,656 bp) were amplified by PCR using the primer combinations OsNAS1-BamHI-FOR (5<sup>0</sup> -AGG ATC CAT GGA GGC TCA GAA CCA AGA GGT CG-3<sup>0</sup> ) and OsNAS1-HindIII-REV (5<sup>0</sup> -AAA GCT TCA TAA TAT AGT GCG CCT GAT CGT CCG GCT GT-3<sup>0</sup> ), and HvNAATb-BamHI-FOR (5<sup>0</sup> -AGG ATC CAT GGC CAC CGT ACG GCC AGA GAG CGA CG-3<sup>0</sup> ) and HvNAATb-HindIII-REV (5<sup>0</sup> - AAA GCT TCT AGC AAT CAT CGC TCG CTC GAA TTT CTC-3<sup>0</sup> ), respectively. The products were transferred to the pGEM-T Easy vector (Promega, Madison, Wisconsin, USA) for sequencing and verification. The OsNAS1 and HvNAATb cDNAs were inserted at the BamHI and HindIII sites of expression vector pAL76 (Christensen and Quail, 1996), which contains the maize UBI1 promoter and first intron, and an Agrobacterium tumefaciens nos transcriptional terminator. A separate vector harboring the constitutive cauliflower mosaic virus 35S promoter (CaMV35S) was used to provide the hygromycin phosphotransferase (hpt) selectable marker (Christou and Ford, 1995).

### Rice Transformation and Growth Conditions

Rice embryos were isolated from Oryza sativa L. (cv EYI 105) mature seeds and grown on MS medium containing 2.5 mg L −1 2,4-dichlorophenoxyacetic acid (2,4-D) as in Sudhakar et al. (1998). After 7 days, embryos were incubated on highosmoticum MS medium (0.2 M mannitol, 2.5 mg L−<sup>1</sup> 2,4- D) for 4 h (Sudhakar et al., 1998; Valdez et al., 1998), and then bombarded with gold particles carrying the transgenes and the hpt selectable marker (Christou et al., 1991). Bombarded embryos were selected on MS medium supplemented with 2.5 mg L−<sup>1</sup> 2,4-D and 30 mg L−<sup>1</sup> hygromycin, and callus pieces were transferred sequentially to shooting and rooting medium containing hygromycin (Sudhakar et al., 1998; Valdez et al., 1998).

Regenerated plantlets were transferred to pots filled with substrate (Traysubstract, Klasmann-Deilmann GmbH, Geeste, Germany) and grown flooded in large trays in growth chambers at 26◦C, 900 µmol m−<sup>2</sup> s <sup>−</sup><sup>1</sup> PPFD PAR with a 12/12 h light/dark regime and 80% relative humidity. Plants were watered with 100 µM Fe(III)-EDDHA (Sequestrene 138 Fe G-100; Syngenta Agro SA, Madrid, Spain) until flowering, and then self-pollinated. The Fe(III)-EDDHA solution in the trays was replaced every week. T<sup>0</sup> plants were grown to maturity, T<sup>1</sup> seeds were harvested and the resulting plants were crossed over two generations to generate a homozygous T<sup>3</sup> population. T<sup>3</sup> seeds from the transgenic lines were germinated on <sup>1</sup>/<sup>2</sup> MS medium containing 30 mg L−<sup>1</sup> hygromycin, whereas WT seeds were germinated on <sup>1</sup>/<sup>2</sup> MS medium without hygromycin. Five-day-old seedlings from wild type and transgenic lines were transferred to pots filled with substrate as described above, and maintained until the T<sup>4</sup> seeds had matured. Sampling of the T<sup>3</sup> flag leaf was performed to confirm the expression of the genes of interest and T<sup>4</sup> seeds were harvested to study the metal quantitative distribution in different seed tissues.

The anatomical denominations in this study comply with the monograph by Hoshikawa (1989). Grains from the same panicle were harvested, and some of them were stored at 4◦C until metal localization analyses and others de-husked by hand, to avoid metal contamination from the de-husking machine. Then, brown seeds were separated into embryo and endosperm, in both cases maintaining the corresponding aleurone layer, using new stainless steel razor blades and binocular magnifying glasses. For each genotype, embryo and endosperm samples were obtained pooling materials from 50–100 and 10 seeds, respectively, and 3– 4 replications were used. Samples were ground in liquid N<sup>2</sup> with ceramic mortar and pestle until a fine powder was obtained, and aliquots were stored at −80◦C until analysis.

### RNA Blot Analysis

fpls-09-01190 August 18, 2018 Time: 11:57 # 4

Total RNA was isolated from the flag leaf (T<sup>3</sup> generation) using the RNeasy Plant Mini Kit (Qiagen), and 20-µg aliquots were fractionated on a denaturing 1.2% agarose gel containing formaldehyde before blotting. The membranes were probed with digoxigenin-labeled partial OsNAS1 or HvNAATb cDNAs at 50◦C overnight, using DIG Easy Hyb (Roche Diagnostics, Mannheim, Germany). After washing and immunological detection with anti-DIG-AP (Roche Diagnostics) according to the manufacturer's instructions, CSPD chemiluminescence (Roche Diagnostics) was detected on Kodak BioMax light film (Sigma–Aldrich, St Louis, MO, United States).

#### Analysis of Metals, Nicotianamine and 2 0 -Deoxymugineic Acid in the Embryo and Endosperm

Fifty mg of ground embryo or endosperm tissue from grains of the same panicle were dried at 60◦C and digested with ultrapure 21% HNO<sup>3</sup> (Trace Select Ultra, Fluka) and 6% H2O<sup>2</sup> (Suprapur, Merck) for 55 min at 190◦C in an Ethos 1 microwave oven (Milestone Srl., Sorisole, Italy). Digests were analyzed (3 independent replicates) for Fe, Mn, Cu, and Zn by inductively coupled plasma mass spectrometry (ICP-MS), using an Agilent 7500ce (Agilent, Santa Clara, CA, United States) and monoelemental standard solutions for ICP-MS (Inorganic Ventures, Christiansburg, VA, United States). Recoveries were 98.3, 95.8, 97.0, and 95.0% for Fe, Mn Cu, and Zn, respectively, and limits of detection were 20, 2, 2 and 20 µg L−<sup>1</sup> for Fe, Mn Cu, and Zn, respectively. Concentrations are expressed as µg metal g tissue DW−<sup>1</sup> .

Nicotianamine and DMA were extracted from 50 mg of ground embryo or endosperm tissue (3–4 independent replicates) with 300 µL Type I water containing 18 µL of 1 mM nicotyllysine -used as internal standard- following the procedure previously developed for rice seeds (Banakar et al., 2017a). The extracts were analyzed for NA and DMA using an Alliance 2795 HPLC system (Waters, Mildford, MA, United States) coupled to a time-of-flight mass spectrometer (MS-TOF; MicrOTOF, Bruker Daltonics, Bremen, Germany) equipped with an electrospray (ESI) source. For a detailed description of the method, see Banakar et al. (2017a). The limit of quantification (LOQ), defined as the concentration giving a signal to noise ratio of 10, was 1.0 µg g −1 tissue FW for NA and DMA. All ligand concentrations are expressed as µg NA (or DMA) g tissue FW−<sup>1</sup> .

### Imaging Elemental Distribution in Seed Sections

### Sample Preparation

Thin sections (50–70 µm-thick) were obtained from fully developed, de-husked rice grains of the different genotypes using a vibrating blade microtome (VT1000 S, Leica Microsystems GmbH, Wetzlar, Germany), following the protocol described by Johnson et al. (2011). Seeds were glued (with Loctite Super Glue-3, Barcelona, Spain) to the excised bottom of a 1.5 mL plastic Eppendorf tube, and blades used were Chrome Platinum (Bic, Clichy, France). Vibratome parameters were a blade movement speed of 0.2 mm s−<sup>1</sup> and a vibration frequency of 70 Hz. A 100 µm-thick Kapton polyimide film (DuPont, Des Moines, IA, United States) was used to hold the tissue section as cutting proceeded, to minimize endosperm fragmentation (Johnson et al., 2011). Longitudinal dorso-ventral seed sections were used for optical microscopy, Perl's staining and LA-ICP-MS analysis. Sections were transferred to synchrotron adhesive tape (Leica), attached to glass slides, observed with a stereomicroscope (MZ16, Leica) and images taken with the Leica Application Suite V3.5. Sections were stored at 4◦C until LA-ICP-MS analysis.

### Perl's Prussian Blue Staining

Seed sections (60 µm-thick) were used immediately to localize Fe using Perls staining. Sections were incubated with a solution containing 2% K4[Fe(CN)6] and 2% HCl for 15 min at room temperature. This staining technique allows for the detection of labile Fe forms, including Fe complexes with NA and citrate, Fe hydroxides and inorganic Fe, as the blue compound ferric ferrocyanide (Roschzttardtz et al., 2009; Rios et al., 2016). Stained sections were observed with a stereomicroscope (MZ16, Leica) and images taken with the Leica Application Suite V3.5.

### LA-ICP-MS Analysis

Rice seed sections (60 µm-thick) adjacent to those used for Perl's staining were placed on synchrotron adhesive tape (Leica) and directly analyzed using a Laser Ablation (LA) system (LSX-213, Teledyne Cetac Technologies, Omaha, NE, United States) coupled to an ICP-MS instrument (Element II, Thermo Fischer Scientific, Waltham, MA, United States). Preliminary analyses were first carried out by driving LA straight lines through whole longitudinal dorso-ventral seed sections, and intense ICP-MS signals for the elements of interest were observed only in the embryo, with the endosperm providing very low or no signal. Optimized LA settings allowed for distinguishing embryo structures and obtaining signals of good intensity for the different elements, but the analysis time required for a single whole seed section was longer than 15 h. Therefore, the LA-ICP-MS analyses had to be restricted to the embryo and neighboring endosperm.

Optimized settings used for the LA-ICP-MS analysis (e.g., laser spot diameter and scan speed) are shown in **Table 1**. The net intensity of the signal obtained for each element ( <sup>31</sup>P, <sup>32</sup>S, <sup>55</sup>Mn, <sup>56</sup>Fe, <sup>63</sup>Cu, <sup>64</sup>Zn) was normalized using that of <sup>13</sup>C as an internal standard for quantification purposes. Quantification of the selected elements was carried out using two certified reference materials (CRMs) for calibration: the rice flour standards NIST 1568b (National Institute of Standards and Technology, United States) and NCS ZC73028 (LGC Standards, UK). Powdered CRMs were pressed to pellets 5 mm in diameter using a laboratory press (applying 2 t for 5 min) and subsequently analyzed using the same experimental conditions optimized for the seed sections. Three ablation lines were performed for each CRM along the pellet, and the resulting normalized intensity signals, together with respective elemental concentrations were used to build calibration curves. The linear regression equations

and coefficients for <sup>31</sup>P, <sup>32</sup>S, <sup>55</sup>Mn, <sup>56</sup>Fe, <sup>63</sup>Cu, <sup>64</sup>Zn calibration curves are shown in **Supplementary Table S1** and the calibration curve obtained for <sup>32</sup>S is shown as an example in **Supplementary Figure S1**. Quantitative two-dimensional images of the elemental distributions in seed sections were created using the software packages Origin <sup>R</sup> (OriginLab, Northampton, MA, United States) and ImageJ (NIH, Bethesda, WA, United States). Each data point (or pixel) was converted from intensities into concentrations, using the calibration curve for each element of interest. Then, quantitative elemental map images were first obtained by processing concentration data with the software Origin <sup>R</sup> . Secondly, the concentration data for Fe and Zn were also processed with the software ImageJ <sup>R</sup> , which permits obtaining images with a higher resolution; these maps were obtained for the WT and one line each from the three transgenic types. In the images processed with ImageJ, different concentration scales were used for each section to highlight differences in metal localization and concentration along the seed structures.

Limits of detection (LODs) were calculated by using the 3s criterion (3sb/S), where s<sup>b</sup> is the standard deviation of 5 independent measurements of the blank value in counts per s and S is the sensitivity for the corresponding analyte isotope obtained by measuring the CRM NIST 1568b. At the selected operating conditions and using the CRM pellets, LODs were (in µg g−<sup>1</sup> ) 0.77 for <sup>31</sup>P, 11 for <sup>32</sup>S, 0.10 for55Mn, 0.62 for <sup>56</sup>Fe, 0.049 for <sup>63</sup>Cu, and 0.13 for <sup>64</sup>Zn, with the higher LOD for S being due to the relatively high background. The detection limits in the rice seed sections were higher, likely due to differences between the thin seed sections and the pressed pellets used for rice flour standards.

### Statistical Analysis

Statistical analysis was carried out with SPSS Statistics (v.22, IBM, New York, NY, United States) using (i) ANOVA test (P ≤ 0.05) to determine differences between data from transgenic and WT plants and (ii) bivariate Pearson correlation to determine whether significant linear relationships existed between the



FIGURE 1 | RNA blot analysis showing transgene expression in the leaf tissue of wild-type (WT) and six independent transgenic lines, two expressing OsNAS1 (lines N1 and N2), two expressing HvNAATb (lines D1 and D2) and two co-expressing OsNAS1 and HvNAATb (lines ND1 and ND2). rRNA, ribosomal RNA. Plants were grown under nutrient-sufficient conditions, and total RNA was isolated from flag leaves at physiological maturity.

concentrations of metal ligands and metals in the embryo and endosperm.

### RESULTS

We co-transformed rice mature seed-derived embryos with separate constructs harboring rice NAS (OsNAS1) and/or barley NAAT (HvNAATb), along with the selectable marker hpt. RNA blot analysis using rRNA isolated from flag leaf tissue identified lines expressing OsNAS1 alone (N1 and N2), HvNAATb alone (D1 and D2) and OsNAS1 and HvNAATb together (ND1 and ND2) (**Figure 1**). There was considerable variation in the transgene expression levels among the lines used: the expression of OsNAS1 was higher in N1 than in N2, the expression of HvNAATb was much higher in D2 than in D1, and the expression of both OsNAS1 and HvNAATb were higher in ND2 than in ND1. Brown seeds from these plants were harvested (T<sup>4</sup> seeds), separated into embryo and endosperm, in both cases including their corresponding aleurone layer, and analyzed for NA, DMA and metals.

### Nicotianamine and 2<sup>0</sup> -Deoxymugineic Acid Concentrations in the Embryo and Endosperm

The embryo and endosperm of the WT and transgenic lines were analyzed by HPLC-ESI-MS to determine the concentrations of NA and DMA (**Figure 2**). This study is, to the best of our knowledge, the first to report NA and DMA concentrations in the rice embryo, since previous studies have only used polished and/or unpolished seeds, without analyzing embryos. In WT seeds, the NA concentration was 6.0 ± 1.5 µg g−<sup>1</sup> FW in the embryo and below the LOQ (marked as BLQ in **Figure 2**) in the endosperm, whereas the DMA concentrations were 24.2 ± 1.4

and ND2). Plants (WT and T<sup>3</sup> transgenic lines) were grown under nutrient-sufficient conditions, and the WT and T<sup>4</sup> seeds were harvested at physiological maturity. Asterisks indicate significant differences between WT and transgenic plants as determined by Student's t-test (P < 0.05). Values shown are means ± SE, n = 3–4. BLQ: below limit of quantification.

and 13.5 ± 1.0 µg g−<sup>1</sup> in the embryo and endosperm, respectively (**Figure 2**). Therefore, the embryo was richer in NA and DMA than the endosperm; the DMA/NA ratios were approximately 4 and very high in the embryo and endosperm, respectively (**Supplementary Figure S2**). DMA was also reported to be more abundant than NA in other rice compartments, including seeds (polished and unpolished; Masuda et al., 2009; Lee et al., 2011; Masuda et al., 2012; Trijatmiko et al., 2016), roots and leaves (Masuda et al., 2009), as well as phloem and xylem saps (Kato et al., 2010; Ando et al., 2013).

In transgenic lines, changes in NA and DMA concentrations in the embryo and the endosperm were observed when compared to the WT (**Figure 2**). The embryo NA concentrations of lines N1, N2 and ND1 were 15-, 2.5-, and 3.0-fold higher, respectively, whereas endosperm NA concentrations were well above the LOQ (29.1 ± 4.8, 6.4 ± 0.1, and 16.8 ± 0.3 µg g−<sup>1</sup> , respectively). In contrast, in lines D1, D2 and ND2, the embryo NA concentrations were lower than those of the WT (although not significantly at P ≤ 0.05). The embryo DMA concentrations of lines N1, D1, ND1 and ND2 were 3.5-, 4.2-, 4.1-, and 12 fold higher, respectively, than those in the WT. The endosperm DMA concentrations in lines N1, N2, ND1 and ND2 were also much higher than those in the WT (10.5-, 6.2-, 9.2-, and 19.0-fold, respectively), whereas they increased 2-fold in D2 and decreased by 71% in D1.

As a result of these changes, the DMA/NA ratios in the embryo of N1 and N2 (1 and 3, respectively) were lower than those in the WT (**Supplementary Figure S2**). Conversely, in D1, D2, ND1, and ND2 the DMA/NA ratios in the embryos were higher than in the WT, 34, 10, 7 and 185, respectively. On the other hand, the DMA/NA ratios in the endosperm were very high in D1, D2, and ND2 (in these three lines the NA concentrations were below the LOQ), whereas in N1, N2 and ND1 they were much lower, 5, 13 and 7, respectively.

### Metal Micronutrient Concentrations in the Embryo and Endosperm

The embryo and endosperm of the WT and transgenic lines were analyzed by ICP-MS to determine the concentrations of Fe, Zn, Mn, and Cu (**Figure 3**). In the WT, the Fe concentrations were 99 ± 6 and 20 ± 1 µg g−<sup>1</sup> DW in the embryo and endosperm, respectively. In lines N1, N2, ND2, and ND1 the Fe concentrations were significantly higher both in the embryo (1.3 to 2.1-fold increases over the WT values) and the endosperm (1.7- to 2.9-fold increases). In contrast, in D1 and D2 the Fe concentrations in the embryo and endosperm were not affected significantly, with the exception of a 33% decrease in the endosperm of D1.

The Zn concentrations in the WT were 102 ± 2 and 21 ± 1 µg g −1 in the embryo and endosperm, respectively (**Figure 3**). In the embryo, Zn concentrations increased significantly over the WT values in N1 (1.3-fold) and N2 (1.2-fold), whereas they decreased by 40% in ND2. In the endosperm, Zn concentrations increased over the WT values in N1 (2.5-fold), N2 (1.8-fold), and ND1 (2.5 fold).

The Mn concentrations in the WT were 76 ± 1 and 16 ± 2 µg g −1 in the embryo and endosperm, respectively (**Figure 3**). Significant changes in the Mn concentration in the embryo were only observed in N2, ND1 (a 18% decrease in both lines) and D1 (a 1.3-fold increase). In the endosperm, the only significant change in Mn concentration was found in N1, which showed a 1.4-fold increase when compared to the WT.

The Cu concentrations in the WT were 10 ± 1 and 7 ± 1 µg g −1 in the embryo and endosperm, respectively (**Figure 3**). In the embryo, the only significant changes in Cu concentrations were in N2 (a 1.3-fold increase), and ND1 and ND2 (2-fold increases). In the endosperm, the only changes found in Cu concentrations were in N1 (a 27% decrease), and N2 and ND1 (1.2 and 1.4-fold increases, respectively).

### Correlations Between NA, DMA and Metal Micronutrients in the Embryo and Endosperm

The relationships between the concentrations of metal ligands and metals in the embryo and endosperm were studied by bivariate Pearson correlation analysis, using all data (in nmol g <sup>−</sup><sup>1</sup> DW) from the WT and transgenic lines (**Supplementary Table S2**). Correlations found were different in the endosperm and embryo. In the embryo, DMA was positively correlated with Cu (r = 0.610; P ≤ 0.05), whereas NA was positively correlated with Fe and Zn (r = 0.643 and 0.647, respectively; in both cases significant at P ≤ 0.05). In the endosperm, there were highly significant and strong positive correlations between DMA and Fe (r = 0.702; P ≤ 0.01) and DMA and NA (r = 0.943; P ≤ 0.01), whereas no significant correlation was found between DMA and other metals (**Supplementary Table S2**). There was no correlation between NA and metals, whereas several correlations between metals were found, including Mn vs. Zn (r = 0.782; P ≤ 0.01), Mn vs. Cu (r = 0.725; P ≤ 0.01) and Fe vs. Zn (r = 0.566; P ≤ 0.05). Correlations between Fe and Zn have been found in previous studies (Lombi et al., 2011; Anuradha et al., 2012; Banakar et al., 2017b; Kampuang et al., 2017), and are expected due to the fact that both metals share mechanisms for uptake, short- and long-distance transport in the plant and intracellular trafficking.

In some cases, parameters measured in the embryo showed correlations with those measured in the endosperm (**Supplementary Table S2**). There were strong positive correlations between NA in the endosperm and DMA in the embryo (r = 0.986; P ≤ 0.01), and negative correlations between endosperm Fe and Cu, and embryo Cu (r = −0.644 and −0.604, respectively; in both cases at P ≤ 0.05).

### Perl's Staining of Rice Seed Sections

The Perl's staining of longitudinal dorso-ventral sections of WT and transgenic seeds reveal the accumulation of Fe in the embryo region and the aleurone layer, whereas the blue color was absent in the endosperm (an optical image of the WT seed is shown in **Figure 4A**, and the Perl's stain is shown in **Figure 4B**). The distribution of Fe in the embryo differed among genotypes studied. In the WT, Fe was accumulated mainly in the epithelium, and at lower levels in the scutellum and some parts of the aleurone layer. In N1 (overexpressing only OsNAS1) the whole embryo had more labile Fe than the WT. This increase in labile Fe was marked in the root primordia and scutellum, and also visible in the epithelium, the tip of the leaf primordia and the aleurone layer that covers the embryo and endosperm. While the pattern of Fe over-accumulation was similar among lines overexpressing OsNAS1, in ND2 (co-expressing OsNAS1 and HvNAATb) there was a much higher Perl's stain than in N1, when OsNAS1 was overexpressed alone. In D2 (expressing only HvNAATb) there was an over-accumulation of labile Fe in specific areas of the embryo, mainly in the root primordia and scutellum, although this effect was much less intense than in N1, whereas the Perl's stain in the epithelium and aleurone layer did not differ from the WT.

### Quantitative Images of the Elemental Distributions in Rice Seed Sections Obtained by LA-ICP-MS

The quantitative two-dimensional images obtained for Fe, Zn, Mn, Cu, P and S distribution in the seeds are shown in **Figures 5**, **6**, with the lowest and highest concentrations being represented in dark blue and red, respectively. First, maps are presented together with the corresponding optical images of the same sections, using the same scale for all genotypes (**Figure 5**). The match between elemental maps and optical images allows for the allocation of elemental concentrations to the different embryo structures. Also, high resolution maps were drawn using different concentration scales for each sample to depict optimal contrast (**Figure 6**).

This is the first time quantitative LA-ICP-MS elemental map images (Fe, Zn, Mn, Cu, P, and S) have been obtained for the embryo structures of WT and biofortified rice seeds. Previous studies applying LA-ICP-MS imaging to rice seeds did not provide quantitative images for Fe, and those for Zn had only a low resolution (Wirth et al., 2009; Basnet et al., 2014, 2016). The first remarkable aspect in the elemental distribution was the

preferential accumulation of most elements in the embryo, with the element concentrations in the endosperm being generally lower, with the only exceptions of Cu and S (**Figure 5**). This is in agreement with previous results reported by using semiquantitative, high-resolution techniques such as synchrotronbased X-ray fluorescence spectroscopy (Lombi et al., 2009; Takahashi et al., 2009; Wirth et al., 2009; Johnson et al., 2011; Iwai et al., 2012; Lu et al., 2013; Kyriacou et al., 2014) and secondary ion mass spectrometry (Kyriacou et al., 2014). Since most seed parts are quite heterogeneous, their composition are best described in terms of concentration ranges for each element. g g

lines) were grown under nutrient-sufficient conditions. WT and T<sup>4</sup> seeds were harvested at physiological maturity.

In the case of Fe, there were large differences in concentrations between seed parts, with the distribution being quite different in the WT and some of the transgenic lines, with the exception of the endosperm, where Fe concentrations were below 10 µg Fe g−<sup>1</sup> for all genotypes (**Figures 5**, **6**). In the WT, the highest Fe concentrations were found in the epithelium and root primordium (70–300 µg Fe g−<sup>1</sup> ), whereas lower concentrations were found in the aleurone layer (10–140 µg Fe g−<sup>1</sup> ), leaf primordium (25–100 µg Fe g−<sup>1</sup> ) and scutellum (25–75 µg Fe −1 ) (**Figures 5**, **6**). In N1 and N2, the Fe concentration was increased in the scutellum (to the range 150–650 µg g−<sup>1</sup> in both lines), leaf primordium (150–950 and 150–350 µg g−<sup>1</sup> in N1 and N2, respectively), root primordium (150–400 and 150–600 µg −1 in N1 and N2, respectively) and aleurone layer (70–300 µg Fe g−<sup>1</sup> in both lines) (**Figure 5**; see also N2 in **Figure 6**). Lines D1 and D2 showed lower Fe concentrations in most embryo tissues than those found in the WT, with D1 being more affected than D2 (**Figure 5**; see also D2 in **Figure 6**). Iron concentrations decreased in D1 and D2 in the scutellum (below 50 µg g−<sup>1</sup> ) and

HvNAATb (lines ND1 and ND2). Plants (WT and T<sup>3</sup> transgenic lines) were grown under nutrient-sufficient conditions, and the WT and T<sup>4</sup> seeds were harvested at physiological maturity and dehusked. 60 µm-thick seed sections were used for LA-ICP-MS analysis. Elemental images were obtained processing LA-ICP-MS data with the software Origin <sup>R</sup> . Color scales represent the concentrations for each element, with the lowest ones in dark blue and the highest ones in red. Depending on the element, the scale bar indicates the elemental concentrations in µg g−<sup>1</sup> (Fe, Zn, Mn, and Cu) and in mass fraction percentage (S and P). For a given element, the same scale was used in all genotypes. Pictures shown in the first row are optical images of the sections subjected to LA-ICP-MS analysis. AL, aleurone layer; LP, leaf primordium; RP, root primordium; SC, scutellum; SE, starchy endosperm; EP, epithelium.

the root primordium (<50 µg g−<sup>1</sup> and to 50–100 µg g−<sup>1</sup> in D1 and D2, respectively), whereas decreased in the leaf primordium and epithelium only in D1 (<50 and 50–100 µg g−<sup>1</sup> , respectively) and increased in the aleurone layer only in D1 (50–250 µg g−<sup>1</sup> ). Lines ND1 and ND2 showed large Fe concentration increases in the epithelium (200–900 µg g−<sup>1</sup> ; **Figure 5**; see also ND1 in **Figure 6**), compared to the values found in the WT and the other four transgenic lines. Other relevant changes in Fe distribution in ND1 and ND2 were: (i) increases in root primordium Fe concentrations over those in the WT but not always over those in N2 (up to 200–900 and 100–400 µg g−<sup>1</sup> , respectively), (ii) increases in scutellum Fe concentrations over those in the WT but only slightly higher than those in N1 and N2 (150–550 and 100–350 µg g−<sup>1</sup> , respectively), and (iii) decreases in the leaf primordium Fe concentrations (<50 µg g−<sup>1</sup> ) below those in the WT.

In the case of Zn, large differences in concentrations were also observed between the different seed parts, with modifications in the distribution in the transgenic genotypes when compared to the WT, whereas Zn could not be detected in the endosperm of any genotype (**Figures 5**, **6**). The WT seed showed the highest Zn concentration in the leaf primordium (500–1100 µg g−<sup>1</sup> ), followed by the scutellum and epithelium (50–200 µg g−<sup>1</sup> ) and the aleurone layer (<100 µg g−<sup>1</sup> ). Lines N1 and N2 showed lower concentrations of Zn than those found in the WT in the leaf primordium (300–800 and 50–1100 µg g−<sup>1</sup> , respectively), whereas in the root primordium, scutellum, epithelium and aleurone layer the Zn concentrations increased only in N2 (50– 1500, 200–550, <500 and <200 µg g−<sup>1</sup> , respectively) when compared with the WT. In D1 and D2, Zn concentrations in both the leaf primordia (<600 and 200–500 µg g−<sup>1</sup> , respectively) and scutellum (50–100 and <75 µg g−<sup>1</sup> , respectively) were lower than those in the WT. Other changes in Zn distribution when compared with the WT were different for these two lines: Zn increased in D1 in the root primordia and epithelium (500–1000 and 150–350 µg g−<sup>1</sup> , respectively) and decreased in D2 (<75 µg g −1 in both tissues). In ND1 and ND2 Zn concentrations in the leaf primordia (300–700 µg Zn g−<sup>1</sup> , respectively) were lower than those in the WT, and similar to those found in N1 and N2. Also, moderate increases in the Zn concentrations in the root primordium and scutellum were observed in the two double transgenic lines, especially when compared with those in N1.

Manganese could not be detected in the endosperm of any genotype, and was generally localized in high concentrations in the leaf and root primordia in all genotypes (100–400 and 100– 130 µg g−<sup>1</sup> , respectively), with concentrations in the epithelium and aleurone layer being also similar (<40 and 50–150 µg g−<sup>1</sup> , respectively; **Figure 5**). The only exception was ND1, which

showed higher Mn concentrations in all tissues (100–500 µg Mn g −1 ; **Figure 5**).

Copper was the less abundant micronutrient of those investigated, and differences in distribution between lines were less marked (**Figure 5**). In the WT, this element was mostly found in the aleurone layer, scutellum and leaf primordium (concentrations in the range of 5–15 µg Cu g−<sup>1</sup> ), with Cu concentrations in the rest of the tissues being ≤10 µg g−<sup>1</sup> , with the exception of the inner endosperm, where Cu could not be detected. In N1, N2, ND1 and ND2 some changes in Cu distribution were observed: increases in Cu concentrations in the root primordia (9–12, 8–25, 10–30 µg g−<sup>1</sup> in N1, ND1 and ND2), leaf primordia (9–40 µg g−<sup>1</sup> in N1), and aleurone layer (10–30, 8–25, 8–25 µg g−<sup>1</sup> in N2, ND1 and ND2). In contrast, in D1 and D2 no changes were observed in the Cu distribution.

Phosphorus was present with similar distribution and concentrations in all genotypes used (**Figure 5**). This element was mainly located in the embryo and the aleurone layer (at concentrations between 1 and 2%) and was not detected in the endosperm. Sulfur was present in the whole embryo and the endosperm, and it was mainly located in the leaf and root primordia and the aleurone layer (at concentrations of approximately 0.2%), without any consistent difference among genotypes (**Figure 5**). It is also worth to remark the gradient of S concentrations, from high in the aleurone layer to low in the endosperm.

### DISCUSSION

Enhancing NA synthesis via genetic transformation has been shown to increase the concentrations of Fe and Zn in rice seeds (Nozoye, 2018). However, it was still not known whether this was due to the enrichment in NA itself or to the subsequent enrichment in its derivative DMA, and whether the changes in the relative levels of both ligands may have an effect on the partitioning of metals between the embryo and endosperm. In this study we analyzed the concentrations of NA, DMA and metals in the embryo and endosperm of WT rice and six transgenic lines, overexpressing OsNAS1 and/or expressing barley NAAT (HvNAATb), which provided contrasting levels of DMA and NA. This allows for outlining three different DMA/NA scenarios for metal distribution in the rice seed (**Figure 7**) that are discussed below. Results show that increasing DMA alone led to Fe enrichment in the embryo and endosperm, whereas increasing DMA in combination with NA led to Fe and Zn enrichment in both tissues (**Figure 7**).

### First DMA/NA Scenario

In a first scenario, an enhanced NA level would not be fully depleted because of the limited capacity to use NA for DMA synthesis (**Figure 7**). Lines complying with this scenario were those having an enhanced expression of OsNAS1 alone (N1 and N2), or in combination with a low expression of barley NAATb (ND1), and showed consistent enrichments in NA, DMA, Fe and Zn in the endosperm, and also to enrichments of NA, DMA and Fe in the embryo (**Figures 1–3**). In the endosperm of these lines, the increases in DMA and NA concentrations (6- to 10-fold and 6- to 29-fold, respectively) were much larger than those found for Fe and Zn (1.7- to 2.0-fold and 1.7- to 2.5-fold, respectively), and occasionally accompanied of moderate changes in other metals (e.g., a 1.4-fold increase for Cu in line ND1). Other rice transgenics complying with this DMA/NA scenario are those overexpressing OsNAS1-3 (Lee et al., 2009b) or HvNAS1 alone (Masuda et al., 2009), as well as OsNAS2 in combination with SoyferH1 (Trijatmiko et al., 2016), which showed concentration increases in polished seeds (endosperm) of 2- to 33-fold for DMA, 5- to 32-fold for NA, 2.0- to 7.5-fold for Fe and 2.2- to 3.8-fold for Zn. The same scenario occurs when overexpressing OsNAS1 in combination with HvNAATb (Banakar et al., 2017b), leading to increases in polished seeds of 33-, 160-, 4.0-, and 4.1 fold for DMA, NA, Fe, and Zn concentrations, respectively; that study also shows an increased abundance of NA and DMA in leaves and roots, which promotes Fe and Zn uptake, root-toshoot translocation, and finally seed loading. Other transgenics also showed concomitant Fe and Zn concentration increases, but the DMA and/or NA concentrations were not determined (Wirth et al., 2009; Boonyaves et al., 2017; Wu et al., 2018).

The elemental images of these transgenic lines (N1, N2, and ND1) confirmed that embryos were enriched in Fe, and also showed changes in the metal distribution pattern, with increases in Fe concentrations in the leaf primordium, scutellum and root primordium (**Figures 5**, **6**). The mobilization of Fe and Zn in the rice seed involves transport from the endosperm near the embryo to the epithelium, scutellum and then to the leaf and root primordia (Takahashi et al., 2009). In N1 and N2 embryos, the increases in NA and DMA concentrations (**Figure 2**) and low DMA/NA ratios (**Supplementary Figure S2**) would allow for the formation of Fe(II)-NA in addition to Fe(III)-DMA, with Fe transport occurring via YSLs such as OsYSL9/OsYSL2. OsYSL9 works with Fe(II)-NA and Fe(III)-DMA and is expressed in the endosperm adjacent to the embryo and scutellum (Senoura et al., 2017), whereas OsYSL2 functions with Fe(II)-NA [not with Fe(III)-DMA] and is expressed during seed development in the whole embryo (Koike et al., 2004) and in mature seeds in the epithelium, vascular bundle of the scutellum and leaf primordium (Nozoye et al., 2007).

In contrast, the embryos of ND1 were also enriched in DMA and NA (**Figure 2**) but had a slight increase in the DMA/NA ratio (**Supplementary Figure S2**). These embryos also showed an accumulation of Fe in the epithelium, scutellum and root primordium, but Fe in the leaf primordium was reduced (**Figures 5**, **6**). Since in this genotype Fe(III)-DMA would be favored over Fe(II)-NA, this reduction suggests that Fe transport to the leaf primordium occurs as Fe(II)-NA via OsYSL2, a transporter specific for Fe (and Mn) complexes with NA (but not with DMA) localized in the embryo (Koike et al., 2004).

The Zn distribution pattern in the embryo was also altered in this scenario when compared to the WT (**Figures 5**, **6**), especially in the leaf primordium, the structure with the highest Zn concentration. In the N1 and ND1 embryos, the extremely high levels of DMA + NA (6- and 4-fold higher than in the WT, respectively) resulted in Zn depletion not only of the leaf primordium, but also the root primordium when

NA was as abundant as DMA (in N1) (**Figures 5**, **6**). In contrast, the slight increase of DMA+NA in N2 (1.4-fold) resulted in smaller decreases in the Zn concentrations in the leaf primordium and increased Zn concentrations in the scutellum and root primordium. A large abundance of Zn chelators would diminish the pool of free Zn(II) ions, therefore limiting its availability for transport via OsZIP4 and/or OsIRT1 throughout the embryo, and more specifically toward the meristematic tissues where this metal tends to accumulate massively. OsZIP4 is expressed in the vascular bundle of the scutellum and the leaf and root primordium (Takahashi et al., 2009, 2011), whereas OsIRT1, which transports Zn in addition to Fe (Lee and An, 2009), is also expressed in embryo structures (Nozoye et al., 2007).

### Second DMA/NA Scenario

In a second DMA/NA scenario, an enhanced NA level would be depleted by an enhanced capacity to use NA for DMA synthesis (**Figure 7**). The line complying with this scenario, ND2, had an enhanced expression of OsNAS1 in combination with a high expression of HvNAATb, and showed enrichments only for DMA and Fe, both in the endosperm and embryo, whereas NA and Zn concentrations in the embryo decreased (by 76 and 40%, respectively) and Cu concentrations increased (2.0-fold), without any change in the endosperm concentrations of NA, Zn, Mn and Cu (**Figures 1–3**). In this line, the endosperm enrichments for DMA and Fe (19- and 2.9-fold, respectively) were much larger than those found in the lines of the first scenario (see above). This scenario also occurs with the constitutive expression of the strict Fe(III)-DMA transporter HvYS1 (Murata et al., 2006), which leads to increases in the concentrations of both DMA (2.3-fold) and Fe (2.1-fold) in polished seeds, without affecting the concentrations of NA, Zn and Mn (Banakar et al., 2017a). In a fully opposite DMA/NA scenario, the Osnaat1 mutant shows a large increase in NA accompanied by a large DMA depletion, resulting in an stimulated Fe(II) acquisition and a seed enrichment in Fe (1.8- and 3.8-fold increases in unpolished and polished seeds, respectively) but not Zn (Cheng et al., 2007).

The elemental images of ND2 seeds confirms that embryos were markedly enriched in Fe, and also showed changes in the metal distribution pattern, with an accumulation of Fe in the epithelium, scutellum and root primordium, and a depletion of Fe in most of the leaf primordium (**Figure 5**). This supports that the transport of Fe from the endosperm near the embryo to the epithelium, scutellum and root primordium can be via OsYSL9, mediated by Fe(III)-DMA in addition to Fe(II)-NA, since the extremely high levels of DMA and major depletion of NA (**Figure 2**) would strongly favor the formation of Fe(III)- DMA over Fe(II)-NA. On the other hand, the ND2 data also provide further support to the idea that the transport of Fe from the scutellum to the leaf primordium occur as Fe(II)-NA via OsYSL2.

The decrease in Zn and increase in Cu in the embryo in ND2 provides some hints on the partitioning of both metals in the rice seed. The possible Zn transport forms within the

rice grain include free Zn(II) ions, Zn(II)-NA, and Zn(II)- DMA, with the latter being unlikely to be relevant, since the Osnaat1 mutant shows no Zn partitioning phenotype in the grain (Cheng et al., 2007). In a previous study, the activation of OsNAS2 generated a new pool of bio-available Zn in the endosperm, mainly composed of Zn(II)-NA and Zn(II)-DMA (Lee et al., 2011). The excess of DMA in the embryo of ND2 may favor the formation of Zn(II)-DMA, hampering the transport of free Zn(II) ions from the endosperm via the highly selective transporter OsZIP4 (Ishimaru et al., 2005), expressed in the endosperm region adjacent to the epithelium in mature rice seeds (Takahashi et al., 2009). On the other hand, the excess of DMA and the depletion of NA in ND2 will difficult the formation of Zn(II)-NA, hampering transfer via YSLs using this metal chelate. OsYSL15 and OsYSL9 have not been assayed yet in this respect (Inoue et al., 2009; Lee et al., 2009a; Senoura et al., 2017), whereas others, including OsYSL2 (Koike et al., 2004), OsYSL16 (Zheng et al., 2012) and OsYSL18 (Aoyama et al., 2009) do not transport Zn(II)-NA. The Zn distribution pattern in the embryo was also altered in this scenario when compared to the WT (**Figure 5**), mainly affecting the leaf primordium, the structure with the highest Zn concentration. In the ND2 embryos, the extremely high levels of Zn chelators (DMA + NA) (10-fold higher than in the WT) resulted in a Zn depletion in the leaf primordium (**Figure 5**). As indicated above for N1 and ND1 (first DMA/NA scenario), the abundance of Zn chelators would tend to decrease the pool of free Zn(II) ions, therefore limiting transport via OsZIP4 and/or OsIRT1.

Copper in the embryo was increased in ND2, conversely to what occurs with Zn. In this second scenario, which includes high DMA and low NA availability, Cu complexation is favored, since the stability constants are higher for Cu [18.7 for Cu(II)- DMA, Murakami et al., 1989; 18.6 for Cu(II)-NA, Beneš et al., 1983] than for Zn [12.7 for Zn(II)-DMA, Murakami et al., 1989; 15.4 for Zn(II)-NA, Anderegg and Ripperger, 1989]. A likely candidate for Cu delivery to the embryo is OsYSL16, which is highly expressed in all tissues of developing seeds (Lee et al., 2012), and transports Cu(II)-NA and Fe(III)-DMA, but not Cu(II)-DMA, Fe(II)-NA and Zn(II)-NA (Kakei et al., 2012; Zheng et al., 2012). It is also possible that YSL2, YSL9 and YSL18, which are expressed in embryo and/or endosperm during seed development (Koike et al., 2004; Aoyama et al., 2009; Senoura et al., 2017), could be responsible for Cu delivery to the embryo, since YSLs can transport a broad range of substrates [for instance, ZmYS1 transports Fe(III)-DMA, Zn(II)-DMA, Cu(II)- DMA, Fe(II)-NA, Ni(II)-NA and others; Schaaf et al., 2004; Murata et al., 2006].

### Third DMA/NA Scenario

In the third DMA/NA scenario, the lack of sufficient NA replenishment would limit DMA synthesis, in spite of the enhanced capacity to use NA for this purpose (**Figure 7**). Lines complying with this scenario were those expressing HvNAATb alone (D1 and D2), and resulted in decreases in NA, variable changes in DMA and moderate decreases in Fe in the embryo and endosperm (**Figures 1–3**). A low expression of HvNAATb (D1) led to moderate decreases in DMA (26%) and Fe (35%) in the endosperm and to an accumulation of Fe in the aleurone layer (**Figure 5**), whereas in the embryo NA and Fe also decreased moderately (17 and 24%, respectively, in both cases significantly at P ≤ 0.10), DMA increased (3.9 fold) and Fe was depleted in all embryo structures (**Figure 5**). The presence of low levels of DMA in the endosperm would make more difficult to compete with phytic acid present in the aleurone layer, since Fe(III)-DMA and Fe(III)-phytic acid have similar stability constants (18.4 and 18.2, respectively; Murakami et al., 1989; Torres et al., 2005), and consequently, Fe would stay in the aleurone layer, limiting its transport to the inner endosperm and subsequently to the embryo. The high HvNAATb expression in D2 caused moderate increases in DMA (1.9-fold) in the endosperm and moderate decreases in NA and Fe in the embryo, with the Fe distribution pattern being unaffected.

### CONCLUSION

When the transgenic approach results in increases in the DMA concentration alone or in combination with NA (second and first DMA/NA scenarios, respectively), the prevalent mechanisms appear to be those based on Fe(III)-DMA, which enhance Fe transport and storage in the endosperm, likely using YSL transporters. When increases in DMA occur in combination with NA increases (first DMA/NA scenario), an additional mechanism based on Zn(II)-NA appears to be elicited, which boosts Zn transport and storage in the endosperm. However, when the transgenic approach results only in minor changes in the DMA levels (third DMA/NA scenario) there are no effects on the metal status in the seed. This knowledge can help designing future strategies for biofortification strategies in rice, using the selectivity of the different ligands and transporters. It should be kept in mind that in high-NA/DMA grains the bioavailability of Fe for mammals and humans is improved even when the Fe concentrations are unchanged (Zheng et al., 2010; Eagling et al., 2014). Our study demonstrates that a better understanding of transgenic plant phenotypes, using in-depth localized quantification of the targeted nutrients and related metabolites in plant tissues, will facilitate the application of more refined strategies for biofortification of staple crops.

### AUTHOR CONTRIBUTIONS

PC, BF, and AÁ-F conceived and designed the experiments. RB obtained the plant material. PD-B performed the HPLC-ESI-MS(TOF) analysis, obtained the seed sections, and performed Perl's staining. SR-M and BF performed the LA-ICP-MS analysis. PD-B prepared and analyzed the results, and drafted the manuscript. TC, RP, and RB analyzed critically the results. AÁ-F, JA, BF, and PC wrote, reviewed, and edited the paper. All the authors read and approved the final manuscript.

### FUNDING

This work was supported by the grants of the Spanish Ministry of Science, Innovation and Universities (AGL2016-75226-R, BIO2014-54426-P and AGL2017-85377-R, all co-financed with FEDER), Aragón Government (Group A09\_17R) and Generalitat de Catalunya (Grant 2017 SGR 828). PD-B was supported by a MINECO-FPI contract. RB was supported by a Ph.D. fellowship from the University of Lleida. SR-M was supported by a research contract from the Fundación Universidad de Oviedo (FUO-069-17). BF was supported by a MINECO research contract (RYC-2014-14985; "Ramón y Cajal Program").

### REFERENCES


### ACKNOWLEDGMENTS

The authors acknowledge J. J. Rios (CEBAS-CSIC, Murcia, Spain) for providing technical advice on Perl's staining.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018.01190/ full#supplementary-material

in monocotyledonous plants. Transgenic Res. 5, 213–218. doi: 10.1007/ BF01969712


reveals single-gene strategies for effective iron- and zinc-biofortification of rice endosperm. PLoS One 6:e24476. doi: 10.1371/journal.pone.0024476


fluorescence imaging of Fe, Zn, Mn, and Cu. Plant Soil 325, 39–51. doi: 10.1007/ s11104-009-0045-7


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Díaz-Benito, Banakar, Rodríguez-Menéndez, Capell, Pereiro, Christou, Abadía, Fernández and Álvarez-Fernández. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Overexpression of a High-Affinity Nitrate Transporter OsNRT2.1 Increases Yield and Manganese Accumulation in Rice Under Alternating Wet and Dry Condition

Bingbing Luo1,2† , Jingguang Chen2,3† , Longlong Zhu1,2, Shuhua Liu1,2, Bin Li1,2 , Hong Lu1,2, Guoyou Ye<sup>3</sup> , Guohua Xu1,2 and Xiaorong Fan1,2 \*

<sup>1</sup> State Key Laboratory of Crop Genetics and Germplasm Enhancement, Nanjing, China, <sup>2</sup> Key Laboratory of Plant Nutrition and Fertilization in Low-Middle Reaches of the Yangtze River, Ministry of Agriculture, Nanjing Agricultural University, Nanjing, China, <sup>3</sup> CAAS-IRRI Joint Laboratory for Genomics-Assisted Germplasm Enhancement, Agricultural Genomics Institute in Shenzhen, Chinese Academy of Agricultural Sciences, Shenzhen, China

### Edited by:

Huixia Shou, Zhejiang University, China

### Reviewed by:

Yuanhu Xuan, Shenyang Agricultural University, China Dong Liu, Tsinghua University, China Chuang Wang, Huazhong Agricultural University, China

> \*Correspondence: Xiaorong Fan xiaorongfan@njau.edu.cn

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 14 December 2017 Accepted: 25 July 2018 Published: 15 August 2018

### Citation:

Luo B, Chen J, Zhu L, Liu S, Li B, Lu H, Ye G, Xu G and Fan X (2018) Overexpression of a High-Affinity Nitrate Transporter OsNRT2.1 Increases Yield and Manganese Accumulation in Rice Under Alternating Wet and Dry Condition. Front. Plant Sci. 9:1192. doi: 10.3389/fpls.2018.01192 Nitrate and manganese (Mn) are necessary elements for the growth and development of rice in paddy soil. Under physiological conditions, we previously reported that the uptake of Mn in roots can be improved by the addition of external nitrate but not ammonium. To investigate the mechanism(s) of this phenotype, we produced plant lines overexpressing OsNRT2.1 and assessed Mn uptake under alternating wet and dry (AWD) and waterlogged (WL) conditions. Under AWD condition, we observed a 31% reduction in grain yields of wild type (WT) plants compared to WL condition. Interestingly, the overexpression of OsNRT2.1 could recover this loss, as OsNRT2.1 transgenic lines displayed higher grain yields than WT plants. We also observed 60% higher grain Mn in the transgenic lines in AWD condition and approximately 30% higher Mn in the grain of transgenic lines in WL condition. We further found that the overexpression of OsNRT2.1 did not alter Mg and Fe in the seeds in either growth condition. The reasons for the increased Mn content in OsNRT2.1 transgenic seeds in AWD condition could be explained by the elevated expression of OsNRAMP family genes including OsNRAMP3, OsNRAMP5, and OsNRAMP6 in node I, the panicleneck, and the flag leaves. The mechanism(s) underpinning the upregulation of these genes requires further investigation. Taken together, our results provide a new function of OsNRT2.1 in improving rice yields and grain Mn accumulation during water-saving cultivation patterns. This represents a new strategy for maintaining yield and improving food quality in a sustainable agricultural system.

Keywords: rice, OsNRT2.1, manganese uptake, yield, nitrate

### INTRODUCTION

Trace elements play a vital role in plant growth and development (Yan et al., 2006). All organisms require trace levels of manganese (Mn) for survival due to its necessity during plant metabolism and its participation in several important pathways (Socha and Guerinot, 2014) including the oxygen-evolving complex (OEX) of photosystem II (PS II). In addition, Mn plays an important

**65**

role during phosphoenolpyruvate carboxykinase activation and liquid metabolism (Dziwornu et al., 2018). Thus, it is required for photosynthesis indirectly by repressing thylakoid synthesis. In addition, manganese superoxide dismutase (MnSOD) is the major mitochondrial antioxidant defense enzyme (Shen, 2015) and Mn is a co-factor/activator of many enzymes involved in the catalysis of oxidation reduction, decarboxylation and hydrolytic reactions (Marschner, 1995; Xu et al., 2007).

Mn deficiency is a global problem in agriculture (Hebbern et al., 2005). Mn deficient plants are more vulnerable to cold stress and infections by pathogens, leading to decreased crop yields (Marschner, 1995; Hebbern et al., 2005). Addressing this issue is problematic as Mn2<sup>+</sup> rapidly oxidizes when supplemented into fertilizers. In this regard, further knowledge of molecular mechanisms that can enhance Mn delivery are required. Several Mn transporters contribute to the uptake, transport and maintenance of Mn homeostasis in plants. The NRAMP family was shown to participate in Mn transport during early plant discoveries. In Arabidopsis, AtNRAMP1 localizes to the plasma membrane and displays root-specific expression where its function is to coordinate the absorption of Mn from soil (Cailliatte et al., 2010). OsNRAMP5 is mainly involved in Mn uptake and accumulation in rice and its silencing significantly reduces Mn accumulation in shoots (Yang et al., 2014). OsNRAMP3 is expressed in the node and regulates Mn transport and tissue distribution in response to environmental changes (Yamaji et al., 2013a). OsNRAMP6 distributes to the plasma membrane and transports Mn and Fe, maintaining their balance in cells (Peris-Peris et al., 2017). In rice, Mn homeostasis is controlled by the YSL2/6 gene. OsYSL2 can promote the long-distance transport of Mn (Koike et al., 2004; Ishimaru et al., 2010). OsYSL6 belongs to the Mn-nicotianamine (NA) transporter family and is required for the detoxification of high concentrations of Mn (Sasaki et al., 2011). In addition, the CAX proteins belong to the Ca2+/cation antiporter (CaCA) superfamily (Emery et al., 2012) and are potentially involved in Mn2+/H<sup>+</sup> exchange to export Mn from the cytosol (Connorton et al., 2012).

Nitrogen (N) is an essential element for plant growth and development, especially for crops. Generally, N is absorbed by plants in the form of ammonium (NH<sup>4</sup> <sup>+</sup>) and nitrate (NO<sup>3</sup> <sup>−</sup>), but nitrate easily dissolves in water and is therefore lost to the environment (Jin et al., 2015). Roots acquire NO<sup>3</sup> − via transporters distributed throughout the whole plant (Xu et al., 2012). Plants adapt to the differing NO<sup>3</sup> <sup>−</sup> concentrations in soil by exploiting two forms of NO<sup>3</sup> <sup>−</sup> uptake, including low-affinity transporters (NRT1/NPF) and high-affinity NO<sup>3</sup> − transporters (NRT2) (Crawford and Glass, 1998). Particularly for rice plants, we previously identified a high-affinity NO<sup>3</sup> − transport system. The OsNRT2 gene family was found to play an important role during N uptake and translocation, requiring their partner protein NAR2 to perform this function, besides OsNRT2.3b (Tang et al., 2012; Xu et al., 2012; Chen et al., 2016b; Chen Z.C. et al., 2017; Fan et al., 2016; Chen J.G. et al., 2017).

Simultaneously, Mn can influence NO<sup>3</sup> <sup>−</sup> reductase activity and is associated with photosynthesis in plants (Botrill et al., 1970; Gong et al., 2011). Mn also influences N metabolism and regulates protein synthesis (Jiang, 2006). Studies have shown that the arabidopsis chl1-5 mutant lines display reduced NO<sup>3</sup> <sup>−</sup> uptake and a loss of AtIRT1 expression, which is responsible for Cd uptake into root cells (Muños et al., 2004; Lux et al., 2011). Fe deficiency was also shown to inhibit N metabolism in the roots and leaves of cucumber plants (Borlotti et al., 2012). These effects suggest that NO<sup>3</sup> <sup>−</sup> influences the uptake of trace elements in plants. In this study, we hypothesized that a close relationship between N and Mn in plants exists. We used transgenic rice overexpressing OsNRT2.1 to examine how the different forms of N influence Mn uptake and accumulation in grain.

### MATERIALS AND METHODS

### Plant Materials and Growth Conditions

We amplified the OsNRT2.1 (AB008519) ORF (primers are displayed in **Supplementary Table S1**) using cDNA obtained from Oryza sativa L. ssp. Japonica cv. Nipponbare. PCR products were cloned into the pMD19-T vector (TaKaRa Biotechnology, Dalian, China) and the expression vector pTCK303 containing a ubiquitin promoter. Positive clones were verified by restriction digest analysis and DNA sequencing. Next, the binary vector pUbiquitin-OsNRT2.1 was introduced into A. tumefaciens (strain EHA105), which was used to transform the rice embryonic callus as previously described (Ai et al., 2009). Hygromycin-resistant T0 generation transgenic rice plants were transplanted to soil and grown to obtain seeds in fields (Tang et al., 2012). Three independent T4 generation lines overexpressing OsNRT2.1 were used for further experiments.

Firstly, rice seedlings were selected and cultured in 1 mM (NH4)2SO<sup>4</sup> as the main source of N in nutrient solution (pH 5.5) for 1 month. Other elements and trace elements were supplied in IRRI (International Rice Research Institute) nutrient solution containing 0.35 mM K2SO4, 0.3 mM KH2PO4, 1 mM MgSO4, 1 mM CaCl2, 0.5 mM Na2SiO3, 20 µM H3BO3, 9 µM MnCl2, 20 µM EDTA-Fe, 0.77 µM ZnSO4, 0.32 µM CuSO4, and 0.39 µM (NH4)6Mo7O24. Rice were planted in a growth room (Thermoline Scientific Equipment Pty. Ltd., Smithfield, NSW, Australia) at 30◦C during the day and 22◦C at night with 16-h light/8-h of darkness. The light intensity was 400 µmol m−<sup>2</sup> s −1 and the relative humidity was 65– 70%. Wild type (WT) rice were then transferred to 0.25 or 1.25 mmol/L Ca(NO3)<sup>2</sup> and 0.25 or 1.25 mmol/L (NH4)2SO<sup>4</sup> nutrient solution, respectively, for 2 weeks (**Figure 1**). In **Figures 2** and **4**, WT and overexpression lines were transferred to 0.5 mM NH<sup>4</sup> <sup>+</sup>/NO<sup>3</sup> <sup>−</sup> nutrient solution for 2 weeks. For each line and treatment, four biological repeats were performed.

In field experiments, rice were planted in Nanjing, Jiangsu; a subtropical monsoon climate zone. The characteristics of the soil and N supply were as previously described (Chen et al., 2016b). For waterlogged (WL) treatment, rice fields were watered daily to maintain WT and transgenic lines in a flooded state. For alternating wet and dry treatment (AWD), lines were planted into fields and watered for a week, to keep the soil moist.

### Southern Blot Analysis

Transgene cope numbers were identified by southern blot analysis. Briefly, genomic DNA was extracted from the leaves of WT and transgenic lines and digested with HindIII and EcoRI. Digested DNA was separated on 1% (w/v) agarose gels, transferred to a Hybond-N<sup>+</sup> nylon membrane and hybridized using the hygromycin-resistant gene.

### RNA Extraction and qPCR Analysis

Total RNA was extracted from 100 mg of tissue using TRIzol (Invitrogen, Carlsbad, CA, United States). Total RNA concentrations were assessed by UV spectrophotometry (Eppendorf, Bio-photometer, Germany). RNA (2 µg) was reverse transcribed into cDNA using HiScript Reverse Transcriptase (Vazyme, Nanjing, China) according to the manufacturers protocol. Four biological repeats were performed for each qPCR reaction, using OsActin as a reference gene. Primers were designed to detect OsNRT2.1, OsNRT2.3a, OsNRT2.4, OsNAR2.1, OsNRAMP3, OsNRAMP5, OsNRAMP6, OsIRT1, and OsMGT1 and are listed in **Supplementary Table S2**. PCR amplification was performed using SYBR qPCR Master Mix (Vazyme, Nanjing, China). PCR reactions were performed under the following parameters: 95◦C for 30 s, followed by 40 cycles of 95◦C for 10 s, 60◦C for 30 s, and 72◦C for 10 s.

#### Determination of the <sup>15</sup>N-NH<sup>4</sup> <sup>+</sup>/NO<sup>3</sup> − Influx Rate in Different Rice Lines

Rice seedlings of WT and OsNRT2.1 transgenic rice plants were planted in IRRI solution containing 1 mM NH<sup>4</sup> <sup>+</sup> for 2 weeks and N starved for 3 days. Plants were first transferred into 0.1 mM CaSO<sup>4</sup> for 1 min, then to complete nutrient solution containing either 0.5 mM <sup>15</sup>NH<sup>4</sup> <sup>+</sup> or 0.5 mM <sup>15</sup>NO<sup>3</sup> <sup>−</sup> (atom% <sup>15</sup>N: 99%) for 5 min and finally to 0.1 mM CaSO<sup>4</sup> for 1 min (Duan et al., 2007). The <sup>15</sup>N influx rate was calculated according to methods described by Tang et al. (2012).

### Assessment of Dry Weight, Total N and Metal Ion Accumulation

To investigate the links between OsNRT2 function and metal ion uptake, we investigated the levels of metal elements using the ICP-OES method in OsNRT2.1 transgenic lines and Mn elements in OsNRT2.3a/b transgenic lines. The creation and identification processes of bO-1, bO-2, bO-8 for OsNRT2.3b transgenic lines and aO-1, aO-2 for OsNRT2.3a transgenic lines were performed as previously described (Fan et al., 2016).

Fresh WT or transgenic lines were harvested at the rice mature stage (n = 4) and heated at 105◦C for 30 min. Panicles, flag leaves, second and third leaves, sheaths and stems were then dried for 3 days at 75◦C. Rice obtained from hydroponic experiments was divided into shoots and roots only. Dry weights were recorded as biomass values.

Using the Kjeldahl method (Li et al., 2006), total N accumulation was assessed in the different plant areas through multiplying the N concentration by the corresponding biomass. Dried samples were wet-digested in concentrated HNO<sup>3</sup> at 120◦C until no brown nitrogen oxide gas was emitted. When the samples became transparent, they were further digested with HClO<sup>4</sup> at 180◦C. Samples were then diluted with ultrapure water and the

concentrations of metal elements in the digestates were analyzed using ICP-OES (iCAP 6300).

### Statistical Analysis

All data were analyzed using the Tukey's test of one-way analysis of variance (ANOVA). Statistically significant differences at the p < 0.05 level (one-way ANOVA) between transgenic and WT and/or between other treatments were assessed. All statistical evaluations were performed using IBM SPSS Statistics version 20 software (SPSS Inc., Chicago, IL, United States).

### RESULTS

### Assessment of Mn Absorption Under Different N Treatments

Wild type rice seedlings were planted under different conditions of N supply. Symptomatically, the roots of rice seedlings were better in 0.5 mM NO<sup>3</sup> <sup>−</sup> than in 0.5 mM MH<sup>4</sup> +, 2.5 mM NO<sup>3</sup> <sup>−</sup> and 2.5 mM NH<sup>4</sup> <sup>+</sup> conditions (**Figure 1A**). Statistical analysis showed that the dry weights of the plant roots in 0.5 mM NO<sup>3</sup> <sup>−</sup> condition were significantly increased (**Figure 1B**). Low concentration NO<sup>3</sup> <sup>−</sup> could promote root elongation and increase root hairs (Kiba and Krapp, 2016). For the whole plant, the dry weight was best in 2.5 mM N, with no differences between 2.5 mM NO<sup>3</sup> <sup>−</sup> and 2.5 mM NH<sup>4</sup> <sup>+</sup> observed (**Figure 1C**). Next, the total N in rice seedlings was investigated. Total N content in rice seedlings with 2.5 mM N supply was higher than that of the 0.5 mM N supply (**Supplementary Figure S1**). Rice seedlings planted in 2.5 mM NH<sup>4</sup> <sup>+</sup> nutrient solution, displayed the best outcome (**Supplementary Figure S1**).

Simultaneously, the Mn concentration and content of rice roots in 0.5 mM NO<sup>3</sup> <sup>−</sup> was found to increase more than other conditions. However, shoots were lowest in 2.5 mM NH<sup>4</sup> <sup>+</sup> solution (**Figures 1D,E**). The Mn content of rice seedlings in NO<sup>3</sup> <sup>−</sup> solution was higher than in NH<sup>4</sup> <sup>+</sup> using the same N concentrations (**Figure 1F**). In addition, the expression of the nitrate transporters OsNRT2.1/OsNRT2.3 were up-regulated by external NO<sup>3</sup> <sup>−</sup> and the expression of OsNAR2.1 increased in 0.5 mM NO<sup>3</sup> <sup>−</sup>/2.5 mM NO<sup>3</sup> − compared to NH<sup>4</sup> <sup>+</sup> treatments in the different tissues (**Supplementary Figures S2A–C**). The expression of Mn transporters OsNRAMP3/OsNRAMP5/OsNRAMP6 also increased following NO<sup>3</sup> <sup>−</sup> treatment compared with NH4<sup>+</sup> treatment. Taken together, these results reveal that both Mn uptake and OsNRAMP3/OsNRAMP5/OsNRAMP6 expression are increased by NO<sup>3</sup> <sup>−</sup>. Therefore, NO<sup>3</sup> <sup>−</sup> positively regulates the absorption of Mn in rice.

### Assessment of the Expression Patterns of OsNRT2s and OsNAR2.1 in the Roots of Transgenic Lines

Firstly, transgenic lines were identified by southern blot analysis and RT-PCR. The data showed that three transgenic lines were one copy insertions and OsNRT2.1 was overexpressed to approximately five-fold higher mRNA levels in roots and shoots under normal N conditions (1.25 mM NH4NO<sup>3</sup> supply) (**Supplementary Figure S4** and Chen et al., 2016a). WT and transgenic OsNRT2.1 lines were planted in 0.5 mM NO<sup>3</sup> <sup>−</sup>/NH<sup>4</sup> + nutrition solution, respectively. RT-PCR was performed to confirm the gene expression patterns of the two families of NO<sup>3</sup> <sup>−</sup> transporters in WT and OsNRT2.1 transgenic lines under different N supplies. OsActin was used as a reference gene for comparison. Total RNA was extracted from the rice roots of the different lines. Under conditions of low concentration (0.5 mM) of NH<sup>4</sup> <sup>+</sup> and NO<sup>3</sup> <sup>−</sup>, the expression of OsNRT2.1 in transgenic lines increased 4.5-fold and 5.7-fold, compared to WT (**Figures 2A,B**). No differences in the relative expression of other NO<sup>3</sup> <sup>−</sup> transporters OsNRT2.3a/OsNRT2.4

between transgenic and WT lines or between NO<sup>3</sup> <sup>−</sup> and NH<sup>4</sup> <sup>+</sup> treatments were observed (**Figures 2A,B**). However, the expression levels of OsNAR2.1 increased approximately 80% in transgenic lines in 0.5 mM NO<sup>3</sup> <sup>−</sup>, but not in 0.5 mM NH<sup>4</sup> <sup>+</sup> (**Figures 2A,B**). The total N content of the three transgenic lines was higher than WT in the roots and the shoots under 0.5 mM NO<sup>3</sup> <sup>−</sup> conditions, with no differences in the NH<sup>4</sup> <sup>+</sup> solution observed (**Supplementary Figure S5**). Taken together, these results show that OsNRT2.1 expression is enhanced in the transgenic rice. In addition, the expression of OsNRT2.1 and OsNAR2.1 is enhanced in all transgenic lines, allowing an efficient transfer of NO<sup>3</sup> <sup>−</sup> in 0.5 mM NO<sup>3</sup> − conditions.

#### NH4<sup>+</sup> and NO<sup>3</sup> <sup>−</sup> Influx Rates in WT and OsNRT2.1 Transgenic

To confirm the influence of OsNRT2.1 on high-affinity root NO<sup>3</sup> <sup>−</sup> influx into intact plants, short-term nitrate absorption was assessed by transferring all the lines to either 0.5 mM <sup>15</sup>NH<sup>4</sup> <sup>+</sup> or 0.5 mM <sup>15</sup>NO<sup>3</sup> <sup>−</sup> for 5 min. Under 0.5 mM <sup>15</sup>NH<sup>4</sup> <sup>+</sup> treatment condition, the three transgenic lines displayed no significant differences to WT (**Figure 3A**). However, OsNRT2.1 transgenic lines were enhanced by 19% compared to WT during NO<sup>3</sup> − influx (**Figure 3B**). In addition, the effects of overexpression on rice growth under different forms of N supply were studied by comparing the total N concentration and content in different parts of the rice plants. The total N of the transgenic lines did not significantly differ in the roots and shoots compared to WT lines in 0.5 mM NH<sup>4</sup> <sup>+</sup> solution (**Supplementary Figures S5A,B**). However, the total N content of the roots and shoots of the transgenic rice plants was enhanced by 97% and 36%, respectively, compared to WT lines in 0.5 mM NO<sup>3</sup> <sup>−</sup> conditions (**Supplementary Figure S5E**). Total N concentrations in the shoots did not differ from WT (**Supplementary Figure S5D**). These results show that the overexpression of the high-affinity nitrate transporter OsNRT2.1 improves NO<sup>3</sup> <sup>−</sup> uptake in 0.5 mM NO<sup>3</sup> <sup>−</sup>, compared to WT.

### Mn Concentration of Shoots and Roots of Transgenic Plants Under Different N Treatments

The transferability of Mn is weak. From hydroponic experiments (**Figure 4**), we tested the Mn concentration of shoots and roots in rice when planted in different nutritive forms of N. We found that the dry weight of roots and shoots increased by 66 and 29%, respectively, in transgenic lines relative to WT lines in 0.5 mM NO<sup>3</sup> <sup>−</sup> solution (**Figure 4B**). However, dry weights did not significantly differ in 0.5 mM NH<sup>4</sup> <sup>+</sup> (**Figure 4A**). Simultaneously, Mn concentrations of roots and shoots in the overexpression lines were also enhanced by 43% and 47%, respectively, in 0.5 mM NO<sup>3</sup> <sup>−</sup> solution, but not in 0.5 mM NH<sup>4</sup> <sup>+</sup> (**Figures 4C,D**). From **Figure 4** and **Supplementary Figure S3**, we reasoned that this was due to the OsNRT2.1 gene transferring NO<sup>3</sup> <sup>−</sup> into the rice, increasing total N, Mn uptake and accumulation in 0.5 mM NO<sup>3</sup> <sup>−</sup> condition. These results indicate that Mn assimilation by OsNRT2.1 is NO<sup>3</sup> − uptake dependent, and that the overexpression of OsNRT2.1 does not only increase NO<sup>3</sup> <sup>−</sup> uptake to enhance total N, but also promotes Mn absorption in rice in low NO<sup>3</sup> − condition.

### Effects of Different Irrigation Conditions on N and Mn Concentrations in Grain

Rice typically grows in anaerobic flooded fields, which exist mainly in the form of NH<sup>4</sup> <sup>+</sup>-N. Conversely, NO<sup>3</sup> <sup>−</sup> is present mainly in aerobic uplands (Stitt, 1999). To simulate hydroponic conditions in the presence of different N treatments, we designed a field experiment under different irrigation conditions and investigated OsNRT2.1 function on rice grains in the field. From the assessment of seed morphology, seeds of WT under alternating wet and dry (AWD) condition were shorter than other seeds (**Figure 5A** and **Supplementary Figure S6**). Compared to other field treatments, we found that the grain weight of WT in AWD condition was approximately 31% lower than waterlogged

(WL) condition, with no differences in the transgenic lines observed (**Figure 5B**).

In addition, no evident differences in all lines in WL condition were observed. However, the grain weight of transgenic rice plants was approximately 26% higher compared to WT weights in AWD condition (**Figure 5A**). The 1000-grain weight displayed a similar pattern to the grain weights (**Figure 5B**). We also tested the total N concentration of the seeds under different field treatments. Interestingly, WT and transgenic lines were higher in AWD condition compared to waterlogged condition, and the total N concentration of the transgenic seeds also increased by 15% compared to WT in the AWD field (**Figure 5C**). However, the total N concentration of the husk in the overexpression lines was lower than that of WT in the AWD field, whilst no differences in all lines from the WL field were observed (**Supplementary Figure S7A**). As higher levels of N were transferred into the seeds of transgenic lines in the AWD field, their seed weights were higher than WT.

Simultaneously, the Mn concentrations in the seeds of transgenic lines in AWD condition were enhanced when compared to WL condition. No differences were observed for the different field conditions in WT lines (**Figure 5D**). In addition, the husk of grain displayed similar results in terms of Mn concentrations (**Supplementary Figure S7B**). The concentration of Fe and Mg in seeds and husk appeared to vary irregularly (**Supplementary Figure S7**). This presented the unity of the Mn element.

These results demonstrate that rice planted in AWD condition displays higher total N and Mn concentrations in grain, particularly for OsNRT2.1 transgenic lines. We extracted total RNA from the culm of all lines planted in the two types of irrigated field. From **Figures 5F,G**, the relative expression of OsNRT2.1 in transgenic lines was higher than WT lines in WL and AWD conditions. However, OsNAR2.1 expression was enhanced 2.8-fold only in AWD field relative to WT. Therefore, the soil of AWD primarily existed in NO<sup>3</sup> <sup>−</sup> form to enhance NO<sup>3</sup> <sup>−</sup> uptake through increased OsNRT2.1 expression, leading to the induction of OsNAR2.1 expression. As the relative expression of OsNRT2.1 and OsNAR2.1 increase following AWD treatment, NO<sup>3</sup> <sup>−</sup> uptake may further improve Mn uptake compared to the WL field.

### Assessment of the Expression of Related Genes, Total N and Mn Accumulation During Maturity Stages in AWD Conditions

To understand mechanism(s) of how OsNRT2.1 improves total N and Mn accumulation at the mature stage in AWD field, we extracted total RNA from the different areas of rice (**Supplementary Figure S9**) and assessed the expression of OsNRT2.1, OsNAR2.1, and Mn transporters-OsNRAMP3, OsNRAMP5, and OsNRAMP6 (Yamaji et al., 2013a; Yang et al., 2014; Peris-Peris et al., 2017).

From **Figure 6**, the expression of the related nitrate genes-OsNRT2.1 and OsNAR2.1 in the three transgenic lines were higher in the panicle-neck, flag leaves, flag leaves sheaths and node I compared to WT rice. The panicle-neck connects vegetative and reproductive organs. Flag leaves are functional leaves for transferring nutrients. Studies have reported that

high-affinity nitrate OsNRT2.1 requires its partner protein OsNAR2.1 to transfer nitrate (Feng et al., 2011; Yan et al., 2011; Tang et al., 2012). Accordingly, the expression of OsNRT2.1 and its partner protein OsNAR2.1 increased in the Panicle-neck and in the functional leaves at maturity. NO<sup>3</sup> <sup>−</sup> was transferred to the panicle to enhance total N accumulation in seeds, and further improve grain yields.

Interestingly, we found that the expression of the Mn transporters: OsNRAMP3, OsNRAMP5, and OsNRAMP6 were also upregulated in transgenic lines. These genes displayed similar expression patterns to NO<sup>3</sup> <sup>−</sup> transporters (**Figures 6E–G**). In particular, the expression of OsNRAMP3 and OsNRAMP6 in the transgenic lines increased by 87 and 311% in comparison to WT in the node I, respectively (**Figure 6H**). Node I represents the junction of the vascular system connecting the leaves, stems and panicles. Therefore, Mn transporter genes-OsNRAMP3 and OsNRAMP6 preferentially transport Mn to flag leaves and the panicle during the late stages of plant growth in rice. We found that the biomass of transgenic and WT lines displayed no significant differences at maturity (**Supplementary Figure S10A**). The NO<sup>3</sup> <sup>−</sup> concentrations of the different plant areas (except for leaves in the overexpression lines) were higher than WT (**Supplementary Figure S10D**). However, total N accumulation did not differ in various parts of the plants, and Mn showed an irregular trend without flag leaves (**Supplementary Figure S10**). These results suggest that total N and Mn are transferred to grains from vegetative organs at maturity. We further assessed Fe and Mg content in various parts of the different lines, in which we observed no differences (**Supplementary Figure S11**). When the relative expression of OsIRT1 and OsMGT1 that represent Fe and Mg related genes (Lee and An, 2009; Chen Z.C. et al., 2017) were analyzed, the expression patterns were also inconsistent in diverse areas of the transgenic rice plants (**Figure 7**).

Taken together, these data suggest that the improvement of Mn concentration in OsNRT2.1 lines was due to the increased expression of Mn transporters, but no effects on other metal elements were observed.

## DISCUSSION

Nitrate and Mn are essential nutrients in plants, and it has been reported that Mn deficiency decreases N uptake and metabolism (Gong et al., 2011). Excessive NO<sup>3</sup> <sup>−</sup> was shown to enhance Cd uptake in Thlaspi caerulescens (Xie et al., 2009) and wheat (Li et al., 2011). In addition, crosstalk between mineral elements exists. Wang et al. (2015) reported that Al improves Mn uptake and accumulation in rice roots, However, these tactics do not enhance the security of crops for human consumption because they do not increase the accumulation of beneficial elements in plant. In this study, the main objective was to investigate how interactions between Mn and NO<sup>3</sup> <sup>−</sup> influence rice growth and nutrient accumulation in roots, leaves tissues and grain.

We found that NO<sup>3</sup> <sup>−</sup> improves Mn uptake in rice (**Figure 2**). When Mn concentrations were assessed in OsNRT2.1/OsNRT2.3a/b transgenic lines planted in normal field which was WL condition, respectively (**Supplementary Figure S3**), we observed increased Mn in the seeds and husk of OsNRT2.1 (**Supplementary Figures S3A,D**), but no enhanced uptake in OsNRT2.3a/b transgenic lines (**Supplementary Figures S3B,C,E,F**). Given these data, we investigated the pattern of Mn accumulation in OsNRT2.1 transgenic lines in further detail, under differing conditions of N supply and field

FIGURE 6 | Relative expression of related genes in different arears of WT and transgenic plants in AWD fields. Total RNA was isolated from (A,E) panicle-neck, (B,F) flag leaves, (C,G) flag leaves sheaths, and (D,H) Node I of WT and transgenic lines. Error bars: standard error (n = 4 plants). Different letters indicate a significant difference between WT and overexpression lines (P < 0.05, one-way ANOVA).

conditions. This information is important as rice typically grows in anaerobic flooded fields, in which N exists mainly in the form of NH<sup>4</sup> <sup>+</sup>, as opposed to aerobic uplands where the major form of N is NO<sup>3</sup> <sup>−</sup> (Stitt, 1999). We found that OsNRT2.1-regulates NO<sup>3</sup> <sup>−</sup> uptake in roots, which in turn increases Mn root entry. This increases the Mn concentration in rice grain in the presence of low concentrations of NO<sup>3</sup> <sup>−</sup> and under AWD condition. Thus, enhancing NRT2.1-mediated NO<sup>3</sup> <sup>−</sup> uptake represents an attractive mechanism of increasing Mn accumulation in food.

#### Effects of NO<sup>3</sup> <sup>−</sup> and NH<sup>4</sup> <sup>+</sup> Nutrition on Mn Accumulation

We demonstrate that NO<sup>3</sup> <sup>−</sup> nutrition promotes Mn assimilation in plants to higher levels than NH<sup>4</sup> <sup>+</sup> nutrition (**Figure 1**). Thus

Mn availability in nutrient solutions is influenced by the type of N-nutrient treatments. In 0.5 mM NH<sup>4</sup> <sup>+</sup>, Mn uptake in the roots did not differ in OsNRT2.1 transgenic lines (**Figure 4C**). However, Mn uptake drastically increased in the transgenic lines in 0.5 mM NO<sup>3</sup> <sup>−</sup> (**Supplementary Figure S5** and **Figure 4D**). It was recently shown that NO<sup>3</sup> <sup>−</sup> uptake induces external alkalization, reducing Fe/Mn concentrations by enhancing the levels of H2O<sup>2</sup> in rice (Chen et al., 2018). In this study, we performed hydroponic experiments in MES buffered nutrient medium (to control pH) and nutrient treatments were replaced every 2 days. Furthermore, the Mn content in grain from field experiments should not be influenced by pH as the rhizosphere ranges from pH 5.5 to 6.0 in paddy soil (Pan et al., 2016). Thus, any effects of soil alkalization were excluded. We thus hypothesize that NO<sup>3</sup> <sup>−</sup> upregulates the expression of Mn transporters, including OsNRAMP3, OsNRAMP5, and OsNRAMP6 to increase Mn uptake and accumulation (**Figure 1** and **Supplementary Figure S2**). We verified the expression of OsNRT2.1 and OsNAR2.1 under NH<sup>4</sup> + and NO<sup>3</sup> <sup>−</sup> conditions in transgenic OsNRT2.1 lines and found that OsNRT2.1 was unaffected by the different N forms. However, the expression of its partner protein OsNAR2.1 was significantly up-regulated in 0.5 mM NO<sup>3</sup> <sup>−</sup> (**Figure 2**). The OsNRT2.1 lines could still promote NO<sup>3</sup> <sup>−</sup> uptake (**Figure 3**) into the different tissues compared to WT plants (**Supplementary Figure S10D**). Thus, the up-regulation of OsNAR2.1 expression in NO<sup>3</sup> − condition (**Figure 2B**) promotes NO<sup>3</sup> <sup>−</sup> uptake in transgenic plants compared with WT (**Supplementary Figure S10D**). The observation that the upregulation of OsNRT2.1/OsNAR2.1 is favorable to the transport of NO<sup>3</sup> <sup>−</sup> in plants and improves rice yield, is consistent with our previous findings (Chen et al., 2016b; Chen J.G. et al., 2017). As we did not observe enhanced expression of either OsNRT2.3a or OsNRT2.4 lines in NO<sup>3</sup> <sup>−</sup> condition, we speculate that the regulation of OsNAR2.1 differs from other OsNRT2 genes according to the plant NO<sup>3</sup> − content (**Supplementary Figure S2**, Yan et al., 2011; Wei et al., 2018).

### AWD vs. WL Conditions in WT vs. Transgenic Lines

In field experiments, the grain weight of WT lines under AWD condition decreased by 31% compared to WL condition. The NO<sup>3</sup> <sup>−</sup> concentrations in OsNRT2.1 transgenic lines also differed across plant areas in AWD condition and which were higher than WT lines (**Supplementary Figure S10D**). However, the total N concentration did not significantly differ across the lines (**Supplementary Figure S10C**). The total N of seeds in transgenic lines increased compared to WT (**Figure 5**). Thus, the overexpression of OsNRT2.1 improves NO<sup>3</sup> <sup>−</sup> uptake and assimilation efficiency to increase N accumulation in grain, leading to enhanced grain yields. In hydroponic experiments, the overexpression of OsNRT2.1 also enhanced NO<sup>3</sup> <sup>−</sup> uptake (**Figure 3B**) and N accumulation (**Figure 5D**), maintaining plant grain yields in WL condition. This is because in AWD condition, a high concentration of dissolved oxygen is present, which can influence nitrification by nitrifying bacterial, or chemical oxidation for the conversion of NH<sup>4</sup> <sup>+</sup> to NO<sup>3</sup> <sup>−</sup> at the root surface (Li et al., 2008; Steffens et al., 2011). Thus, under AWD condition, NO<sup>3</sup> <sup>−</sup> plays an important role in N accumulation and contributes to enhanced grain yields. However, for WT plants, the capacity to uptake NO<sup>3</sup> <sup>−</sup> is limited; and thus, grain yields are dramatically reduced. This observed loss of grain in WT type rice can likely be explained by a multitude of mechanisms.

Surprisingly, we found that under AWD, Mn in the grain of transgenic plants was greatly increased compared to WL condition, but in WT rice, no changes were evident (**Figure 5E**). In addition, in WL condition, Mn levels also increased in the transgenic lines compared to WT (**Figure 5E**). We observed no differences in Mn concentrations in other parts of the plant under AWD condition (**Supplementary Figure S10**). Thus, higher levels of Mn were transported to grain and accumulated (**Figure 5**). This explains the improvement in seed quality, emergence, and seeding growth observed, as the positive effects of Mn on these processes is well documented (Dimkpa and Bindraban, 2016). The seeds of OsNRT2.1 overexpression lines not only increased in their total N accumulation, but enhanced Mn content was also observed (**Figure 5**). The length/width of these seeds were also better than WT (**Figure 5A** and **Supplementary Figure S6**), demonstrating that Mn plays an important role in increasing crop nutritional quality, crop yield and biomass production. Other metal elements such as Mg and Fe were not influenced by OsNRT2.1 overexpression (**Supplementary Figures S8**, **S11**).

### Enhanced Expression of Mn Transporters Explains Enhanced Mn Uptake in Transgenic Lines

We verified gene expression profiles in the organs responsible for grain filling and discovered that the expression of OsNRT2.1 and OsNAR2.1 were enhanced in the panicle-neck, flag leaves and sheaths (**Figure 6A**). In the same plant areas, the expression of OsNRAMP5 and OsNRAMP6 increased in the OsNRT2.1 lines. Interestingly, OsNRAMP3/6 expression was enhanced in node I (**Figure 6B**). The expression of related genes involved in Mg and Fe uptake were also altered by OsNRT2.1 overexpression (**Figure 7**). It is understood that node I is a junction of vasculatures that link leaves, stems and panicles and so is important for the transport of nutrient elements into grain (Yamaji and Ma, 2009, 2014; Yamaji et al., 2013a). Transporters responsible for the delivery of minerals into seeds have been reported, including OsYSL16 for Cu (Zheng et al., 2012), OsHMA2 for Zn and Cd (Yamaji et al., 2013b) and AtNIP6;1 that is expressed in the node region for B distribution (Tanaka et al., 2008). Accordingly, the majority of these genes are also strongly expressed in node I (Tanaka et al., 2008; Yamaji and Ma, 2009, 2014; Zheng et al., 2012; Yamaji et al., 2013a).

### CONCLUSION

Taken together, we show that AWD treatment can induce the expression of NO<sup>3</sup> <sup>−</sup> and Mn transporters in grain filling

organs which increases the accumulation of N and Mn in grain. NO<sup>3</sup> <sup>−</sup> uptake in OsNRT2.1 transgenic lines can improve Mn accumulation, however, the Mn concentration does not increase in the seeds and husk of OsNRT2.3a/b overexpression lines, which also display increased NO<sup>3</sup> <sup>−</sup> uptake compared to WT lines (Fan et al., 2016). Thus, the mechanism(s) linking NO<sup>3</sup> <sup>−</sup> and Mn in OsNRT2.1 overexpressing plants differ from other OsNRT2 overexpression lines and is worthy of further investigation. From our findings, we propose a new application to improve both N and water efficiency in agricultural systems and demonstrate how high OsNRT2.1 expression improves Mn content in rice grain.

### AUTHOR CONTRIBUTIONS

BiL, JC, and XF conceived the study, analyzed the data, and drafted the manuscript. BiL, LZ, and SL cultivated the rice materials and collected the rice samples. LZ, BL, and HL extracted RNA and performed the qRT-PCR experiments. BL and LZ participated in field and material management. BiL and JC conducted the statistical analysis of raw data. XF, GX, and GY revised the manuscript. All authors read and approved the final manuscript.

### FUNDING

This study was financially supported by China National Key Program for Research and Development (2016YFD0100700), National Natural Science Foundation (Grant No. 31372122), Jiangsu Science Fund for Distinguished Young Scholars (Grant No. BK20160030), the Transgenic Project (Grant 2016ZX08001003-008).

### ACKNOWLEDGMENTS

We are also grateful for the Anhui Provincial Natural Science Foundation of China Science Foundation of China (No. 1608085MC59) and Major Special Science and Technology Project of Anhui Province (No. 16030701102).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018.01192/ full#supplementary-material

FIGURE S1 | Assessment of the total N-content under different N treatments. (A) total N concentration, (B) total N content of roots and shoots from different N treatments, (C) total N content of whole plants. 0.5/2.5A: 0.5 mM/2.5 mM NH<sup>4</sup> + as an N source; 0.5/2.5 N: 0.5/2.5 mM NO<sup>3</sup> <sup>−</sup> as an N source. Error bars: standard error (n = 4 plants). Different letters indicate a significant difference between N treatments (P < 0.05, one-way ANOVA).

FIGURE S2 | Expression pattern of NO<sup>3</sup> <sup>−</sup> transporters and Mn transporters under different N treatments; total RNA was isolated from WT rice supplied with 0.5/2.5N: 0.5/2.5 mM NO<sup>3</sup> <sup>−</sup> and 0.5/2.5 A:0.5/2.5 mM NH<sup>4</sup> <sup>+</sup> as an N source for 2 weeks. (A) relative expression of OsNRT2.1/OsNRT2.3/OsNAR2.1 and (D) OsNRAMP3/OsNRAMP5/OsNRAMP6 in leaves; (B) relative expression of

OsNRT2.1/OsNRT2.3/OsNAR2.1 and (E) OsNRAMP3/OsNRAMP5/OsNRAMP6 in sheath; (C) relative expression of OsNRT2.1/OsNRT2.3/OsNAR2.1 and (F) OsNRAMP3/OsNRAMP5/OsNRAMP6 in roots. Error bars: standard error (n = 4 plants). Different letters indicate a significant difference between different N treatments (P < 0.05, one-way ANOVA).

FIGURE S3 | Mn concentration in seeds and husk of OsNRT2.1/OsNRT2.3b overexpression lines. Mn concentrations in seeds of (A) OsNRT2.1 overexpression lines, (B) OsNRT2.3b overexpression lines and (C) OsNRT2.3a overexpression lines. Mn concentration in husk of (D) OsNRT2.1 overexpression lines, (E) OsNRT2.3b overexpression lines and (F) OsNRT2.3a overexpression lines. b-O1/2/8: three OsNRT2.3b overexpression lines; a-O1/2: two OsNRT2.3a overexpression lines. Error bars: standard error (n = 4 plants). Different letters indicate a significant difference between N treatments (P < 0.05, one-way ANOVA).

FIGURE S4 | Identification of transgenic lines. (A) Southern blot of genomic DNA isolated from WT and transgenic plants. Hybridization was performed using a hygromycin gene probe. P, positive control; M, marker. Extraction of total RNA from roots and shoots of WT and transgenic lines and qRT-PCR results under. (B) M: DNA molecular-weight marker II, DIG – labeled; P: positive controls. Error bars: standard error (n = 4 plants). Different letters indicate a significant difference between N treatments (P < 0.05, one-way ANOVA).

FIGURE S5 | Comparison of total N/Mn concentrations and content of transgenic plants at different nitrogen supply levels. (A–C) Under 0.5 mM NH<sup>4</sup> <sup>+</sup> treatments, (A) total N concentration, (B) total N content and (C) Mn content of roots and shoots. (D–F) Under 0.5 mM NO<sup>3</sup> <sup>−</sup> treatments, (D) total N concentration, (E) total N content and (F) Mn content of roots and shoots. Error bars: standard error (n = 4 plants). Different letters indicate a significant difference between WT and overexpression lines (P < 0.05, one-way ANOVA).

FIGURE S6 | Assessment of the length and width of seeds in different lines under WL and AWD treatments. (A) Seeds lengths (mm), (B) seed widths (mm). Error bars: standard error (n = 4 plants), 15 repeats. Different letters indicate a significant difference between WT and overexpression lines (P < 0.05, one-way ANOVA).

FIGURE S7 | Effects of different irrigation conditions on Mn concentrations in rice husk. Under WL and AWD, (A) total N concentration and (B) Mn concentration of rice husk were assessed. Error bars: standard error (n = 4 plants). Different letters indicate a significant difference between the irrigation conditions of all lines (P < 0.05, one-way ANOVA).

FIGURE S8 | Effects of different irrigation conditions on other elements in rice seeds. Under WL and AWD, Fe and Mg concentrations of husk (A,B) and seeds (C,D). Error bars: standard error (n = 4 plants). Different letters indicate a significant difference between WT and overexpression lines in different irrigation conditions (P < 0.05, one-way ANOVA).

FIGURE S9 | Diagram of RNA sampling in WT and transgenic plants.

FIGURE S10 | Effect of transgenic lines on total N/Mn content in vegetative organs under AWD conditions. (A) Dry weight of different parts in all lines, (B) Total N concentration, (C) Total N content, (D) NO<sup>3</sup> <sup>−</sup> concentration, (E) Manganese concentration, and (F) Manganese content from different parts of all lines. Error bars: standard error (n = 4 plants). Other leaves: second and third leaves. Different letters indicate a significant difference between WT and overexpression lines (P < 0.05, one-way ANOVA).

FIGURE S11 | Concentration of other elements in different parts of transgenic lines under AWD conditions. (A) Mg concentration, (C) Mg content, (B) Fe concentration, (D) Fe content. Error bars: standard error (n = 4 plants). Different letters indicate a significant difference between WT and overexpression lines (P < 0.05, one-way ANOVA).

FIGURE S12 | Correlation analysis between expression of OsNRT2.1 and total nitrogen/nitrate concentration in flag leaves of wt and transgenic lines. (A) Linear Analysis of relative expression of OsNRT2.1 and total N concentration. (B) Linear Analysis of relative expression of OsNRT2.1 and nitrate concentration.

TABLE S1 | Primers used to amplify the OsNRT2.1 open reading frame.

TABLE S2 | Primers used for quantitative real-time polymerase chain reaction.

### REFERENCES

fpls-09-01192 August 13, 2018 Time: 20:0 # 11



metal ATPase OsHMA2. Plant Physiol. 162, 927–939. doi: 10.1104/pp.113.21 6564


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Luo, Chen, Zhu, Liu, Li, Lu, Ye, Xu and Fan. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Elemental Profiling of Rice FOX Lines Leads to Characterization of a New Zn Plasma Membrane Transporter, OsZIP7

Felipe K. Ricachenevsky1,2,3 \*, Tracy Punshon<sup>3</sup> , Sichul Lee<sup>4</sup> , Ben Hur N. Oliveira<sup>1</sup> , Thomaz S. Trenz<sup>1</sup> , Felipe dos Santos Maraschin<sup>5</sup> , Maria N. Hindt<sup>3</sup> , John Danku<sup>6</sup> , David E. Salt<sup>6</sup> , Janette P. Fett1,5 and Mary Lou Guerinot<sup>3</sup>

<sup>1</sup> Programa de Pós-Graduação em Biologia Celular e Molecular, Centro de Biotecnologia, Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil, <sup>2</sup> Departamento de Biologia, Centro de Ciências Naturais e Exatas, Universidade Federal de Santa Maria, Santa Maria, Brazil, <sup>3</sup> Department of Biological Sciences, Dartmouth College, Hanover, NH, United States, <sup>4</sup> Center for Plant Aging Research, Institute for Basic Science (IBS), Daegu, South Korea, <sup>5</sup> Departamento de Botânica, Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil, <sup>6</sup> School of Biosciences, University of Nottingham, Loughborough, United Kingdom

### Edited by:

Edgar Peiter, Martin Luther University of Halle-Wittenberg, Germany

#### Reviewed by:

Scott Aleksander Sinclair, Ruhr-Universität Bochum, Germany Massimiliano Corso, Free University of Brussels, Belgium Damien Blaudez, Université de Lorraine, France

\*Correspondence:

Felipe K. Ricachenevsky felipecruzalta@gmail.com

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 21 February 2018 Accepted: 04 June 2018 Published: 03 July 2018

#### Citation:

Ricachenevsky FK, Punshon T, Lee S, Oliveira BHN, Trenz TS, Maraschin FS, Hindt MN, Danku J, Salt DE, Fett JP and Guerinot ML (2018) Elemental Profiling of Rice FOX Lines Leads to Characterization of a New Zn Plasma Membrane Transporter, OsZIP7. Front. Plant Sci. 9:865. doi: 10.3389/fpls.2018.00865 Iron (Fe) and zinc (Zn) are essential micronutrients required for proper development in both humans and plants. Rice (Oryza sativa L.) grains are the staple food for nearly half of the world's population, but a poor source of metals such as Fe and Zn. Populations that rely on milled cereals are especially prone to Fe and Zn deficiencies, the most prevalent nutritional deficiencies in humans. Biofortification is a cost-effective solution for improvement of the nutritional quality of crops. However, a better understanding of the mechanisms underlying grain accumulation of mineral nutrients is required before this approach can achieve its full potential. Characterization of gene function is more time-consuming in crops than in model species such as Arabidopsis thaliana. Aiming to more quickly characterize rice genes related to metal homeostasis, we applied the concept of high throughput elemental profiling (ionomics) to Arabidopsis lines heterologously expressing rice cDNAs driven by the 35S promoter, named FOX (Full Length Over-eXpressor) lines. We screened lines expressing candidate genes that could be used in the development of biofortified grain. Among the most promising candidates, we identified two lines ovexpressing the metal cation transporter OsZIP7. OsZIP7 expression in Arabidopsis resulted in a 25% increase in shoot Zn concentrations compared to non-transformed plants. We further characterized OsZIP7 and showed that it is localized to the plasma membrane and is able to complement Zn transport defective (but not Fe defective) yeast mutants. Interestingly, we showed that OsZIP7 does not transport Cd, which is commonly transported by ZIP proteins. Importantly, OsZIP7-expressing lines have increased Zn concentrations in their seeds. Our results indicate that OsZIP7 is a good candidate for developing Zn biofortified rice. Moreover, we showed the use of heterologous expression of genes from crops in A. thaliana as a fast method for characterization of crop genes related to the ionome and potentially useful in biofortification strategies.

Keywords: zinc, ZIP transporter, rice, fox lines, synchrotron x-ray fluorescence, ionomics, biofortification

### INTRODUCTION

fpls-09-00865 June 30, 2018 Time: 16:14 # 2

Zinc (Zn) is an essential micronutrient for plant nutrition and development, being a catalytic and structural co-factor in a large number of enzymes and regulatory proteins, including transcription factors (Marschner, 1995; Maret, 2009). However, Zn can become toxic in concentrations above a certain threshold. Fe participates in Fenton chemistry, generating reactive oxygen species. However, Zn competes with other ions for binding sites, and can become toxic in concentrations above a certain threshold (Clemens, 2001; Briat, 2002). Thus, plants have to keep Zn concentration within a narrow range for proper function. Many proteins are dedicated to Zn homeostasis, including organ and tissue partitioning as well as subcellular compartmentalization (Ricachenevsky et al., 2015).

Zn deficiency is the one the most widespread mineral nutritional disorders in humans, second only to Fe deficiency. Conservative estimates suggest that 25% of the human population is at risk of becoming Zn deficient (Maret and Sandstead, 2006). A diet composed mainly of milled cereal grains, common among poor populations, increases the risk of mineral deficiencies because staple foods, including rice (Oryza sativa), have low concentrations of Zn (as well as Fe) in edible tissues (Gomez-Galera et al., 2010). Biofortification, the increase of nutrient concentrations in edible portions of crops before harvesting, has been proposed as a cost-effective solution for micronutrient malnutrition (White and Broadley, 2005; Murgia et al., 2013).

Rice is a staple food for nearly half of the world's population<sup>1</sup> and a model species for monocots, making it an obvious candidate for biofortification efforts. A recent screening of a large diversity panel of rice genotypes indicated that it is possible to breed for Zn concentrations in seeds (Pinson et al., 2014). However, rice grains have lower concentration of Zn compared to other cereals, and thus genetic engineering tools might be useful to generate biofortified plants (Kennedy and Burlingame, 2003; Pfeiffer and McClafferty, 2008). In order to devise strategies for increasing micronutrient concentrations in grains, it is necessary to understand how plants acquire, distribute and store Fe and Zn within their tissues, which proteins are involved in each step and which would be good candidates for targeted, molecular breeding approaches. Despite the knowledge accumulated in recent years (Sperotto et al., 2012; Ricachenevsky et al., 2015), functional characterization of genes related to Fe and Zn homeostasis in rice is slower in comparison to the model species Arabidopsis thaliana. Due to genome size and complexity, and availability of protocols for genetic transformation and mutant generation, gene characterization in other crops such as wheat, maize and barley is even more time-consuming (Wu et al., 2015). Thus, strategies for fast, medium to high-throughput gene characterization would help to identify promising candidates for biofortification.

Heterologous expression of crop genes in A. thaliana has been used in several studies to demonstrate gene function. In order to allow high-throughput analyses of interesting phenotypes, more than 30,000 independent A. thaliana lines over-expressing rice genes were developed, and named FOX lines (Full-length Over-eXpressor Arabidopsis lines; (Kondou et al., 2009; Sakurai et al., 2011). These lines have been successfully used for characterization of genes associated with several processes in plants including responses to fungal and bacterial pathogens (Dubouzet et al., 2011), tolerance to abiotic stresses (Yokotani et al., 2008, 2009), enzyme characterization (Higuchi-Takeuchi et al., 2011; Anders et al., 2012) and in metabolomics profiling (Albinsky et al., 2010). More recently, similar approaches were used to describe stress-related genes in the halophyte Eutrema salsugineum (Ariga et al., 2015).

In this work, we characterized a collection of Arabidopsis FOX lines expressing rice genes using ionomics techniques in order to demonstrate their feasibility for the rapid functional characterization of crop genes with potential use in biofortification strategies. By combining elemental profiling by inductively-coupled plasma mass spectrometry (ICP-MS; (Salt et al., 2008) and Synchrotron X-Ray fluorescence (SXRF; Punshon et al., 2013), we described lines that express a new plasma membrane Zn transporter of the Zn-regulated, ironregulated transporter-like protein (ZIP) family from rice, OsZIP7. Our work demonstrates that ionomics of A. thaliana lines heterologously expressing rice cDNAs is a useful method for the rapid characterization of genes involved in regulation of the ionome, an approach that should also be feasible for other crops.

### MATERIALS AND METHODS

### Plant Materials and Growth Conditions

For ionomics profile screening, Rice FOX lines and Col-0 WT seeds were sown and cultivated as described (Lahner et al., 2003), with minor modifications. After 3 days at 4 ◦C for stratification, trays were kept in a climate-controlled growth room with 10 h of light (90 µmol.m−<sup>2</sup> s −1 )/14 h dark, humidity of 60% and temperature ranging from 19 to 22◦C. Twelve plants of each genotype, including WT Col-0, were cultivated for 6 weeks, and were watered twice a week with 0.25X Hoagland solution using 10 µM Fe-HBED [N,N<sup>0</sup> -di(2-hydroxybenzyl) ethylenediamine- N,N<sup>0</sup> -diacetic acid monohydrochloride hydrate; Strem Chemicals, Inc.] as the Fe source.

For growth in axenic conditions, seeds were sterilized for 15 min in 1.5% sodium hypochloride with 0.05% SDS, washed five times in sterile H2O and stratified at 4◦C for 3 days. Sterile 0.1% agar was used to suspend seeds, which were sown using a pipette onto plates made with full strength Gamborg's B5 media plus vitamins, 1 mM MES [2-(N-morpholino)ethanesulfonic acid], 2% sucrose and 0.6% agar. After 5 days, seedlings were transferred to minimal media containing 2 mM MES, 2 mM Ca(NO3)2.4H2O, 0.75 mM K2SO4, 0.65 mM MgSO4.7H2O, 0.1 mM KH2PO4, 10 µM H3BO3, 0,1 µM MnSO4, 50 nM CuSO4, 5 nM (NH4)6Mo7O<sup>24</sup> and 50 µM Fe-EDTA. ZnSO<sup>4</sup> was added to a final concentration of 50 nM in control conditions, or at indicated concentrations. Seedlings were analyzed after 15 days

<sup>1</sup>http://www.fao.org/rice2004/en/rice-us.htm

of growth and plates were kept at 22◦C with 16 h of light/8 h of dark in growth chambers.

For ICP-MS analyses of seed samples, plants were grown on soil in a growth room at 22◦C with 16 h of light/8 h of dark. Seeds were collected from five plants of each genotype and analyzed by ICP-MS as above.

### Elemental Analyses by ICP-MS

Elemental concentration analyses of leaf samples were performed as described (Lahner et al., 2003), with the minor modification that the plants were grown in soil for 6 weeks. Sample handling and preparation was performed as described (Lahner et al., 2003). Data was normalized across different trays Col-0 values, which were present in each tray. All data is publicly available at www. ionomics.org for download.

For ICP-MS analyses of shoots and roots of axenically grown plants, metals were desorbed from samples for 10 min on ice with cold 5 mM CaSO4, 1 mM MES, pH 5.7, for 5 min on cold 5 mM CaSO4, 10 mM EDTA, 1 mM MES, pH 5.7 and then washed twice with cold ultrapure water (Haydon et al., 2012). Sample processing was performed as above. We used 12 replicates por line for the initial screening, five replicates per line for shoots and roots, and eight replicates for seeds ICP-MS analyses.

### Subcellular Localization

For protoplast preparation, A. thaliana Col-0 plants were grown in soil in a growth chamber at 22◦C with 12 h of light/12 h of dark. After 4 weeks, approximately 25 leaves were detached and had their abaxial epidermis removed following digestion by the tape-sandwich method (Wu et al., 2009). Macerozyme and cellulase treatment and protoplast recovery from the remaining leaf mesophyll were performed as described (Yoo et al., 2007).

For subcellular localization in protoplasts, the OsZIP7 coding sequence lacking the stop codon was amplified using specific primers (**Supplementary Table S2**) and cloned into the pENTR/D-TOPO entry vector. Subsequently, LR recombination was performed into a vector containing a C-terminal fusion with YFP, pEarleyGate101, generating pEarleyGate101-OsZIP7. High concentrations of the final construct were prepared using PureYieldTM Plasmid Midiprep from Promega <sup>R</sup> . AtAHA2-RFP construct (Kim et al., 2001) was used as plasma membrane localization control.

Protoplast transfection was performed as described (Yoo et al., 2007). Because of the large size of the pEarleyGate101- OsZIP7 construct, 20 µg of DNA were used. For AtAHA2-RFP, 10 µg were used. For visualization of YFP and RFP signals, a Nikon Eclipse Ti inverted microscope stand was used, and image capture and processing was performed with Nikon Elements software.

Nicotiana benthamiana plants were grown in a growth chamber at 24◦ C in a 16/8 h light/dark cycles until leaves were fully expanded for agroinfiltration. Transient expression in N. benthamiana leaves was performed as described previously (Sparkes et al., 2006). Agrobacterium tumefaciens (EHA105 strain) carrying pEarleyGate101-OsZIP7 binary vector was coinfiltrated with Agrobacterium tumefaciens carrying pBIN20/PM-CK binary vector, which contains the coding sequence of PIP2A of A. thaliana fused in frame with cyan fluorescent protein (CFP; Nelson et al., 2007), in an optical density ratio of 1:1. Plasmolysis was induced using a 20% NaCl hypertonic solution. Fluorescence microscopy was performed under an Olympus FV1000 confocal laser-scanning microscope, using YFP and CFP filters. Images were captured with a high-sensitivity photomultiplier tube detector. Due to confocal microscope limitation in both co-localization experiments, we obtained subsequent images showing fluorescence signals from the same cells, without the signal overlay.

### Yeast Assays

A full-length version of OsZIP7 was amplified using specific primers (**Supplementary Table S1**) and cloned into the pDR195 vector using XhoI and BamHI sites. As a control, AtIRT1 was also amplified and cloned into pDR195 using XhoI and BamHI sites.

Yeast strains BY4743 (MATa/α his311/his311 leu210/ leu210 LYS2/lys210 met1510/MET15 ura310/ura310), ZHY3 (MATα ade6 can1 his3 leu2 trp1 ura3 zrt1::LEU2 zrt2::HIS3) and DEY1453 (MATa/MATα ade2/ADE2 can1/can1 his3/his3 leu2/leu2 trp1/trp1 ura3/ura3 fet3-2::HIS3/fet3-2::HIS3 fet4- 1::LEU2/fet4-1::LEU2) were grown in YPD media pH 5.3 (DEY1453 was grown in pH 4 to increase Fe availability) and transformed with pDR195, pDR195-OsZIP7, or pDR195-AtIRT1 by the LiOAc/PEG method (Gietz and Schiestl, 2007). Selection of transformants was performed in SD media without uracil (SD –ura: 6.7 g/L yeast nitrogen base without amino acids, supplemented with 2% glucose, 0.1% casamino acids, 0.01% adenine, and 0.01% tryptophan), pH 5.3. Colonies were grown overnight in liquid SD -ura media, diluted to OD<sup>600</sup> 1.0, 0.1, 0.01 and 0.001, and spotted onto plates. To test for Cd toxicity, SD ura was amended with CdCl<sup>2</sup> at given concentrations. To test for Zn deficiency, no Zn was added, and 10 µM ZnCl<sup>2</sup> was added to control plates. To test for Fe deficiency, the pH was raised to 6.0, and compared to control plates at pH 5.3. Pictures were taken after 3–5 days of growth.

### Synchrotron X-Ray Fluorescence

For microtomography of seed, tomograms were collected at the bending magnet beamline X26A at the National Synchrotron Light Source, Brookhaven National Laboratory. µ-SXRF seed analyses were performed as described (Kim et al., 2006), using Col-0 and OsZIP7-FOX1 seeds deved from plants grown simultaneously. Elemental abundances (weight fraction) were calculated for the fluorescence measurements as described (McNear et al., 2005).

### Statistical Analyses

For ionomics profile comparison between FOX lines and Col-0, we used intra-tray comparisons (i.e., each line had their profile compared to Col-0 plants growing in the same tray). Concentration values for a given element (x) were considered outliers when x > Q75% + 1.5 × Q75% − Q25% or x < Q25% − 1.5 × Q75% − Q25%, where Q75% − Q25% represents 50% of the values observed (i.e., between the 1st and 3rd quartile). Statistical significance was accessed using

the Wilcoxon–Mann–Whitney test and the Benjamini–Hochberg correction. All other data were subjected to ANOVA and means were compared by the Tukey HSD test.

### RESULTS

### Rice FOX Lines Selection and Elemental Analysis

To perform an informed selection of Rice FOX lines, we searched the rice genome<sup>2</sup> for predicted proteins with similarity to proteins described in the literature as involved in Zn and Fe homeostasis in plants. We used sequences from known genes families as queries, such as ZIP (Zinc-Regulated/Iron-Regulated Transporter Protein; Eide et al., 1996), YSL (Yellow Stripe-Like; Lee et al., 2009), ZIFL (Zinc-Induced Facilitator-Like; Haydon and Cobbett, 2007; Ricachenevsky et al., 2011), MTP (Metal Tolerance Protein; Ricachenevsky et al., 2013b), NRAMP (Natural Resistance Associated Macrophage Protein; Sasaki et al., 2012), OPT (Oligopeptide Transporter; Stacey et al., 2008), VIT (Vacuolar Iron Transporter; Zhang et al., 2012), FER (Ferritins; Stein et al., 2009), PCS (Phytochelatin Synthase; Li et al., 2007), transcription factors of the NAC (Non-Apical Meristem/Arabidopsis Transcription Activation Factor/Cup-Shaped Cotyledon) stress-related subfamily (Ricachenevsky et al., 2013a), IRO2 (Iron-related transcription factor 2; Ogo et al., 2006), and enzymes of the phytosiderophore biosynthetic pathway (Deoxymugineic acid synthase – DMAS; Bashir et al., 2006). Characterized genes for each family cited above were selected and used as queries to search the rice genome. All rice gene products showing at least 30% similarity to query sequences were compiled and used as queries to search the Rice FOX

<sup>2</sup>http://rice.plantbiology.msu.edu/

Database<sup>3</sup> (Sakurai et al., 2011). We identified 42 lines expressing 24 different rice genes, comprising 13 different gene families (**Supplementary Table S1**). Fifteen genes were expressed in two or more of the FOX lines, while nine were expressed in a single line (**Figure 1**).

All FOX lines were grown under the same conditions alongside WT Col-0, in soil amended with subtoxic concentrations of trace elements, watered with Hoagland solution, and after 6 weeks leaves were collected to quantify 20 elements by ICP-MS (Lahner et al., 2003). Comparing the ionomics profiles of each FOX line with WT, we sought to find statistically significant differences in elemental concentrations (**Figure 1**). We found two lines expressing OsZIP7 that showed a consistent 25% increase in leaf Zn concentration each (**Figure 1**, lines K11313\_OsZIP7 and K27616\_OsZIP7). It is important to note the initial screen was performed in segregating FOX lines. We would expect changes in elemental profiles of FOX lines to be dominant, as they are a result of heterologous expression using 35S promoter. Since we analyzed 12 individual plants per line, we expected to find 3 wild types on average for each line, which would allow to detect significant changes in the ionome. Indeed, we demonstrated the feasibility of performing such a screen in FOX lines before the additional time required for isolating homozygous lines. We further confirmed the elemental profile phenotype of OsZIP7-FOX lines in the next generation (hemizygous lines; **Figure 2**) and decided to further characterize the molecular function of OsZIP7.

### OsZIP7 Can Complement Yeast Cells Defective in Zn Uptake

We expressed the OsZIP7 full-length coding sequence in different yeast mutant strains to assess its metal transport ability. When introduced into the Zn uptake-defective zrt1zrt2 mutant, OsZIP7 was able to rescue growth in low Zn medium (**Figure 3**). When

<sup>3</sup>http://ricefox.psc.riken.jp/

expressed in the Fe uptake-defective strain fet3fet4, however, OsZIP7 did not restore growth in high pH medium, which lowers Fe availability (**Supplementary Figure S1**), indicating that OsZIP7 is able to transport Zn but not Fe. This is in contrast to a previous report of OsZIP7 as an Fe transporter (Yang et al., 2009). We also transformed the wild-type strain BY4743 and tested whether OsZIP7 increases cadmium (Cd) toxicity, indicative of Cd transport ability. When growing in media containing 50 µM Cd, both OsZIP7 and empty vector-transformed yeast were able to grow, while AtIRT1-transformed cells grew to a lesser extent (**Figure 3C**). Therefore, we concluded that the OsZIP7 protein is likely to function as Zn transporter, but not as an Fe or Cd transporter.

### OsZIP7 Is Localized at the Plasma Membrane in A. thaliana Protoplasts and N. benthamiana Epidermal Cells

In order to determine the subcellular localization of OsZIP7, we transiently expressed an OsZIP7-YFP construct in A. thaliana protoplasts, either alone or co-transfected with AHA2-RFP, a known plasma membrane marker. The OsZIP7-YFP signal was observed in a pattern that indicated plasma membrane localization (**Figure 4**). When co-expressed with the AHA2-RFP control, (Kim et al., 2001), expression of OsZIP7-YFP and AHA2- RFP were localized in a similar pattern, although it is possible that OsZIP7 also localized to internal membranes (**Figure 4**).

We also transiently expressed the OsZIP7-YFP construct in Nicotiana benthamiana epidermal cells. N. benthamiana leaves were co-agroinfiltrated with the plasma membrane marker PIP2A-CFP (cyan fluorescent protein, Nelson et al., 2007). OsZIP7-YFP and PIP2A-CFP localization is highly similar in cells co-expressing both constructs (**Figure 5**). When plasmolyzed, colocalization of OsZIP-YFP with the plasma membrane marker was also evident (**Figure 5**). Plasma membrane localization is consistent with our yeast complementation results, since OsZIP7

complemented the zrt1zrt2 mutant (**Figure 3**), which lacks two ZIP plasma membrane transporters (MacDiarmid et al., 2000). Thus, OsZIP7 is likely to be a Zn transporter localized at the plasma membrane.

### Expression of OsZIP7 in Arabidopsis Leads to Enhanced Zn Sensitivity and Disruption of Zn Root-to-Shoot Partitioning

To gain more information on OsZIP7 function, we tested the Zn sensitivity of two independent homozygous OsZIP7-FOX lines (OsZIP7-FOX1, derived from FOX lines K11313, and OsZIP7-FOX2, derived from FOX line K27616) grown on media containing excessive Zn levels. When both OsZIP7-FOX lines were grown at control conditions, we observed similar growth compared to wild type lines (WT; **Figure 6A**). However, at 100 µM Zn, both OsZIP7-FOX lines showed decreased growth, with significantly decreased root length and shoot fresh weight compared to wild type. At 200 µM Zn, OsZIP7-FOX lines were stunted, with short roots and small shoots (**Figure 6A**). Root length was 30–35% decreased in OsZIP7-FOX lines compared to wild type in 100 and 200 µM Zn, while shoot fresh weight was about 40% decreased in 100 µM Zn and 60– 65% in 200 µM Zn (**Figures 6B,C**). Thus, we concluded that OsZIP7 expression in Arabidopsis leads to increased sensitivity to Zn.

We also quantified elemental concentration by ICP-MS in roots and shoots of wild type and OsZIP7-FOX1 plants under the same conditions, as well as in plants grown at 50 µM (the highest non-toxic Zn concentration, in our growth conditions). Zn concentrations were significantly higher in leaves of OsZIP7-FOX plants grown in media containing 50, 100, and 200 µM Zn compared to wild type (**Figure 7A**). When comparing Zn concentrations in roots of WT and OsZIP7- FOX1, the opposite effect was observed, with OsZIP7-FOX1

having lower Zn concentrations than wild type, especially under 200 µM Zn, in which OsZIP7-FOX1 root Zn concentrations were only 40% of wild type (**Figure 7B**). To clarify the change in Zn partitioning caused by expression of OsZIP7-FOX1, we compared the shoot-to-root ratio of WT and OsZIP7 plants. Clearly, ectopic expression of OsZIP7 throughout the plant led to increased root-to-shoot translocation of Zn (**Figure 7C**). Interestingly, changes in Fe concentrations were also seen in both roots and shoots of OsZIP7-FOX1 plants: roots of OsZIP7- FOX1 plants had higher Fe concentrations than in the WT when grown under 100 and 200 µM Zn, and shoots had higher Fe concentrations when grown on 200 µM Zn (**Supplementary Figure S2**).

### OsZIP7 Over-Expression Leads to Zn Accumulation in Seeds

Because we are interested in good candidates for biofortification of the edible parts of plants, we decided to investigate the effect of OsZIP7 expression on Arabidopsis seed metal accumulation and distribution. We performed ICP-MS elemental quantification of WT and OsZIP7-FOX seeds from both lines. Zn concentration was 20–25% higher in the OsZIP7-FOX plants than in WT (**Figure 8A**), an increase similar to what was observed in leaves of soil-grown plants by ICP-MS (**Figure 2**). Interestingly, we also observed a small but significant decrease in Cd concentration in OsZIP7-FOX seeds, especially in OsZIP7-FOX2 line, a trait that is desirable when considering OsZIP7 as a candidate for biofortification (**Supplementary Figure S3**). The same trend was observed for Cu concentration (**Supplementary Figure S3**).

We also used synchrotron X-ray fluorescence (SXRF) microtomography to directly visualize metal distribution and abundance in seeds. Zn was clearly more abundant in OsZIP7- FOX seeds compared to WT (about twice as much), but there were no changes in distribution (**Figure 8B**). Abundance of other elements (K, Ca, Mn, Fe, and Cu) did not vary, or varied only slightly (**Figure 8C**). As we have not observed changes in these elements concentration by ICP-MS, except for OsZIP7- FOX2 line which had a slightly decrease in Cu (**Supplementary**

**Figure S3**), it is possible that the observed differences are seedto-seed variation. These results indicate that OsZIP7 constitutive expression increases Zn concentration in Arabidopsis seeds and slightly reduces Cd concentration, indicating OsZIP7 is a good candidate for Zn biofortification.

### DISCUSSION

### Coupling Rice FOX Lines and Ionomics Profiling Is a Fast Method for Identification of Metal-Related Genes From Rice

Numerous proteins have been described as having a role in metal homeostasis in plants, including transporters, transcription factors and enzymes (for reviews, Hindt and Guerinot, 2012; Sinclair and Kramer, 2012; Sperotto et al., 2012; Ricachenevsky et al., 2015). However, gene characterization in crops is not as fast as in A. thaliana. There is need to translate the information from models to agronomically important plants. In rice, many transporters already annotated in the genome do not have an assigned molecular function, and characterization of possible targets for biofortification in other economically relevant cereals such as corn (Zea mays), sorghum (Sorghum bicolor), and wheat (Triticum aestivum) is difficult. Thus, the use of heterologous systems for high-throughput characterization of genes from species that are slower to cultivate or especially difficult to transform is attractive.

The use of Rice FOX lines was successful to describe proteins involved in several processes (Yokotani et al., 2008, 2009; Albinsky et al., 2010; Dubouzet et al., 2011; Higuchi-Takeuchi et al., 2011; Anders et al., 2012). In this work, we highlight the feasibility of using FOX lines coupled with ionomics profiling for characterization of metal-related genes from crop species, such as rice. Besides OsZIP7, which we discussed in detail, our screen identifies other examples of interesting lines that might be studied in depth to understand their role in the regulation of the ionome.

The results reported here are derived from a subset of FOX lines selected because they contain cDNAs from gene families involved in metal homeostasis. Similar focused approaches have successfully identified genes from E. salsugineum that confer heat or heat and salt stress tolerance when expressed in A. thaliana, in which 78 and 433 lines were tested, respectively

(Higashi et al., 2013; Ariga et al., 2015). In these studies, T2 generations were also screened for stress tolerance. Thus, it is clear that phenotyping of FOX lines and similar tools can be performed even without isolation of homozygous lines, since it is expected that altered phenotypes would be dominant. It should be considered that an unbiased screen (i.e., not focused on selected metal transporters) lines expressing heterologous genes could lead to the identification of previously unknown regulators of the ionome. Moreover, it should also be noted that constitutive expression of a gene in a heterologous system might overcome regulatory mechanisms that might modulate protein activity (i.e., post-transcriptional regulation) in their native environment, increasing the chances of identifying interesting genes that could otherwise be regulated by transcriptional and post-transcriptional mechanisms.

### OsZIP7 Is a New Zn Transporter

The first member of the ZIP (Zinc-regulated/Iron-Regulated Protein) family of transporters described was AtIRT1 (Eide et al., 1996), followed by characterization of several ZIP members in Arabidopsis, rice, corn, barley (Hordeum vulgare) among other species (Lee et al., 2010a,b; Li et al., 2013; Milner et al., 2013; Tiong et al., 2014). Plants harbor many ZIP genes in their genomes, with as many as 16 loci in the genomes of some Poaceae (Tiong et al., 2015). ZIP transporters are known for having broad substrate specificity: AtIRT1 is able to transport Zn+<sup>2</sup> , Fe2+, Mn2+, Cd2+, Co2+, Ni2+, and Fe3<sup>+</sup> (Korshunova et al., 1999) while its rice ortholog OsIRT1 transports Fe2+, Zn2+, and Cd2<sup>+</sup> (Ishimaru et al., 2006; Lee and An, 2009). AtIRT2 and AtIRT3 transport Fe2<sup>+</sup> and Zn+<sup>2</sup> , but not Mn2<sup>+</sup> or Cd2<sup>+</sup> (Vert et al., 2001; Lin et al., 2009). In Arabidopsis, others ZIPs are commonly Zn2<sup>+</sup> or Zn2<sup>+</sup> and Mn2<sup>+</sup> transporters, with AtZIP7 also being able to transport Fe2<sup>+</sup> (Milner et al., 2013). Moreover, ZIP proteins characterized in plants are mostly localized to the plasma membrane, which also seems to be true for OsZIP7 based on our data (**Figures 4**, **5**). OsZIP7 is the closest rice homolog of barley HvZIP7 and maize ZmZIP7 (Tiong et al., 2014; Li et al., 2016) and Arabidopsis AtIRT3 and AtZIP4 (Li et al., 2013; Tiong et al., 2015). Of these, HvZIP7 and AtIRT3 had their subcellular localization determined to be at the plasma membrane (Lin et al., 2009; Tiong et al., 2014).

Here we have shown that OsZIP7 was able to complement the Zn-deficient zrt1zrt2 yeast mutant, but not the Fe-deficient fet3fet4 (**Figure 3**). OsZIP7 has been indicated as the rice ortholog of barley HvZIP7, which was recently characterized as a Zn transporter (Tiong et al., 2014, 2015). HvZIP7 was localized to the plasma membrane and increased plant Zn root-to-shoot translocation when compared to WT controls in over-expressing barley plants (Tiong et al., 2014). This is consistent with our observation that OsZIP7 expression in A. thaliana under the control of 35S promoter led to increased Zn concentrations in leaves and seeds and increased root-to-shoot Zn translocation (**Figures 7**, **8**).

Expression of OsZIP7 in Arabidopsis led to increased rootto-shoot Zn translocation when plants are exposed to high Zn in the growth media, with roots of OsZIP7-FOX lines showing lower Zn concentrations compared to WT, whereas shoots have increased Zn concentrations (**Figure 7**). This is similar to what was observed for HvZIP7 over-expression in barley, with plants showing higher Zn concentration in shoots and lower in roots compared to null-segregant lines (Tiong et al., 2014). In both our OsZIP7-FOX lines and in HvZIP7 over-expressing plants, Zn concentrations in shoots and leaves were not changed under control conditions. One possible explanation for these phenotypes is that OsZIP7 expression might increase sink strength in shoots, while also causing increased primary Zn uptake in shoots. Zn xylem-loading transporters such as AtHMA2/AtHMA4 (Hussain et al., 2004) may not limit

Zn translocation to shoots under such conditions. Although the precise mechanism is not clear, OsZIP7 expression in Arabidopsis and HvZIP7 over-expression in barley seem to result in distinct phenotypes compared to over-expression of other ZIP transporters such as OsZIP4, OsZIP5, and OsZIP8 (Ishimaru et al., 2007; Lee et al., 2010a,b). In seeds, however, OsZIP7- FOX lines and HvZIP7 over-expressing lines showed increased Zn concentrations even without excessive Zn in the media, indicating that higher Zn accumulation for biofortification using OsZIP7/HvZIP7 may not require Zn addition (**Figure 8**, Tiong et al., 2014).

Interestingly, two OsZIP7 protein sequences have been reported in rice, differing in only four amino acid positions: OsZIP7, which is characterized in this work, and OsZIP7a, characterized by Yang et al. (2009). Three aminoacid changes are in positions outside transmembrane domains, while one is inside the VII domain. OsZIP7a was shown to not complement Zn-defective yeast mutants (Yang et al., 2009). HvZIP7 failed to complement the yeast strain zrt1zrt2, although several other lines of evidence indicate its function as a Zn transporter (Tiong et al., 2014). Here we showed that OsZIP7 is able to rescue the zrt1zrt2 yeast mutant phenotype to some extent (**Figure 3**). Considering the increased Zn sensitivity of Arabidopsis expressing OsZIP7 (**Figure 6**), these results suggest that OsZIP7 is a Zn transporter. It is possible that OsZIP7 is a low-affinity Zn transporter, as suggested for its closest homologous gene from barley (HvZIP7; Tiong et al., 2014). Interestingly, OsZIP7a has been described as able to complement fet3fet4, suggesting it could transport Fe (Yang et al., 2009). We did not observe fet3fet4 complementation when fet3fet4-expressing OsZIP7 was cultivated in high pH media, while AtIRT1-expressing yeast was able to grow (**Figure 3**), indicating that OsZIP7 does not transport Fe. However, it is still possible that OsZIP7 transports Fe. One hypothesis is that transport is dependent on pH, with high pH decreasing transport function. Despite that, our results support that OsZIP7 is a Zn transporter.

The FOX lines expressing OsZIP7 showed increased Fe concentrations in roots and shoots upon high Zn concentration in the growth media (**Supplementary Figure S2**). This may indicate that OsZIP7 might transport Fe, although it is not clear why Fe concentrations would increase only under high Zn. Heterologous expression in Arabidopsis of OsZIP7 maize ortholog, ZmZIP7, led to increased Fe and Zn concentrations in all tissues, and concomitant upregulation of the Fe uptake regulon, including AtIRT1 (Li et al., 2016). Conversely, HvZIP7 over-expression in barley does not change Fe concentrations in either shoots or roots, even when plants are cultivated under high Zn in the growth media (Tiong et al., 2014). A possible explanation is that

high Zn concentrations induced Fe-deficiency and Fe uptake genes, leading to increased root and shoot Fe concentrations (**Supplementary Figure S2**). Thus, it is more likely that OsZIP7 is not able transport Fe. Still, future work should address if OsZIP7 and OsZIP7a differ in their substrates and if the four distinct aminoacids can change metal specificity (Yang et al., 2009).

### Potential of OsZIP7 for Zn Biofortification of Rice Seeds

OsZIP7 expression in Arabidopsis increased Zn concentration in seeds by 25% (**Figure 8**). Similarly, HvZIP7 over-expression in barley led to significant increase in Zn concentration in grains, with no changes in other elements (Tiong et al., 2014), whereas expression of ZmZIP7 in Arabidopsis led to increased Zn and Fe concentrations in seeds. Interestingly, we have also observed a decrease in Cd concentration of 12–24% in seeds (**Supplementary Figure S3**), indicating that OsZIP7 overexpression in rice could lead to increase Zn in grains without concomitantly increasing Cd levels. From a biofortification perspective, that makes OsZIP7 a good candidate for genetic engineering, since Cd co-transport when manipulating Zn and Fe transport such as the ZIP family members should be considered (Slamet-Loedin et al., 2015). OsZIP7 is highly expressed in developing grains in rice plants (**Supplementary Figure S4**).

Several different genes have been used to improve Zn concentration in rice grains, and increases have been moderate so far (for a review, see Ricachenevsky et al., 2015). Two successful transgenic approaches involved activation tagging or over-expression of nicotianamine synthase (NAS) genes (Johnson et al., 2011; Lee et al., 2011). Presumably, increased levels of nicotianamine in these plants facilitate Zn loading in the phloem and translocation to grains, but increased available Zn for translocation might lead to further accumulation. Thus, there is still potential to increase Zn levels. Either OsZIP7 overexpression as a single transgene, combined with OsNAS2 of expressed in specific cell types such as the endosperm could be promising to generate biofortified rice in the future.

### CONCLUSION

We have demonstrated that Arabidopsis lines generated to heterologously express rice genes are useful for fast screening genes that are involved in metal homeostasis when combined with elemental analyses. We have also molecularly characterized OsZIP7, a Zn plasma membrane-localized transporter from rice. Based on our results, OsZIP7 is a good candidate for overexpression in rice to generate lines that are able to accumulate Zn in their seeds.

### AUTHOR CONTRIBUTIONS

FKR, TP, DES, JPF, and MLG designed the experiments. FKR, TP, and MNH performed the experiments. FKR, TP, BHNO, and JD performed the analyses. FKR, TP, SL, BHNO, TST, FSM, MNH, DES, JPF, and MLG wrote the manuscript. All authors approved the manuscript.

### FUNDING

The authors would like to thank FAPERGS (Fundação de Amparo à pesquisa do Estado do RS) and CAPES (Coordenação de Aperfeiçoamento de Pessoal de Nível Superior) for funding. This work was supported by a grant from the National Science Foundation's Plant Genome program to DES, MLG, FKR, and JPF (DBI 0701119). Use of NSLS facility was supported by the Department of Energy under Contract DE-AC02-98CH10886. Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) granted a fellowship to FKR and a research grant to JPF.

### ACKNOWLEDGMENTS

The authors would like to thank Brett Lahner and Elena Yakubova for technical assistance for plant growth and ICP-MS analyses, and Prof. Dr. Marcia Maria Auxiliadora Naschenveng Pinheiro-Margis for giving access to laboratory facilities. The authors would also like to especially thank John Danku, which sadly passed away during the preparation of the manuscript. John premature death is a great loss to science and the field of ionomics.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018.00865/ full#supplementary-material

FIGURE S1 | Yeast phenotype complementation assays. Empty pDR195 vector, pDR195-OsZIP7 or pDR195-AtIRT1 constructs were transformed into yeast cells. Liquid cultures were diluted as indicated before plating. (A) Fe-uptake defective strain fet3fet4 transformed with each construct growing under Fe-sufficient (pH 5.3) or Fe-deficient (pH 6.0) conditions. (B) Wild type strain BY4743 transformed with each construct growing under control or 50 µM Cd conditions.

FIGURE S2 | Metal concentrations in shoots (A–D) and roots (E–H) of Col-0 (white bars) and OsZIP7-FOX Arabidopsis plants (gray bars) grown for 15 days in Minimal Media containing 50, 100, or 200 µM Zn (n = 5). Concentrations of Mn (A,E), Fe (B,F), Cu (C,G), and Cd (D,H) are shown. Different letters show significant differences by ANOVA and Tukey HSD.

FIGURE S3 | Metal concentrations in seeds of Col-0 and OsZIP7-FOX1 and OsZIP7-FOX2 plants (n = 8). Concentrations of Mn (A), Fe (B), Cu (C), and Cd (D) are shown. Different letters show significant differences by ANOVA and Tukey HSD.

FIGURE S4 | Rice OsZIP7 transporter expression pattern based on data from the eFP Browser public database (http://bar.utoronto.ca/efp\_rice/cgi-bin/ efpWeb.cgi).

TABLE S1 | Rice FOX lines used in this work.

TABLE S2 | Primers used in this work.

### REFERENCES

fpls-09-00865 June 30, 2018 Time: 16:14 # 11


reveals single-gene strategies for effective iron- and zinc-biofortification of rice endosperm. PLoS One 6:e24476. doi: 10.1371/journal.pone.0024476


computed microtomography to image metal compartmentalization in Alyssum murale. Environ. Sci. Technol. 39, 2210–2218. doi: 10.1021/es0492034


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Ricachenevsky, Punshon, Lee, Oliveira, Trenz, Maraschin, Hindt, Danku, Salt, Fett and Guerinot. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Genetic Basis and Breeding Perspectives of Grain Iron and Zinc Enrichment in Cereals

Ana Luisa Garcia-Oliveira<sup>1</sup> \*, Subhash Chander <sup>2</sup> , Rodomiro Ortiz <sup>3</sup> \*, Abebe Menkir <sup>1</sup> and Melaku Gedil <sup>1</sup>

1 International Institute of Tropical Agriculture, Ibadan, Nigeria, <sup>2</sup> Department of Genetics & Plant Breeding, Chaudhary Charan Singh Haryana Agricultural University, Hisar, India, <sup>3</sup> Department of Plant Breeding, Swedish University of Agricultural Sciences, Alnarp, Sweden

Micronutrient deficiency, also known as "hidden hunger," is an increasingly serious global challenge to humankind. Among the mineral elements, Fe (Iron) and Zn (Zinc) have earned recognition as micronutrients of outstanding and diverse biological relevance, as well as of clinical importance to global public health. The inherently low Fe and Zn content and poor bioavailability in cereal grains seems to be at the root of these mineral nutrient deficiencies, especially in the developing world where cereal-based diets are the most important sources of calories. The emerging physiological and molecular understanding of the uptake of Fe and Zn and their translocation in cereal grains regrettably also indicates accumulation of other toxic metals, with chemically similar properties, together with these mineral elements. This review article emphasizes breeding to develop bioavailable Fe- and Zn-efficient cereal cultivars to overcome malnutrition while minimizing the risks of toxic metals. We attempt to critically examine the genetic diversity regarding these nutritionally important traits as well as the progress in terms of quantitative genetics. We sought to integrate findings from the rhizosphere with Fe and Zn accumulation in grain, and to discuss the promoters as well as the anti-nutritional factors affecting Fe and Zn bioavailability in humans while restricting the content of toxic metals.

Keywords: biofortification, cereals, iron, zinc, micronutrient deficiency, toxic risks

## INTRODUCTION

From the 1950s onwards, the advancement in science and technology together with concerted efforts of international and national agricultural organizations has resulted in significant gains in world food production widely referred to as the "Green Revolution" (Ortiz, 2011). Globally, the availability of sufficient quantities of food is not only a simple achievement of the Green Revolution but has also helped to avert large-scale famines and social and economic upheavals (Khush, 1999). Without the Green Revolution, crop yields in Asia and Latin America would be at least 20% less, food prices would be up 19%, calorie consumption would be down by about 5%, and the number

### Edited by:

Felipe Klein Ricachenevsky, Universidade Federal de Santa Maria, Brazil

### Reviewed by:

Ümit Bari ¸s Kutman, Gebze Technical University, Turkey Hamid Khazaei, University of Saskatchewan, Canada

#### \*Correspondence:

Ana Luisa Garcia-Oliveira a.oliveira@cgiar.org Rodomiro Ortiz rodomiro.ortiz@slu.se

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 16 January 2018 Accepted: 11 June 2018 Published: 02 July 2018

#### Citation:

Garcia-Oliveira AL, Chander S, Ortiz R, Menkir A and Gedil M (2018) Genetic Basis and Breeding Perspectives of Grain Iron and Zinc Enrichment in Cereals. Front. Plant Sci. 9:937. doi: 10.3389/fpls.2018.00937

**89**

**Abbreviations:** Fe, Iron; Zn, Zinc; CGIAR, Consultative Group on International Agricultural Research; QTL, Quantitative Trait Locus; MAS, Marker-aided Selection; MAB, Marker-aided Breeding; ICP-OES, Inductively Coupled Plasma-Optical Emission Spectrometry.

of malnourished children would be up by at least 2% (Evenson and Gollin, 2003). The widespread adoption of these technologies has made it possible to improve the per capita calorie consumption in different continents, especially in the developing world. Overall, the Green Revolution has paid rich dividends in food grain production, particularly cereals, that have led to a significant reduction in the proportion of undernourished people worldwide; however, the problem of malnutrition or lack of quality food still persists, leading to an economic burden for society (Pingali, 2012).

Anemia is the most common human nutritional malaise, resulting from iron (Fe) deficiency and affecting 32.9% people worldwide; meanwhile, zinc (Zn) deficiency affects 17% of the world's population, with the highest risk occurring in sub-Saharan Africa and South Asia (Wessells et al., 2012; Kassebaum et al., 2014). In the twenty-first century, there are strong concerns worldwide regarding the ability to produce nutritionally rich food because cereals are inheritably poor in essential micronutrients. Moreover, owing to a burgeoning human population and industrialization, this situation may be further compounded by the production of cereals in areas with low mineral phytoavailability (White and Broadley, 2009). Thus, there is an urgent global need to cope with the problem of micronutrient deficiencies that contribute to what is referred to as "hidden hunger" and affect at least 2 billion people (or 1 out of 3), mostly in sub-Saharan Africa, South Asia, and Latin America (FAO et al, 2015).

Diversity in diets in order to provide adequate micronutrient consumption is difficult to achieve in the developing world where resource-poor people cannot afford a variety of different foods. For example, many of the relatively cheap and staple crops such as cereals (**Table 1**), roots [cassava (Manihot esculentaz)], tubers [sweet potato (Ipomoea batatas) and yam (Dioscorea spp.)], and plantain (Musa spp.) that play a very important role in the daily diets of resource-poor people lack high enough amounts of micronutrients (Gibson et al., 2010). Hence, malnutrition and poor health affect these people, who may suffer from blindness or stunting, and sometimes even face death. To overcome this "hidden hunger," medical supplements and fortification have been pursued (Underwood, 2000). In fact, food fortification has a long history of use in industrialized countries and relies on the addition of micronutrients to processed foods. However, food fortification tends to have a rapid but less sustainable impact, because various safety, technological, and cost considerations may place constraints on such interventions (Allen et al., 2006). Furthermore, such interventions do not always reach the desired target populations (Pfeiffer and McClafferty, 2007). By increasing the micronutrient content of energy-rich crops, micronutrient intakes among the poor can be increased, thereby leading to decreases in the prevalence of micronutrient deficiencies.

Biofortification is a strategy that involves the use of plant breeding or agronomic practices to increase the density of essential nutrients in the edible part of staple crops that may help to combat deficiencies among poor people who survive on main staples such as cereals (www.harvestplus.org). Agronomic biofortification is a fertilizer-based approach that relies on soil and/or foliar application of micronutrients either alone or in combination with other fertilizers. It is well-established that a Zn fertilizer strategy is an effective way to biofortify cereal crops with Zn, but recurrent cost is involved (Cakmak and Kutman, 2018). By contrast, genetic biofortification is a seed-based approach that complements agronomic biofortification and also current intervention methods such as supplementation and fortification of foods consumed daily. The aim of this strategy is to enhance the content and bioavailability of micronutrients such as minerals and vitamins in crops through plant breeding, thereby impacting favorably the diets of targeted populations, particularly the resource poor worldwide (Bouis and Welch, 2010). Because of its cost effectiveness, the 2008 Copenhagen Consensus ranked biofortification fifth for combating the world's greatest challenges (http://www.copenhagenconsensus. com/publication/second-copenhagen-consensus-biofortification -best-practice-meenakshi). Biofortifying staple crops through plant breeding is therefore a key option to improve micronutrient deficiency in human diets (Bouis and Saltzman, 2017).

### RELEVANCE OF FE AND ZN TO HUMAN HEALTH

Among the mineral nutrients required by humans for their well-being, Fe and Zn play vital roles in numerous metabolic processes and are required in trace amounts by plants as well as animals (Welch and Graham, 2004). For instance, Fe is a well-known essential component of hemoglobin and myoglobin, which are involved in oxygen transport and storage. The greatest effects of Fe deficiency anemia are seen in females during adolescence and pregnancy. Likewise, Fe deficiency affects children's cognitive development until adolescence, and it increases their susceptibility to infectious diseases and mortality (Oliver and Gregory, 2015).

Similarly, Zn is also an essential cofactor for many enzymes and regulatory proteins, and it plays an important role in DNA as well as RNA synthesis and gene expression. Children show stunted growth and neurobehavioral difficulties resulting from Zn deficiency, which may increase the incidence and severity of diarrhea, among other conditions (Nriagu, 2007). Furthermore, Zn deficiency seems to be significantly related to anemia associated with Fe deficiency because Zn controls Fe absorption in the intestines (Chang et al., 2010; Graham et al., 2012). Hence, Fe and Zn are acknowledged as outstanding micronutrients owing to their importance in global public health.

### PHYSIOLOGY, GENETICS, AND MOLECULAR ASPECTS OF FE AND ZN ENRICHMENT IN CEREAL GRAINS

For genetic biofortification, a better understanding of the key steps of mineral nutrient transport from the rhizosphere to grains is needed, which involves coordination of complex physiological steps such as acquisition of Fe and Zn in roots (uptake), subsequent long-distance transport from roots to shoots, and further redistribution toward the developing seeds (Zhao and McGrath, 2009; Carvalho and Vasconcelos, 2013). Although Fe and Zn are known to accumulate in grains, further insights regarding the underpinning physiology and genetics are yet to be revealed. Soil redox potential and pH affect uptake of Fe by roots and Zn accumulation in grains. Fe is mostly available in the rhizosphere as low solubility Fe3<sup>+</sup> oxyhydrates, while Fe is oxidized in aerobic soils with high pH, thus occurring as insoluble ferric oxides. Free ferric Fe from the oxides become available at low pH for further uptake by roots (Lindsay and Schwab, 1982).

Plants have developed different strategies for Fe uptake from the rhizosphere: Strategy I involving ferrous Fe2<sup>+</sup> (non-Poaceae) and Strategy II utilizing ferric Fe3<sup>+</sup> (Poaceae), referred to as reducing and chelating strategies, respectively, or a combination of strategies I and II (Connorton et al., 2017). Poaceae family members such as rice (Oryza sativa L.), maize (Zea mays L.), and wheat (Triticum aestivum L.) follow Strategy II, and their root epidermis secretes phytosiderophores (PSs) that form stable Fe(III) chelates in the rhizosphere (Roberts et al., 2004). TOM1 (Transporter of Mugineic acid family phytosiderophores)/ZIFL4 belongs to the major facilitator superfamily (MFS) that exports Fe3+-PS chelates in rice and barley (Nozoye et al., 2011). From the rhizosphere, these Fe(III)-PS complexes can be taken up into root cells by Yellow Stripe-Like proteins (YSLs). The maize oligopeptide transporter YS1 is the founding member of the YSL family, which facilitates Fe3+-PS complex uptake from the rhizosphere, and subsequently the role of YSL15 has also been confirmed in rice (Curie et al., 2001; Inoue et al., 2009). Besides Strategy II, both rice and barley have a functional homolog of IRT1 (Iron-Regulated Transporter 1) that allows direct uptake of Fe2<sup>+</sup> from the rhizosphere, thus clearly showing the different uptake strategies for Fe2<sup>+</sup> and Fe3<sup>+</sup> in these cereal crops.

Soil pH significantly influences Zn acquisition and uptake from the rhizosphere by roots because Zn binds tightly to soil elements and plant cell wall parts under high pH. However, under anaerobic conditions in soils, additional factors such as soil redox potential, total sulfur content, and soluble bicarbonate also affect the availability of Zn (Impa and Johnson-Beebout, 2012). As is also the case for Fe(III), which exhibits even lower solubility, Zn solubilization in the rhizosphere is thought to occur via plantmediated acidification and secretion of low molecular weight organic chelator (Sinclair and Krämer, 2012). Details regarding the role and contribution of Zn acquisition by the plant remain unknown. The uptake of Zn may occur as a divalent cation (Zn2+) or as a Zn-PS complex formed with PSs known as Fe3+ chelators, which are secreted by roots of the plant (von Wiren et al., 1996). ZIP-like transporters may take up Zn as noted in Strategy I plants (Ramesh et al., 2003). Thus, modification of rhizosphere chemistry through root architecture and by secretion of more root exudates that can alter soil pH could be the first promising target for improving Fe and Zn acquisition in cereal roots.

After roots acquire Fe and Zn, their translocation to the shoot and further movement to other vegetative organs depend of several steps, going through symplast, xylem, and phloem. Physiological studies have indicated that chelating molecules such as citrate, nicotianamine, and mugineic acid play a vital role in symplast heavy metal homoeostasis including Fe and Zn. Both minerals move through the xylem into the shoot, where Zn can move as free ions or in a complex with organic acids, while Fe is chelated to organic compounds of low molecular weight that are subsequently translocated by the xylem and phloem to other plant organs (Rellan-Alvarez et al., 2010; Lu et al., 2013). In plants, leaves are the most important sink tissue for these minerals where they are required in the plastids and mitochondria for numerous enzymes essential for photosynthesis and other cellular metabolic processes (Gupta et al., 2015). The FDR3 gene has an important role in transporting Fe (Green and Rogers, 2004); while for Zn transport, heavy metal ATPase (HMA), a member of the P1B ATPase family, is a likely candidate for performing this task (Eren and Arguello, 2004: Hussain et al., 2004). At the molecular level, a plethora of genes associated with influx and efflux transporters has been discovered and extensively characterized in plants, including cereals, and which are involved in translocation of these minerals (Kobayashi and Nishizawa, 2012; Ricachenevsky et al., 2015; Vasconcelos et al., 2017). Moreover, transcription factors have also been identified that regulate the genes involved in the uptake of Fe and Zn and synthesis of PSs in cereals. Despite the tremendous improvement in understanding those components participating in translocation, it remains difficult to define precisely the contribution of each of the components in the metal movement flux for each translocation step.

Plants remobilize and move nutrients from vegetative source organs into seeds during the filling of grains (Waters and Sankaran, 2011). Hence, the amount of both Fe and Zn in the cereal grain is dependent on the former physiological processes: firstly, their acquisition from the soil by roots, and secondly, transportation to the shoots and further remobilization of stored minerals from leaves when they senesce at grain filling. Despite the large amount of these minerals in the vegetative tissues of cereals, their remobilization from leaves is an important process (from senescence to grain filling), contributing to accumulation in the seeds. Fe and Zn accumulate throughout cereal seeds, being primarily concentrated in the aleurone and embryo parts and to lesser extent in the endosperm, except in rice and barley where Zn appears to be less strictly confined to the aleurone than Fe (Persson et al., 2009). From a biofortification perspective, the heterogeneous distribution of these essential mineral elements in cereal grains further complicates the situation for their efficient loading into the core endosperm (Cakmak et al., 2010). Fe, Zn, copper (Cu), and manganese (Mn) are micronutrients that primarily accumulate in the seed aleurone layer, where phytic acid (the main form of Pi storage in seeds) is a strong chelator of metal cations, binding them to form phytate, a salt of inositol phosphate (Raboy, 2009). However, recent studies have shown that Fe is mainly associated with phytic acid, while Zn is bound to proteins, which clearly suggests that Fe and Zn have a different speciation in cereal grain tissues (Persson et al., 2009; Kutman et al., 2010). Future research, hence, is needed to elucidate the molecular aspects of bivalent metal speciation including Fe and Zn in different tissues of seeds for efficient Fe and Zn biofortification strategies in cereals (Persson et al., 2016).

### BREEDING PERSPECTIVE OF FE AND ZN BIOFORTIFICATION IN CEREALS

To address the widespread prevalence of micronutrient deficiency, especially pro-vitamin A, Fe, and Zn, the Consultative Group on International Agricultural Research (CGIAR; https:// www.cgiar.org) HarvestPlus in collaboration with international and national research institutes emphasized biofortification of staple food crops as a cost-effective, easily applicable, and sustainable approach to benefit low-income households. This may complement other efforts aimed at reaching rural populations in developing countries (www.harvestplus.org). Considering the average content of Fe and Zn in cereal grains and their retention after processing, as well as addressing the issues related to the bioavailability of these mineral nutrients, HarvestPlus established target levels of these nutritionally important traits in cereal grains (**Table 1**). The initial screening of a large amount of crop germplasm suggested the existence of substantial genetic variation for these traits in cereal crops and their wild relatives. Besides the complex nature of these traits, the assays required to measure micronutrient content in plant samples are tedious and costly.

While plant breeding has been pursued significantly to achieve biofortification in staple crops, the success in breeding for Fe and Zn biofortification in cereal crops lags behind the development of pro-vitamin A enriched cultivars of staple crops (Andersson et al., 2017). The possible reason for this is a better understanding of the carotenoid biosynthesis pathway in plants, especially in maize, that has led to the deployment of functional markers for pro-vitamin A biofortification in maize (Gebremeskel et al., 2018). Although, there have been significant advances in elucidating the mechanisms related to Fe and Zn homeostasis in model plants, detailed understanding is still lacking. It is noteworthy to highlight that several genes controlling Fe and Zn homeostasis in cereal grains—particularly rice—have been characterized, but their role in genotypic variation for the accumulation of these minerals in the grain remains unclear. Hence, a more holistic breeding approach is required for Fe and Zn biofortification of cereal grains that emphasizes the genetic

TABLE 1 | Information and assumptions used to set target levels for mineral nutrient content in grains of biofortified staple cereals by CGIAR HarvestPlus.


Considering 90% retention of both Fe and Zn after processing, and 5 and 25% bioavailability for Fe and Zn, respectively, except Fe in rice grain where bioavailability is 10% (adapted from Bouis and Welch, 2010).

enhancement of the contents of these minerals in cereal grains together with the factors that determine their bioavailability in humans such as inhibitors and/or enhancers (**Figure 1**). Further, there is the need to be cautious regarding inadvertent enhancement of non-essential/certain toxic elements, such as cadmium (Cd), in cereal grains.

### EXPLORING GENETIC VARIABILITY FOR FE AND ZN ENHANCEMENT IN CEREAL GRAINS

A pre-requisite for breeding for a specific trait is the availability of its genetic variation within the target gene pool. The task is somewhat complex while breeding for Fe and Zn biofortification in cereal grains because their concentration in the grain depends on various physiological processes. Plant breeders rely on additive genetic effects, transgressive segregation, and heterosis for improving desired traits when enough genetic variation exists. Recently, the genetic variability for these minerals in cereals, particularly maize, rice, wheat, barley (Hordeum vulgare L.), sorghum (Sorghum bicolor L.), and pearl millet (Pennisetum glaucum L.), which are the six most important crops and represent 89% of all cereal production worldwide, was reviewed, and the existence of significant genetic differences for these minerals was reported (Teklic et al., 2013; Goudia and Hash, 2015; Gregory et al., 2017). A survey of 1,400 improved maize genotypes and 400 landraces maintained at the genebank of the International Maize and Wheat Improvement Center (CIMMYT, El Batan, México) indicated about four- to six-fold variation for grain Fe and Zn (Bänziger and Long, 2000). Among the tropical-adapted maize inbreds of the International Institute of Tropical Agriculture (IITA, Ibadan, Nigeria), the best inbreds exhibited 32 to 78% more grain Fe and 14 to 180% more grain Zn over their trial mean (Menkir, 2008). Similarly, several-fold variation for grain Fe and Zn in disomic hexaploid bread wheat has also been reported (Velu et al., 2014; Goudia and Hash, 2015). Correspondingly, substantial variation for these minerals in rice grain has also been reported among different cultivars; however, grain polishing removed up to 50% of the Fe from the brown rice grain (Gregorio et al., 2000; Prom-u-thai et al., 2007). About, two-fold higher Zn concentration but slightly lower Fe concentration was reported in indica rice compared with japonica rice (Yang et al., 1998). Significantly lower Fe and Zn contents were found in the seed of modern cultivars of rice than in landraces (Anandan et al., 2011), thus arguing that breeders failed in introducing quality improvement, particularly for micronutrients, because they gave priority to other traits such as size, shape, and appearance of grain, milling quality, and cooking features. However, a notable aspect of the lower content of these nutrients in the seed of modern cultivars compared with landraces/germplasm may be the yield dilution effect; the total grain nutrient content may not differ significantly between landraces and modern cultivars, and part of this effect could be ascribed to higher grain yield in modern cultivars (Pfeiffer and McClafferty, 2007; McDonald et al., 2008). Therefore, grain yield

must be kept in mind when discussing breeding solutions in cereals biofortification.

Simple and reliable phenotyping is always preferred by breeders, but an extensive survey of the literature pertaining to the existence of genetic variability for grain Fe and Zn contents in cereal crops clearly suggests that accurate measurement of these mineral nutrients is a challenging task. For the measurement of Fe and Zn content in plants and related material, a wide range of analytical methods is available ranging from semi-quantitative [Perl's Prussian blue and diphenyl thiocarbazone-based dithizone] to fully quantitative [atomic absorption spectrometry, inductively coupled plasma-optical emission spectrometry (ICP-OES), ICP-mass spectrometry, near-infrared reflectance spectrophotometry, X-ray fluorescence spectrometry, elemental distribution maps secondary ion mass spectrometry, synchrotron X-ray, fluorescence spectroscopy, micro-X-ray fluorescence spectroscopy, and Laser-induced breakdown spectroscopy], which can differ substantially with respect to the many attributes describing method performance (Ihnat, 2003; Pfeiffer and McClafferty, 2007). Therefore, the choice of analytical method would depend on the purpose and precision required in estimation. Alternatively, non-destructive quantitative techniques could be the choice method from a breeding perspective, because initial screening for grain Fe and Zn content together with other mineral elements in an appropriately large number of breeding lines can be obtained with minimal or no sample preparation, thereby enabling the discarding of progenies with the lowest content of these mineral elements. There are numerous possibilities for introducing variation in the results of different studies. Firstly, sensitivity of the method used for the quantification of Fe and Zn contents. Secondly, improper postharvest handling of the samples has also been observed to give erroneous results while estimating grain micronutrient concentrations. Furthermore, it is noteworthy that the variability among microenvironments for Fe and Zn may be significant, and most of the research presenting extremely high or low values of these mineral nutrients is based on single-year data; thus, the results are affected by a significant confounded influence

of the sampling and environment. Additionally, extremely high trial mean values of these nutrients reported in some studies appear to be affected by the prior use of manure at some locations, because the level of these nutrients in cereal grains, particularly Zn content, can be even lower when grown in infertile/Zndeficient soils (Cakmak et al., 2010; Xu et al., 2011; Velu et al., 2014).

### THE VALUE OF WILD RELATIVES FOR FE AND ZN BIOFORTIFICATION IN CEREALS: OPPORTUNITIES FOR GENETIC GAIN

Modern cereal cultivars have a lower concentration of Fe and Zn in grains than landraces. This is because breeding has been mainly aimed at increasing grain yield or improving host plant resistance, among other target traits, instead of also improving the micronutrient concentration in grain. Utilization of landraces or crop wild relatives for genetic gain is not a new concept. The genetic variability for content of micronutrients is becoming acknowledged as a desired trait of crop wild relatives, particularly for rice and wheat.

The wild Triticum and Aegilops species have very high grain Fe and Zn contents when compared with both bread and durum wheat (Cakmak et al., 2000; Ortiz-Monasterio and Graham, 2000; Chhuneja et al., 2006; Rawat et al., 2009) as well as synthetic amphiploids (Calderini and Ortiz-Monasterio, 2003). Wild species such as Triticum boeoticum, Triticum monococcum, Triticum dicoccoides (wild emmer), Aegilops tauschii, and Aegilops speltoides were found to have substantially higher levels of these minerals (two- to three-fold) in their grain than modern wheat cultivars (Rawat et al., 2009; Xu et al., 2011; Velu et al., 2014). Compared with the alternative durum allele, recombinant chromosome substitution lines (RSLs) with T. dicoccoides carrying the Gpc-B1 allele had a 12, 18, and 38% higher concentration of Zn, Fe, and protein content, respectively (Cakmak et al., 2004). High concentrations of Fe, Mn, and Zn in grain were stable across sites (Distelfeld et al., 2007). Hence, T. dicoccoides seems to be an interesting source for enhancing both protein and essential mineral content and concentration in wheat cultigens. Similarly, wild accessions of rice such as Oryza rufipogon, Oryza nivara, Oryza latifolia, and Oryza officinalis seem to be assets in rice improvement, showing higher values for Fe and Zn content than cross-bred cultivars (Banerjee et al., 2010; Anuradha et al., 2012).

### IDENTIFICATION OF MOLECULAR MARKERS FOR GRAIN FE AND ZN BIOFORTIFICATION IN CEREALS

The finding of quantitative trait loci (QTLs) led to dissection of complex multigenic traits that were difficult to improve through crossbreeding before the progress made in DNA-aided analysis. QTL mapping for mineral nutrients in cereal grains has allowed the identification of many QTLs for both Fe and Zn (**Table 2**). Most of these QTLs, with a few exceptions, do not seem to be stable across sites. Furthermore, QTL mapping has also clearly indicated the role of epistasis in expression of these traits in cereal grains through interactions with other loci (**Table 2**).

Unfortunately, there is no literature indicating so far a success story for marker-aided selection (MAS) for improving Fe and Zn in cereal grains, but some progress has been made that has laid the foundation stone toward breeding for Fe and Zn biofortification in cereals using MAS. For instance, some of the QTLs identified for Fe and Zn are co-localized, thereby suggesting common mechanisms for their transport. Furthermore, some QTLs for these mineral nutrients are also co-localized with those for other mineral elements such as phosphorus (P) and calcium (Ca), or other agronomically important traits including grain protein content and grain weight (**Table 2**). Fine mapping of candidate genes related to various QTLs could be a further step for developing biofortified germplasm.

Rice is a model plant for cereal genetics. Chromosome 11 of rice bears a QTL for Zn concentration in the grain, which seems to be associated with OsNAC5—a transcription factor that appears to be related with the remobilization of Zn from green tissues to the seed (Lu et al., 2008; Sperotto et al., 2009, 2010). In unpolished rice grains, 10 candidate genes known for Fe and Zn homeostasis were localized in the QTL regions whereas another six candidate genes were close to QTLs on chromosomes 3, 5, and 7, respectively (Anuradha et al., 2012). Based on these results, Anuradha et al. (2012) emphasized the importance of candidate genes OsYSL1 and OsMTP1 for Fe; OsARD2, OsIRT1, OsNAS1, and OsNAS2 for Zn; and OsNAS3, OsNRAMP1, heavy metal ion transport, and APRT for both Fe and Zn biofortification of grain in rice. Recently, Norton et al. (2014) also found several QTLs for grain Zn and other elements in diverse rice genotypes using genome-wide association mapping, but the known Zn-related genes were not found in these regions, thereby showing the novelty of their results.

The first QTL for grain Fe and Zn in wheat was found by Joppa et al. (1997), who mapped a major QTL (Gpc-B1) for grain protein content to chromosome 6BS in a population of recombinant inbred lines (RILs) that derived after crossing "Langdon" (LDN)—a durum wheat cultivar—and DIC6B—a chromosome substitution LDN line including wild emmer wheat. Subsequently, the Gpc-B1 locus was also found to be related to high concentrations of both Fe and Zn, as well as with fast leaf senescence. The dissection of the Gpc-B1 locus by positional cloning revealed that the gene underlying the Gpc-B1 locus encodes NAM1, which is a NAC transcription factor that belongs to a protein group that includes "No Apical Meristem" (NAM) in Arabidopsis thaliana (Uauy et al., 2006; Distelfeld et al., 2007). The ancestral wild wheat allele NAM-BI leads to fast senescence and enhances the remobilization of nutrients from the leaves to the developing grains. Modern wheat cultivars have instead a non-functional NAM-BI. Both Fe and Zn can be manipulated together because of the co-localization of their QTLs (Shi et al., 2008), whose mapping was facilitated by using RILs or diverse double-haploid (DH) populations (**Table 2**). The identification and tagging of DNA markers related to both traits provides an aid for crossbreeding, thereby accelerating biofortification for Fe and Zn in grains of cereals.

TABLE 2 | Main effect and epistatic quantitative trait loci (QTLs) associated with Fe and Zn accumulation in different tissues and their co-localization with other traits in cereal crops reported by different groups.


(Continued)

### TABLE 2 | Continued


MQTL, main effect QTL; EQTL, epistatic QTL; GFe, grain iron; GZn, grain zinc; GPhy, grain phytic acid; LFe, leaf Fe; LZn, leaf Zn; Zneffi, Zn efficiency; ShZn, shoot Zn; ClZn, clum Zn; FlZn, flag leaf Zn; YlFe, young leaf Fe; YlZn, young leaf Zn; CobF, cob Fe; CobZn, cob Zn; GbioFe, grain bioavailable Fe; DH, double haploid; IL, inbred line; RIL, recombinant inbred line; Fn, segregating offspring.

### BREEDING FOR ENHANCEMENT OF FE AND ZN BIOAVAILABILITY: ROLE OF INHIBITORS AND PROMOTERS

The ultimate goal of the breeding for Fe and Zn biofortification in cereals is to satisfy the requirement of the human body for these minerals. Thus, the bioavailability of these minerals should be measured according to the cereal-based foods consumed rather than as their quantity in the cereal grains. Considering the low bioavailability of these minerals, it seems to be difficult to meet this demand alone by enhancing the grain Fe and Zn content in cereals. Hence, Fe and Zn should be easily absorbable in the intestines—a difficult task due to inhibitors (e.g., phytic acid) or promoters such as prebiotics enhancing their absorption in the gut—to ensure their effective availability from cereal-based diets (Roberfroid, 2007; White and Broadley, 2009; Dwivedi et al., 2014).

Phytic acid is an effective chelator of positively charged elements such as Ca, Fe, Mn, magnesium (Mg), potassium (K), and Zn, which after human or animal consumption binds to these minerals in the intestines forming mixed salts that are further excreted, thus resulting in mineral deficiency in human populations (Ali et al., 2010). Although phytate is considered an inhibitor of Fe and Zn bioavailability and therefore referred to as an anti-nutritional trait in cereal grains, it may have some health benefits such as being an antioxidant or anticarcinogen (Schlemmer et al., 2009). Furthermore, the important role of phytic acid has also been noted in plant traits such as seedling vigor or protection of seeds against oxidative stress during their lifespan (Doria et al., 2009). Hence, the existence of a minimum concentration of phytic acid in the cereal grains is still under scientific debate from health as well as crop performance perspectives.

Nonetheless, various low-phytic acid (lpa) mutants have been found in barley, maize, rice, and wheat exhibiting 50 to 95% reduced phytic acid P (Rasmussen and Hatzack, 1998; Raboy et al., 2000; Pilu et al., 2003; Shi et al., 2003; Guttieri et al., 2004; Liu et al., 2007). However, the pleiotropic effects of these lpa mutations resulted in significant grain yield loss and also affected other agronomic traits such as poor seed germination along with low grain weight and starch accumulation, and poor plumpness, among other characteristics (Raboy et al., 2000; Pilu et al., 2003; Guttieri et al., 2006; Zhao et al., 2008). Thus, seeking available variability will assist in finding new genetic mechanisms that reduce phytate and avoid any grain yield penalty in cereals. About two-fold variation for seed phytate concentration has been observed in wheat and rice (Liu et al., 2006; Stangoulis et al., 2007). Interestingly, two QTLs for seed phytate concentration have been identified so far in rice; one each on chromosomes 5 and 12 accounting for phenotypic variance of 24 and 15%, respectively (Stangoulis et al., 2007). Genetic markers nearby these QTLs should be used for testing their efficacy as aids for selecting low-phytate lines.

With the growing awareness about diet-related health problems, the presence of health-promoting natural compounds in staple foods, which was earlier considered of minor importance, has attracted greater attention in the food industry. Prebiotics are a group of carbohydrates that are known to confer benefits for human health by selectively promoting the growth or activity of gut microbiota (Dwivedi et al., 2014). Thus, prebiotics in cereal grains should be taken into account for enhancing Fe and Zn bioavailability while undertaking biofortification. To date, scarce research has reported the influence of prebiotics on the absorption of these mineral nutrients in humans and the prevalence of the natural variation and inheritance of these compounds in cereal grains. There is, however, significant genetic variability for inulin concentration in the grains of maize and rice, both of which have lower inulin concentration that those of rye and wheat (Genc et al., 2005; Huynh et al., 2008a). Similarly, substantial genetic variation has also been reported in grain fructan content ranging from 0.7 to 2.9, 3.6 to 6.4, and 0.9 to 4.2% of grain dry weight in the different genotypic lines and cultivars of wheat, rye, and barley, respectively (Boskov-Hansen et al., 2003; Huynh et al., 2008a; Nemeth et al., 2014). There is a relatively high level of low molecular weight soluble dietary fiber in wheat. It includes fructan, which was found in a double mutant sweet wheat (SW) line; however, seeds were severely shrunken and shriveled, and had reduced kernel weight (Shimbata et al., 2011). Nevertheless, the SW mutant can be utilized in breeding programs as a novel source to raise grain fructan levels.

Among cereal crops, genetic mapping studies have been mainly performed in wheat for concentrations of grain prebiotics such as fructan, inulin, and arabinoxylan (Huynh et al., 2008b; Falcon, 2011; Nguyen et al., 2011). A total of five, four, and two QTLs explaining 2–27, 3–19, and 15–20% of phenotypic variation were detected in wheat for grain fructan, inulin, and arabinoxylan concentrations, respectively. Some epistatic QTLs were additionally detected for grain fructan and arabinoxylan concentration, although, their contributions were limited (Huynh et al., 2008b; Nguyen et al., 2011). Despite this, two QTLs each for fructan (6D and 7A), inulin (2BL.2 and 5BS), and arabinoxylan content (2A.1 and 4D.1) were major QTLs (PVE > 10%), suggesting molecular breeding to improve prebiotics significantly in grains of wheat. Recently, Huynh et al. (2012) mapped the fructan biosynthetic pathway gene coding for the enzyme sucrose:sucrose-1-fructosyltransferase (1-SST), which corresponds to the position of a major QTL on wheat chromosome 7A that affects the accumulation of grain fructan (Huynh et al., 2008b). Thus, identification of candidate genes underlying these QTLs would provide a basis for functional analysis and for the development of DNA markers that may assist molecular breeding with the aim of increasing prebiotic concentrations in the grain.

### BREEDING FOR HARMONY BETWEEN QUALITY AND SAFETY OF CEREAL GRAIN

Besides food quality, food safety is also a "hot" topic that encourages scientists to engage in research related to health risks after consuming non-essential metals such as Cd and lead (Pb), and/or metalloids (arsenic, As), which have no beneficial role in plants, animals, or humans (Khan et al., 2015). Among these non-essential heavy metals, Cd particularly is known as highly phytotoxic, having a very low toxicity threshold level, and as a carcinogen, which is a great threat to human health. Nearly 27% of dietary Cd exposure is contributed by grain or grain products (Guttieri et al., 2015). Similarly, arsenic is also carcinogenic and can pose a serious threat to human health even at low concentrations. Moreover, the presence of high concentrations of these non-essential elements in cereal straw is still menacing because cereal straw is mainly used as livestock feed and thus these toxic elements may enter into the human food chain via contaminated meat or milk.

Soil is a natural source of heavy metals, and their elevated concentration in soil can occur either naturally or through anthropogenic activities such as urban and industrial activities as well as from agricultural practices. These toxic metals contamination is a non-reversible accumulation process due to their long estimated half-life in soil. Thus, accumulation of toxic metals in cereal grains impacts significantly on nutritional quality and crop safety. Generally, metals commonly enter plants as divalent cations. It has been reported that increasing accumulation of Fe and Zn in seeds leads to a higher accumulation of Cd, which chemically resembles Fe and Zn. Thus, uptake of Cd in roots and then translocation to seeds appears to occur inside plants along nutrient translocation pathways (Krämer, 2009). The first overlapping QTLs for essential and non-essential metals were identified in the Zn/Cd hyperaccumulator Arabidopsis halleri, and the candidate gene underlying the major QTL was identified as AbHMA4 (Heavy Metal ATPase 4) (Hanikenne et al., 2008). Subsequently, HMA2 was determined to contribute to Cd and Zn translocation in rice (Clemens et al., 2013).

The concentrations of essential mineral nutrients and nonessential metals in grains appear to be independently regulated because some independent grain Cd accumulation loci have been reported in cereals, such as the Cdu1 locus on 5BL in durum wheat (Knox et al., 2009) and one major QTL on 5AL in bread wheat (Guttieri et al., 2015). The identification of causal genes underlying these QTLs will provide more biological insights into Cd accumulation in cereal grains. Similarly, rice genotypes having dysfunctional OsNRAMP5 (Ishikawa et al., 2012) showed a substantial decrease in Cd uptake by roots, as well as Cd content in the straw and grain, but without decreasing the uptake of Fe by the roots, shoots, and straw (Ishimaru et al., 2012; Sasaki et al., 2012). These results suggest that a low grain Cd cereal cultivar can be developed without reducing the concentration of essential mineral nutrients through marker-aided breeding (MAB). Recent research has emphasized the importance of wild relatives for breeding high grain Fe and Zn in cereals crops. Nonetheless, possible pleiotropic effects of the introgression of elevated mineral nutrients need to be investigated by ICP–MS, thereby facilitating joint selection.

### OUTLOOK

Globally, the committed efforts by CGIAR HarvestPlus have led to the integration of essential micronutrients as a core activity in the breeding programs of almost all major cereal crops. Considering the complex genetic mechanism of Fe and Zn accumulation in cereal grains, eradication of these mineral nutrient deficiencies by increasing their levels in cereal grains through conventional breeding is simply too difficult. In the post-genomic and computational systems biology era, the combination of high-throughput genomics and robust statistical analysis, particularly QTL mapping studies, has helped to dissect the molecular basis of natural diversity for complex quantitative traits in a better way. Recent molecular mapping studies clearly indicate the co-localization of QTLs for Fe and Zn with those for other potentially toxic metals such as Cd, Pb, and As. Available knowledge can be used to design targeted crosses for MAB targeting cereal cultivars with high levels of Fe and Zn.

Undoubtedly, QTLs detected only for Fe or Zn have also revealed that plants may be able to differentiate between

### REFERENCES


nutrients and chemically similar toxin ions. Although no information is available so far about the enhancement of toxic metals in cereal grains through cross breeding, there is fear of inadvertent breeding for these non-essential metals that are toxic to both plants and animals even in low concentrations. Moreover, modifications in the accumulation of these toxic elements that are of concern for food safety are rarely determined during research on mineral nutrients. Thus, utilization of natural genetic variation for these mineral nutrients through a molecular breeding approach seems to be more attractive in the future. Furthermore, existence of substantial genetic variability for Fe and Zn bioavailability inhibitors and promoters also offers good opportunities to increase the bioavailable forms of these mineral nutrients in cereal grains. Genes accounting for this variability have rarely, however, been found and, therefore, are not yet being used in breeding; however, this also seems to be a promising approach for the near future. Hence, Fe and Zn bioavailability from cereal grains may be improved through breeding by accumulating either anti-nutrient agents or prebiotics. Furthermore, both functional and genetic evidence along with genome sequencing will provide means for gaining more insights regarding the emerging biofortification genomics.

### AUTHOR CONTRIBUTIONS

AG-O conducted the literature survey and together with SC wrote the first draft. RO edited and together with AG-O, SC, MG, and AM improved the manuscript writing. All authors read and approved the final manuscript.


enhancing iron and zinc content in wheat. Genet. Resour. Crop Evol. 56, 53–64. doi: 10.1007/s10722-008-9344-8


zinc concentration in diploid A genome wheat. J. Hered. 100, 771–776. doi: 10.1093/jhered/esp030


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Garcia-Oliveira, Chander, Ortiz, Menkir and Gedil. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Applications of New Breeding Technologies for Potato Improvement

Amir Hameed1†, Syed Shan-e-Ali Zaidi <sup>2</sup> \* † , Sara Shakir 2† and Shahid Mansoor <sup>2</sup> \*

<sup>1</sup> Department of Bioinformatics and Biotechnology, Government College University, Faisalabad, Pakistan, <sup>2</sup> Agricultural Biotechnology Division, National Institute for Biotechnology and Genetic Engineering, Faisalabad, Pakistan

### Edited by:

Felipe Klein Ricachenevsky, Universidade Federal de Santa Maria, Brazil

### Reviewed by:

Sunette M. Laurie, Agricultural Research Council of South Africa (ARC-SA), South Africa Felipe Dos Santos Maraschin, Universidade Federal do Rio Grande do Sul (UFRGS), Brazil

#### \*Correspondence:

Syed Shan-e-Ali Zaidi shan.e.ali@outlook.com Shahid Mansoor shahidmansoor7@gmail.com

### †Present Address:

Amir Hameed, Akhuwat-Faisalabad Institute of Research, Science and Technology, Faisalabad, Pakistan Syed Shan-e-Ali Zaidi, Gembloux Agro-Bio Tech, University of Liège, Gembloux, Belgium Sara Shakir, Boyce Thompson Institute, Ithaca, NY, United States

### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 01 December 2017 Accepted: 11 June 2018 Published: 29 June 2018

### Citation:

Hameed A, Zaidi SS, Shakir S and Mansoor S (2018) Applications of New Breeding Technologies for Potato Improvement. Front. Plant Sci. 9:925. doi: 10.3389/fpls.2018.00925 The first decade of genetic engineering primarily focused on quantitative crop improvement. With the advances in technology, the focus of agricultural biotechnology has shifted toward both quantitative and qualitative crop improvement, to deal with the challenges of food security and nutrition. Potato (Solanum tuberosum L.) is a solanaceous food crop having potential to feed the populating world. It can provide more carbohydrates, proteins, minerals, and vitamins per unit area of land as compared to other potential food crops, and is the major staple food in many developing countries. These aspects have driven the scientific attention to engineer potato for nutrition improvement, keeping the yield unaffected. Several studies have shown the improved nutritional value of potato tubers, for example by enhancing Amaranth Albumin-1 seed protein content, vitamin C content, β-carotene level, triacylglycerol, tuber methionine content, and amylose content, etc. Removal of anti-nutritional compounds like steroidal glycoalkaloids, acrylamide and food toxins is another research priority for scientists and breeders to improve potato tuber quality. Trait improvement using genetic engineering mostly involved the generation of transgenic products. The commercialization of these engineered products has been a challenge due to consumer preference and regulatory/ethical restrictions. In this context, new breeding technolgies like TALEN (transcription activator-like effector nucleases) and CRISPR/Cas9 (clustered regularly interspaced palindromic repeats/CRISPR-associated 9) have been employed to generate transgene-free products in a more precise, prompt and effective way. Moreover, the availability of potato genome sequence and efficient potato transformation systems have remarkably facilitated potato genetic engineering. Here we summarize the potato trait improvement and potential application of new breeding technologies (NBTs) to genetically improve the overall agronomic profile of potato.

Keywords: CRISPR, genome editing, nutritional quality, potato, TALEN

## INTRODUCTION

The rising food demand in a populating world will require a proportional increase in the food source. In contrary, several factors like climatic change, industrialization, and urbanization have overburdened the existing agriculture lands and food resources (Badami and Ramankutty, 2015). Other factors causing food decline include various biotic and abiotic stresses continuously affecting

**102**

crops worldwide. With the technological advancements and joint public-private partnership, several crops with enhanced nutritional profile have been developed using the existing gene pool (Ricroch and Henard-Damave, 2016; Ma X. et al., 2017).

Potato (Solanum tuberosum L.), a solanaceous food crop, is ranked fourth among the major staple crops after maize, rice, and wheat. It can provide more carbohydrates, proteins, minerals, and vitamins per unit area of land and time as compared to other potential food crops (Zaheer and Akhtar, 2016). In addition to being a raw marketable product, potato is largely used in industry for making processed food products, alcohol, starch, animal feed and for biofuel production (Scott and Suarez, 2012; Liang and McDonald, 2014). Short crop duration and wide climatic adaptability have facilitated potato to spread across diverse geographical borders from its South American origin. Today, more than three thousand potato cultivars are widely distributed in more than 125 countries, particularly under temperate, subtropical and tropical regions covering a major economic share in the global agricultural market (Birch et al., 2012). For the last two decades, potato cultivation and utilization have also been notably increased in developing countries such as Bangladesh, India, and China (Zaheer and Akhtar, 2016).

In terms of nutrition, potato is a complex source of nutrients (vitamins, carotenoids, anti-oxidant phenolics, proteins, magnesium etc.), and some anti-nutrients (primarily glycoalkaloids). On average, potato tubers contain 77% water, 20% carbohydrates, and less than 3% of proteins, dietary fiber, minerals, vitamins and other compounds (Zaheer and Akhtar, 2016). Comprehensive information regarding the tuber composition of different potato cultivars is described by Burlingame et al. (2009). In low-income food-deficit countries (http://www.fao.org/countryprofiles/lifdc/en/), potato could replace other high-priced foods and can be sustainably used as a cheap food giving enough calories (93 kcal/100 g tuber) to sustain a normal life (Burlingame et al., 2009). The global importance of potato is unquestionable and to commercialize its role in defeating food-shortage, poverty, and predominantly malnutrition, United Nations celebrated 2008 as the "International Year of the Potato" (http://www.fao.org/ potato-2008/en/).

Several breeding and molecular approaches have been employed for trait improvement in potato. Conventional breeding techniques for potato improvement are directed to increase yield, processing, and storage-quality (Halterman et al., 2016). Potato breeders incorporated resistance against early and late blight disease by crossing hybrid lines with wild species (S. brevidens and S. bulbocastanum) which inherited resistance against fungal pathogens (Naess et al., 2000; Tek et al., 2004). Although conventional breeding has been successfully employed for targeted trait improvement with less intraspecific variability, the progress is relatively slow and limited due to the phenotypic characterization of leading individuals in successive generations. In addition, the search of useful genetic variability in wild relatives could be laborious and its introgression in cultivated variety can be another challenging task. High heterozygosity and tetraploid nature of the potato genome (Consortium, 2011) are major drawbacks in breeding efforts to improve potato because of allelic suppression at each breeding cross (Lindhout et al., 2011). Other factors may include intra-species incompatibilities and inbreeding depression that causes failure in trait incorporations in polyploid crops through conventional breeding.

In this context, new breeding technologies (NBTs) offer a leading hand for trait improvement in crop plants and provide a platform for precise and robust plant genome editing. These NBTs include, but are not limited to, the cutting-edge genome editing approaches like clustered regularly interspaced short palindromic repeats/CRISPR associated 9 (CRISPR/Cas9), transcription activator-like effector nucleases (TALENs) and zinc-finger nucleases (ZFNs) (Jinek et al., 2012; Schaart et al., 2016; Weeks et al., 2016). Although developed recently, CRISPR system has been effectively employed for trait improvement of several economically important crops like wheat, maize, rice, cassava, cotton, soybean, and potato (Puchta, 2017). The introduced traits include herbicide tolerance, fungal/bacterial/viral disease resistance, drought tolerance, and increased shelf life, leading to overall improved quality and production. The working methodology and the anticipated role of these NBTs in plant genetic engineering have been extensively reviewed (Bortesi and Fischer, 2015; Mahfouz et al., 2016; Schiml and Puchta, 2016; Puchta, 2017; Weeks, 2017; Zaidi et al., 2017a,b, 2018). The current review provides a comprehensive information on different genetic approaches, including NBTs, that have been successfully employed to enhance the nutritional value of potato (**Tables 1**, **2**). Moreover, we summarize the data on transgenic potato commercialized so far (**Table 3**) and the major concerns associated with their regulatory approvals.

### CONSTRAINS TO POTATO PRODUCTIVITY AND QUALITY

The sustainable potato production faces a number of challenges due to biotic stresses (viruses, bacteria, fungal, insect pests) and abiotic stresses (drought, salinity, temperature, frost and postharvest problems, i.e., accumulation of reducing sugars during cold storage).

### Diseases and Insect Pests Affecting Potato

Most of the potato diseases are due to the diverse prevalence of phytopathogens of which viruses are of prime importance. Cultivated potato is susceptible to around 40 different viral and 2 viroid species (Salazar, 1996). Among the dominating viruses, Potato virus Y (PVY, genus; potyvirus), Potato leafroll virus (PLRV, genus; polerovirus), and Potato virus X (PVX, genus; potexvirus) are probably the most diverse and devastating viruses infecting potato worldwide (Fletcher, 2012; Hameed et al., 2014; Steinger et al., 2014). Viral diseases appear as necrotic strains on leaves/tubers, mosaic, and overall stunted growth to plant, leading to reduced yield and poor-quality tubers. Moreover, several bacterial diseases (soft rot/blackleg caused by Dickeya solani, common scab caused by Streptomyces scabies) (Buttimer et al., 2017), and fungal diseases (late blight caused by Phytophthora infestans, powdery scab caused by Spongospora subterranea; Arora et al., 2014; Balendres et al., 2016) are



Transgenesis: Introducing an exogenous gene "transgene" into a living organism so that the organism will stably exhibit a new property and transmit that property to next generation.

also severely deteriorating potato quality worldwide. Late blight affected potato plants exhibit water-soaked leaves having necrotic lesions and irregular colored tissue in tubers making them hard, dry and more susceptible to other microbial diseases.

Virus resistance in potato has been engineered through different approaches ranging from simple plant breeding to advanced genetic engineering. Transgenic approaches to engineer virus resistance in potato are seemed to be more appropriate than conventional breeding due to its polyploid nature making difficulties for the introgression of resistance genes. Thus, RNA interference (RNAi)- mediated resistance targeting viral coat protein (CP) region has been demonstrated in potato, where single or multiple RNA viruses have been targeted with different success levels; such as PVY- resistance (Missiou et al., 2004); PVY, and PLRV-resistance (Chung et al., 2013); and PVX, PVY, and Potato virus S (PVS)-resistance (Hameed et al., 2017). The current scenario of GM potatoes being commercialized in some countries encompasses viral resistant potatoes generated through genetic engineering (Mathur et al., 2017).

Potato pests cause direct damage to potato crop in the forms of necrosis, deformations of plant tissues and/or indirect damage by facilitating the pathogen dispersal, especially for viruses. Important destructive insects affecting potato include Colorado potato beetle (Leptinotarsa decemlineata) (Casagrande, 2014), peach-potato aphid (Myzus persicae) (Bass et al., 2014), potato tuber moth (Phthorimaea operculella) (Liu et al., 2018) etc. The extensive use of chemical insecticides on insect pests has led to the evolution of insecticide-resistance in particular insects, thus posing alarming threats. To effectively control their incidence in planta, genetic engineering has offered some promising solutions like introgression of insecticidal proteins/toxins (Palma et al., 2014), RNAi-mediated insect resistance (Zhang J. et al., 2017) and CRISPR-Cas9-mediated crop protection (Douglas, 2017) etc. In


TABLE 2 | Applications of some new breeding technologies for potato trait improvement.

potato, RNAi technology was used to engineer resistance against Colorado potato beetle (Zhang et al., 2015). Sap transmitted RNAi reagents (long double-stranded (ds) RNAs in chloroplasts) resulted in more than 80% of the reduced expression of insect targeted genes (β-actin gene) and triggered a lethal RNAi response destructive to its larvae (Zhang et al., 2015).

### Agronomic Attributes Affecting Potato

Farming systems comprising of different agronomic attributes like tillage, nutrient management, and crop rotation significantly affect potato tuber productivity and quality. Due to its shallow root-system, potato needs a fair supply of nutrient inputs to maintain its tuber vigor and yield (Alva et al., 2011). Research has shown the influence of different farming practices on tuber quality parameters like tuber dry mass accumulation, enhanced nutrient/mineral concentration and yield improvements (Brazinskiene et al., 2014; Tein et al., 2014; Nyiraneza et al., 2015). Through adopting a potato-legume crop rotation, (Qin et al., 2017) observed a positive influence on soil microbiota coupled with significantly improved tuber yield up to 19% when compared with the continuous cultivation of potato crop only. Integrated crop rotations with an exogenous supply of organic and mineral [nitrogen (N), phosphorous (P), and potassium (K)] fertilizers significantly influenced potato tuber N, nitrate, magnesium (Mg) and P concentrations when compared with non-fertilized controls (Tein et al., 2014). Leonel et al. (2017) analyzed five potato cultivars for their tuber chemical composition in response to different concentrations of available P supplemented with uniform cultural practices. Potato tubers fertilized with increased P exhibited a significant positive influence of tuber dry matter and protein/starch contents and a lower concentration of total sugar contents (Leonel et al., 2017). The chemical composition of potato tubers is a prerequisite for determining the nutritional and processing quality of industrial perspectives. Understanding the importance of organic products, Lombardo et al. (2017) evaluated the nutritional value of organic vs. conventionally grown potatoes. Field trials of yellow-fleshed potato cultivars growing under organic cultivation produced high-quality tubers having enhanced concentrations of phenolics, reduced nitrate and a more attractive tuber flesh color (Leonel et al., 2017).

### Climatic and Soil Factors Affecting Potator

A number of abiotic stresses ranging from soil to climate significantly affect the potato productivity and quality during its growth and/or after harvest. Potato cultivation performs better under cool condition (19◦C) and is vulnerable to high temperatures (Kim et al., 2017). A fairly low temperature promotes the first tuber set and sudden elevations in temperature during this early tuberization significantly affect tuber yield and size (Zhou et al., 2017). A drastic reduce in potato tuber yield ranging from 3 to 11% per 1◦C rise in temperature was observed across various geographical locations (Fleisher et al., 2017; Kim et al., 2017). During the early growth stage of potato, low freezing and/or frost attacks severely damage the young plantlets and could reduce the tuber yield and quality (Chang et al., 2014).

Drought and salinity are other important abiotic stresses having adverse effects on potato cultivation. Potato cultivars respond differentially to drought conditions and mostly exhibit various physiological and morphological changes in tuberization



\*Unknown. R&D, Research and Development; EPA, U.S. Environmental Protection Agency; FDA, Food and Drug Administration; EC, European Commission; EU, Europe; CPB, Colorado Potato Beetle; PLRV, Potato leafroll virus; Cry3A and Cry2a1, Bacillus thuringiensis genes; VInv, Vacuolar acid invertase; Asn1, Asparagine synthetase-1 gene; Rpi-vnt1, Late Blight resistance gene from wild potato (Solanum venturii); AmAI, Amaranthus hypochondriacus1; GDP, Arabidopsis thaliana <sup>L</sup>-galactose phosphorylase; GBBS, Granule bound starch synthase; BBSRC, Biotechnology and Biological Sciences Research Council; WUR, Wageningen University and Research Centre; NIPGR, National Institute of Plant Genome Research.

and plant growth (Chang et al., 2018). The major impact of this water stress has been recorded in the form of reduced tuber yield due to a loss of internal water pressure during tuber bulking and maturation (Stark et al., 2013). Salinity, another acute abiotic stress causes many inhibitory effects on plant growth and development (Parihar et al., 2015). Salinity stress in potato severally affects its productivity by causing enhanced oxidative stress, reduced photosynthesis and significantly reduced tuber yield. Research efforts through interspecific breeding in potato resulted in improved tolerance to salinity and oxidative stresses (Jbir-Koubaa et al., 2015).

### Post-harvest Factors Affecting Potato

The post-harvest storage of potato tubers is another criterion for determining their end-product processing quality. Usually, this long-term storage is accompanied with various storage diseases like soft rot, black dot, and Fusarium dry rot which significantly reduces the tuber quality and unfit it for further processing (Usall et al., 2016). Soil-transmitted black dot, caused by Colletotrichum coccodes imparts brown necrotic lesions/stains on tuber skin and promotes their rapid decay (Brierley et al., 2015). Storage temperature and durations are two important factors determining the tuber susceptibility to black dot disease and could be managed to prevent the quality losses in potato (Peters et al., 2016). The tuber harvesting date also influences tuber quality during the long-term storage. Makani et al. (2015) observed the storage quality of potato cultivars in response to harvest time and subsequent storage. The results showed that the tubers harvested at full maturity retained their quality during storage in contrary to early harvested tubers (less mature), which exhibited a significant loss in tuber dry matter and ascorbic acid contents (Makani et al., 2015). Potato dormancy characteristics are other challenging factor determining the tuber quality during storage. Dormancy is an innate ability to sustain sprouting for a time and after its natural breakage, sprouting starts which cause various quality issues. Dormancy could be regulated in potato tubers through the topical applications of phytohormones, such as ethylene which has the ability to suppress bud formation/sprouting (Sonnewald and Sonnewald, 2014).

Potatoes are usually subjected to cold-storage (4–8◦C) in order to ensure a continuous supply to consumers/markets throughout the year. This cold-storage is accompanied by elevated levels of reducing sugars in the tuber, a phenomenon termed as "Cold-Induced Sweetening: CIS" (Bhaskar et al., 2010). During CIS, tuber starch content is biochemically converted to sugars (sucrose) through the cohesive activity of several hydrolytic enzymes (Sowokinos, 2001). The elevated sucrose is subsequently transported inside a vacuole where it is further reduced to glucose and fructose through the activity of a host gene (vacuolar acid invertase, VInv) (Sowokinos, 2001; Bhaskar et al., 2010). The CIS affected tubers when used as feedstock for high-temperature processing gives rise to the accumulation of a dark brown, bitter tasting product, i.e., acrylamide. The rising acrylamide contents in food products is a huge concern to global food safety as well as to end-chain consumers (Vinci et al., 2012). Several reports depict the alarming levels (up to 70%) of acrylamide in food products that mainly come through the intake of fries, chips and other fried potato products (Pedreschi et al., 2014; McCombie et al., 2016; Esposito et al., 2017).

### ENHANCING NUTRIENT CONTENTS IN POTATO

For the last two decades, several efforts have been conducted to improve the nutritional traits of potato. The following section describes the information regarding nutrient enhancement in potato and is summarized in **Table 1**.

### Increased Protein Content

The risk of protein deficiency is more in the countries where people take protein-deficit diet as a staple food (Chakraborty et al., 2010). Unfortunately, the cultivated potato contains fewer proteins (0.85–4.2%) lacking lysine, tyrosine, and some other essential amino acids (Burlingame et al., 2009). To deal with this limitation, scientists have engineered potato with enhanced protein content through constitutive expression of tuber-specific gene, Amaranthus hypochondriacus1 (AmA1) (Chakraborty et al., 2000, 2010). The AmA1 gene encodes for a seed protein, albumin: a non-allergic protein containing essential amino acids and considered safe for human/animal consumption [safety accredited by the World Health Organization (WHO)]. In transgenic potato, the enhanced protein (albumin) localizes inside cytoplasm/vacuole. The tubers of seven engineered potato cultivars showed an increased protein content up to 60% as compared to controls (Chakraborty et al., 2010). In addition to increased protein content, the transgenic potato also showed an accelerated rate of photosynthesis that ultimately increased the total biomass/yield of plants. Recently, methionine content (an essential amino acid involved in multiple cellular pathways) was significantly increased in transgenic potato cultivar (cv.) Desirée (Kumar and Jander, 2017). By using RNAi technology, overexpression of an exogenous gene Arabidopsis thaliana cystathionine γ-synthase (AtCGS), along with the suppression of a host gene S. tuberosum methionine γ-lyase (StMGL), resulted in nearly a double concentration of free methionine inside transgenic tubers as compared to control tubers (Kumar and Jander, 2017). Moreover, the experimental studies of engineered plants showed no morphological and yield differences when compared with control plants. Other studies were also conducted to increase the protein content in potato but met with limited success and yield penalties (Zeh et al., 2001; Dancs et al., 2008; Rinder et al., 2008; Galili and Amir, 2013).

### Increased Vitamin and Carotenoid Contents

Several studies have been made to increase the vitamin content in potato, for example, expressing an exogenous gene, A. thaliana <sup>L</sup>-galactose phosphorylase (GDP) showed a 3-fold increase in ascorbate contents (vitamin C) (Bulley et al., 2012). Carotenoids are phytonutritive, anti-oxidative, lipophilic compounds (precursors to vitamin) present in many fruits and vegetables (Dellapenna and Pogson, 2006), and provide nutritional benefits in terms of increased vitamin uptake. Introduction of cauliflower Orange (Or) gene has shown a net increase in carotenoid content (pro-vitamin A) in coldstored tubers (Li et al., 2012). Among other carotenoids (Lutein, zeaxanthin, violaxanthin, neoxanthin), β-carotene concentration is considerably low in potato (Ezekiel et al., 2013). RNAi approach was utilized to silence the β-carotene hydroxylase (bch) gene that showed a significant increase in β-carotene and lutein contents in the tubers (Van Eck et al., 2007). Another study reported a 20-fold increase in tuber carotenoid contents through expressing three bacterial genes involved in carotenoid biosynthesis (Diretto et al., 2007). Similarly, transgenic potato cv. Taedong Valley was produced, over-expressing GLOase gene (L-gulono-γ-lactone oxidase from rat cells) that showed an enhanced (141%) content of <sup>L</sup>-Ascorbic acid (vitamin C) (Upadhyaya et al., 2010).

### Increased Calcium Content

Being nutritious with several other elements, the cultivated potato is a poor source of Ca (Weaver et al., 1999). To address this deficiency, Park et al. (2005) utilized a transgenic approach through expressing an exogenous gene, Arabidopsis H+/Ca2<sup>+</sup> transporter (sCAX1) in potato cv. Russet Norkotah. The regenerated plants expressing sCAX1 gene showed a significant increase (up to 3-fold) of Ca contents in tuber as compared to controls. Field trials and morphological data from three consecutive crop generations proved the stable integration of enhanced Ca trait with no alteration in tuber yield and other growth/morphological characters. Potato with enhanced Ca contents could be potentially used as a dietary source, more specifically in countries where potato is a staple food (Park et al., 2005).

### Increased Phenolic Contents

In potato, 80% of the phenolic compounds are present in the form of caffeoyl quinic acids (CQAs) (Brown, 2005). Recently, Li et al. (2016) conducted a study to increase the CQAs content in potato tubers. Tuber specific constitutive expression of an exogenous gene, flavonol-specific transcriptional activator (AtMYB12: derived from A. thaliana) showed a significant increase (>3-folds) of CQAs and total flavonoid content. Importantly, they utilized a selectable marker-free approach to facilitate the downstream regulatory approvals (Daniell, 2002). The increased phenolic contents being imposing health benefits also induce some antimicrobial properties to plants, particularly with reduced fungal infections (Li et al., 2016).

### Increased Starch Contents

Potato tubers are a rich source of dietary starch and can provide a significant calorie intake in food-deficit countries. Starch is primarily composed of two structural components, amylose, and amylopectin, which are biosynthesized through cohesive actions of several enzymes. Extensive studies have been conducted to improve the digestible amylose content in potato by engineering different steps of starch biosynthesis pathway (Schwall et al., 2000; Hofvander et al., 2004). Tuber specific RNA silencing of two host genes (SBEI, SBEII), involved in starch branching pathway, resulted in the generation of potato with enhanced amylose content (Andersson et al., 2006). Recently, amylose contents were significantly increased (28–59%) in non-genetically modified potatoes by introducing a recessive allele (gene marker: IAm) from wild potato (S. sandemanii) into cultivated potato (S. tuberosum) through marker-assisted crossing (Krunic et al., 2018).

Plant-based oils are promising for the near future as a potential feedstock for a renewable energy. Currently, biofuel research is more focused on engineering crops with enhanced oil contents through genetic manipulation in lipids/triacylglycerol (TAG) synthesis pathways (Vigeolas et al., 2007; Vanhercke et al., 2014; Zale et al., 2016). Contrary to high starch content, oil (lipids) concentration is very low in potato tubers. Some recent studies have shown the enhanced TAG content in the engineered potato tubers (Hofvander et al., 2016; Liu et al., 2017). Tissuespecific constitutive expression of three genes (WRI1, DGAT1, and OLEOSIN) resulted in a 100-fold increase in TAG content in tuber as compared to controls (Liu et al., 2017). However, this TAG increase was also accompanied by a depletive nutritional effect in terms of significantly reduced starch (amylose) and accumulated sugar (sucrose) levels. Further exploration of this mechanism revealed a better understanding of negative impacts of TAG accumulation on tuber amylose and phosphate contents, as well as needs to optimize genetic engineering for particular traits (Mitchell et al., 2017).

### REDUCTION OF ANTI-NUTRIENT CONTENTS IN POTATO

Another strategy to improve the nutritional quality of food is by reducing the anti-nutrient elements. There is no lethal toxicity reported with the consumption of potato as food (Zaheer and Akhtar, 2016), however, some anti-nutrient elements like steroidal glycoalkaloids (SGAs) (0.071–175 mg/100 g), primarily α-solanine, and α-choconine accumulate in tubers during crop maturation (Burlingame et al., 2009). These SGAs when present in higher amounts in food may cause neuro-toxic and/or nutrient absorption problems (Itkin et al., 2013). Biosynthesis of SGAs involves a concurrent expression of two key enzymes, uridine 5' diphosphate, and glycosyltransferase which biochemically react with cholesterol, sugars, and other nitrogenous compounds to build up the glycoalkaloid molecules (Itkin et al., 2011). RNAimediated silencing of the host gene, Glycoalkaloid metabolism 4 (GAME4) in potato showed a significant decrease (up to 74 fold) in SGAs content in leaves and tubers (Itkin et al., 2013). Importantly, many wild species of potato produce high levels of SGAs naturally (Gregory et al., 1981), therefore breeders must be careful to map the SGAs-gene (s) linkage with the desired traits when using the wild germplasm as a genetic resource.

Usually, raw potato is processed (fried, baked, mashed, microwaved) into various food products like snaps, fries, chips, etc. prior to eating (Zaheer and Akhtar, 2016). Sometimes, potato processing results in hazardous compounds causing obesity, cardiovascular diseases, and/or neurotoxicity. The first report of acrylamide, a potential neurotoxin and carcinogenic element, presence in potato fried products raised a debate among food regulatory authorities and processing industry (Tareke et al., 2002). Since then, several studies were conducted to explore the acrylamide formation during "Maillard Reaction," a reaction among tuber asparagine contents, reducing sugars (primarily glucose and fructose) and free α-radicals present in cooking oil during high-temperature processing of potato (Stadler et al., 2002; Friedman, 2003; Vinci et al., 2012).

Contrary to cultivated tetraploid potato (S. tuberosum), some diploid species of wild potato are naturally resistant to CIS. Through quantitative trait loci (QTL) mapping and other molecular studies, scientists have identified some recessive genes associated with CIS resistance in wild species. Potato breeders tried to incorporate these genes in cultivated potato in order to reduce the CIS effect but met with limited success. Therefore, different molecular strategies have been applied to reduce the formation of acrylamide through indirectly mediating the CIS mechanism at the cellular level (Bhaskar et al., 2010; Li et al., 2013; Zhu et al., 2016; Hameed et al., 2018). Tuberspecific constitutive expression of VInv gene in anti-sense binary constructs resulted in significant reduction of reducing sugar content in cold-stored tubers. High-temperature processing of food products derived from these transgenic lines showed an 8-fold decrease in acrylamide content as compared to controls (Ye et al., 2010). In another study, RNAi-mediated simultaneous silencing of potato asparagine synthetase genes (StAS1 and StAS2) and VInv gene significantly reduced the CIS process as well as asparagine content in transgenic potato cv. Russet Burbank (Zhu et al., 2016). Tubers derived from these CIS resistant transgenic lines showed a significant reduction (15 fold) of acrylamide content in fried potato products. The firstgeneration biotech potato (Simplot's Innate TM) was engineered to have lower reducing sugars levels and reduced asparagine contents to address the acrylamide forming problems during potato processing (Halterman et al., 2016).

### NEW BREEDING TECHNOLOGIES USED FOR INCREASING NUTRITIONAL QUALITY OF POTATO

Gene pyramiding in polyploid crops using conventional breeding is a difficult, laborious, and time-consuming (Weeks, 2017). In potato, several breeding efforts have been made for particular trait improvement using wild species germplasm but met with limited success (Carputo and Barone, 2005). The presence of four copies (alleles) of genes in the tetraploid (2n = 4x = 48) genome of cultivated potato (S. tuberosum) makes it difficult for researchers/breeders to precisely edit the genome using conventional breeding tools (Consortium, 2011). Thus, NBTs such as CRISPR/Cas9, TALENs, and ZFNs offer great potential for expediting genome editing in a more precise and time-saving way (Mahfouz et al., 2016; Petolino et al., 2016; Schiml and Puchta, 2016).

In case of potato, both TALENs (Sawai et al., 2014; Nicolia et al., 2015; Clasen et al., 2016), and CRISPR/Cas9 (Butler et al., 2015; Wang et al., 2015) technologies have been efficiently utilized for precise genome editing (**Table 2**). In 2014, the first attempt of utilizing TALENs technology in potato genome editing paved the way for the next technology shifts. Sawai et al. (2014) utilized TALENs approach in potato (S. tuberosum cv. Sassy) to silence a host gene, Sterol side chain reductase 2 (StSSR2) that is predominantly involved in cholesterol biosynthesis and a precursor to many toxic SGAs formations. Transgenic expression of TALENs constructs generated a site-specific mutation of variable size (nucleotide deletion/insertion) in four alleles of StSSR2 gene. The transgenic potato with knock-out StSSR2 activity showed a significant reduction of SGAs contents without affecting plant growth, thus eliminating an anti-nutritional factor (Sawai et al., 2014). In another study, protoplast delivery of TALENs constructs resulted in a significant mutation frequency (7–8%) at targeted gene loci, i.e., Acetolactate synthase (ALS) in transgenic potato cv. Desirée (Nicolia et al., 2015). Sequencing analysis of ALS-mutated lines confirmed the targeted protein disruption either through amino acid substitutions, truncations, and/or frameshift mutations and importantly mutated lines showed no phenotypic differences compared to controls (Nicolia et al., 2015). Wang et al. (2015) reported the use of CRISPR/Cas9 system for inducing efficient targeted mutagenesis in potato. Agrobacterium-mediated transformation of cells with CRISPR constructs resulted in efficient site-specific mutation in host gene, Auxin/indole-3-acetic acid (StIAA2) engineered for altered Aux/IAA protein expression (Wang et al., 2015). In a later study, Butler et al. (2015) targeted Acetolactate synthase1 (ALS1) gene in potato through the Agrobacterium-mediated delivery of CRISPR/Cas reagents. Importantly, they utilized both tetraploid (S. tuberosum cv. Desirée) and diploid (MSX914-10-X914-10) potato as explant for their experiments. Stable expression of CRISPR/Cas reagents resulted in site-specific mutations (ranged from 3 to 60%) in ALS alleles and were stably heritable (87–100%) in successive diploid and tetraploid potato generations (Butler et al., 2015).

To improve the nutritional value of potato tubers, TALENs technology has been used to interrupt the VInv activity in order to reduce the accumulation of reducing sugars during cold-induced storage (CIS) (Clasen et al., 2016). Protoplastmediated transformation of potato cv. Ranger Russet with TALENs constructs resulted in knockout of VInv alleles in transformed plants (Clasen et al., 2016). Interestingly, 5 out of 18 transformed lines showed a nearly complete silencing of VInv gene having minimal or no detectable CIS activity. Furthermore, high-temperature processing (fried chips) of transgenic derived tubers resulted in light brown products having a significantly lowered level of dietary acrylamide. Importantly, in downstream characterization, few of transgenic lines showed a complete absence of TALENs sequences, thus offering a transgene-free approach.

Other efforts to incorporate nutritional traits in potato include starch alterations using TALENs (Kusano et al., 2016) and CRISPR/Cas9 (Andersson et al., 2017). Through designing a novel delivery system, termed "Emerald–Gateway TALEN system," Kusano et al. (2016) targeted a host gene, Granule-bound starch synthase (GBSS) in potato for site-specific mutation. GBSS is predominantly involved in amylose biosynthesis during starch granulation. Its disruption may reduce the amylose content and amylose/amylopectin ratio (Zeeman et al., 2010), and thus might affect starch quality in potato tubers. Agrobacterium-mediated transformation of potato cells with TALEN constructs resulted in three types of stable mutations in regenerated lines, dominantly having a deletion mutation (63 nucleotides deletion) (Kusano et al., 2016). Recently, CRISPR/Cas9 technology has been used to efficiently silence the GBSS in potato cv. Kuras (Andersson et al., 2017). Protoplast transformation with CRISPR/Cas9 constructs resulted in a site-specific mutation in all (four) alleles of GBSS gene in 2% of transformed lines. Full knock-out of targeted genomic sites resulted in complete loss of the GBSS activity and yielded an altered starch quality in transgenic tubers when compared to controls (Andersson et al., 2017). Genomewide analyses coupled with transcriptomics, proteomics and metabolomics will further dissect the molecular basis of starchrelated traits of potato and could facilitate/accelerate the starch modifications by using NBTs. The production of high-quality starch in potato may be of current research interest to meet the demands of food and industrial sectors.

In another study, Butler et al. (2016) utilized a geminivirus replicon (GVR) vector for delivering sequence-specific nucleases (SSNs) to target the potato herbicide tolerance gene ALS1 and regenerated transformants carrying a point mutation in ALS1 gene were confirmed for herbicide susceptibility. Forsyth et al. (2016) reported the targeted integration of transgene into a pre-selective, transcriptionally active site of potato genome using TALEN system coupled with a molecular marker, i.e., the mutated Acetolactate synthase (ALS) gene. Potato Ubi7 (constitutively expressing gene) was selected as a target for TALEN and after its functional confirmation in a transient system (N. benthamiana), Agrobacterium-mediated transformation was used to develop transgenic potatoes (S. tuberosum cv. Ranger Russet) (Forsyth et al., 2016). Importantly, the molecular confirmation of transgenic lines showed a single copy of transgene in most of the regenerated events. This could help in downstream transgenic characterization by reducing the workload of generating multiple independent lines for random transgene insertions. Their work established the efficacy of TALENs for achieving a more precise and site-specific genome editing in potato for trait incorporation. However, stable transformation of TALEN reagents carrying bacterial genes (TALE DNA binding domain from Xanthomonas) in plants may trigger the GM concerns having transgenes as codified by the regulatory authorities.

Recently, transient expression of TALENs, delivered through non-viral Agrobacterium-mediated transformation, yielded targeted mutations in two potato cultivars, Russet Burbank and Shepody (Ma J. et al., 2017). The infiltrated TALEN constructs were meant to induce mutations in two different host genes, i.e., (i) 1,4-alpha-glucan branching enzyme (SBE1), (ii) Vacuolar invertase 2 gene (StvacINV2). The regenerated lines were confirmed for targeted chromosomal mutation through deep sequencing (Illumina), that revealed three types of induced mutations having dominantly deletion mutations in both of cultivars. TALEN technology through agroinfiltrations could be effectively used to induce targeted mutation for improving some elite potato cultivars (Ma J. et al., 2017). **Figure 1** illustrates a schematic model of NBTs application for incorporating desired modifications in the potato genome to enhance nutritional improvements.

ZFNs are first-generation genome editing nucleases engineered to make DSBs through fusions of artificial, sequencespecific zinc finger proteins with the nonspecific DNA cleavage domains of the FokI restriction endonuclease (Kim and Kim, 2014). The applications of ZFNs for genetic engineering has been limited to crops like tobacco (Cai et al., 2009), Arabidopsis (Zhang et al., 2010), and soybean (Curtin et al., 2011) and not in potato and other horticultural crops (Gaur et al., 2018). The limited examples of ZFNs-mediated genome modification in plants are might be due to some disadvantageous such as low success rate (∼24%), low or variable mutation rate (∼10%), high off-target effects, and technically difficult in designing feasibility (Xiong et al., 2015; Zhang H. et al., 2017). These challenges have greatly narrowed the spectrum of ZFNs technology for adoption by the scientific community.

The studies discussed in this section provide comprehensive information regarding the utility of some NBTs for potato genome editing. Although, the applications of these tools are unlimited in the context of genetic engineering, the selection of suitable genomic targets and efficient editing tool is a critical prerequisite to get the desired goals. For example, transient expression of TALEN/CRISPR system could incorporate desired traits without stable integration of the transgene. Edited crops having non-detectable foreign DNA/RNA could face less opposition in regulatory/public clearance and could seem in line with their natural variants. NBTs could offer promising solutions to engineer complex genomic traits involving several molecular pathways like synthesis of starch, proteins, vitamins, etc., that otherwise would require tedious multistep engineering using conventional techniques. Other application may include functional studies of uncharacterized genes in potato using NBTs that could facilitate more precise and site-specific mutation at targeted loci. Furthermore, the induced mutations could be mapped by utilizing various next-generation sequencing (NGS) techniques. This could save the time for estimating off-site targeting effects that may or may not phenotypically appear later during crop growth. Conclusively, NBTs could be effectively used to engineer a number of nutritional traits in potato like enhanced protein content (Chakraborty et al., 2010), vitamin C content (Bulley et al., 2012), β-carotene level (Li et al., 2012), and others etc.

### GM POTATO COMMERCIALIZED SO FAR: RISK ASSESSMENTS AND REGULATIONS

The expansion of biotech crops over last two decades has firmly established the role of genetic engineering in modulating various agronomical, environmental and predominantly health-related traits in plants. Today, more than two billion hectares of agricultural land is under cultivation of biotech crops, which signifies its importance and adaptability to meet future challenges through generating useful phenotypes in plants (Parisi et al., 2016). Most of the GM potato cultivars commercialized so far include trait incorporations such as resistance to viruses and other phytopathogens (Ricroch and Henard-Damave, 2016). The first GM potato appeared in the market in 1995 was named "NewLeaf " by Monsanto <sup>R</sup> , which was genetically engineered using a toxin Bt gene to generate resistance against Colorado beetle (Leptinotarsa decemlineata) (Kilman, 2001). Another engineered potato variety appeared in March 2010; a GM potato "Amflora," developed by BASF Plant Science and aimed at improved amylopectin content (waxy tuberous starch) for the processing industry, was approved by the European Commission (Lucht, 2015; Zaheer and Akhtar, 2016). A total of 23 GM potato lines are in the regulatory approval process (10 at precommercial, 11 at regulatory, and 2 at advance development stages) that has been engineered for various agronomical and quality related traits (Parisi et al., 2016). In 2017, the U.S. Environmental Protection Agency (EPA) and the Food and Drug Administration (FDA) has approved the cultivation of three GM potatoes (InnateTM Second-Generation; developed by Simplot corp. <sup>R</sup> ) meant to resist fungal (late blight) infections, and acrylamide formation (https://durangoherald. com/articles/140336-u-s-approves-3-types-of-geneticallyengineered-potatoes?wallit\_nosession=1). Importantly, these GM potatoes were developed using various modern breeding tools and got regulatory approvals. A number of other potato varieties engineered for various agronomic traits are in the pipeline or subjected to biosafety and field trials in different

countries (Ricroch and Henard-Damave, 2016; **Table 3**). The development and commercialization of GM crops is a huge challenge for scientists and regulatory domains due to multiple technical, ethical and social/public limitations (Podevin et al., 2012). In addition, the socioeconomic benefit of utilizing these GM products is a big constraint to their producer and consumer adaptations. Still, the costs related to GM development/approval usually exceed \$35 million (Smyth et al., 2017), which seemed to be unsuitable for many low-income public-private institutions. This limits the interest of investing in GM technology in many developing countries (Pérez-Massot et al., 2013). The public acceptance in adopting GM-labeled products further makes it questionable for legislation, government and/or environmental authorities. With technological advancement, the NBTs could be used to generate a highly specific genetic modification that is indistinguishable

FIGURE 1 | A schematic diagram of new breeding technologies (NBTs) application for editing potato genome for nutritional improvement. (A) Clustered regularly interspaced short palindromic repeat/CRISPR associated9 (CRISPR/Cas9) system. Expression of constructs containing a single guide RNA (sgRNA) and Cas9 endonuclease will result in the assembly of sgRNAs and Cas9 nuclease to make a sgRNA/Cas9 complex. The designed sgRNA having sequence complementarity will bind specifically to a targeted site on genomic DNA and sgRNA/Cas9 complex will cleave 3' upstream of PAM (protospacer adjacent motif) sequence; shown by black scissors. This cleavage will result in double-stranded brakes (DSB) in targeted genome. (B) Transcription activator-like effector nucleases (TALENs) system. The TALE array contains a highly conserved (33–34 nt) DNA binding domain having repeat variable di-residues (RVDs) at positions 12 and 13 to guide the target-specific binding. Nuclease activity is performed by domains containing FokI endonucleases to produce DSBs. These DSBs are normally repaired by host-mediated DNA repair mechanisms which might results in targeted mutation and end in either gene disruption, correction or addition. The black circles having white text (1,2) represent the CRISPR/Cas9/TALENs cleavage of two host genes (vacuolar invertase, VInv; sterol side chain reductase, SSR2). (C) VInv is primarily involved in bioconversions of sucrose to fructose and glucose inside cell vacuole, precursors of acrylamide formation. (D) Biosynthesis of steroidal glycoalkaloids in plant cell from cycloartenol which is mediated by the activity of host SSR2 gene. NBTs-mediated targeting of host genes will result in reduced formation of anti-nutrients (acrylamide and steroidal glycoalkaloids) inside tubers and thus result in the improved quality of potato tubers. The proposed challenges (rectangles) by using these technologies might result in some questions such as society and regulation regimes' approval for editing food crop, off-site targeting effects on plants, the presence of any transgene, biosafety trails to check health-related issues, and the potential risks of horizontal gene transfer by using these GM crops. These questions need to be addressed while before using some NBTs.

from natural variants/mutants and therefore significantly reduce the GM concerns (Wolt et al., 2016). For example, Cellectis <sup>R</sup> (a multinational biotech company) developed a transgenefree potato in 2014 engineered for improved processing traits (Wolt et al., 2016). They used TALEN technology to introduce a base deletion in potato genome through protoplast transformation of an exogenous genetic material from some plant pest (Phytophthora infestans). Phenotypic and molecular confirmation of regenerated products showed no detection of any exogenous material in segregating generation, thus, being

transgene-free, APHIS (a USDA regulatory domain under the Plant Protection Act: CFR 7) did not take it under regulatory process (Wolt et al., 2016).

### CONCLUSION AND FUTURE PROSPECTS

Research focusing on food safety and security can provide substantial ways to meet up the rising food demands, especially in the food-deficit countries. The rapid development of plant genetic engineering has provided new exciting tools to generate crops with enhanced yield and nutritional traits. In this context, potato crop has enormous potential to contribute to food security as it could provide low-cost, high energy food at sustainable basis (Zaheer and Akhtar, 2016). Several studies have demonstrated the incorporation of nutritional traits in potato such as enhanced protein content (Chakraborty et al., 2010), vitamin C content (Bulley et al., 2012), β-carotene level (Li et al., 2012), triacylglycerol (Hofvander et al., 2016), tuber methionine (Kumar and Jander, 2017), and amylose content (Krunic et al., 2018; **Table 1**). Other research priorities are given to reduce anti-nutritional compounds in potatoes such as steroidal glycoalkaloids (Itkin et al., 2013), acrylamide (Clasen et al., 2016) and other food toxins (Hajeb et al., 2014; **Table 1**). Recently, the emergence of NBTs such as TALENs, ZFNs, CRISPR/Cas9 etc. has provided opportunities for a robust, precise, and site-specific genome editing to introduce important agronomical traits in various crop plants (Mahfouz et al., 2016; Weeks, 2017).

Within the context of potato genome editing, ongoing research is focused on utilizing NBTs to incorporate important traits (**Table 2**). However, most of these efforts generated end products having transgenic tags, being questioned by food safety, legislation, and extensive consumer opposition. To circumvent these regulatory barriers, NBTs research should now focus on generating transgene-free products, specifically in case of food crops (Wolt et al., 2016). Since, in vegetatively propagated crops like potato, the procedure for transgene removal in subsequent generations through segregation is time-consuming, the utilization of agroinfiltration and protoplast transformation to deliver NBTs' reagents provide a rational procedure for transgene-free potato production (Bortesi and Fischer, 2015). The CRISPR/Cas9 approach can be utilized to incorporate nutritional improvement in potato coupled with late blight resistance through transient expression of transcription factor (StWRKY1) in a transgene-free method (Yogendra et al., 2015). Other research priorities could focus on eliminating allergen compounds in potato such as alkaloids, glycoprotein patatin etc. (Zaheer and Akhtar, 2016). In addition, incorporation of abiotic (environmental, salinity, drought, temperature) stress resistance coupled with increased nutrition could facilitate potato to acclimatize in diverse agro-ecological zones, thus

### REFERENCES


impeding food-shortage in less fertile/water deficit agricultural lands. The introduction of pest resistance into commercial cultivars would reduce the pesticide applications, thus impeding the environmental pollution. Further expansion of nutritional studies can set some preliminary values to justify the health benefits of potato-derived foods. Research efforts are needed to mitigate the mechanisms of nutrient-loss, such as copigmentation and to enhance the health-promoting components such as antioxidants and phytochemicals in commercial cultivars of potato.

The availability of potato genome sequence (www. potatogenome.net) has facilitated the comparative genomic analyses to identify the genes useful for improving several agronomically important traits like tuberization, loss of bitterness, and diseases resistance (Hardigan et al., 2017; Li et al., 2018). The NBTs offer fast-track development of commercial potato cultivars such as Russet Burbank, Désirée, Kathadin etc. with superior traits such as improved nutrition, biotic and abiotic stress tolerance, and enhanced yield. However, to achieve such goals, it is paramount to acknowledge that not a single GE approach is sufficient to incorporate all the desired traits, rather an integration of NBTs coupled with well-established conventional breeding techniques will be needed. Here, we believe that the future of GM potato is reliant not only on some consumer-oriented traits such as fortified nutrition, enhanced flavor and appearance, but also on some industrial traits such as enhanced starch quality, and reduced CIS activity, which will ultimately enhance the marketability and long-term acceptability of GM potato.

### AUTHOR CONTRIBUTIONS

SM provided the outlines of the review and contributed the key ideas. AH, SS and SZ wrote the manuscript and prepared the figures. SM, AH, and SZ worked on and improved the original draft and figures. The manuscript was approved by all co-authors.

### ACKNOWLEDGMENTS

The authors would like to thank Dr. Ghulam Mustafa for valuable suggestions, and critically reviewing this manuscript.


in the cultivated potato. Proc. Natl. Acad. Sci. U.S.A. 114, E9999–E10008. doi: 10.1073/pnas.1714380114


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Hameed, Zaidi, Shakir and Mansoor. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Investigation of Baseline Iron Levels in Australian Chickpea and Evaluation of a Transgenic Biofortification Approach

Grace Z. H. Tan<sup>1</sup> , Sudipta S. Das Bhowmik <sup>1</sup> , Thi M. L. Hoang<sup>1</sup> , Mohammad R. Karbaschi <sup>1</sup> , Hao Long<sup>1</sup> , Alam Cheng<sup>1</sup> , Julien P. Bonneau<sup>2</sup> , Jesse T. Beasley <sup>2</sup> , Alexander A. T. Johnson<sup>2</sup> , Brett Williams <sup>1</sup> and Sagadevan G. Mundree<sup>1</sup> \*

*<sup>1</sup> Centre for Tropical Crops and Biocommodities, Queensland University of Technology, Brisbane, QLD, Australia, <sup>2</sup> School of Biosciences, The University of Melbourne, Melbourne, VIC, Australia*

#### Edited by:

*Michael A. Grusak, Children's Nutrition Research, Agricultural Research Service, United States Department of Agriculture, United States*

#### Reviewed by:

*Marta R. M. Lima, University of California, Davis, United States Hamid Khazaei, University of Saskatchewan, Canada*

\*Correspondence: *Sagadevan G. Mundree sagadevan.mundree@qut.edu.au*

#### Specialty section:

*This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science*

Received: *13 February 2018* Accepted: *24 May 2018* Published: *14 June 2018*

#### Citation:

*Tan GZH, Das Bhowmik SS, Hoang TML, Karbaschi MR, Long H, Cheng A, Bonneau JP, Beasley JT, Johnson AAT, Williams B and Mundree SG (2018) Investigation of Baseline Iron Levels in Australian Chickpea and Evaluation of a Transgenic Biofortification Approach. Front. Plant Sci. 9:788. doi: 10.3389/fpls.2018.00788* Iron deficiency currently affects over two billion people worldwide despite significant advances in technology and society aimed at mitigating this global health problem. Biofortification of food staples with iron (Fe) represents a sustainable approach for alleviating human Fe deficiency in developing countries, however, biofortification efforts have focused extensively on cereal staples while pulses have been largely overlooked. In this study we describe a genetic engineering (GE) approach to biofortify the pulse crop, chickpea (*Cicer arietinum* L.), with Fe using a combination of the chickpea nicotianamine synthase 2 (*CaNAS2*) and soybean (*Glycine max*) ferritin (*GmFER*) genes which function in Fe transport and storage, respectively. This study consists of three main components: (1) the establishment for baseline Fe concentration of existing germplam, (2) the isolation and study of expression pattern of the novel *CaNAS2* gene, and (3) the generation of GE chickpea overexpressing the *CaNAS2* and *GmFER* genes. Seed of six commercial chickpea cultivars was collected from four different field locations in Australia and assessed for seed Fe concentration. The results revealed little difference between the cultivars assessed, and that chickpea seed Fe was negatively affected where soil Fe bioavailability is low. The desi cultivar HatTrick was then selected for further study. From it, the *CaNAS2* gene was cloned and its expression in different tissues examined. The gene was found to be expressed in multiple vegetative tissues under Fe-sufficient conditions, suggesting that it may play a housekeeping role in systemic translocation of Fe. Two GE chickpea events were then generated and the overexpression of the *CaNAS2* and *GmFER* transgenes confirmed. Analysis of nicotianamine (NA) and Fe levels in the GE seeds revealed that NA was nearly doubled compared to the null control while Fe concentration was not changed. Increased NA content in chickpea seed is likely to translate into increased Fe bioavailability and may thus overcome the effect of the bioavailability inhibitors found in pulses; however, further study is required to confirm this. This is the first known example of GE Fe biofortified chickpea; information gleaned from this study can feed into future pulse biofortification work to help alleviate global Fe deficiency.

Keywords: pulse biofortification, iron, genetic modification, nicotianamine synthase, soybean ferritin, crop improvement, chickpea (Cicer arietinum L.)

### INTRODUCTION

Iron (Fe) deficiency has long been recognized as one of the most common micronutrient deficiencies in the world. Afflicting both developing and developed nations, it is the cause of more than 60% of global anemia cases (WHO, 2008; Alvarez-Uria et al., 2014). To combat this problem several strategies have been developed such as dietary diversification, supplementation, food fortification, and crop development. Amongst these, the development of crops with increased Fe concentrations and/or bioavailability (also known as "biofortification") has garnered great interest due to its sustainability, cost-effectiveness, and accessibility of products to vulnerable populations (Nestel et al., 2006).

Biofortification can be achieved through breeding or genetic engineering (GE), and this has been performed in various crop species. The focus thus far, however, has mostly been on starchy staples like as cereals (e.g., rice, wheat, pearl millet) and root crops (e.g., potato, cassava) (HarvestPlus, 2015). Naturally, it is in these species that the greatest advances have been made. For instance, more than three-fold increase in Fe concentration has been reported in biofortified pearl millet and its effectiveness in combating Fe deficiency anemia has been verified via feeding trials (Cercamondi et al., 2013; Finkelstein et al., 2015).

Aside from the aforementioned staples, recent years have seen growing interest in pulses as targets for Fe biofortification. Pulses are defined as leguminous crops harvested solely for dry grain (FAO, 1994), and most serve as important secondary staples, particularly with their high protein content (Iqbal et al., 2006); it is this latter feature that also complements the existing biofortification work in cereals. The pulse biofortification effort is relatively young compared to cereals but there has been considerable progress, notably in the common bean. Several biofortified varieties have been generated from the HarvestPlus breeding programs, with up to 94% enhancement in seed Fe concentration achieved (Katsvairo, 2015). The work has also progressed to feeding trials which have yielded promising results (Cercamondi et al., 2013; Kodkany et al., 2013; Finkelstein et al., 2015). This success has paved the way for advances in other pulses. Studies on Fe accumulation traits have been performed on cowpea (Fernandes Santos and Boiteux, 2015), chickpea (Diapari et al., 2014), pea and lentil (Ray et al., 2014), and test trials are currently underway for some of them (HarvestPlus, 2016).

Thus far, most, if not all, of this work has been focused on breeding while GE remains unexplored. As such, there are no established GE strategies for pulses, though some lessons can be drawn from the work in cereals. One of the most successful examples to date is the GE Fe biofortified rice (Orzya sativa), in which seed Fe concentration was increased by 7.5-fold with no yield penalty (Trijatmiko et al., 2016). The strategy targeted the three core processes of Fe metabolism—uptake, translocation, and storage—through constitutive overexpression of the rice Nicotianamine Synthase 2 gene (OsNAS2) gene and seed-specific expression of the soybean (Glycine max) Ferritin (GmFER) gene.

NAS catalyzes the biosynthesis of nicotianamine (NA), a nonproteogenic chelator of divalent transition metals that facilitates translocation of said metals in plants (Scholz et al., 1992). In graminaceous species, it is also a precursor for the mugeneic acid (MAs) family of phytosiderophores which contribute to both Fe uptake from soil and in planta translocation (Higuchi et al., 1999). When constitutively overexpressed in rice cv. Nipponbare, the OsNAS2 gene caused a four-fold increase in grain Fe concentration (Johnson et al., 2011).

Ferritin (FER), on the other hand, is an Fe storage protein that allows for safe sequestration of Fe in a soluble and bioavailable form. When overexpressed in the seed, GmFER has been demonstrated to increase seed Fe concentration by up to threefold in several plant species (Goto et al., 1999). Excessive expression, however, may lead to disproportionate sink strength, resulting in altered sequestration of Fe in source tissues and the development of Fe deficiency symptoms (Van Wuytswinkel et al., 1999; Qu et al., 2005; Masuda et al., 2013b). This problem can be rectified by increasing Fe uptake and translocation capacities, such as through the co-expression with NAS (Masuda et al., 2013a). In this case, a synergistic effect was also achieved, producing greater enhancement in seed Fe concentration (Wirth et al., 2009; Trijatmiko et al., 2016).

Whether this strategy will have a similar effect on Fe concentration when applied to a pulse crop is uncertain. However a major advantage is its potential effect on Fe bioavailability. Both NAS and GmFER have been linked to increased Fe biovailability (Davila-Hicks et al., 2004; Lönnerdal et al., 2006; Zheng et al., 2010), a feature not usually accorded to other commonly used Fe metabolism genes. This is particularly relevant to pulse biofortification given the inherently high levels of antinutrients like phytic acid which inhibit Fe absorption in the gut (Sandberg et al., 1989; Hemalatha et al., 2007; Petry et al., 2014).

For this study, target species is chickpea (Cicer arietinum). The second most important pulse crop in the world with an annual production exceeding 14.2 million tons (FAO, 2016). The bulk of the chickpea crop is currently grown and consumed in India where human Fe deficiency is prevalent, however, continued population growth is likely to result in increased demand for chickpea in Africa and other parts of Asia (Rao et al., 2010; Akibode and Maredia, 2012). Fe concentrations in chickpea has been found to range from 3 to 14.3 ppm (Wood and Grusak, 2007), though due to the presence of naturally occurring inhibitors, only a small fraction is bioavailable (Hemalatha et al., 2007). Both iron concentration and bioavailability is subject to genotype and environmental effects, and to date, detailed studies of such effects are limited to populations in India (Upadhyaya et al., 2016) and Canada (Diapari et al., 2014). No such information is available for Australian populations, and part of this study would therefore serve to partially fill in this gap.

Overall, the main aim of this study was to biofortify chickpea by GE to overexpress NAS and GmFER. This body of work consisted of three parts: (1) assessing the macroand micro-elemental composition of six modern Australian chickpea cultivars and identifying a suitable cultivar for Fe biofortification research, (2) cloning and expression analysis of an endogenous chickpea NAS gene termed CaNAS2, and (3) constitutive overexpression of the CaNAS2 gene and constitutive expression of soybean GmFER in chickpea as a novel GE approach to produce Fe biofortified chickpea.

### MATERIALS AND METHODS

### Plant Material

For the elemental composition analysis of commercial chickpea, three kabuli cultivars (Genesis090TM, KalkeeTM, and PBA Monarch) and three desi cultivars (PBA Boundary, CICA0912, and PBA HatTrick) were used. Seed samples were obtained from field trials at four locations within Queensland—Billa Billa, Warra, Roma, and Kingaroy—and from the seed company Grainland in Moree, New South Wales. All seeds were produced during the 2014 winter growing season. Information on the cultivation sites and conditions during the growing period, where available, is listed in Supplementary Table 1. The soil types of the field locations were provided by Dr Yash Chauhan from the Agricultural Production Systems sIMulator (APSIM) database.

For the gene expression analyses and chickpea transformation, PBA HatTrick seeds were purchased from the seed company Grainland in Moree, New South Wales.

### Plant Growth Conditions

All in vitro cultivation was performed in growth cabinets set at 24 ± 1 ◦C, under fluorescent lights with a 16 h light/ 8 h dark cycle.

For glasshouse cultivation, temperatures were maintained at 21 ± 1 ◦C and 61 ± 1.5% relative humidity. Natural lighting was used except during dusk, when artificial lighting was then turned on to complete a 16 h light/ 8 h dark cycle. An average natural light intensity of around 450 ± 1 µmoles s-1 m-2 of cloudy and sunny day prevailed during the growth period. Seeds were first germinated in Plugger's potting mix, before transplanting to 400 × 250 mm pots containing a 1:1 mixture of University of California (UC) mix and Searles <sup>R</sup> Premium potting mix. The recipe for the UC mix consists of 80 kg sand, 120 kg peat, and 100 kg sand, peat, and gravel, supplemented with 400 g blood and bone, 100 g Micromax micronutrients, 40 g KSO4, 40 g KNO3, 400 g superphosphate, 300 g hydrated lime, and 1,200 g dolomite.

Plant were watered with 100 mL every 2 days via an automated watering system. After ∼3 months, when at least 80% of the pods have filled, watering was ceased in preparation for harvesting. Harvesting was done approximately 3 weeks thereafter, or when the plants have completely dried. All seeds were de-husked by hand and stored in paper envelopes at 4◦C until planting or analysis.

### Elemental Analysis

All samples were cleaned, freeze-dried, and milled prior to analysis. A minimum of three biological replicates were used per transgenic event.

For leaf tissue, milled samples were pressed into 5 mm diameter pellets and analyzed via LA-ICP-MS (laser ablation inductively-coupled mass spectroscopy) using an Agilent 8,800 Inductively Coupled Plasma Mass Spectrometer attached with an ESI 193 nm Excimer Laser. The laser was set at a pulse width of 4 ns, spot size of 85 microns, and scan speed of 10 microns/s. At least three lines scans were used for each sample as technical replicates.

For whole seed analysis, acid digestion was performed on milled samples. Briefly, 2 mL HNO<sup>3</sup> and 0.5 mL H2O<sup>2</sup> were added to 200–300 mg of milled sample, vortexed, and allowed to stand overnight at room temperature. Following digestion, the tubes were shaken at 200 rpm for 20 min, incubated at 80◦C for 30 min, then 125◦C for 2 h. Upon cooling to room temperature, the volume was made to 25 mL using MilliQ water and the samples agitated at 300 rpm for 5 min. Undissolved material (e.g., silicates) was settled for 60 min. The settled extract was then filtered and analyzed via ICP-OES (inductively coupled plasma optical emission spectroscopy) using a Perkin Elmer Optima 8300 DV Inductively Coupled Plasma Optical Emission Spectrometer. Three technical replicates were prepared per sample.

For analysis of trace element distribution in the seed, 100 seeds were imbibed in MilliQ water for 20 h. The seeds were then separated into the seed coat, cotyledons, and radicle. To measure approximate distribution of mass, the weight of the individual parts of 10 seeds were taken. Tissues of the same type were then pooled and processed for analysis like the whole seed. Three technical replicates were prepared per sample.

### Designation and Bioinformatics Analysis of Chickpea NAS2 Gene

Four chickpea NAS amino acid sequences (XP\_004495658.1, XP\_004487761.1, XP\_004488704.1 and XP\_004494544.1) were retrieved from the NCBI database. The ortholog with highest similarity to the amino acid sequence encoded by the rice OsNAS2 gene (LOC\_Os03g19420) was designated as the CaNAS2 gene coding for protein XP\_004495658.1 (Supplementary Table 2). The three additional NAS genes were named as the following: CaNAS3 coding for XP\_004487761.1 protein, CaNAS4 coding for XP\_004494544.1 protein, and CaNAS1 for XP\_004488704.1 protein respectively. The four NAS amino acid sequences were aligned based on amino acid conservation using the Geneious Pro 5.6.6, as per the settings described by Bonneau et al. (2016) (CLUSTALW—cost matrix BLOSUM62,


TABLE 1 | Summary of Fe, Zn, and P concentrations in kabuli and desi cultivars grown at different locations.

*All values are expressed as ppm. Data are expressed as mg/100 g and presented as a mean of all cultivars collected from that site. For each cultivar per site, n* = > *3. Values with different superscript letters indicate a significant difference at p* < *0.05.*

threshold 1). A blastn using the genomic sequences of the four CaNAS genes was performed against Cicer arietinum (cv. kabuli, CDC Frontier)—CDS database (https://legumeinfo. org) to identify chromosomal location. Several bioinformatics tools were then used to predict characteristics of the enzyme encoded by CaNAS2: the theoretical isoelectric point (pI) and molecular weight were calculated using the Compute pI/Mw tool on ExPASY (Bjellqvist et al., 1993, 1994; Gasteiger et al., 2005); the hydrophobicity profile of the protein was assessed using ProtScale (Gasteiger et al., 2005) and potential transmembrane sections were identified using TMpred (http://www.ch.embnet. org/software/TMPRED\_form.html). A check for motif sequences was conducted using ScanProsite (de Castro et al., 2006) and MOTIF Search GenomeNet<sup>1</sup> Phobius (Käll et al., 2004) and iPSORT (Bannai et al., 2001, 2002) was used to identify potential signaling peptides.

### Phylogenetic Analysis of NAS Proteins

A progressive pairwise alignment was performed using full length amino acid sequences of 58 NAS proteins (Supplementary Table 3) using default settings of Geneious alignment (global alignment with free end gaps, Blosum62, gap open penalty 12, gap extension penalty 3)—Geneious Pro 8.1.7 software. Once the protein alignment, a phylogenetic analysis which generated an unrooted tree was conducted as described in Beasley et al. (2017).

### RNA Extractions and Quantitative RT-PCR

Total RNA was isolated from frozen tissues using RNeasy Mini kits (QIAGEN) following the manufacturer's instructions. A 500 ng aliquot of total RNA was treated with RQ1-DNAse (Promega) and the absence of contaminating DNA confirmed via PCR. cDNA was then synthesized using SuperScriptTM IV Reverse Transcriptase (ThermoFisher Scientific). The synthesized cDNA was used for qualitative RT-PCR and quantitative RT-PCR.

For qualitative RT-PCR, each reaction comprised of 5 µL of 2X GoTaq green (Promega), 0.25 µL each of 10µM forward and reverse primers, 0.6 µL of DMSO, and 1 µL of undiluted cDNA as the template. MilliQ water was added to reach a final reaction volume of 10.6 µL. The PCR program used was as such: initial denaturation at 95◦C for 3 min, followed by 30 cycles of denaturation at 95◦C for 30 s, annealing at 48–60◦C (depending on primers) for 30 s, and extension at 72◦C. Extension time was set at 1 min per 1 kbp of the final product size. A final extension was done at 72◦C for 5 min.

For qPCR, a 1:30 dilution of cDNA was used as template. The latter was performed on a CFX384 TouchTM Real-Time PCR Detection System (BIO-RAD) using the SYBR Green PCR Master Mix kit (Applied Biosystems). A primer concentration of 300 nM was used, and the primer sequences are as listed in Supplementary Table 4. The housekeeping genesEF1α, and GAPDH were included in each run to serve as internal controls; their primer sequences are as published by Garg et al. (2010). All housekeeping genes were confirmed to be stable under the experimental conditions used. Three biological replicates were used, from each of which three technical replicates were

<sup>1</sup>Available online at http://www.genome.jp/).



*Data is presented as a mean* ± *SD. All cultivars from had n*=*3 except for HatTrick from NSW, which has n* = *5. Values sharing the same superscript letters indicate groups that are not significantly different at p* < *0.05 when tested with one-way ANOVA, using Tukey's HSD post-hoc test.*

prepared. The qPCR program used was as follows: initial denaturation at 95◦C for 10 min, followed by 45 cycles of denaturation at 95◦C for 10 s, annealing at 60◦C for 30 s, and slow ramping of 0.5◦C/min from 65 to 90◦C for the melt curve.

### Fe Deficiency Experiment

PBA Hattrick seeds were sterilized and germinated on half strength Murashige and Skooge (MS) media. Seed coats were removed post-germination and the seedlings were acclimatized for 4 days in tap water. Twenty four-week old seedlings of approximately the same size and developmental stage were transferred to a hydroponics system in a growth cabinet, with 10 replicates per set-up. Later, 1 month old plants were treated with full-strength Hoagland solution with or without Fe-EDTA. Each setup contained ∼600 mL of Hoagland solution which was topped up every 3 days. The Hoagland solution was replaced with MilliQ water every third top-up to dilute any accumulated salts. Visible chlorosis in the Fe-deprived plants was observed after 4 weeks of treatment and samples were collected 2 weeks thereafter. Three plants of similar conditions and growth stage were selected from each treatment and the following tissue types collected: mature leaf, stem, cotyledon, and root. Senescent leaf and chlorotic leaf were also collected from the Fe-sufficient and Fe-deficient plants respectively. All samples were snap-frozen in liquid nitrogen immediately after collection and stored in −80◦C until RNA extraction.

### Cloning of CaNAS2 and Construction of NAS-GmFER Overexpression Vectors

A binary vector using the pOPT-EBX backbone was constructed to constitutively overexpress the CaNAS2 gene and constitutively express the GmFER gene (**Figure 1**). Included in the T-DNA region was the selectable marker gene neomycin phosphotranferase II (NPTII) which confers resistance to the antibiotics geneticin and kanamycin. All cloning primers used are listed in Supplementary Table 5. The CaNAS2 gene was cloned from chickpea (cv HatTrick) genomic DNA, with primers designed from the predicted sequence in the Genbank database, accession number XM\_004495601. Restriction sites were added to the ends via site-directed mutagenesis using high fidelity PCR (Phusion <sup>R</sup> , NEB). The amplified fragment was cloned into a pGEM <sup>R</sup> -T Easy, then subcloned


TABLE 3 | Pearson's correlation coefficient between the different trace elements in PBA HatTrick.

*For Billa Billa, Roma, Warra, and Kingaroy, samples were n* = *3 while NSW was n* = *5. Grain samples from four locations within Queensland (Billa Billa, Warra, Roma, and Kingaroy) and a seed supplier from New South Wales were analyzed by ICP-OES. Values marked with an* \**Indicate a significant correlation between two elements (p* < *0.05).*

in a 5′ to 3′ direction to a pGEM <sup>R</sup> -T Easy vector with a cassette containing a Nos promoter and CaMV 3′ UTR. Following this the NosP-CaNAS-CaMV 3′ UTR was digested and ligated to a pOpt-EBX-GmFER backbone to form the complete vector. To generate the pOpt-EBX-GmFER backbone, GmFER was cloned from a synthesized fragment (accession no. NM\_001250105.2). The amplified fragment was cloned into a pGEM <sup>R</sup> -T Easy vector, then subcloned into the pOpt-EBX backbone containing a CaMV 35s promoter and Nos terminator.

To ensure integrity and correct orientation of each gene and component, sequence verification was performed after each cloning step in the above process. The final verification was performed on the completed construct, which was then transformed into electrocompetent Agrobacterium tumefaciens strains Agl1.

### Generation and Molecular Characterisation of Transgenic Chickpea

Agrobacterium-mediated transformation of the desi cultivar, PBA Hattrick, followed the protocol developed by Sarmah et al. (2004) with a few modifications. Briefly, half-embryonic axes were prepared from imbibed seeds. Additional injury was inflicted to the cut surface of the radicle using a sterile 26 gauge needle dipped in Agrobacterium strain AGL-1 harboring the expression vectors. The explants were immersed in Agrobacterium for an hour, followed by co-cultivation in B5 media for 72 h. Following co-cultivation, the explants were transferred to regeneration and selection medium 1 (MS media containing 500 µg/L of BAP, 500 µg/L of kinetin, 50 µg/L of NAA, 200 mg/L of kanamycin and 25 mg/L of meropenem). Shoots obtained in first round of regeneration and selection medium were further selected by subsequent subculturing in the regeneration and selection medium 2 (MS media containing 500 µg/L of BAP, 500 µg/L of kinetin, 200 mg/L of kanamycin and 25 mg/L of meropenem) every 14–21 days. Up to eight rounds of selection were done to obtain putative GE events. Any explants that exhibited proliferative shoot growth during that duration were isolated and considered an individual GE event. Upon reaching an appropriate size, shoots from such multiplying clumps were grafted onto non-GE rootstocks grown on half-strength MS media. Grafts were allowed to set for up to 3 weeks before acclimatization.

Acclimatized plants were screened via PCR of gDNA for the genes of interest. To avoid false negatives caused by potential chimerism, a pooled sample consisting of leaves from every branch was used. Primers used for PCR screening are as listed in Supplementary Table 6. Two PCR-positive T<sup>0</sup> events were propagated to maturity for two generations to obtain sufficient T<sup>3</sup> seed for the experiments described in this paper. Null segregants from every generation were also maintained to serve as negative controls.

### NA Quantification

Freeze-dried seeds from three different plants of the same transgenic event were pooled and milled to form a bulked flour sample, from which four technical replicates were drawn for NA quantification. Liquid chromatography-mass spectrometry (LC-MS) was used to quantify 9-fluorenylmethoxycarboxyl chloride (FMOC-Cl) derivatized NA on an LC 1290 series coupled to a 6490 series triple quadrupole MS (Agilent Technologies Inc.) using established protocols (Selby-Pham et al., 2017). In short, a combined methanol (100%) and 18 M H2O extraction (5 µL) of pulverized chickpea flour (25 mg) was combined with sodium borate buffer (pH = 8, 1 M, 10 µL), EDTA buffer (pH = 8, 50 mM, 10 µL), and fresh FMOC-Cl solution (50 mM, 40 µL). After incubation (60◦C, 700 rpm, 15 mins), the solution mixture was quenched via the addition of formic acid (pH = 4, 5%, 8.9 µL). Chromatography was performed using a reversephase column (Zorbax Eclipse XDB-C18, HS 2.1 × 100 mm 1.8 Micron, Agilent Technologies Inc.) with aqueous (0.1% v/v FA in dH2O) and organic (0.1% v/v FA in acetonitrile) mobile phases.

### Statistical Analysis

All statistical analysis was performed on Minitab statistical software (Arend, 2010) using one-way ANOVA. The Tukey HSD


*Data are presented as a mean* ± *SD. All cultivars from had n* = *3 except for HatTrick from NSW, which has n* = *5. Values sharing the same superscript letters indicate groups that are not significantly different at p* < *0.05 when tested with one-way ANOVA, using Tukey's HSD post-hoc test.*

FIGURE 2 | Distribution of macro and micro elements in chickpea (cv HatTrick) seed expressed as a percentage of the total element content in the three main tissue types present in the grain. Values for the elemental profile was derived from a bulked flour sample produced from pooling tissue from 100 seeds. Three technical replicates were used. Mass distribution in the seed was calculated from 10 biological replicates.

represent the YXXΦ and the LL, IL, or ML motifs.

this tree are *Arabidopsis thaliana* (AtNAS), *Hordeum vulgare* (HvNAS), *Lotus japonicus* (LjNAS), *Medicago truncatula* (MtNAS), *Oryza sativa* (OsNAS), *Thlaspi caerulescens* (TcNAS), *Solanum lycopersicum* (SlNAS), *Zea mays* (ZmNAS), and *Triticum aestivum* (TaNAS). Black nodes (•) represent weak bootstrap values (<75%). The scale bar corresponds to branch length and longer branches correspond to greater numbers of nucleic acid polymorphisms along the sequence.

test was used in the analysis the different chickpea cultivars, while Dunnett's test was used in the analysis of the GE chickpea.

### RESULTS

### Mineral Composition of Chickpea Cultivars and Identification of Factors Influencing Seed Fe Concentration

Fe concentrations in chickpea were found to range from 3.36 to 52.0 ppm, with no significant differences between the cultivars, though average values were slightly higher in the kabuli types compared to the desi (**Tables 1**, **2**). The highest values were noted in the kabuli cultivar Genesis090TM, while the lowest values were mostly found in PBA HatTrick, though the difference to other desi cultivars was negligible (**Table 2**). This, in combination with the availability of established transformation protocols for the cultivar, made HatTrick the choice candidate for further work.

Between locations, similar mineral profiles were observed amongst samples grown in Billa Billa, Roma, and Warra, which had vertosol-type soils. In contrast, samples obtained from Kingaroy contained less Fe. Also, unique to this locality was a high Mn concentrationto Fe ratio, which appears to have produced the negative correlation between the two elements

(**Table 3**). No other negative correlations were observed between Fe and other elements. On the other hand, the strongest positive correlations were found between Fe, Zn, and P. Zn and P concentrations in particular. Unlike Fe however, greater differences in Zn and P were observed between the locations than between the genotypes (**Table 4**).

### Cotyledons Serve as the Primary Store for Fe in PBA Hattrick Seeds

Amongst the different part of the seed, the radicle was found to have the highest Fe concentration at 95.0 ppm, followed by the cotyledons at 50.0 ppm. Due to its small mass however, its contribution only 3% to the total seed Fe content. The cotyledons on the other hand, constituted the bulk of the seed mass and contained 90% of the seed's total Fe. It was also the main store for all the other elements tested. The exception to this was calcium and manganese—the bulk of the former was found in the seed coat, while the latter was almost equally divided between the seed coat and cotyledons (**Figure 2**).

### Legume NAS Homologs Form Distinct Branches Among the Non-graminaceous Orthologues

Concerning their chromosomal locations, the CaNAS2 and CaNAS4 genes were located on chromosomes 4 (Ca4) and 3

(Ca3) respectively, while both CaNAS1 and 3 were located on chromosome 1 (Ca1). All four CaNAS genes were found to be to consist of a single exon. CaNAS1, 2, 3, and 4 coded for 285, 306, 311, and 318 amino acids respectively. In the CaNAS1 protein, a longer N-terminal and a shorter C-terminal was seen compared to the other three CaNAS homologs.

In the four CaNAS amino acid sequences, several highly conserved regions were noted (**Figure 3**). Of these, the YXX8 (Y refers to tyrosine, X to any amino acid residue, and 8 to bulky hydrophobic residues) and di-leucine (LL; leucine may be substituted with isoleucine) motifs were known to be conserved amongst the NAS homologs. Amongst some of the legume NAS sequences however, a variation of the LL motif was observed where the first leucine was substituted by methionine. This was seen in the CaNAS3 protein sequence, as well as in MtNAS2 from Medicago truncatula and LjNAS2 from Lotus japonicus (Supplementary Table 3).

Phylogenetic analysis of the 58 NAS proteins revealed a clear distinction between graminaceous and non-graminaceous sequences (**Figure 4**). Two clades were present in the former and were consistent with a prior report (Bonneau et al., 2016). With the latter, three subgroups for legumes were observed these were defined as subgroups 1, 2, and 3. Subgroup 1 consisted of includes CaNAS1 from chickpea and MtNAS1 from Medicago truncatula. Subgroup 2 consisted of CaNAS3, MtNAS2, and LjNAS2 (from Lotus japonicus). Lastly, Subgroup 3 consisted of CaNAS2 and 4, MtNAS3 and 4, and LjNAS1. The presence of CaNAS2/MtNAS3 and CaNAS4/MtNAS4 in the same branch is most likely due to genome duplication.

Further bioinformatic analysis of the CaNAS2 protein indicated an approximate molecular weight and pI of 34.36kDA 5.52 respectively. The enzyme was mostly hydrophilic with a

potential transmembrane domain at position 126–151. A noncytoplasmic localisation was predicted, though no apparent signaling peptides were detected.

### CaNAS2 Is Expressed in Various Vegetative Tissues Under Fe Sufficient Conditions

The expression pattern of CaNAS2 was examined under Fesufficient and Fe-deficient conditions. The Fe-deficient plants used in this study were observed to be paler green than the Fe sufficient controls, with severe chlorosis in the young leaves. No nodules were observed in either Fe-deficient or Fesufficient plants. CaNAS2 transcripts were detected in all the tissue types tested, though the levels were largely influenced by Fe status, with an overall downregulation under Fe deficiency. Gene expression in the Fe deficient plants, where detected, was generally low, and comparable across all tissues (**Figure 5**). Similar levels were also detected in the senescing leaf of Fesufficient plants. By contrast, other Fe sufficient tissues exhibited markedly higher expression, particularly in the stem, cotyledons, and roots. A 16-fold difference was seen between stems of the Fe-sufficient and Fe deficient plants, while expression was only detected in Fe-sufficient cotyledons. For root tissue however, no data could be obtained for the Fe deficient plants due to the consistently poor quality of the extracted RNA.

### Transgenic Chickpea Highly Express Both Transgenes and Have Increased Fe in Leaf Tissue and Increased Nicotianamine in Seed Tissue

Both of the regenerated GE events (6.1 and 6.14) that were propagated to the T<sup>3</sup> generation were confirmed via Southern blot to have single transgene integration sites. Expression analyses showed a 49- and 93-fold increase in CaNAS2 expression in events 6.6 and 6.14, respectively, compared to null segregant controls (**Figure 6B**). Expression analysis showed an 18- and 30 fold increase in expression of GmFER in events 6.6 and 6.1, respectively, compared to null segregant controls (**Figure 6A**). Agronomic performance of the T<sup>3</sup> GE events in the glasshouse was generally comparable to the null controls and no significant differences were seen in terms of morphology and the other parameters measured (**Figures 7**, **8**).

Two differences were observed with respect to micronutrient composition of the leaf tissue of events 6.6 and 6.14 compared to null segregant controls (**Figure 9**). Event 6.6 had significantly higher leaf Fe concentration which was 1.39-fold higher than the null segregant control. Event 6.14 had significantly lower leaf Zn concentration which was 1.4-fold lower than the null segregant control. No significant differences for Fe, Zn, or Mn concentrations were observed in the seed of events 6.6 and 6.14. Seed from events 6.6 and 6.14 contained significantly higher

flowering/pod-filling stage.

NA concentration which was nearly 2-fold higher than the null segregant control (**Figure 10**).

## DISCUSSION

### Conditions Affecting Seed Fe Concentrations in Chickpea and the Selection of cv PBA Hattrick for Fe Biofortification Research

Australian-grown chickpeas were previously reported to contain up to 140 ppm of Fe in their seed, though average values were ∼50 ppm (Petterson and Mackintosh, 1994). Such average values appear to be the norm globally, with similar results reported in chickpea from other countries (Meiners et al., 1976; Jambunathan and Singh, 1981; Thavarajah and Thavarajah, 2012; Diapari et al., 2014; Ray et al., 2014). That similar average values for seed Fe concentrations were obtained for the six cultivars used in our

study, indicating that the cultivars used fell within that global norm.

using Dunnett's post-hoc test (p 0.05).

A past study by Ray et al. (2014) has reported seed Fe concentration to be influenced firstly by environmental conditions, then by genotype. Our observations were only partially consistent with that report, perhaps due to the smaller number of locations and cultivars we examined. Between the cultivars and most locations, no major differences were noted. The major environmental effect on seed Fe was only observed where soil Fe bioavailability was potentially compromised, such as with the Kingaroy samples which had lower seed Fe concentrations. Past records have shown Kingaroy ferrosols to be acidic with a high manganese to Fe ratio, and such conditions have been documented to inhibit Fe uptake, sometimes to the point of chlorosis (Twyman, 1951; Tanaka and Navasero, 1966; Alvarez-Tinaut et al., 1980). In our study, this inhibition was asymptomatic; no Fe deficiency symptoms were reported by the growers, and the effect was only apparent upon assessment of the seed mineral profile. This is potentially problematic where biofortification efforts are concerned as attempts to increase Fe concentrations may be unknowingly hijacked by adverse soil conditions.

An environmental effect was also apparent in the Zn and P concentrations, both of which were positively correlated with Fe and affect seed nutritional value. With Zn, significant variations were noted between all sites regardless of cultivar, corresponding with observations by other authors who also reported significant year to year variations, even with seed from the same sites (Diapari et al., 2014; Ray et al., 2014). Management practices may explain at least part of the higher degree of Zn variation compared to Fe. Application of Zn fertilizers is a recommended practice in chickpea cultivation due to the risk of Zn deficiency in most Australian soils (Norton, 2013; Pulse Australia, 2016). The effects of Zn fertilization on grain Zn concentration however, are unpredictable and may differ between seasons (Akay, 2011). This variability is compounded by the effect of other management regimes like the application of P fertilizer. Aside from directly affecting grain P concentration (Saastamoinen, 1987), studies in pearl millet and wheat have highlighted a negative impact of P fertilizer on seed Zn concentration (Buerkert et al., 1998; Ryan et al., 2008). The precise reason behind it is uncertain, though it has been attributed to altered zinc uptake and the dilution effect caused by increased yields. The former was deemed the more likely, given the absence of adverse effects on seed Fe concentration (Ryan et al., 2008), though regardless, the implications on the nutritional quality of the seeds are still considerable. Seed P is primarily stored as phytate (Lolas et al., 1976; Griffiths and Thomas, 1981; Ravindran et al., 1994), a potent inhibitor of Fe and zinc bioavailability (Turnlund et al., 1984; Sandberg et al., 1989, 1999), and its levels can be considered a crude indicator of micronutrient bioavailability.

In terms of the genetic effect, only a slight influence was seen. Few significant differences were found amongst the cultivars assessed, though kabuli cultivars had marginally higher Fe concentrations than the desi. This lack of difference is likely a product of the breeding process. Currently, no information exists for micronutrient accumulating traits in existing germplasm. As micronutrient accumulation is often accompanied by yield penalties (Garvin et al., 2006; Ficco et al., 2009; Diapari et al., 2014), it is likely that such traits may have been bred out of the current cultivars as breeding efforts in Australia have primarily focused on yield, abiotic stress, and biotic stress resistance with no consideration for nutritional value (Pulse Breeding Australia, 2017). Consequently, reintroduction of Fe-accumulation traits may prove challenging, though the difficulty can be alleviated with modern biotechnology.

For this purpose the cultivar with the lowest Fe concentrations, PBA HatTrick, was identified as a suitable candidate for Fe biofortification. The benefits of this choice are manifold. PBA HatTrick is a popular choice amongst growers due to its high yield and resistance to phytophthora root rot. As a desi cultivar, it also has great potential for widespread dissemination, as desi constitutes 90% of the Australian chickpea export and therefore the bulk of the international market (Pulse Australia, 2016). With Fe localized primarily to the cotyledon, which is the main product, enhancements in Fe concentration will reach the consumer regardless of the form in which the seed is consumed. Bioavailability, however, may be a concern as phosphorus (and by extension, phytate) co-localizes with Fe to the cotyledons. This may perhaps be addressed with appropriate biofortification strategies and choice of target genes, one of which, for the purposes of this study, has identified as the novel CaNAS2 gene.

### CaNAS2 Has a Potential Housekeeping Role Under Fe Sufficient Conditions

In our study, CaNAS2 grouped with the other dicot sequences in a separate clade to monocot sequences. This dichotomy between the dicot and monocot sequences is consistent with the findings of other authors (Hakoyama et al., 2009; Filipe de Carvalho et al., 2012), and is likely reflective of the differing physiological roles of NAS between the two. However whether this is conclusive remains to be seen due to the limited number of experimentally verified homologs; monocot sequences used in this study were of graminaceous origins, and the inclusion of the non-graminaceous homologs may potentially alter the existing arrangment. Nonetheless recent evidence have revealed functional grouping amongst NAS homologs (Bonneau et al., 2016) in terms of roles in development and Fe deficiency. Assuming this is universal amongst higher plants, such grouping will allow for accurate prediction of the function of closely related homologs. This accuracy however, remains subject to the availability of sequences that can be extrapolated from–such is evident in this study. For example, nodule-specific expression of LjNAS2 (Hakoyama et al., 2009) was neither mirrored in the closely related MtNAS2 (Medicago truncatula Gene Expression Atlas, 2014), nor in the any of the other legume NAS used in this study. Whether the third member of Subgroup 2, CaNAS3, will be nodule-specific is uncertain—this could not be investigated in our study due to the absence of nodules as the plants were not inoculated with Rhizobium. Still, given that the substitution of the di-leucine motif by a methionine is a trait unique to Subgroup 2, it is plausible that some functional specialization is present. Further investigation and discovery of more nodulespecific homologs may shed more light on this.

Functional specialization may also be present in Subgroup 1, to which CaNAS1 and MtNAS1 belong. The nature of this specialization is still inconclusive. Subgroup 1 was the most divergent from other NAS proteins used in the phylogenetic analysis and no discernible trend could be seen between members. No orthologue of CaNAS1 was found in Lotus japonicus, perhaps due to the genome evolution in legumes (Wang et al., 2017). As with Subgroup 2, further study is required before conclusive statements may be made.

In the interim, only Subgroup 3 bears enough information for reasonable inference of function. Subgroup 3 homologs are notable for their widespread expression in various vegetative tissues (Hakoyama et al., 2009; Medicago truncatula Gene Expression Atlas, 2014), and such is also seen in CaNAS2 and 4 (**Figure 5**, Supplementary Figure 1). While the expression sites vary between homologs, ranging from roots, to leaves, and cotyledons, a common feature is the expression in the stem. Using LjNAS1 as a reference, this pattern may indicate a housekeeping role in the systemic redistribution of Fe (Hakoyama et al., 2009) which, in the case of CaNAS2, occurs under Fe-sufficient conditions. It is also likely that the expression of NAS in the diverse sites may operate at different scales, given the involvement of NA in long and short-distance translocation. Expression in the stem may serve to feed NA into the vascular tissue and symplast for systemic transport, while expression in the other locations may provide NA for more localized translocation.

Concerning the movement and localization of CaNAS2 within an intracellular context, the results predicted a noncytoplasmic, and potentially vesicular, localization. The accuracy of this however, is contentious. The YXXΦ and LL motifs conserved in the NAS family have been linked to maintenance of enzyme structure, and vesicular localization and movement (Nozoye et al., 2014a). Studies of various NAS homologs have yielded conflicting results. For example, vesicular localization have only been confirmed in OsNAS2, ZmNAS1 and ZmNAS2, while ZmNAS3 and the AtNAS family, localized to the cytoplasm (Mizuno et al., 2003; Nozoye et al., 2014a,b). It was proposed by Nozoye et al. (2014b) that vesicular localization is required for DMA synthesis, hence its specificity to the graminaceous homologs. With chickpea lacking in that regard, its localization pattern is likely to be more similar to that of AtNAS, though more studies are required to confirm this.

### Constitutive Expression of CaNAS2 and GmFER Does not Increase Seed Fe Concentration but Is Likely to Increase Seed Fe Bioavailability

As demonstrated in this study, constitutive overexpression of the endogenous CaNAS2 gene combined with constitutive expression of the GmFER gene in chickpea resulted in higher leaf Fe in one event, and lower leaf Zn in the other event, with no apparent effects on yield or morphology in either event. Seed Fe, Zn and Mn concentrations were not changed in either event. These results suggest that high expression of the CaNAS2 and GmFer transgenes is not an effective strategy for improving the micronutrient composition of chickpea grain. Potentially a better strategy in the future would be to use a seed-specific promoter to drive GmFer expression in conjunction with CaNAS2 overexpression. Indeed, this method has been demonstrated to be extremely effective in rice, producing a 7.5 fold increase in the Fe concentration of polished seeds with no yield penalty (Trijatmiko et al., 2016).

Due to high levels of inhibitory compounds (i.e., phytic acid), the Fe bioavailability in pulses is low relative to other crops (Hemalatha et al., 2007). As NA is a known promoter of Fe bioavailability, doubling the concentration of NA in chickpea flour may increase Fe bioavailability without alterations to seed mineral concentration (Zheng et al., 2010; Eagling et al., 2014). Future in vitro Fe bioavailability studies utilizing the Caco-2 cell line assay are needed to confirm increased Fe bioavailability in high-NA chickpea events.

### AUTHOR CONTRIBUTIONS

SM, BW, SD, and AJ conceived and designed the project. GT, SD, HL, AC, TH, MK, AJ, JPB, and JTB designed and performed the experiments. GT, JTB, and JPB data analysis. GT, JTB, JPB, and AJ wrote the paper.

### ACKNOWLEDGMENTS

We would like to thank the Australian Government for the funding through the Tropical Pulses for Queensland project. We also thank the following people: Dr TJ Higgins for his invaluable guidance on chickpea transformation, Queensland Department of Agriculture and Fisheries, and Dr Yash Chauhan for provision of seed material, Waite Analytical Services, Dr Bulukani Mlalazi, Dr Charlotte Allen, and Miss Karine Moromizato for their guidance and assistance with trace elemental analyses.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018. 00788/full#supplementary-material

## REFERENCES


collection of Italian durum wheat cultivars. Field Crops Res. 111, 235–242. doi: 10.1016/j.fcr.2008.12.010


metabolite levels during the iron deficiency response of rice. Rice 10:14. doi: 10.1186/s12284-017-0152-7


**Conflict of Interest Statement:** AJ is an editor for the Improving the Nutritional Content and Quality of Crops: Promises, Achievements, and Future Challenges topic of Frontiers in Plant Science.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Tan, Das Bhowmik, Hoang, Karbaschi, Long, Cheng, Bonneau, Beasley, Johnson, Williams and Mundree. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Biofortification of Cereals With Foliar Selenium and Iodine Could Reduce Hypothyroidism

### Graham Lyons\*

School of Agriculture, Food and Wine, University of Adelaide, Glen Osmond, SA, Australia

Concurrent selenium and iodine deficiencies are widespread, in both developing and developed countries. Salt iodisation is insufficient to ensure global iodine adequacy, with an estimated one-third of humanity at risk of hypothyroidism and associated iodine deficiency disorders (IDD). Agronomic biofortification of food crops, especially staples such as cereals, which are consumed widely, may be an effective component of a food system strategy to reduce selenium and iodine malnutrition. Iodine and selenium are needed in the optimum intake range for thyroid health, hence joint biofortification makes sense for areas deficient in both. Foliar application is recommended as the most effective, efficient, least wasteful method for selenium and iodine biofortification. Currently, selenium is easier to increase in grain, fruit, and storage roots by this method, being more phloem mobile than iodine. Nevertheless, strategic timing (around heading is usually best), use of surfactants and co-application with potassium nitrate can increase the effectiveness of foliar iodine biofortification. More research is needed on iodine transporters and iodine volatilisation in plants, bioavailability of iodine in biofortified plant products, and roles for nano selenium and iodine in biofortification. For adoption, farmers need an incentive such as access to a premium functional food market, a subsidy or increased grain yield resulting from possible synergies with co-applied fertilisers, enhancers, fungicides, and insecticides. Further research is needed to inform these

aspects of foliar agronomic biofortification. Keywords: biofortification, cereals, deficiency, hypothyroidism, iodine, iodine deficiency disorders (IDD),

### INTRODUCTION

selenium, wheat

Malnutrition is the main cause of global human mortality, with over 50% of deaths attributed to diet-related diseases. Micronutrient deficiencies, notably iron (Fe), zinc (Zn), selenium (Se), iodine (I), and certain vitamins are widespread globally, affecting about 60% of the world's population, and in many areas multiple deficiencies occur (Lyons and Cakmak, 2012). Dysfunctional food systems fail to provide optimum nutrition to populations, especially to vulnerable sub-groups such as infants, children, and pregnant and nursing women (White and Broadley, 2009). This has been exacerbated by high-yielding Green Revolution cereal varieties with grain often less micronutrient-dense than previously (Smolen et al., 2016a).

#### Edited by:

Alexander Arthur Theodore Johnson, University of Melbourne, Australia

#### Reviewed by:

Elizabeth Pilon-Smits, Colorado State University, United States Michael A. Grusak, USDA-ARS Children's Nutrition Research Center, United States

\*Correspondence: Graham Lyons graham.lyons@adelaide.edu.au

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 18 January 2018 Accepted: 15 May 2018 Published: 08 June 2018

#### Citation:

Lyons G (2018) Biofortification of Cereals With Foliar Selenium and Iodine Could Reduce Hypothyroidism. Front. Plant Sci. 9:730. doi: 10.3389/fpls.2018.00730

**132**

Biofortification of staple crops to achieve higher micronutrient concentrations in edible parts represents a food system strategy to address dietary deficiencies, with the potential to reach the neediest of the population (Haug et al., 2007; Bouis and Welch, 2010; Lyons and Cakmak, 2012). This approach, which links a nutritious agriculture with human health, can be more effective and sustainable than provision of food supplements (Lyons, 2014).

Previous research suggests that genetic biofortification (plant breeding and genetic engineering) may be more suitable for increasing pro-vitamin A carotenoids and Fe, whereas an agronomic (fertiliser) strategy may be more effective for Zn, Se, and I (Genc et al., 2004; Cakmak, 2008; Bouis and Welch, 2010; Lyons and Cakmak, 2012). Transgenics may play an important role in micronutrient biofortification (White and Broadley, 2009), as shown by the high-Fe variant of the popular IR64 rice variety (Trijatmiko et al., 2016). Biofortification using conventional breeding or transgenics is a long-term process. Furthermore, the success of genetic biofortification of Se and I depends largely on their plant available concentrations in the soil solution. In most soils, plant available Se, for example, comprises only about 2.5% of total Se (Tan et al., 2002). Agronomic and genetic biofortification are hence complementary (White and Broadley, 2009; Lyons and Cakmak, 2012).

If minerals such as Fe, Zn, Se, and I can be increased in staple foods, population status of these minerals can be increased without behavioural change (Bouis and Welch, 2010). Hence widely consumed cereals, especially wheat, provide a suitable vehicle for increasing population Se status using agronomic biofortification (Broadley et al., 2006; White and Broadley, 2009; Fairweather-Tait et al., 2011).

The iodothyronine deiodinases D1, D2, and D3, which are selenoenzymes, control thyroid hormone turnover and hence are crucial in thyroid gland metabolism. Selenium supply is prioritised to the thyroid under conditions of Se restriction. Concurrent deficiencies of Se and I may exacerbate hypothyroidism (Schomburg and Kohrle, 2008; Fairweather-Tait et al., 2011; Kohrle, 2013; Gashu et al., 2016), and low Se status increases risk of goitre, especially in women (Rasmussen et al., 2011; Schomburg, 2012; Wu et al., 2015). The more severe the Se deficiency, the less effective is I supplementation in alleviating goitre (Zimmermann et al., 2000; Drutel et al., 2013; O'Kane et al., 2018). Moreover, Se-dependent glutathione peroxidases protect the thyroid against oxidative stress, for example, due to excess I (Schomburg and Kohrle, 2008; Schomburg, 2012; Drutel et al., 2013; Kohrle, 2013).

Hypothyroidism is not the only pathological condition that can be exacerbated by concurrent I and Se deficiencies: myxoedematous cretinism, whose aetiology requires I and Se deficiency accompanied by a goitrogen (for example, TGF-beta, thiocyanates from cassava, Fusarium toxins in wheat), exists in parts of Tibet and the Democratic Republic of Congo (Contempre et al., 1992; Schomburg and Kohrle, 2008; Christophersen et al., 2012; Kohrle, 2013). In myxoedematous cretinism, hypothyroidism persists despite I supplementation (Contempre et al., 1992). Where both deficiencies occur, it is important to normalise I intake and status first, before supplementing with Se. If Se is supplemented first, hypothyroidism can worsen in the short term (Contempre et al., 1992).

This mini-review will focus on research on agronomic biofortification of cereals with Se and I, and explore the proposal that simultaneous application of these micronutrients has the potential to reduce hypothyroidism and related iodine deficiency disorders (IDD) in areas with concurrent Se and I deficiencies (**Figure 1**).

### SELENIUM

### Profound Influence on Human Health With a Variable Distribution

The importance of Se to human health, in terms of its key roles in the thyroid, brain, heart, and gonads, along with heavy metal-binding, antioxidant, anti-cancer, anti-bacterial, and antiviral activity, is indicated by its status as the only micronutrient to be specified in the human genome, as selenocysteine, the twenty-first amino acid (Rayman, 2000, 2002). Its deficiency is also linked to several diseases, including the osteoarthropathy, Kashin-Beck disease (KBD), which is still prevalent in parts of China, including the Loess Plateau in Shaanxi Province, and Tibet. Aetiological factors for KBD include Fusarium mycotoxins in infected grain, organic acids in drinking water, low dietary Se, and gene polymorphisms (Fairweather-Tait et al., 2011; Bissardon et al., 2017).

Although much less common than Se deficiency, Se toxicity can occur, for example in Enshi in the Chinese province of Hubei, when selenosis, characterised by hair loss and thickened nails, occurred, particularly from 1961 to 1964. It was caused by eating crops grown on high-Se soil (Yang et al., 1983). Daily recommended intake of Se is mostly 40–75 µg/day globally, with <30 µg/day inadequate and >900 µg/day potentially harmful; however, tolerable upper limits have been set lower, in the range of 400–450 µg/day for the United Kingdom, United States, Canada, EU, Australia, and New Zealand (Fairweather-Tait et al., 2011). Selenium's interplay with human physiology is complex and integral, and deficiency, sufficiency, and toxicity span a relatively narrow range of Se intake and status (Vinceti et al., 2009; Fairweather-Tait et al., 2011; Winkel et al., 2012).

Selenium delivery in a food system depends mainly on the levels of plant available Se in soils used for agriculture. Selenium is ubiquitous but of uneven plant-availability, hence its variability in populations and their sub-groups. It is estimated that up to a billion people are deficient in Se (Combs, 2001; Haug et al., 2007; Winkel et al., 2012; Ros et al., 2016). The element's availability in soils depends on soil pH, redox potential, cation exchange capacity, and levels of Fe, sulphur, aluminium, and carbon (Broadley et al., 2006; Chilimba et al., 2011; Christophersen et al., 2012; Winkel et al., 2012).

### Agronomic Biofortification: Foliar Selenate More Efficient

Selenium is well suited to agronomic biofortification of food crops. In the selenate form, it is readily taken up by plants

growing on most soils, then transported throughout the plant, accumulating in edible parts. In cereals, it is converted mostly into selenomethionine, which is well represented in grain endosperm, hence Se can be abundant and bioavailable in milled products such as white flour and polished rice (Lyons and Cakmak, 2012).

Selenium form is important for effective biofortification. Most studies have shown selenate (where Se exists in its highest oxidation state, +6) to be easily the most effective form when applied to the soil and usually more effective than selenite (Se +4) when applied as a foliar (Broadley et al., 2006; Boldrin et al., 2013; Mao et al., 2014; Ros et al., 2016; Xiong et al., 2018). A recent review found selenate to be 33 times more effective overall than selenite (Ros et al., 2016). In many soils, selenite is rapidly adsorbed on clay colloids, rendering it poorly available to plants. Dry climate, low organic matter, high temperature, high soil pH, and aeration are likely to increase the selenate: selenite ratio in the soil and hence the availability of Se to plants (Christophersen et al., 2012).

In Finland, the use of Se (selenate) fertilisers commenced on a national scale in 1984, resulting in a fourfold increase in dietary Se intake and doubling of the plasma/serum Se concentrations of the study population. There were concerns that the addition of Se in this manner may have long-term environmental effects. In California, for example, drainage water collected from an irrigated area overlaying a high-Se shale resulted in deaths and malformations in fish and aquatic birds at the Kesterson reservoir in the 1980s (Hartikainen, 2005). A study of lake and ground water in Finland in 1992 found no differences in Se concentration in water from lakes in agricultural and nonagricultural areas. Ground water samples were variable in Se (33–260 µg/l), partly explained by different Se concentrations of bedrock and sediments, and some leaching from fertilisers (as indicated by correlations with phosphorus and nitrogen) in certain areas (Mäkelä et al., 1995). The ongoing Finnish biofortification programme demonstrates the relative safety, effectiveness, ease, and cost-efficiency of this strategy (Eurola et al., 1990; Broadley et al., 2006; Haug et al., 2007; Winkel et al.,

2012; Ros et al., 2016). This model could be applied to other low-Se countries, like Malawi (Chilimba et al., 2012).

Nevertheless, Se soil biofortification is a relatively wasteful process. The recovery of soil-applied Se in wheat grain varies from 5 to 32%, with an estimated average of about 12% (Eich-Greatorex et al., 2007; Haug et al., 2007; Broadley et al., 2010; Lyons and Cakmak, 2012; Ros et al., 2016). Selenium is a valuable, mostly non-renewable resource, which should be conserved (Haug et al., 2007; Ros et al., 2016).

Foliar application has usually been found to be more efficient than soil application for Se (Ylaranta, 1984; Mao et al., 2014; Winkel et al., 2015; Ros et al., 2016; Gupta and Gupta, 2017). Foliar application not only obviates the soil factors that can reduce the effectiveness of soil agronomic biofortification of Se, but also reduces possible environmental Se accumulation (see above) as less Se is applied per hectare. Timing of foliar Se and I application is important, with the best effect usually obtained between booting and early milk stage, with heading, when green leaf cover is maximised, the "best bet" for an effective single application. A recent meta-analysis enables estimation of the amount of selenate needed to increase grain Se from 7 to 100 µg/kg: 30–60 g/ha soil-applied selenate, and 4.5–10 g/ha foliar selenate. This study found foliar-applied Se to be on average eight times more efficient than soil-applied Se (Ros et al., 2016). In tropical/sub-tropical countries where Se fertilisers are unavailable, leaves of the Drumstick tree (Moringa oleifera), which has exceptional ability to take up and accumulate Se, can provide useful levels of Se, even when grown on soils that provide little Se to most other plants (Lyons et al., 2015).

### IODINE

### Iodised Salt Needs Help to Fix Global Iodine Insufficiency

Iodine is essential to humans, being required for synthesis of thyroid hormones, which are essential for human development and health. Requirement is in the range 90–250 µg/day. Inadequate I is one of the major micronutrient deficiencies, leading to a range of clinical and social issues known as IDD. The classic symptom of I deficiency is an enlarged thyroid, known as goitre (Zimmermann et al., 2008). The safe upper limit of I intake is estimated at 1000–1100 µg/day; chronic intakes above this level can increase risk of Graves disease (Surks et al., 2004; Leung et al., 2015). Like Se, plant-available I is unevenly distributed (**Figure 1**): the sea is an important source, hence I in food systems usually declines with distance from it. Inland, high rainfall, mountainous areas are notoriously deficient in I. The overall global average soil I concentration is 2.6 mg/kg (Hetzel, 1989; Watts et al., 2010), but I concentration in plants grown on I-deficient soils may be as low as 10 µg/kg, compared with 1 mg/kg in plants on an I-replete soil (Hetzel and Pandav, 1994).

Although the number of countries designated as I deficient halved in the decade to 2014 (Gonzali et al., 2017), I deficiency remains prevalent, affecting an estimated 33% of humanity (Fuge and Johnson, 2015). Marginal I status is even present in developed countries, including England, Germany, Italy, and Australia (Andersson et al., 2012). It is apparent that iodisation of salt is insufficient to ensure overall I adequacy. Contributing factors include lack of availability of iodised salt for all households, food manufacturers not using iodised salt, volatilisation of I during food transport, storage, and cooking (on average, 20% of I in iodised salt is lost during cooking), and in many countries salt consumption has declined due to public health measures to reduce hypertension (Winger et al., 2008; White and Broadley, 2009; Comandini et al., 2013; Medrano-Macias et al., 2016; Smolen et al., 2016a; Cakmak et al., 2017). Most terrestrial foods are low in I. Strategies complementary to the iodised salt programme are needed, such as production of I-rich plants (White and Broadley, 2009; Comandini et al., 2013; Medrano-Macias et al., 2016; Smolen et al., 2016a; Cakmak et al., 2017). Vegetables biofortified with foliar I showed a high I stability during cooking (Comandini et al., 2013). Genetic approaches may be productive, for example metabolic engineering to reduce the problem of I volatilisation (Gonzali et al., 2017).

### Agronomic Biofortification: Foliar Iodate More Effective, but Easier to Biofortify Leaves Than Fruits, Roots, Grains, and Seeds

To address I insufficiency, researchers have urged the WHO to move beyond an iodised salt focus to a broader food system strategy that includes I biofortification of a range of vegetables (Smolen et al., 2016a). A case study of introducing I via agriculture was a spectacular success in Xinjiang province in north-west China. Potassium iodate was dripped into irrigation canals and resulted in a threefold increase in soil I levels, a twofold increase in I in wheat straw, increases in animal and poultry production, and in humans a 50% reduction in infant mortality and virtual elimination of IDD. Benefits were evident up to 7 years later (Jiang et al., 1997; Duxbury et al., 2015).

Most studies have shown that iodate is more suitable than iodide for biofortification (Mackowiak and Grossl, 1999; Dai et al., 2006 Lawson et al., 2015; Medrano-Macias et al., 2016; Smolen et al., 2016b; Cakmak et al., 2017). Iodate is also more likely than iodide to promote plant growth and less likely to be phytotoxic (Borst-Pauwels, 1961; Blasco et al., 2008). Iodide is more available than iodate in solution culture, while under field conditions it is more subject to cumulative losses (Lawson et al., 2015).

Iodine in plants, unlike Se, is transported mostly (but not entirely: see below) in xylem tissue (Mackowiak and Grossl, 1999), hence it is relatively easy to biofortify leaves, and thus leafy vegetables such as cabbage, lettuce, spinach (Smolen et al., 2014, 2016a,b). It is more difficult to increase I levels in grain or storage roots/tubers (Mackowiak and Grossl, 1999; Hurtevent et al., 2013; Mao et al., 2014; Medrano-Macias et al., 2016; Gonzali et al., 2017). Hence there are more published articles to date on

I biofortification of vegetables than for cereals. These vegetable articles provide valuable knowledge of I behaviour in plants that can be applied to cereals.

## Evidence for Phloem Mobility Supports Iodine Biofortification for Cereals

A comprehensive study that included glasshouse and field trials of cereals (wheat, rice, maize) in Pakistan, Brazil, Thailand, and Turkey, showed that foliar-applied I can increase grain I (Cakmak et al., 2017). For example, in a pot trial, wheat grain I was increased from 21 to 296 µg/kg using two applications (at heading and early milk stage) of potassium iodate (0.065%) plus a non-ionic surfactant (0.05%) and potassium nitrate (1%). The surfactant and potassium nitrate had an additive effect in enhancing I biofortification. In a field trial in Brazil, potassium iodate (0.05%) applied twice increased grain I from 8 to 485 µg/kg. Other studies also found that surfactants increased the efficiency of foliar micronutrient biofortification (Lawson et al., 2015; Gonzali et al., 2017). Foliar I biofortification was most effective for wheat, followed by rice, then maize (Cakmak et al., 2017).

The study of Cakmak et al. (2017) adds to recent evidence of phloem mobility of I in wheat (Hurtevent et al., 2013) and vegetables (Kiferle et al., 2013; Smolen et al., 2014; Li et al., 2017). The mechanism of potassium nitrate's enhancement of leaf absorption and possibly translocation to grain of I may relate to the chemical similarity of nitrate and iodate and is worthy of investigation (Cakmak et al., 2017).

### BIOFORTIFICATION OF CEREALS WITH SELENIUM AND IODINE COULD REDUCE IODINE DEFICIENCY DISORDERS

### Combined Selenium and Iodine Foliar Biofortification: A Promising Strategy for Many Areas

In the extensive parts of Sub Saharan Africa, China, South America, Europe, and New Zealand with concurrent Se and I deficiencies (**Figure 1**) (Hetzel and Pandav, 1994; Oldfield, 1999; Gashu et al., 2016), foliar agronomic biofortification with both Se and I may be effective in increasing the supply of both micronutrients in food systems (Smolen et al., 2016a). Resulting health benefits would be likely to include reduced incidence and prevalence of hypothyroidism with its consequent spectrum of IDD and myxoedematous cretinism.

The suitability of foliar Se application for cereal grain biofortification, irrespective of soil type, was discussed above, while the findings of Cakmak et al. (2017) for foliar I biofortification of cereals are promising. Given the observed enhancement of I biofortification provided by potassium nitrate, trials to assess its effect on Se foliar biofortification may also be fruitful.

In view of the optimum molar ratio of I:Se, which is in the range of 4.4–8.8:1 (with an average around 6) in the human diet, calculated from the RDIs of 150–250 µg/day for I and 55– 65 µg/day for Se (Smolen et al., 2016a), plausible target levels of I and Se in cereal grain could be 1.0 and 0.25 mg/kg, respectively. Toxic effects can be expected at chronic Se intakes in livestock feed/human food that exceed 1 mg/kg (Hartikainen, 2005). There is an agreeable symmetry in a joint biofortification concept for Se and I, their importance for the thyroid notwithstanding, given their juxtaposition on the Periodic Table.

### Could Se+I Foliar Biofortification of Cereals Be Attractive to Farmers?

For agronomic biofortification to become commercial, it needs to benefit both producers and consumers (Bouis and Welch, 2010; Cakmak et al., 2010). Cereal yield is unlikely to be increased by Se and/or I application (Lyons and Cakmak, 2012; Ros et al., 2016), therefore fertiliser containing Se and I may need to be subsidised (Chilimba et al., 2012), or a biofortified product could attract a premium price as a desirable functional food.

Although considered to be non-essential to plants, Se and I can be beneficial. For example, Se addition increased biomass in mungbean (Phaseolus aureus) (Malik et al., 2010) and turnip (Brassica rapa var. rapa) (Xiong et al., 2018), increased seed production in canola (Brassica rapa) (Lyons et al., 2009), and improved quality and shelf-life of vegetables and fruits (Puccinelli et al., 2017). Selenate and selenite at low doses increased growth and sulphur accumulation in wheat seedlings, but these effects were not seen in grain (Boldrin et al., 2016). Iodine use in agriculture has been reviewed by Medrano-Macias et al. (2016). Iodine is involved in various plant physiological and biochemical processes (Gonzali et al., 2017). Benefits include growth enhancement, increased nitrogen uptake, increased sugars and amino acids, improved seed viability, and increased tolerance to salinity and heavy metals via induction of antioxidants including ascorbate, glutathione, and superoxide dismutase (Borst-Pauwels, 1961; Medrano-Macias et al., 2016; Gonzali et al., 2017).

Potential benefits from applying Se and I, including increased growth and product quality, together with the convenience and economy of combining them with strategic fertiliser, fungicide and insecticide applications, could make Se+I biofortification commercially viable for farmers.

### Further Research Needed

Research needed on combined Se+I biofortification includes evaluation of potential enhancers, including salicylic acid, a phytohormone-like compound, which improved tomato fruit biofortification with I (Smolen et al., 2015), the pineal gland hormone melatonin, which is also present in plants and can act as a synergist with antifungal agents (Zhang et al., 2017), the carrier dimethyl sulfoxide, which increased the effectiveness of foliar Fe application (Leonard, 2006) and synergists such as potassium nitrate (Cakmak et al., 2017). Trials of co-application with fungicides and insecticides are also recommended, due to promising earlier findings (Mahmoud et al., 1996; Costa et al., 2003; Zhang et al., 2003; Hanson et al., 2004). Knowledge of I transporters in plants is incomplete (White and Broadley, 2009; Gonzali et al., 2017), and such research could include I volatilisation studies (Gonzali et al., 2017). An efficient single application of Se+I will be more acceptable to farmers than multiple applications.

Nanotechnologies in agriculture are attracting interest (De Rosa et al., 2010; Liu and Lal, 2015). Bioavailable biogenic elemental Se (BioSe), for example, is widespread in the microbial environment (Winkel et al., 2012). For roles in foliar biofortification, Se and I nanoparticles need to be well characterised, including particle size: stomatal openings are about 20 nm in diameter, thus movement of particles larger than this is problematic (Alshaal and El-Ramady, 2017).

More bioavailability studies that examine losses of Se and I from biofortified cereals during milling and during various cooking methods are also required, along with speciation of I in biofortified cereals.

### REFERENCES


### AUTHOR CONTRIBUTIONS

GL researched, wrote, and checked the manuscript.

### ACKNOWLEDGMENTS

This article was written while the author was working on the project Field Testing of Sodicity- and Salinity-Tolerant Oat Varieties, supported by the South Australian Grain Industry Trust Fund (SAGIT). The author and colleagues' earlier agronomic biofortification field trials using selenium and iodine were supported by HarvestPlus, The Grains Research and Development Corporation (Australia), The University of Adelaide, SAGIT, Northwest A&F University (Shaanxi, China), and the International Centre for Tropical Agriculture (CIAT, Cali, Colombia). This article is dedicated to Drs. Robin Graham, Ross Welch, and Howarth Bouis, the founders of HarvestPlus. Dr. Bouis was awarded jointly the 2016 World Food Prize.

through fertilizer strategy. Plant Soil 418, 319–335. doi: 10.1007/s11104-017- 3295-9


fpls-09-00730 June 6, 2018 Time: 16:18 # 6


and K. Moran (Norcross, GA: International Plant Nutrition Institute), 97–122.


fpls-09-00730 June 6, 2018 Time: 16:18 # 7


**Conflict of Interest Statement:** The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Lyons. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fpls-09-00730 June 6, 2018 Time: 16:18 # 8

# Should Heavy Metals Be Monitored in Foods Derived From Soils Fertilized With Animal Waste?

Rafael da Rosa Couto<sup>1</sup> \*, Jucinei J. Comin<sup>2</sup> , Monique Souza<sup>2</sup> , Felipe K. Ricachenevsky 3,4 , Marcos A. Lana5,6, Luciano C. Gatiboni <sup>7</sup> , Carlos A. Ceretta<sup>8</sup> and Gustavo Brunetto<sup>8</sup>

<sup>1</sup> Técnico em Agroecologia, Instituto Federal Catarinense, Rio do Sul, Brazil, <sup>2</sup> Agroecossistemas, Centro de Ciências Agrárias, Universidade Federal de Santa Catarina, Florianópolis, Brazil, <sup>3</sup> Departamento de Biologia, Agrobiologia, Universidade Federal de Santa Maria, Santa Maria, Brazil, <sup>4</sup> Biologia Celular e Molecular, Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil, <sup>5</sup> Department of Crop Production Ecology, Swedish University of Agricultural Sciences, Uppsala, Sweden, <sup>6</sup> SUSLAND, Leibniz Centre for Agricultural Landscape Research, Müncheberg, Germany, <sup>7</sup> Ciência do Solo, Centro de Ciências Agroveterinárias, Universidade do Estado de Santa Catarina, Lages, Brazil, <sup>8</sup> Ciência do Solo, Departamento de Solos, Universidade Federal de Santa Maria, Santa Maria, Brazil

Keywords: heavy metals, animal waste, soil fertilization, residue management, seed contamination

#### Edited by:

Jose M. Garcia-Mina, Universidad de Navarra, Spain

#### Reviewed by:

Heitor Cantarella, Instituto Agronômico de Campinas (IAC), Brazil Lourdes Hernandez-Apaolaza, Universidad Autonoma de Madrid, Spain

\*Correspondence:

Rafael da Rosa Couto rrccouto@hotmail.com

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 21 February 2018 Accepted: 15 May 2018 Published: 05 June 2018

#### Citation:

da Rosa Couto R, Comin JJ, Souza M, Ricachenevsky FK, Lana MA, Gatiboni LC, Ceretta CA and Brunetto G (2018) Should Heavy Metals Be Monitored in Foods Derived From Soils Fertilized With Animal Waste? Front. Plant Sci. 9:732. doi: 10.3389/fpls.2018.00732 Heavy metals (HM) represent a large group of elements with atomic density >5 g cm−<sup>3</sup> or atomic number >20 (Saidur et al., 2017), among which some are essential to plants, such as iron (Fe), zinc (Zn), copper (Cu), nickel (Ni), and manganese (Mn). However, HMs may be contaminants and/or pollutants, depending on the concentration in soils.

HMs such as Cu, Zn, Ni, and chromium (Cr) are essential to human beings, and biofortification approaches to improve levels of some elements in plant edible parts are underway (Bouis et al., 2012; Ricachenevsky et al., 2015). However, these HMs may be toxic if accumulated, and may only be ingested in very small quantities (EPA-U.S. Environmental Protection Agency, 1995; FAO-Food Agriculture Organization of the United Nations, 1995; Tchounwou et al., 2012). On the other hand, Pb, Cd, As, and Br are not essential and can be toxic even at low concentrations (Tchounwou et al., 2012). The safe daily intake level for As, Cd, Cr, Cu Ni, Pb, and Zn is 20, 300, 1500, 4, 20, 40, 300 µg kg−<sup>1</sup> of body weight per day, respectively (EPA-U.S. Environmental Protection Agency, 1993). These levels are based on the degree to which the element may cause disturbance, the capacity of the body to accumulate the element and the weight of the individual who is ingesting it (Abbasi et al., 2013). However, when HMs are ingested for long periods, even at doses considered safe, they can cause harmful effects, known as chronic intoxication (Jorge Mendoza et al., 2017; Li et al., 2017).

The increase in total HM concentration and their chemical forms in soils can occur naturally due to atmospheric deposition, weathering of rocks, and anthropic activities such as mining, deposition of ash from coal burning, application of pesticides in plants, addition of mineral and organic fertilizers, among others (Guilherme et al., 2005). HM accumulation in the soil is typically assessed by indicators such as Geo-accumulation index (Igeo) (Equation 1) (Müller, 1979) and Enrichment Factor (EF) (Equation 2) (Abbasi et al., 2013) that allow the identification of the presence and the intensity of deposition of anthropogenic contaminants in topsoil.

$$Age = \log\_2\left(\frac{\lfloor \frac{\lfloor C\_{\rm u} \rfloor}{1.5} \* \lfloor B\_{\rm u} \rfloor}{}\_{\ast}\right) \tag{1}$$

where: Cn is the measured concentration in the soil for the metal n, Bn is the background value for the metal n, and the factor 1.5 is used because of possible variations of the background data due to lithological variations.

$$EF = \frac{\left[\frac{metal}{RE}\right]sample}{\left[\frac{metal}{RE}\right]control} \tag{2}$$

where: RE is the value of metal, adopted as Reference Element.

High HM concentrations in soils may cause intoxication upon inhalation, contact with the skin, indirect ingestion of soil and intake of fruits, vegetables, grains, and their byproducts (Zheng et al., 2010; Chabukdhara and Nema, 2013; Chen et al., 2016; Jiang et al., 2017). Plants grown in soils contaminated/polluted with HM tend to absorb, accumulate, transport, and redistribute larger amounts of HM. This is likely due to the presence of nonselective essential element transporters. For instance, iron high affinity transporter IRT1 of the model plant Arabidopsis thaliana, which is necessary for Fe acquisition under iron deficiency, is known to also transport Zn, Mn, Ni, Co, and Cd, possibly leading to metal toxicity under Fe deficiency (Korshunova et al., 1999; Barberon et al., 2014; Ricachenevsky et al., 2018). In rice, IRT1 might also transport Zn and Cd (Lee and An, 2009). Arsenic uptake is also performed by phosphate transporters (as arsenate) or by silicon transporters (as arsenite), which are not able to distinguish between these elements (Kochian et al., 2015). Thus, non-selective transport leads to accumulation of toxic elements, which might end up accumulating in grains or other harvested parts, and may change nutrient abundance and distribution (Punshon et al., 2018). These agricultural products containing high HM concentration might then be used for human consumption directly or indirectly through the intake of processed foods (Hariri et al., 2015; Avkopashvili et al., 2017).

To assess the risk of ingestion of a particular HM over the life of an individual, it is necessary to consider the period of ingestion. Therefore, indexes have been established to verify the risk that certain elements, such as HMs, could cause to human beings (Abbasi et al., 2013). A few examples of these indexes are the Health Risk Index (HRI), Target Hazard Quotient (THQ) and Target Cancer Risk (TCR) (Equation 5) (EPA-U.S. Environmental Protection Agency, 2010).

$$HRI = \frac{(C\_n \ltimes D\_n)}{\left(R \lhd D \lhd B \, W\right)}\tag{3}$$

where: Cn, total concentration of the metal in edible plant organ (mg kg−<sup>1</sup> ); Dn, daily intake (g day−<sup>1</sup> ); BW, average body weight (kg); RfD, reference dose (EPA-U.S. Environmental Protection Agency, 2010).

$$HQ = \frac{\left(C\_n \times D\_n \times 10^{-3} \,\mathrm{\chi EF\_r \times ED\_{tot}}\right)}{RfD \mathrm{xB}W\_a \mathrm{xAT\_n}}\tag{4}$$

where: EF<sup>r</sup> , exposure frequency (days); EDtot, exposure duration (years); ATn, average exposure time to non-carcinogenic heavy metals (e.g., EDtot x 365 days/year).

$$TCR = \frac{\left(C\_n \varkappa D\_n \varkappa 10^{-3} \text{\AA} \text{\textdegree C} \text{\textdegree S}\_0 \text{\textdegree E} F\_r \text{\textdegree} D\_{tot}\right)}{\left(BW\_a \varkappa A T\_n\right)} \tag{5}$$

where: CPS0, carcinogenic potential (µg g−<sup>1</sup> day−<sup>1</sup> ).

The effects of HM accumulation in soil, excess uptake by plants, and the risks that HM-contaminated foods can promote to human beings are commonly reported in mining regions (Qing et al., 2015; de Souza et al., 2017; Li et al., 2017). As example, the release and drifting of dust from coal mines in the Qingshui River basin (China) has resulted in pollution of arable soils. Despite the knowledge associated to the deposition of HMs, few studies approach the increase of HM concentration in different edible plant organs cultivated on soils subjected to a long history of animal waste application.

Different environmental agencies have established acceptable levels of HM in food. FAO and EPA-USA established maximum levels for Cu, Zn, Cd, Pb, Cr, and Ni in crop grains of 20, 50, 0.1, 0.2, 1, and 0.04, respectively. However, studies on soils subjected to the addition of urban sludge and animal residues reported increased HM concentration above these limits in grains, fruits, and vegetables (Suarez-Tapia et al., 2017; Zhang et al., 2017). The use of wastewater for irrigation in Iran containing 0.06, 0.010, 0.01, 0.010, and 0.010 mg kg−<sup>1</sup> of Cu, Zn, Cd, Pb, Cr, and Ni, respectively, caused the accumulation of Cd, Cr, and Pb in wheat and corn grains above the limits established by the EPA. Health risks to adults and especially children by Cu, Cd, and Cr intake in corn and wheat grains were also reported (Asgari and Cornelis, 2015). Animal waste contains HM derived from drugs or feed (Gunkel-Grillon et al., 2015; Couto et al., 2016).

It is worth mentioning that soils with frequent application of organic wastes typically have higher HM concentrations than those described in studies where negative effects of excess HM on edible plant organs and human health risk have been reported, indicating that we might be underestimating the contamination of foods derived from such areas **(Table 1)**. Studies that address the effects of increasing HM concentration in soils subjected to long-term animal waste application and consequent changes of HM concentration in edible plant organs are still scarce. Although organic fertilization recommendations exist both for conventional and organic production systems, the application of organic residues is often carried out indiscriminately in regard to HMs content, increasing their concentration in soils and likely increasing of HM concentration in edible plant organs (Couto et al., 2016; Suarez-Tapia et al., 2017; Zhang et al., 2017). A very important aspect is that in organic production systems, organic residues (including animal manure) are the main—if not the only—source of nutrients for the crops. Considering that organic production systems currently occupy 42 million hectares worldwide (FIBL, 2017), with a global growth rate of 4.5% per year, the risk of HM contamination in the food systems is present, especially in some regions of the world. In Brazil, Japan, and the European Union, the growth rate of the area cultivated under the organic system is 30, 13, and 8% per year, respectively (FIBL, 2017). This emergent risk indicates that the organic residues that will be used as source of nutrients for the crops needs to be assessed in terms of HM concentration.

In Brazil, the applications of pig slurry, cattle slurry, and pig deep litter for 10 years in sandy soil with low organic matter content under no-till increased Ni, Cu, and Zn concentrations in shoots and grains of corn and wheat (da Rosa Couto et al., 2018).


TABLE 1 | Heavy metal contents in soils and vegetables

 of diverse areas of the world and risks to human health.

The applications of organic wastes (pig slurry, cattle slurry, and pig deep litter) and mineral fertilizers also increased the values of HRI and THQ for Br and Zn, presenting health risks to adults and especially children who have lower body weight (da Rosa Couto et al., 2018). They also report that Cu concentrations in corn grains of plants grown in soil with application of pig deep litter and cattle slurry were 2.7 and 2.2 mg Cu kg−<sup>1</sup> , respectively. On the other hand, Zn concentrations in corn grains of plants grown in soil with application of pig deep litter, pig slurry, and cattle slurry were 26, 31, and 23 mg Zn kg−<sup>1</sup> , respectively. In the grains of wheat grown in soil with the application of pig deep litter, pig slurry, and cattle slurry, concentrations of Cu were 6.0, 6.0, and 4.5 mg kg−<sup>1</sup> , respectively, and Zn were 96, 95, and 84 mg kg−<sup>1</sup> , respectively. Thus, Cu and Zn concentrations in grains of corn and wheat grown in soil with a long history of application of organic wastes were higher than those found in plants grown in the control soil or even with the application of mineral fertilizer. This justifies the monitoring of concentrations of elements in grains of plants grown in soils with a long history of organic waste application, especially in soils with low capacity for adsorption of elements, such as sandy soils with low organic matter (Brunetto et al., 2014).

Plants have the potential to absorb and accumulate larger amounts of several heavy metals. In studying heavy metal contents in vegetables fertilized with wastewater in India, Singh et al. (2010) found that the concentrations of Cd in plants varied from >2 to 15 mg kg−<sup>1</sup> , while Pb ranged from >1 to 28 mg kg−<sup>1</sup> and Ni from >1 to 41 mg kg−<sup>1</sup> . The authors verified a risk to consumer health (HRI> 1) through the ingestion of Cd accumulated in radish, cabbage, cauliflower, okra, eggplant wheat and rice; of Pb accumulated in palak, cabbage, cauliflower, Lady's fingers, brinjal, wheat, and rice; and Ni accumulated in palak, cauliflower, wheat, and rice. However, it is important to observe the proportion of vegetables and cereals in the diet, which may change according to the culture of each place and country, causing higher or lower risk.

### REFERENCES


### FINAL CONSIDERATIONS

The use of waste as a source of nutrients in plant production systems, traditional, and organic, is common worldwide, and important strategy for nutrient cycling. However, longterm application of such waste causes the increase of HM concentrations in soils, increasing HM uptake by plants and assimilation in edible organs such as grains, as indicated by the data presented in **Table 1**. As different plant organs can be used in the preparation of numerous products for human consumption, it is necessary to monitor the concentrations of HM in edible plant organs of different species and cultivars fertilized with organic waste. This monitoring can be done through indexes such as Igeo, EF, HRI, THQ, TCR, allowing us to estimate the possible dangers of HMs to the health of children, young adults, and adults who eat food derived from plants grown in soils with a history of animal waste application. Thus, we recommend careful consideration of practices that indiscriminately use animal waste in plant production to avoid HM accumulation and health hazards to consumers. Moreover, strongly indicate that evaluation of metal contamination in foods derived from plants cultivated using animal waste should be commonplace, and further studies of how widespread that is should be conducted by the scientific community.

### AUTHOR CONTRIBUTIONS

RdRC: wrote the first draft of the manuscript. MS: organized the database. JC, LG, CC: senior researchers in the field of soil science. They made specific contributions of the area. FR: senior researchers in the field of plant physiology. They made specific contributions of the area. ML: senior researchers in the field of agroecology. They made specific contributions of the area. All authors contributed to the revision of the manuscript, read and approved the version sent.

to Improve Human Health: A Scientific Review, eds T. W. Bruulsema, P. Heffer, R. M. Welch, I. Cakmak, K. Moran (Norcross, GA: IPNI), 97–122.


health from simultaneous exposure to multiple contaminants in an artisanal gold mine in Serra Pelada, Pará, Brazil. Sci. Tot. Environ. 576, 683–695. doi: 10.1016/j.scitotenv.2016.10.133


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 da Rosa Couto, Comin, Souza, Ricachenevsky, Lana, Gatiboni, Ceretta and Brunetto. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Dynamic Modeling of Silicon Bioavailability, Uptake, Transport, and Accumulation: Applicability in Improving the Nutritional Quality of Tomato

Mari C. López-Pérez <sup>1</sup> , Fabián Pérez-Labrada<sup>1</sup> , Lino J. Ramírez-Pérez <sup>1</sup> , Antonio Juárez-Maldonado<sup>2</sup> , América B. Morales-Díaz <sup>3</sup> , Susana González-Morales <sup>4</sup> , Luis R. García-Dávila<sup>5</sup> , Jesús García-Mata<sup>6</sup> and Adalberto Benavides-Mendoza<sup>1</sup> \*

<sup>1</sup> Departamento de Horticultura, Universidad Autónoma Agraria Antonio Narro, Saltillo, Mexico, <sup>2</sup> Departamento de Botánica, Universidad Autónoma Agraria Antonio Narro, Saltillo, Mexico, <sup>3</sup> Robótica y Manufactura Avanzada, Centro de Investigación y de Estudios Avanzados Unidad Saltillo, Ramos Arizpe, Mexico, <sup>4</sup> Departamento de Horticultura, CONACYT-Universidad Autónoma Agraria Antonio Narro, Saltillo, Mexico, <sup>5</sup> Cosmocel España, Zaragoza, Spain, <sup>6</sup> Cosmocel Brasil, São Paulo, Brazil

#### Edited by:

Raul Antonio Sperotto, University of Taquari Valley, Brazil

#### Reviewed by:

Sergio Esposito, University of Naples Federico II, Italy MCarmen Martinez-Ballesta, Consejo Superior de Investigaciones Científicas (CSIC), Spain

#### \*Correspondence:

Adalberto Benavides-Mendoza abenmen@gmail.com

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 20 January 2018 Accepted: 27 April 2018 Published: 17 May 2018

#### Citation:

López-Pérez MC, Pérez-Labrada F, Ramírez-Pérez LJ, Juárez-Maldonado A, Morales-Díaz AB, González-Morales S, García-Dávila LR, García-Mata J and Benavides-Mendoza A (2018) Dynamic Modeling of Silicon Bioavailability, Uptake, Transport, and Accumulation: Applicability in Improving the Nutritional Quality of Tomato. Front. Plant Sci. 9:647. doi: 10.3389/fpls.2018.00647 Silicon is an essential nutrient for humans, additionally is beneficial for terrestrial plants. In plants Si enhances tolerance to different types of stress; in humans, it improves the metabolism and increases the strength of skeletal and connective tissues as well as of the immune system. Most of the Si intake of humans come from edible plants creating a double benefit: first, because the absorption of Si increases the antioxidants and other phytochemicals in plants, thereby increasing its functional value, and second because the higher concentration of Si in plants increases intake in human consumers. Therefore, it is desirable to raise the availability of Si in the human diet through the agronomic management of Si accumulator species, such as corn, wheat, rice, soybeans, and beans. But also in such species as tomatoes, carrots, and other vegetables, whose per capita consumption has increased. However, there are few systematized recommendations for the application and management of Si fertilizers based on the physicochemical factors that determine their availability, absorption, transport, and deposition in cells and tissues. This study presents updated information about edaphic and plant factors, which determine the absorption, transport, and deposition rates in edible organs. The information was integrated into an estimated dynamic model that approximates the processes previously mentioned in a model that represents a tomato crop in soil and soilless conditions. In the model, on the other hand, was integrated the available information about key environmental factors related to Si absorption and mobilization, such as the temperature, pH, and soil organic matter. The output data of the model were compared against information collected in the literature, finding an adequate adjustment. The use of the model for educational or technical purposes, including the possibility of extending it to other horticultural crops, can increase the understanding of the agronomic management of Si in plants.

Keywords: silicates, nutritional quality, stress tolerance, mathematical models, silicon and health

## INTRODUCTION

On average a human organism contains 1–2 g of Si, being the third most abundant trace element after Fe and Zn. When it is contained in food in adequate quantity, silicon is effectively absorbed by the human organism (Sripanyakorn et al., 2009), transferring to practically all tissues, but concentrating in greater quantity in the connective tissues (O'Dell and Sunde, 1997; Jugdaohsingh, 2007). With a diet rich in vegetables the daily intake of silicon is between 140 and 204 mg Si day−<sup>1</sup> ; however, in western populations with lower consumption of vegetables, the daily intake can range between 20 and 50 mg day−<sup>1</sup> . Silicon is rarely toxic when taken orally (Arora and Arora, 2017), with a recommended maximum intake of 1,500 mg day−<sup>1</sup> (White and Broadley, 2005). On the other hand, the minimum value of Si consumption to achieve some benefits has been determined at 25 mg day−<sup>1</sup> (Nielsen, 2014). After ingestion, most of the absorbed Si is excreted in the urine (Jugdaohsingh, 2007), most likely as orthosilicic acid and/or magnesium orthosilicate.

Plant foods are the primary source of Si in the human diet. This includes grains of cereals (rice, wheat, oats, and barley) and less refined products of cereals, fruits (bananas and apples), vegetables (potato, beet, carrot, bean, spinach, and lentils) (Powell et al., 2005), and beverages such as beer since the Si contained in barley and hops is solubilized during the manufacturing process (Pennington, 1991; Powell et al., 2005; Jugdaohsingh, 2007). Other sources of silicon are meat, fish, milk, and eggs (Nielsen, 1974; Nuurtamo et al., 1980). Drinking water can also be a source of Si depending on the source and the method of processing (Jugdaohsingh, 2007; Sripanyakorn et al., 2009).

In plants Si is not considered an essential element, but it has been found that its inclusion in fertilizer formulations provides higher tolerance to stress (Adrees et al., 2015; Rizwan et al., 2015; Cooke and Leishman, 2016; Luyckx et al., 2017), especially on soilless growing conditions (Epstein, 1994, 2009; Voogt and Sonneveld, 2001). An additional benefit of adding silicon in the fertilization of crops is related to the more significant amount of silicon available to human consumers. In other words, the use of silicon in agricultural production brings a benefit to agricultural producers in the form of stronger and stress-tolerant plants, while for consumers of harvested products it gives an advantage in the way of higher silicon intake in the food.

The use of mathematical models in the mineral nutrition of plants allows to simulate the dynamics of the absorption of water and dissolved ions in response to different internal and external factors (Juárez-Maldonado et al., 2017). The models contribute to the quantitative understanding of the factors involved in the absorption, transport, and accumulation of mineral elements; additionally, they allow to explore different environmental or endogenous situations that modify the nutrition of the plant (Mankin and Fynn, 1996). Regarding Si modeling, (Sakurai et al., 2017) presented a dynamic model of Si absorption and transport for rice. The model was integrated considering the activity of different transporters and the distribution of Si through different nodes in the entire plant; the model was able to predict the dynamic behavior of silicon in the plant successfully. However, in the case of vegetables, there are no models that consider silicon, although there are published models that effectively simulate nutrition with other mineral elements (Juárez-Maldonado et al., 2014b, 2017; Ramírez-Pérez et al., 2018).

The aim of this manuscript was the integration of an estimated dynamic model that approximates the availability, absorption, transport, and accumulation of silicon in a tomato crop in soil and soilless conditions.

### BENEFITS OF SILICON IN HORTICULTURAL PLANTS

The ferns, horsetails, and grasses such as corn, wheat and rice, and sugar cane, are the plants that naturally accumulate more silicon (Liang et al., 2015). However, in the presence of adequate amounts of silicon in the form of Si(OH)4, all plants, including horticultural species such as tomato and cucumber, absorb it. Plants use silicon in a manner not yet well understood to stimulate the antioxidant metabolism, the processes of plant's hardening, defense, and adaptation to environmental factors. Depending on whether they are species that carry out silicification, Si(OH)<sup>4</sup> is concentrated in polymeric form (amorphous hydrated silica) in different cellular and extracellular compartments, finally transforming it using a deposition process dependent on transpiration into insoluble biogenic silica (SiO2.nH2O) which forms structures called phytoliths or opal (Sangster et al., 2001; Katz, 2014; Exley, 2015). The biogenic silica is subsequently incorporated into the soil contributing with 1–3% of the total Si in the soil (Desplanques et al., 2006).

The silicon absorbed by the plants seems to be maintained under constant exchange between the soluble forms (Si(OH)4) and the insoluble fraction (polymeric silicic acid and biogenic silica) (Exley, 2015). Most of the Si deposited as biogenic silica remains as such throughout the life of the plant (Sangster et al., 2001). The soluble part is directly available to humans when they consume plant foods, while the insoluble fraction could perhaps be considered as an integral part of the fiber. The environmental factors and differences between plant species that modify the ratio Si soluble/Si insoluble, which ultimately determines the dietary utility of the product, have not been studied.

Considering that, (i) all plants seem to have the capacity to absorb silicon (Exley, 2015), (ii) and in view of the rise that has taken in recent years the production of vegetables using soilless production systems (Pignata et al., 2017), (iii) in addition to the fact that irrigation water and horticultural substrates provide little bioavailable silicon (Liang et al., 2015), then, it would be advisable to include silicon on a daily basis in fertilizer formulations used in those soils with low Si bioavailability as well as in soilless crops for the production of vegetables under protected conditions (Epstein, 1994, 2001).

The different groups of plants have different capacity to mobilize Si toward their various organs, but practically all absorb the silicon from the soil when it is available in the soil solution or the nutrient solution. The species with low mobilization capacity accumulate it in the roots and stems, while the species with high mobilization capacity accumulate it in stems, leaves, fruits, and seeds. Si appears to be absorbed in the form of Si(OH)<sup>4</sup> by channels belonging to the aquaporins' group. Thus the rate of absorption and transport depends on the flow of water linked to transpiration (Exley, 2015; Sakurai et al., 2017).

The cereals are plants with a high capacity of silicification and therefore represent a significant amount of Si in the diet. However, the silicon contained in cereals will be encountered almost all in the form of insoluble biogenic silica (Sangster et al., 2001) which would be partially dissolved by the acids of the digestive system; on the other hand, in fruits and vegetables, due to their lower silicification capacity, it is expected that there will be more soluble silicon, which theoretically would be more available to be assimilated during intake. Considering the above, it is possible that the fruits of horticultural species such as tomato can be excellent sources of Si for the diet.

Additionally, it is known that in comparison with the dicotyledons, cereals contribute less Ca and Mg (White and Broadley, 2005). Therefore, a diet high in cereals that provides a significant amount of Si on average will contain less Ca and Mg than a mixed diet with base in cereals and dicots. On the other hand, the consumption of vegetables and fruits has grown considerably in recent decades among the human population and it is desirable that species such as tomatoes, eggplants, strawberries, cucumbers, avocados, melons, watermelons, carrots, onions, chilies, pumpkins, among others, contain a higher amount of silicon, considering the double benefit already mentioned of the crop higher tolerance to stress and the contribution of Si to human consumers.

In soilless crops, it is necessary to consider the contribution of Si in fertilizers since irrigation water does not provide enough, only from 5 to 24 mg L−<sup>1</sup> Si (Liang et al., 2015). The lowest recommended concentration of Si in the nutrient solution for plants growing on substrates other than soil is 28 mg L−<sup>1</sup> Si (Epstein, 1994), which can be achieved with 123 mg L−<sup>1</sup> of Na2SiO3.

### EDAPHIC FACTORS THAT DETERMINE THE AVAILABILITY OF SILICON

Si is found in soil as an inert mineral in the form of quartz or aluminosilicates such as micas and feldspars. The weathering of these materials by rainwater, irrigation water, or by the acid metabolites of microorganisms and plant roots produce Si(OH)<sup>4</sup> that under a balanced condition reaches a concentration of up to 1.8 mM (173 mg L−<sup>1</sup> , equivalent to 50.4 mg L−<sup>1</sup> Si). Above this level, reaching 2 mM (192.18 mg L−<sup>1</sup> ), Si(OH)<sup>4</sup> forms hydrated amorphous silica polymers containing Si unavailable for plants (Epstein, 2001; Liang et al., 2015).

The actual value of the concentration of Si(OH)<sup>4</sup> in the soil solution is much lower than 1.8 mM, commonly found between 0.1 and 0.6 mM (9.61–57.66 mg L−<sup>1</sup> ), but with such low values as 0.02 mM (1.92 mg L−<sup>1</sup> ) in very eroded soils (Epstein, 2001; Liang et al., 2015).

The concentration of bioavailable Si in soil solution results from the release rate of Si(OH)4. Bioavailability is dependent on the silicon content of the soil minerals, organic matter, the temperature, the amount of precipitation and the acidity of the soil or soil pore water. The incorporation of Si in plants occurs at a rate dependent on the intensity of the transpiration (Exley, 2015), so that conditions of rapid growth can rapidly decrease the availability of Si in the soil solution (Epstein, 2001; Liang et al., 2015).

Soils of tropical areas where high precipitation occurs as well as calcareous and sandy soils of semi-arid and arid regions with low vegetation provide low quantities of Si to the soil solution, so it is recommended to use fertilizers with Si (Epstein, 1999). An affordable source of Si is siliceous sand that is offered in different granulometries, and that is used in quantities of between 500 and 4,000 kg ha−<sup>1</sup> . On the other hand, Mollisol and Vertisol soils of the temperate and subtropical regions are soils that can provide an adequate amount of silicon (Epstein, 2001; Gérard et al., 2008). However, this has not been corroborated in regards of the actual availability of Si(OH)<sup>4</sup> in soil pore water, since there is little-published information about concentrations, dynamic behavior, and association with other edaphic components of Si in the solution of the soil.

The temperature exerts a substantial impact on the soil solubilization rate of Si. However, the seasonal changes in temperature are significant as a determinant of the Si concentration in the soil solution only in the cold seasons of temperate zones, because the range of temperatures suitable for the growth of a crop is also adequate for the solubilization reactions of silicon in the soil (Sommer et al., 2006). Therefore, the temperature is not considered as a factor subject to management regarding the bioavailability of Si for crop plants. Possible exceptions would be crops in soil mulching and crops grown in greenhouse soil or tunnels. In both cases, soil or substrate temperatures are more stable, and on average higher than those of uncovered soil, in addition to water management more precise in time and quantity, so the bioavailability of Si is expected to be higher.

Another factor regulating the availability of Si(OH)<sup>4</sup> is the pH of the soil pore water, that depends on the pH of the rainwater or irrigation water and is also modified by the respiratory activity and extrusion of organic acids by microorganisms and plant roots (Pérez-Labrada et al., 2016). In fact, the presence of Si induces the synthesis of citric acid in plants (Hernandez-Apaolaza, 2014). The pK1 of Si(OH)<sup>4</sup> is 9.6, which indicates that its bioavailability in a nutrient solution is practically unaltered with pH values lower than 9. In the study of Gérard et al. (2008), there was little impact of pH on the bioavailability of Si in the soil solution, but the study conditions were developed under a very narrow range of pH variation. It will be necessary to collect data in different types of soil, or in soils subjected to treatments that modify its reaction or the pH of the soil pore water, to determine the effect of pH on the concentration of Si(OH)4.

Both a nutrient solution and the soil solution contain components that modify pH and interact with Si. With a pH> 7 that promotes the formation of Fe hydroxides, an adsorption process occurs that causes the polymerization of Si(OH)4. With pH< 6 Si(OH)<sup>4</sup> begins to polymerize on surfaces with minerals containing Fe, while Al3<sup>+</sup> would promote the stabilization of Si(OH)<sup>4</sup> polymers, which would make Si unavailable for plants (Sommer et al., 2006). Considering this, it is possible that the availability of Si in the soil solution is higher with pH values between 6.0 and possibly 7.5 (maybe showing some resemblance to the pattern of bioavailability of P), as long as the soil parent material provides Si in sufficient quantity. Calcareous soils, which naturally have pH values> 8 in the soil solution (Pérez-Labrada et al., 2016), do not provide enough Si (Liang et al., 1994). Thus the fertilizer contributions with Si in crops in calcareous soils are beneficial (Zhang et al., 2017).

Another factor to consider regarding the availability of Si(OH)<sup>4</sup> in the soil pore water is soil organic matter (SOM) and its dissolved forms. SOM have a profound impact on the availability of mineral elements (Diacono and Montemurro, 2010), either directly by chemical processes or indirectly by the promotion of bacteria and fungi that solubilize Si and other elements of soil minerals (Landeweert et al., 2001). An expected effect of SOM would be the adsorption of Al3<sup>+</sup> through organic acids (Rustad and Cronan, 1995), which would decrease the Al-Si association and increase the concentration of Si(OH)<sup>4</sup> available in the soil solution. The organic acids derived from SOM are also agents that promote dissolution in mineral surfaces (Drever and Stillings, 1997) so that in soils with silicon-rich parent materials or agricultural soils with the application of Si fertilizers would be very helpful. In nutrient solutions for soilless crops, the use of organic acids can also be useful to improve the solubility of fertilizers with silicon. On the other hand, indirect evidence is available that indicates that SOM is a factor that increases the bioavailability of Si for crop plants (Ding et al., 2008; Sun et al., 2017), thereby SOM management should be considered to increase the availability of Si for crops.

### THE MODEL

The information in the previous section highlighted the factors that can be subjected to management in a crop, both in soil and soilless, with the purpose of increasing the availability of Si(OH)<sup>4</sup> for plants. In the crops grown in soil, a primary factor is the silicon content of soil's parent material. In the fertile soils of temperate and subtropical zones, Si inputs are rarely required in the fertilizers since the soil will undoubtedly provide the necessary amount. On the other hand, in the calcareous soils of arid and semi-arid regions, and in the soils of tropical regions subject to regimes of intense precipitation, the application of Si with fertilizers will be necessary, but also the consideration of the pH and organic matter management of the soil to ensure adequate availability of Si(OH)4.

In soilless crops, the critical factor to consider will be the concentration of Si in the irrigation water. Values below 28 mg L −1 Si point out the need to provide Si up to a maximum of 50 mg L −1 . The management of pH is the next factor to be considered. However, the data presented indicate that pH management aimed at ensuring the bioavailability of P in nutritive solution (that is, maintaining it between 5.5 and 7) will be adequate.

A Matlab-Simulink model (the archives are included in Supplementary Material) is presented below which allows verifying the impact of different environmental scenarios, both in a soil crop and in a soilless crop, using as a model tomato plants. There is also an example of the use of the software to obtain the estimated impact of the environmental variables on the absorption of Si by the tomato plants. The data presented in the previous parts of the manuscript can be tested in this model by verifying the result regarding the concentration of Si in the plants. The purposes of the use of the model are educational or technical, and from our perspective, the model can be useful in the agronomic management of Si in a tomato crop and, possibly applicable to other horticultural crops.

### Description of the Model

Tomato (Solanum lycopersicum L.) was used as a biological model to describe the distribution of silicon accumulation in the different organs. To describe the effects of the various environmental factors mentioned, the deterministic mathematical model proposed by Tap (2000) and modified by Juárez-Maldonado et al. (2014b) will be used as a basis.

The model consists of six state equations, using as inputs the radiation (PAR, µmol m−<sup>2</sup> s −1 ), temperature (◦C), and CO<sup>2</sup> concentration (µL L−<sup>1</sup> ). The model allows to directly considering the effect of these three variables on the accumulation of silicon in the different organs of the tomato plant.

The scope of the model refers to environmental conditions where intense stress does not prevail since it is assumed that the growth rate of the plants will be a direct function of the irradiance and temperature.

Within the plant, silicon accumulates in different organs depending on the corresponding transpiration rates. Thus, it is necessary to calculate the transpiration by a tomato plant dynamically. For this, the equation 1 is used, based on the fact that a linear correlation can be considered between the biomass accumulated by the tomato plant and its transpiration (Juárez-Maldonado et al., 2014a).

$$Transipation = Biomass \ast plm/t\text{cg} \tag{1}$$

Biomass is the mass of the tomato plant in g m−<sup>2</sup> ; plm is a parameter of the linear model (8.5714); and tcg is the time of crop growth (10279801 s).

Assuming that the tomato plant does not present a substantial accumulation of Si (Liang et al., 2015), the maximum absorption limit was set for the model at 1% (as SiO2) of the dry biomass (Miyake and Takahashi, 1978). In this condition, and while there is an unlimited availability of silicon in the soil solution, the accumulation of maximum total silicon (MSiT) in the plant would be as follows:

$$\text{MSiT} = \frac{Biomass}{100} \ast PD \tag{2}$$

Where PD is the planting density expressed in plants m−<sup>2</sup> , which for this model was established in 3 plants m−<sup>2</sup> . This plant density was used by Juárez-Maldonado et al. (2014b) and provide the best financial margin, high yield, and fruit quality (Peet and Welles, 2005).

The distribution of accumulated silicon in the tomato plant will then follow the different transpiration rates of its organs, that is, leaves> stem> fruits ≥ root. In the particular case of tomato, organ transpiration can be approximated to the following percentages of total transpiration: leaves = 90%, stem = 5%, fruits = 2.5%, and root = 2.5%.

Even though the potential availability of Si(OH)<sup>4</sup> in the soil solution is 192.18 mg L−<sup>1</sup> (Epstein, 1999; Liang et al., 2015), disponibility is affected by temperature, pH, and organic matter content of the soil. In addition to the factors that are modified with agricultural management such as soil moisture and soil profile.

According to the literature, the availability of silicon in soils is directly affected by soil temperature (Epstein, 1999; Liang et al., 2015). Although there is no clear explanation of how this behavior occurs, it is possible to approach it with a third-order model (Equation 3). This is due to the disponibility of silicon is between 8 and 35◦C, being its highest availability at 25◦ C.

$$T\mathfrak{B}\*Temp^3 + T\mathfrak{Z}\*Temp^2 + T1\*Temp + T0\tag{3}$$

Where T3, T2, T1, and T0 are the parameters of the thirdorder model (equivalent to −0.0003; 0.0127; −0.1093; and 0.1674 respectively), and Temp is the 0–30 cm soil temperature (◦C).

Concerning the SOM, it is known that there is a positive correlation with the availability of silicon (Ding et al., 2008; Sun et al., 2017). View from an agricultural perspective, soil is rich in organic matter when it has a concentration of 5%. An adjustment with a Michaelis-Menten function was used to describe the higher availability of Si, due to the effect of SOM. For this, the following Equation (4) was used.

$$Vma\overline{\times} \times \frac{OM}{(Km+OM)}\tag{4}$$

Where Vmax is the parameter of maximum availability of silicon due to organic matter normalized to 1 (Vmax = 1). OM is the amount of organic matter contained in the soil (%, w/w). And Km is the Michaelis-Menten parameter (Km = 2.5).

The pH is also a determining factor in the availability of silicon (Liang et al., 2015). This factor, as well as temperature, is related to the availability of silicon that can be approached to a thirdorder model. The availability of silicon in soil occurs in the pH range from 2 to 9, with a pH of 7 being the highest availability. Therefore, its effect can be described as follows:

$$pH\text{3} \* pH^{3} + pH\text{2} \* pH^{2} + pH\text{1} \* pH + pH\text{0}\tag{5}$$

Where pH3, pH2, pH1, and pH0 are the parameters of the thirdorder model (with values −0.0235, 0.325,−1.1563, and 1.2262, respectively). And pH represents the pH of the soil studied.

Therefore, the Si(OH)<sup>4</sup> available (SiAv) to be absorbed by the tomato plant is described by the following equation:

$$\text{SiAv} = (\text{SiP} - \text{SiWater}) \ast ETem \ast EOM \ast EpH + \text{SiWater} \quad \text{(6)}$$

Where SiP is the maximum amount of silicon in a soil without polymerization [192.18 mg L−<sup>1</sup> Si(OH)4]; ETem represents the effect of temperature on the availability of silicon (Equation 3); EOM describes the impact of organic matter on the availability of silicon (Equation 4); EpH represents the effect of pH on the availability of silicon (Equation 5); and SiWater is the amount of Si(OH)<sup>4</sup> available in the irrigation water. The model supposes that under no condition will be the available Si(OH)<sup>4</sup> be higher than the SiP value.

The accumulated Si (as SiO2) in the tomato plant (SiT) was determined with the silicon [Si(OH)4] available and the transpiration (Equation 1) in the following way:

$$\text{SiT} = \text{SiAv} \ast \text{Transjunction} \ast \text{SiSi} \tag{7}$$

Where SiSi is the fraction of cumulative Si in the plant of the total available Si(OH)<sup>4</sup> (g).

In soilless cultivation conditions, the only source of Si(OH)<sup>4</sup> for the crop will be the content of the irrigation water since there is no such source of soil mineral replacement as in the soil. Therefore, the accumulation of Si in the tomato plant in soilless culture (ASiTSC) will depend entirely on the transpiration of the plant (Equation 1) and the availability of Si(OH)<sup>4</sup> in the irrigation water (SiWater). This relationship is expressed as:

$$\text{ASiTSC} = \text{Transjunction} \ast \text{SiWater} \ast \text{SiSi} \tag{8}$$

As previously described, the availability of silicon in soil depends on three primary conditions: pH, organic matter, and soil temperature. Of these conditions, it is feasible to modify the amount of organic matter or the pH. In the case of temperature, the easiest way would be to use covers as plastic mulches, which could increase the soil temperature by 3–4◦C (Ruíz-Machuca et al., 2015). Therefore, these factors can be considered as crucial factors to the agronomic management of silicon availability (Liang et al., 2015).

### Silicon Accumulation in Tomato

According to the simulations carried out using the proposed model, a soil with pH 7.0 and organic matter content of 6% can obtain the maximum availability of Si(OH)4, which can be > 4,500 mg m−<sup>2</sup> at 15◦C; or > 8,300 mg m−<sup>2</sup> at 25◦C (**Figures 1A,D**). Considering the availability, and two average temperature conditions of soil (15 and 25◦C), the highest availability of silicon is obtained with an average soil temperature of 25◦C (**Figure 1D**). On the contrary, when the organic matter content is low (<1%) along with a pH >8 (like a soil of a semi-arid region), the availability of Si(OH)<sup>4</sup> in the soil can drastically decrease to <20 mg m−<sup>2</sup> at 15◦C (**Figure 1A**), or <34 mg m−<sup>2</sup> at 25◦ C (**Figure 1D**). These results describe the effect of pH, organic matter, and temperature factors on the availability of Si(OH)<sup>4</sup> in soil. In addition to demonstrating the potential sensitivity of the availability of SiOH<sup>4</sup> in the soil to the modifications on any of the conditions mentioned.

The availability of silicon in the soil will directly impact the accumulation of plants grown on it (Epstein, 2001). The higher availability of silicon derived from the factors evaluated (**Figure 1D**), results in a more significant accumulation of Si (as SiO2) based on the dry weight of the tomato plant (up to 1 %), as can be seen in **Figure 1E**. The **Figure 1E** fits well the reported Si concentration in tomato plants grown under pH 5.5 (Miyake and Takahashi, 1978) to pH 8.48 (Gunes et al., 2007).

The same behavior is observed concerning accumulated silicon per plant, reaching up to 13.7 g per plant, that represent the maximum accumulation of silicon for tomato plants under this conditions (**Figure 1F**). However, low availability of silicon in the soil can lead to a small accumulation of silicon in the plant. In the example the conditions of a soil corresponding to the situation of a semi-arid region, there would be an accumulation <1.3 g per plant for both soil temperature conditions (**Figures 1C,F**). This equals to a silicon concentration based on plant dry weight <0.1% (**Figures 1B,E**).

Since silicon accumulates in the different organs of the tomato plant as a function of the rate of transpiration, then the highest accumulation will be observed in the leaves, since they represent 90% of total transpiration. In fruits, a lower accumulation of silicon will be seen, since the rate of transpiration is little compared to that of the leaves (∼5%) (Leonardi et al., 2000). However, the availability of silicon in the soil will ultimately define the accumulation of silicon both in the entire plant and in its various organs.

Contrary to a crop established in soil, in soilless cultivation, e.g., hydroponics, the primary factor that will modify the availability of Si for the plant will be its concentration in the irrigation water used. It has been reported that irrigation water can have a Si content [as Si(OH)4] of 5–20 mg L−<sup>1</sup> , while it is considered that an adequate concentration of Si would be 28 mg L −1 (Epstein, 1994, 1999; Liang et al., 2015). However, the transpiration rate of the crop will finally define the amount of SiOH<sup>4</sup> absorbed and accumulated in the different organs. Since the growth of the plant and the proper distribution of biomass in the various organs will affect the rate of transpiration, then crop growth should be considered as an additional factor that will change the accumulation of silicon in a soilless crop system. Therefore, the environmental factors (PAR, CO2, air temperature) that affect the growth of the crop will, in turn, affect the accumulation of silicon in the different organs.

According to the simulations carried out, a low concentration of CO<sup>2</sup> and a low incidence of PAR generate little accumulation of biomass in the tomato plant (**Figures 2A,D**). This same result is observed when the temperature of the environment changes, 30◦C generates biomass of up to 3,000 g per plant (**Figure 2A**); while 20◦C produces up to 2,240 g per plant (**Figure 2D**), at the highest conditions of PAR and CO<sup>2</sup> concentration. As a consequence, the total transpiration of the plant is modified when

environmental factors changes, and therefore the accumulation of silicon. With an average temperature of 30◦C, the biomass accumulated in the fruits can represent around 60% of the total of the tomato plant, and the leaves less than 10%. However, when the temperature drops to 20◦C, the biomass distribution in the tomato plant changes. Under this condition, the biomass accumulated in the fruits is ∼33%, whereas in the leaves it increases up to 33%.

Since the leaves constitute the largest area of transpiration, modifying the temperature of the environment in which the tomato grows substantially alters the final accumulation of silicon. Therefore, according to the model, a temperature of 30◦C will lead to a lower accumulation of silicon in the tomato plant (**Figure 2B**), while at 20◦C there will be more significant accumulation (**Figure 2E**). The **Figure 2C** fits the reported Si concentration in tomato plants grown under temperatures of 28–32◦C (Cao et al., 2015). The result will be a higher level of silicon in dry weight of the tomato plant at low temperatures (**Figure 2F**), while high temperatures will decrease the concentration considerably (**Figure 2C**).

When considering the availability of silicon in the irrigation water, it can be observed that a condition of low Si content (5 mg L−<sup>1</sup> as Si[OH]4) will result in less concentration and accumulation of silicon in the plant (**Figures 3A,B**). On the contrary, adequate availability of Si (28 mg L−<sup>1</sup> as Si[OH]4) in the irrigation water will result in increased accumulation and concentration of silicon in the plants (**Figures 3C,D**). High accumulation appears to occur regardless of the environmental conditions in which the crop develops, when there is adequate availability of Si in the irrigation water. However, when conditions are favorable for the development of leaves in tomato plants, the maximum concentration of silicon for this species can be reached (**Figure 3D**).

The model presented focuses on the impact of external factors on the growth of tomato plants, under the assumption that as long as exists the availability of Si the plants absorb it and transport it at a rate proportional to the growth rate and transpiration rate. The point highlighted with the model is that the biofortification of the fruits with Si depends on the availability of the element in the form of Si(OH)<sup>4</sup> both in soil and in soilless culture.

Sakurai et al. (2017) developed a successful dynamic model of Si absorption and transport for rice. The model is based on endogenous variables, as the activity of different transporters and the distribution of Si through different nodes in the entire plant. The model presented in this manuscript is focused on exogenous variables, susceptible to agronomic management both in field cultivation as in the greenhouse, and it has been used successfully

(C,D). For the simulation, 12 h of PAR and an average air temperature of 20◦C were considered.

to simulate the absorption of other elements in tomato and other crops (Juárez-Maldonado et al., 2014b; Ramírez-Pérez et al., 2018).

However, it must be taken into account that the presented model is used to describe the accumulation of silicon in the plant under relatively favorable environmental situations. The presence of stresses such as water deficit, salinity, deficiency of mineral nutrients and pathogens, results in loss of precision. With a certain amount of PAR and with a particular temperature regime, the stressed plants would have real biomass lower than the estimated by the model, which means an overestimation of the absorbed silicon.

As far as we know, except those published by (Sakurai et al., 2015, 2017) for monocotyledons, there are no similar dynamic models published about the absorption of silicon in dicots. The model described for the tomato crop is a first preliminar advance that we believe substantially can improve the understanding of some factors that regulate the bioavailability of silicon.

### CONCLUSIONS

After the results obtained from the presented model, the following is proposed:

1) When crops are grown in soil, the bioavailability of silicon can be increased by adding organic matter from organic amendments or humic substances, or by modifying the pH of the soil solution, using organic or inorganic acids, to be the closest to 7.0.

### REFERENCES


## AUTHOR CONTRIBUTIONS

All authors were responsible for processing information and manuscript writing. AB-M, AJ-M, FP-L, and AM-D: Conceptualization; AJ-M, ML-P, LR-P, and AM-D: Model design and implementation; ML-P, FP-L, SG-M, LG-D, JG-M, and AB-M: Manuscript drafting. All authors read and approved the final manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018. 00647/full#supplementary-material


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 López-Pérez, Pérez-Labrada, Ramírez-Pérez, Juárez-Maldonado, Morales-Díaz, González-Morales, García-Dávila, García-Mata and Benavides-Mendoza. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Genetic Basis of Variation in Rice Seed Storage Protein (Albumin, Globulin, Prolamin, and Glutelin) Content Revealed by Genome-Wide Association Analysis

Pingli Chen<sup>1</sup> , Zhikang Shen<sup>1</sup> , Luchang Ming<sup>1</sup> , Yibo Li<sup>1</sup> , Wenhan Dan<sup>1</sup> , Guangming Lou<sup>1</sup> , Bo Peng<sup>1</sup> , Bian Wu<sup>1</sup> , Yanhua Li<sup>2</sup> , Da Zhao<sup>1</sup> , Guanjun Gao<sup>1</sup> , Qinglu Zhang<sup>1</sup> , Jinghua Xiao<sup>1</sup> , Xianghua Li<sup>1</sup> , Gongwei Wang<sup>1</sup> and Yuqing He<sup>1</sup> \*

<sup>1</sup> National Key Laboratory of Crop Genetic Improvement and National Center of Plant Gene Research, Huazhong Agricultural University, Wuhan, China, <sup>2</sup> Life Science and Technology Center, China National Seed Group Co., Ltd., Wuhan, China

### Edited by:

Huixia Shou, Zhejiang University, China

### Reviewed by:

Umesh K. Reddy, West Virginia State University, United States Shahidul Islam, Murdoch University, Australia

> \*Correspondence: Yuqing He yqhe@mail.hzau.edu.cn

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 21 January 2018 Accepted: 18 April 2018 Published: 09 May 2018

#### Citation:

Chen P, Shen Z, Ming L, Li Y, Dan W, Lou G, Peng B, Wu B, Li Y, Zhao D, Gao G, Zhang Q, Xiao J, Li X, Wang G and He Y (2018) Genetic Basis of Variation in Rice Seed Storage Protein (Albumin, Globulin, Prolamin, and Glutelin) Content Revealed by Genome-Wide Association Analysis. Front. Plant Sci. 9:612. doi: 10.3389/fpls.2018.00612 Rice seed storage protein (SSP) is an important source of nutrition and energy. Understanding the genetic basis of SSP content and mining favorable alleles that control it will be helpful for breeding new improved cultivars. An association analysis for SSP content was performed to identify underlying genes using 527 diverse Oryza sativa accessions grown in two environments. We identified more than 107 associations for five different traits, including the contents of albumin (Alb), globulin (Glo), prolamin (Pro), glutelin (Glu), and total SSP (Total). A total of 28 associations were located at previously reported QTLs or intervals. A lead SNP sf0709447538, associated for Glu content in the indica subpopulation in 2015, was further validated in near isogenic lines NIL(Zhenshan97) and NIL(Delong208), and the Glu phenotype had significantly difference between two NILs. The association region could be target for map-based cloning of the candidate genes. There were 13 associations in regions close to grainquality-related genes; five lead single nucleotide polymorphisms (SNPs) were located less than 20 kb upstream from grain-quality-related genes (PG5a, Wx, AGPS2a, RP6, and, RM1). Several starch-metabolism-related genes (AGPS2a, OsACS6, PUL, GBSSII, and ISA2) were also associated with SSP content. We identified favorable alleles of functional candidate genes, such as RP6, RM1, Wx, and other four candidate genes by haplotype analysis and expression pattern. Genotypes of RP6 and RM1 with higher Pro were not identified in japonica and exhibited much higher expression levels in indica group. The lead SNP sf0601764762, repeatedly detected for Alb content in 2 years in the whole association population, was located in the Wx locus that controls the synthesis of amylose. And Alb content was significantly and negatively correlated with amylose content and the level of 2.3 kb Wx pre-mRNA examined in this study. The associations or candidate genes identified would provide new insights into the genetic basis of SSP content that will help in developing rice cultivars with improved grain nutritional quality through marker-assisted breeding.

Keywords: GWAS, storage protein, grain quality, endosperm, nutrition, Oryza sativa L.

### INTRODUCTION

fpls-09-00612 May 7, 2018 Time: 23:5 # 2

Rice is one of the major staple cereal foods and is an important source of total protein in human food. SSP account for approximately 8% of the dry grain weight and are the second most abundant ingredient after starch in rice. Rice has the lowest protein content among cereal grains, but net protein utilization is highest (Juliano, 1992). High-protein rice is likely to increase human nutrition in poor families, especially where rice is a staple food. Therefore, increasing the SSP content has become one of the main breeding objectives in improving nutritional quality in rice.

The SSP in rice can be classified into four fractions: albumin, globulin, prolamin, and glutelin, according to differences in solubility. Glutelin encoded by 15 genes accounts for as much as 80% of the total SSPs and is concentrated in the milled fraction, whereas prolamin, the most evenly distributed protein, accounts for less than 5% (Yamagata et al., 1982). Based on amino acid sequence similarity, glutelins are classified into four subfamilies: GluA, GluB, GluC, and GluD (Kawakatsu et al., 2008). In rice, SSP genes have been cloned and characterized mostly by mutant screening (Ren et al., 2014). Glutelins are synthesized in the rough endoplasmic reticulum as a 57 kDa precursor. Previous studies have identified rice 57H mutants that accumulate relatively high levels of 57 kDa pro-glutelin and have floury/opaque endosperm phenotype (Wang et al., 2009; Ren et al., 2014). Of 57H mutants, only gpa3, Osvpe1, and OsRab5a have been successfully cloned (Wang et al., 2009; Wang Y. et al., 2010; Ren et al., 2014). Prolamins are encoded by a multigene family of 34 gene copies. Based on the molecular mass, prolamins classified into three groups: 10 kDa prolamin (RP10), 13 kDa prolamin (RM1, RM2, RM4, and RM9), and 16 kDa prolamin (RP16) (Yamagata et al., 1982; Kawakatsu et al., 2008). Both albumin and globulin are concentrated in the bran and polishing during milling removes a major portion of these proteins (Shewry, 2007). Globulin is also easily digested (Yamagata et al., 1982; Zhang et al., 2008), and only a limited number of genes has been cloned and characterized (Bhullar and Gruissem, 2013). RA16 and RA17 that are associated with seed allergenic protein have been reported as albumin genes in previous studies (Adachi et al., 1993). The nutritional value of glutelin is higher than prolamin because it has a greater digestive capacity by humans and higher lysine content in rice (Ogawa et al., 1987). However, patients with kidney disease and diabetes need low glutelin diet (Mochizuki and Hara, 2000; Nishimura et al., 2005; Morita et al., 2009). Some proteins that belong to the albumin and globulin are considered to be allergenic. Collectively, emphasis in rice breeding should not only be on the concentration, but also on the quality of rice protein.

Seed storage protein is quantitatively inherited, and is affected by growing environment (Shewry, 2007). QTL mapping based on molecular markers and linkage maps has always been a common method of genetic studies (Li et al., 2014; Peng et al., 2014). Many QTL for crude protein content in rice have been reported (Aluko et al., 2004; Wang et al., 2008; Lou et al., 2009; Yu et al., 2009; Peng et al., 2014), but fewer studies have investigated the individual protein fractions in milled rice. qPC1 (OsAAP6), controlling the natural variation in SSP content, has been cloned using a mapbased cloning strategy (Peng et al., 2014). Zhang et al. (2008) identified 16 QTL for contents of crude protein and the four protein fractions.

Genome-wide association study (GWAS) by means of single nucleotide polymorphism (SNP) has become the method of choice for investigation of the genetics of important traits in Arabidopsis thaliana (Chan et al., 2011), rice (Huang et al., 2010), maize (Xiao et al., 2016), sorghum (Morris et al., 2013), and others (Ogura and Busch, 2015). Although GWAS is widely used in genetic analysis of grain quality traits in rice, such as gelatinization temperature, amylose content, grain appearance, and milling quality (Borba et al., 2010; Huang et al., 2010), few studies have used this approach to investigate total SSP in rice (Huang et al., 2012; Bryant et al., 2013). The analysis of genetic basis of nutritional quality has been reported in maize (Deng et al., 2017). Maize opaque2 (o2) mutation could increase free lysine levels and create the foundation for quality protein maize (QPM) breeding. Combining GWAS and linkage mapping, a gene duplication at the 27-kDa γ-zein locus qγ27 was identified (Liu et al., 2016). qγ27 increases the level of 27-kDa γ-zein gene expression in QPM, which is essential for endosperm modification. GWAS on amino acids has been carried, and 247 and 281 significant loci were identified in two different environments (Deng et al., 2017). However, no studies have been reported about the genetic basis of the four SSP fractions by GWAS in rice.

In this study, we performed GWAS of Alb, Glo, Pro, Glu, and total SSP in milled rice using 527 Oryza sativa accessions grown in two environments with an aim to identify loci involved in the genetics. Haplotype analysis and expression pattern of candidate genes then provided valuable in better understanding the genetic basis of variation in SSP content. The results shown here promote our understanding of the genetic basis of the storage protein groups, should be of use for breeders attempting to improve nutritional quality by means of marker assisted selection.

### MATERIALS AND METHODS

### Plant Materials, Field Experiments, and Trait Measurements

A diverse worldwide collection of 527 O. sativa landraces and elite accessions (Supplementary Table S1) was used in this study. Structural analysis indicated that the entire collection belonged to nine subpopulations: indI, indII, indica intermediate, Tej, Trj, japonica intermediate, Aus, VI, and intermediate (Chen et al., 2014) and are available at the RiceVarMap<sup>1</sup> . The indica subpopulation (indI, indII, and indica intermediate) included 294 accessions, whereas the japonica subpopulation (Tej, Trj, and japonica intermediate) included 155 accessions.

Lines were planted in two environments in Hubei province: Ezhou in 2014 (Env. 1) and Wuhan in 2015 (Env. 2). The sowing dates were 25 May in both years. Seedlings about 25 days old were transplanted to the field. There were three

<sup>1</sup>http://ricevarmap.ncpgr.cn

rows with 10 plants each in each plot. The planting density was 16.5 cm between plants within a row, and 26.4 cm between rows. Field management basically followed recommended practice of agriculture, with fertilizer applied (per hectare) as follows: 48.75 kg nitrogen, 58.5 kg phosphorous, and 93.75 kg potassium as the basal fertilizer; 86.25 kg nitrogen at the tilling stage; and 27.6 kg nitrogen at the booting stage.

At maturity, three plants in the middle of the second row of each accession were harvested and bulked. Dry seeds were threshed in bulk and the rough rice was air-dried, and stored at room temperature for 3 months, and then stored at 4◦C. 50 g of rough rice were dehulled into brown rice using a TR 200 dehuller (Kett, Tokyo, Japan). The embryo and aleurone layer of brown rice were removed into milled rice through a Pearlest mill (Kett, Tokyo, Japan). The rice was ground into flour with a CT 410 Cyclotec mill (FOSS, Hillerod, Denmark), passed through an 80-mesh sieve and stored at −20◦C until the Alb, Glo, Pro, and Glu contents were determined based on previously published previous methods (Kumamaru et al., 1988). Briefly, 0.1 g sample of milled rice flour was placed in a centrifugation tube with 1.0 ml solvent containing 10 mM Tris–HCl buffer (pH7.5) for Alb extraction; 1.0 ml solvent containing 1 M NaCl, for Glo extraction; 60% n-propanol containing 1 mM EDTA-2Na, for Pro extraction; and 0.05 M NaOH for Glu extraction. The mixture was stirred for 2 h at room temperature, and extracts were separated from residues by centrifugation at 12,000 rpm for 15 min at 4◦C The procedure was repeated three times. The extracts were stored at −20◦C until further analysis. The contents of each fraction were determined by the Coomassie brilliant blue G-250 dye-binding method (Bradford, 1976) using bovine serum albumin as a standard, and quantitative analysis was carried out using Infinite M200 (Tecan Group, Männedorf, Switzerland) (Peng et al., 2014). Total SSP was the sum of the Alb, Glo, Pro, and Glu contents. The 2-year field experiment was designed with three replicates per year. The average SSP contents across three replicates within 1 year were used for GWAS. The SSP contents of the 527 O. sativa accessions are listed in Supplementary Table S1. Amylose content was measured as previously described method (Tan et al., 1999).

### Genome-Wide Association Study

All 527 accessions were genotyped via sequencing (Chen et al., 2014). SNP information was available on RiceVarMap (see foot note text 1), a comprehensive database for rice genomics. The physical locations of the SNPs were identified based on the rice annotation version 6.1 of variety Nipponbare from Michigan State University. A total of 3,916,415 SNPs in the whole population; 2,767,159 SNPs in the indica subpopulation; and 1,857,845 SNPs in the japonica subpopulation (minor allele frequency ≥0.05; number of accessions with minor alleles ≥6) was used for GWAS (Chen et al., 2014). A linear mixed model (LMM) was used for detecting associations using Fast-LMM (Lippert et al., 2011). Population structure was controlled using a kinship matrix constructed with all SNPs (Chen et al., 2014). Effective independent SNPs were detected (Li et al., 2012), and were 757,578 in the whole population; 571,843 in the indica subpopulation; and 245,348 in the japonica subpopulation. The thresholds were set at a P-value of 5.0 × 10−<sup>6</sup> to identify significant association signals. To obtain independent association signals, multiple SNPs, exceeding the threshold in a 5 Mb region, were clustered based on an r <sup>2</sup> of LD ≥ 0.25, and SNPs with the minimum P-value in a cluster were deemed to be lead SNPs.

### Statistical Analysis

Based on the standardized disequilibrium coefficients (D'), linkage disequilibrium (LD) was investigated. LD heatmaps were constructed using the TASSEL5.0<sup>2</sup> program and R package "LDheatmap"<sup>3</sup> . Statistical analysis, including a correlation analysis, was conducted using IBM SPSS Statistics 22.0. Differences in SSP values were examined by Student's t-tests. Broad-sense heritability (H<sup>2</sup> ) for each phenotype was estimated using repeatability between 2 years of phenotypic data, calculated as the variance among variety grand means divided by their total phenotypic variance.

### Candidate Genes and Haplotype Analysis

Candidate genes within a 200 kb genomic region ( ± 100 kb from the lead SNP) in the associated loci were selected based on (i) biochemically related proteins or protein clusters; (ii) homologous genes with known function, and (iii) expression profiles. The genotypes of RP6, RM1, Wx, PROLM1, and other three candidate genes in the 527 rice accessions were obtained from the RiceVarMap database (see foot note text 1). The haplotypes were classified according to all SNPs (except sites in intron) including their intragenic region and 2 kb upstream with an MAF > 0.05 in a candidate gene. There were at least five rice accessions in the haplotypes for comparative analysis. One-way ANOVA and Student'st-tests were applied to compare differences in SSP content among all possible haplotype pairs.

### RNA Extraction and Quantitative RT-PCR Analysis

According to the manufacturer's instructions, the total RNA was extracted from rice different tissues using TRIzol reagent (Invitrogen). About 3 µg of RNA sample was processed by RNase-free DNaseI (Invitrogen) and reverse transcribed using M-MLV reverse transcriptase (Invitrogen) with Oligo(dT)15. Quantitative RT-PCR was carried out using Fast Start Universal SYBR Green Master (Rox) superMIX (Roche, Mannheim, Germany) in a ViiA 7 Real-Time PCR system (Applied Biosystems), according to the manufacturer's introductions. Measurements were obtained using the relative quantification method. Actin was used as a reference gene in the qRT-PCR experiments. The experiment was designed with three biological replicates and three technical replicates per material. Error bars indicate standard error. The measurements were obtained using the relative quantification method. The significant difference was analyzed statistically by One-way ANOVA and Student's t-tests. All primers for qRT-PCR analysis are listed in Supplementary Table S2.

<sup>2</sup>https://tassel.bitbucket.io/

<sup>3</sup>https://www.r-project.org/

### RESULTS

fpls-09-00612 May 7, 2018 Time: 23:5 # 4

### Phenotypic Variation and Heritability of SSP Content

The result analysis revealed a large variation in all phenotypes evaluated and the traits appeared to be normally distributed (**Figures 1A–E** and Supplementary Table S3). Average SSP contents in Env. 1 and 2 were 70.9 and 53.7 mg/g; Alb was 2.9 and 3.3 mg/g; Glo was 5.8 and 5.3 mg/g; Pro was 2.8 and 2.1 mg/g; and Glu was 59.4 and 43.5 mg/g, respectively (**Figures 1A–E**). Glu accounted for approximately 80% of total SSP; Alb and Pro each accounted for about 5%; and Glo accounted for about 10% (**Figure 1F**). Compared with other three storage protein contents, the average content of Pro was the lowest, but the variation of Pro was largest in the whole population and each subpopulation in both environments. In four storage protein, Total SSP showed the lowest heritability (29.5%), whereas Pro showed the highest (76.8%) (Supplementary Table S3).

Correlation coefficients between each pair of components, and between components and total SSP were significant and positive in both environments, except for those between Glo and Pro in any environment and between Alb and Glo and between between Alb and Pro in Env. 2 (Supplementary Table S4). High correlations were found only between Glu and total SSP in both environments with coefficients of 0.99 in Env. 1 and 0.98 in Env. 2.

### Genome-Wide Association Study for SSP Contents

We performed GWAS on the entire population and on the indica and japonica subpopulations for each year. The FaST-LMM program reduced the effect of population structure (Yang et al., 2014). Quantile-quantile plots of all five traits for the whole population, and indica and japonica subpopulations are illustrated in **Figure 2** and Supplementary Figures S1, S2. Some associations were detected in different subpopulations, and some of the associations for different traits were in the same chromosomal regions. Any two lead SNPs within a 100 kb region were considered to be a single association locus.

The association analysis for the whole population identified 34 loci (phenotypic variance >10%) associated with three traits with a suggestive threshold value at 5.0E-06 (**Table 1**). Most of them were detected for Alb (21 associations) and Pro (12 associations). Lead SNPs for Alb were widely distributed in the rice genome: chromosome 1, 2, 3, 4, 5, 6, 7, 8, and 9, with chromosomes 6 and 7 having more associations. Associations explained phenotype variation of 10.4–20.9%, with the association on chromosome 6 (sf0601764762) making the largest effect. For Pro, associations accounting for 10.1–34.6% of the phenotypic variance were identified on chromosomes 2, 5, 7, 10, and 11, with chromosome 5 and 7 exhibiting more associations. Only one lead SNP sf0317000156 on chromosome 3 with a phenotype variation of 14.4% was detected for Glo. No lead SNP with phenotype variation of more than 10% was detected for Glu or total SSP.

A large number of peaks (phenotypic variance >10%) were also detected by GWAS in the indica (33 associations) and japonica (40 associations) subpopulations (Supplementary Table S5). For Alb, 23 associations were distributed on all 12 rice chromosomes, except chromosome 3 and 10, but only three associations were identified in the indica subpopulation. Alb associations identified in the indica and japonica subpopulations explained phenotype variation of 10.5– 12.3, and 11.3–41.6%, respectively. Four SNPs, sf0142207782, sf0209990680, sf0605251091, and sf0804866973 individually explaining more than 30% of the Alb variation, were detected in the japonica subpopulation in Env. 1. For Glo, 14 associations were identified on chromosomes 1, 3, 4, 5, 6, 9, 11, and 12, and equal numbers of associations were detected in the indica and japonica subpopulations, explaining 10.3–16.8 and 12.9–17.0% of the variation, respectively. For Pro, there were 23 associations, involving chromosomes 1, 3, 4, 5, 7, 9, 11, and 12. Among them 14 (with phenotype variation of 10.4–38.3%) and 9 (with phenotype variation of 15.5–22.6%) associations were identified in the indica and japonica subpopulations, respectively. Ten associations for Glu were detected on chromosomes 1, 4, 5, 6, 7, 8, 9, and 10, in both subpopulations. For total SSP, only three associations on chromosomes 10 and 11 with phenotype variation of 10.4–14.8% were detected in the indica subpopulation.

Among the 107 associations detected in the whole population and in the indica and japonica subpopulations, 16 were detected in different populations and nine involved two or three different traits (**Table 1** and Supplementary Table S5). Examples include lead SNP sf0519612378 with phenotypic variance of 14.9– 36.6% that was detected in the whole population and indica subpopulation in Env. 1 and 2; and lead SNP sf1022972496 detected in indica subpopulation in Env. 1 was associated with phenotypic variances of 12.0 and 14.8% for traits Glu and total SSP, respectively. Seven associations in the whole population and four associations in the indica subpopulation were detected both in both environments. For Pro, two lead SNPs (sf0514987630 and sf0515211855) in the whole population and one lead SNP (sf0515706446) in the indica subpopulation were detected in both environments. The significance levels of the associations ranged from P = 5.0E-06 to P = 8.6E-16, P = 4.9E-06 to P = 9.0E-08, P = 5.0E-06 to P = 1.3E-17, P = 2.1E-06 to P = 1.4E-08, and P = 4.5E-06 to P = 3.1E-08 in LMM for Alb, Glo, Pro, Glu, and total SSP, respectively, and the most significant association for sf0706363663 located in chromosome 7 (**Table 1**).

### Co-localization of Associated Sites With QTLs Previously Reported and Grain Quality-Related Genes

There were many overlaps between the present associations detected by GWAS and reported QTLs or intervals related to SSP content in rice. A total of 28 associations from this study were located at previously reported QTLs or intervals (shown in **Table 1** and Supplementary Table S5 with corresponding references) of which 10 and eight associations were for Alb and Pro, respectively.

In the indica subpopulation, lead SNP sf0709447538 in a notable hotspot region at the interval of 9.1–9.5 Mb, explaining 12.7% of the Glu variation, was detected on chromosomes 7

in 2015 (**Figures 3A,B** and Supplementary Table S5). Interestingly, we found that the lead SNP sf0709447538 was overlapped with the amino acid content QTLs (7–4, 7–5, and 7–6), identified in a previous study using an F<sup>9</sup> recombinant inbred line population, which derived from a cross between Zhenshan97 (ZS97) and Delong208 (DL208) (Wang et al., 2008; Zhong et al., 2011). To validate the QTL, we developed near-isogenic lines (NILs) (**Figure 3C**). NILs of QTL were developed by successive crossing and backcrossing ZS97 (high protein content) and DL208 (low protein content), three times (BC3) to ZS97. The QTL was selected by two molecular markers MRG186 and MRG4499 (Supplementary Table S2). Self-pollinating the BC3F<sup>1</sup> plants heterozygous for this fragment produced NIL(ZS97) and NIL(DL208). Analysis of NIL(ZS97) and NIL(DL208) showed that NIL(ZS97) was significantly higher Glu than NIL(DL208), which was the same as the phenotype in two parents (**Figure 3D**). The result indicated that the QTL was reliable, which helps further identify the underlying genes and their genetic basis.

On the other hand, 13 associations were detected in regions close to previously identified grain-quality-related genes. Five genes (PG5a, Wx, RM1, RP6, and AGPS2a) were less than 20 kb from lead SNPs. PG5a, RM1, and RP6 were associated for Pro; Wx

(P-value). Lead SNPs in significant peaks are red. The horizontal dotted line indicated the genome-wide significance threshold (P = 5.0E-06). Total, total SSP content.

and AGPS2a was associated with Alb. Glutelin genes GluA1 and OsAAT2 were associated with total SSP in Env. 2. Additionally, several lead SNPs were located close to starch-metabolism-related genes, such as PUL, ISA2, Wx, GBSSII, OsACS6, and AGPS2a. SNPs located close to other reported grain quality-related genes (Prol14, RA17, RA16, and PGL) were also examined using a LMM (**Table 1** and Supplementary Table S5).

### Haplotype Analyses for the Reported Genes RP6 and RM1

The association SNP sf0705739605 for Pro was each <10 kb away from two prolamin genes RP6 and RM1 reported to encode prolamins in rice (Wen et al., 1993; Kawakatsu et al., 2009). In particular, the lead SNP with highly significant P-values (P = 6.5E-11 in Env. 1 and P = 5.4E-16 in Env. 2) was identified in the whole population and indica subpopulation (**Figures 4A–C**, **Table 1**, and Supplementary Table S5). The association explained 24.7 and 26.3% of the phenotypic variances in the whole population in Env. 1 and 2, respectively. Lead SNPs sf0705735351 and sf0705739605 and all polymorphic sites in RP6 and RM1 were in high linkage disequilibrium (in high LD with each other; r <sup>2</sup> = 0.94–0.99) with most polymorphic sites (**Figure 4H**). We performed haplotype analyses for RP6 and RM1 and identified three main haplotypes (Hap1-3) at each locus (Supplementary Table S6). Hap1 and hap2 of RP6 and hap2 and hap3 of RM1 were not identified in japonica. Hap2 of RP6 and hap3 of RM1 had significantly higher Pro than the alternative haplotypes in both

TABLE 1 | Associated single nucleotide polymorphisms (SNPs) identified by linear mixed model (LMM) method in the whole population.


Pop, population; Chr., chromosome; P, P-value estimated in LMM; P.V (%), proportion of phenotypic variance explained; Total, total seed storage protein (SSP) content. <sup>a</sup>Associated SNPs within 100 kb for the same trait are considered the same locus; for Alb, Glo, and Pro, only lead SNPs with P.V (%) more than 10% are shown; bold scripts indicate detections in different populations and italic indicates detection in different traits.bNegative value means the gene is upstream of the SNP site.cRefers to interval or QTL reference.

environments (**Figures 4D,F**). In the region contained coding region and 2 kb upstream of RP6 and RM1, 17 and 26 SNPs were found, respectively (Supplementary Table S6). At the RP6 locus, there were two synonymous SNPs, two non-synonymous SNPs in the exon, three SNPs in the 3<sup>0</sup> untranslated regions, and 10 substitutions in the 2 kb cis-regulatory region. At the RM1 locus, there were five synonymous SNP, no non-synonymous SNPs in the exon, two SNPs in the 5<sup>0</sup> and 3<sup>0</sup> untranslated regions, and 19

plots of LMM for Pro in the all accessions in 2015 (A). (B) Local Manhattan plot surrounding the peak in 2015 on chromosome 7. Arrow indicates the position of the lead peak. The corresponding colors of r 2 represent linkage disequilibrium levels. (C) Plant architectures of near-isogenic lines. (D) Phenotypes of Glu of two parents ZS97, DL208, NIL(ZS97), and NIL(DL208). ∗∗Indicates the differences of Glu between two materials are significant at P < 0.01.

substitutions in the 2 kb cis-regulatory region. To get an overview of the expression profiles of RP6 and RM1, the CREP database<sup>4</sup> , a website that contains the dynamic gene expression profile of indica rice, was searched (Wang L. et al., 2010). The results showed that RP6 and RM1 displayed high-level expressions in endosperm but low-level expressions in other tissues in ZS97 (**Figures 4E,G**). Considering the complexity of population structure and genetic background, we checked the expressions of RP6 and RM1 with different haplotypes in 20 and 34 accessions randomly chosen from indica group, respectively. Using qRT-PCR analysis, The results indicated that in the endosperm at 7 days after pollination (d.a.p.), expression levels of RP6 in hap2 accessions were much higher than those in hap3 accessions (**Figure 4I** and Supplementary Table S7), and expression levels of RM1 in hap3 accessions were much higher than those in hap1 accessions (P < 0.01) (**Figure 4J** and Supplementary Table S7). These results show that two genes might be good candidates for the GWAS locus. In conclusion, two genes had high-level and specific expressions in endosperm of rice, and genotypes with

### Analysis of One Candidate Gene Wx

We also found a highly significant association signal for Alb involving sf0601764762 (P = 2.0E-08 in Env. 1 and P = 8.6E-16 in Env. 2) on chromosome 6 (**Table 1** and **Figures 5A–C**). The lead SNP, explaining 12.9 and 20.9% of the phenotypic variances in the whole population in Env. 1 and 2, respectively, was located in the first intron of Wx (**Table 1** and **Figure 5D**). Other SNPs in Wx showed different LD associations with sf0601764762 (**Figure 5D**). Wx is the most important genetic determinant of amylose content (Tian et al., 2009). We identified eight major Wx haplotypes with 23 SNPs (**Figures 5E,F** and Supplementary Table S6). Hap1, 6, and 7 showed lower Alb values than hap4 and 5 in both environments. The lead SNP sf061764762 is located to the first intron of Wx, the major gene determining starch content. This SNP generates alleles Wx<sup>a</sup> with a normal GT sequence at the 5 0 splice junction of intron 1, and Wx<sup>b</sup> with a G to T mutation in intron 1. Wx<sup>a</sup> and Wx<sup>b</sup> produce a mature 2.3 kb Wx mRNA and a 3.3 kb Wx pre-mRNA, respectively (Wang et al., 1995).

higher Pro exhibited much higher expression levels in indica group in this study.

<sup>4</sup>http://crep.ncpgr.cn/crep-cgi/home.pl

signals of RP6 (E) and RM1 (G) in various tissues of ZS97 based on the microarray data. The y-axis represents the expression signals. (H) Representation of pairwise r 2 values among polymorphic sites in RP6 and RM1. The lines in red represent lead SNP. Expression levels of RP6 (I) and RM1 (J) in the endosperm at 7 days after pollination in indica group. Error bars, SE of 3 replicates. ∗∗Indicates the differences of expression levels between two haplotypes are significant at P < 0.01. Hap, haplotype; HD, heading date; DAP, day after pollination.

We checked the quantity of 2.3 and 3.3 kb Wx RNA with four haplotypes in 35 and 37 accessions randomly chosen from indica and japonica, respectively. Using qRT-PCR analysis, it showed that quantity of 2.3 kb Wx RNA in the endosperm at 7 d.a.p. in hap7 indica accessions were much higher than those in hap2 indica accessions, hap3 japonica accessions and hap5 japonica accessions (P < 0.01). In contrast, quantity of 3.3 kb Wx RNA in hap7 indica accessions were much lower than those in hap2 indica accessions and in hap5 japonica accessions (**Figures 5G,H** and Supplementary Table S7). With highest Alb content among four haplotypes, hap5 had relatively high 3.3 kb Wx RNA quantity and low 2.3 kb Wx RNA quantity. However, with lowest Alb content among four haplotypes, hap7 had relatively low 3.3 kb Wx RNA quantity and highest 2.3 kb Wx RNA quantity. It was found that Wx exhibited higher level quantity of 2.3 kb Wx mRNA in endosperm in ZS97 (hap7) than Minghui 63 (hap2), two indica cultivars (**Figure 5I**). We further compared the quantity of 2.3 and 3.3 kb Wx RNA in various tissues between two varieties Zhonghua 11 (hap2) and ZS97 by quantitative RT-PCR (**Figure 5J**). The results showed that ZS97 had lower 3.3 kb Wx RNA quantity and higher 2.3 kb Wx RNA quantity than Zhonghua 11 in endosperm of 7 and 14 d.a.p. However, Wx displayed very low-level expressions in stem, sheath, and flag leaf in both ZS97 and Zhonghua 11. Then we measured amylose content of the corresponding accessions in 2015 (Supplementary Table S7). The correlation analysis among Alb, amylose content, quantity of 2.3 and 3.3 kb Wx RNA was performed, and the results are presented in **Table 2**. Amylose content had significant correlations with both quantity of 2.3 and 3.3 kb Wx RNA, but amylose content was positively correlated with quantity of 2.3 kb Wx RNA and was negatively correlated with quantity of 3.3 kb Wx RNA, which was consistent with previous study (Wang et al., 1995). Significant and negative correlations were observed between Alb and amylose content or quantity of 2.3 kb Wx RNA (**Table 2**), suggesting that Wx may negatively regulate Alb. However, Alb and quantity of 3.3 kb Wx RNA had no

significant correlations. We speculated that Wx might influence Alb content.

### Analyses of Four Candidate Genes

Association analysis at the haplotype level greatly increases mapping power (Han et al., 2016). We re-detected these grain-quality-related genes mentioned above at the haplotype level (Supplementary Table S8). Four SSP loci (OsAAT2,


Alb of 46 accessions in 2015 was measured; AC, amylose content of 44 accessions in 2015; 2.3 kb, relative expression level of 2.3 kb Wx; 3.3 kb, relative expression level of 3.3 kb Wx. <sup>∗</sup>P < 0.05 and ∗∗P < 0.01.

RA17, RM1, and RP6) and four starch-metabolism-related genes (AGPS2a, ISA2, PUL, and Wx) were detected in both environments; however, GBSSII, GluA1, and RA16 were only detected in one environment. The order of magnitudes of the P-values for these genes was greatly reduced compared with that at the SNP level. To further verify the association possibility, we validated some of candidate genes detected via GWAS by haplotype analyses and expression profiles in public databases.

To determine whether novel functional loci were implicated by GWAS, we investigated SNP sf0515211855 (P = 6.5E-07 in Env. 1 and P = 2.7E-11 in Env. 2), in Chr.5 (14.7- 15.8 Mb) (**Figures 6A–C**). This lead SNP was significantly associated with Pro and explained 8.5 and 13.7% of the phenotypic variance in the whole population in Env. 1 and 2, respectively (**Table 1**). SNPs in the region were grouped into two LD block, and sf0515211855 was in the first block. Within this block, there were a gene cluster of 18 prolamin precursors and a candidate gene encoding an expressed protein (LOC\_Os05g25500) (**Figure 6B**). Within the

represents the expression signals. HD, heading date; DAP, day after pollination.

cluster we focused on PROLM1 (LOC\_Os05g26240), located close to the lead SNP, and identified twelve major haplotypes (Hap1-12) (**Figures 6D,E** and Supplementary Table S6). Hap4, 5, and 6 had significantly higher Pro than other haplotypes in both environments. Microarray data (see foot note text 4) indicated that PROLM1 had extremely high expression enrichment in endosperm of ZS97 at the ripening stage (**Figure 6G**). We obtained five major haplotypes (Hap1-5) for LOC\_Os05g25500 (Supplementary Table S6). Hap3 and 5 had significantly higher Pro content than Hap1 and 4 in both environments (**Figure 6F**). LOC\_Os05g25500 exhibited higher expression levels in endosperm than other tissues of ZS97 (**Figure 6H**). We focused on other two candidate genes LOC\_Os03g29750 (encoding an expressed protein) and LOC\_Os02g13130 (encoding a KH domain-containing protein) for Glo in indica group in 2015 (**Figure 7A**) and Glu in all group in 2014 (**Figure 7B**), respectively. Hap5 of LOC\_Os03g29750 had significantly higher Glo than other haplotypes, and LOC\_Os03g29750 showed very high-level expression in endosperm of ZS97 (**Figure 7C**). Hap4 of LOC\_Os02g13130 had significantly higher phenotypes than other haplotypes. For LOC\_Os02g13130, homologous with maize gene encoding high molecular weight glutenin subunit x, it had high-level expressions in most tissues and organs (**Figure 7D**). These results indicated that these genes might be good candidates for the GWAS locus.

HD, heading date; DAP, day after pollination.

### DISCUSSION

### Phenotypic Variation and Trait Correlation

The contents of the four components of SSP and the total SSP were normally distributed in two environments and influenced by environment in this study (**Figure 1**). The indica subpopulation showed wider variation in Pro than the japonica subpopulation, but narrower variation compared to japonica for Alb. These results indicated that there were probably different genetic bases underlying Alb and Pro in the two subpopulations. The contents and proportions of different components as well as total SSP in the present study were in agreement with those reported by Kumamaru et al. (1988). Glu is highest composition of SSP in rice endosperm and has more essential amino acid required for human (Tanaka et al., 1995), suggested that it is more effectively to improve protein content and nutritional value of rice by increasing Glu than other three SSPs.

In previous studies, different conclusions were reached on the heritability of SSP content in rice. Hillerislambers et al. (1973) reported that the heritability of SSP content was 13.3–37.2% in rice. However, Shenoy et al. (1991) obtained values as high as 71% and considered that it is effective to select protein content phenotypes in early generation in the protein content breeding. Here, different SSP showed different heritabilities (29.5–76.8%) (Supplementary Table S3). It suggested that four SSP may have differences in generation selection. To some extent, Alb with higher heritability was effectively selected in early generation, but it may not be good for Glu and Total with lower heritability. The relationship between different SSP components in rice is complex. In the present study, there were significant positive correlations between Glu and the other three SSPs in both environments (Supplementary Table S4), that was consistent with

reported study (Zhang et al., 2008). These results indicated that Glu and other three SSPs might partially share the common genetic mechanism.

### GWAS for Seed Storage Protein in Rice

The use of high-density, genome-wide SNPs in GWAS not only can detect candidate genes, but also have a comprehensive understanding of the regulatory mechanism of related traits. Among the 107 associations detected in the study, few sites were repeatedly detected in both environments. Compared with Glo and Glu, Alb and Pro have larger variations in the populations (Supplementary Table S5). Similarly, association mapping was more efficient in identifying associations for Alb and Pro with relatively higher heritability than total SSP and Glu with lower heritability (Supplementary Tables S3–S5). On the other hand, we noticed that all associations detected for Alb and Pro explained more than 100% of the phenotypic variance. This suggests that phenotypic variance explained by the interaction between some of these associations might be very large.

Although genome-wide association studies are becoming more sophisticated, it should be noted that association mapping may lead to false positive associations largely caused by population structure. In order to reduce false-positives resulting from genetic structure, we also analyzed the indica and japonica panels separately by LMM. Thirteen associations were adjacent to previously known grain-quality-related genes (**Table 1** and Supplementary Table S5). These results suggested that association mapping was an effective way to find candidate genes for SSPs in rice. And 28 associations were detected in previously reported intervals or QTLs (Aluko et al., 2004; Wang et al., 2008). QTL pro6 was repeatedly detected for Alb (sf0605251091 and sf0601764762) and Glu (sf0602321094). Lead SNPs (sf0709447538 and sf0712842943 for Glu, sf0705739605 and sf0706196757 for Pro) were co-located in intervals 7–4 and 7–5. Although large numbers of QTLs for grain protein content were detected in the past; only one major QTL has been cloned (Peng et al., 2014). For the lead SNP sf0709447538, explaining 12.7% of the Glu variation and located to the same chromosome region as QTL (7–4 and 7–5), we developed NILs to confirm the QTL. The results showed that it was significantly difference in Glu between two NILs and suggested that the QTL was reliable. To further purify the genetic backgrounds and fine map the QTL, it is needed to backcross the NILs for recombinant screening. It has been reported that a quantitative trait locus (qγ27) affecting expression of 27-kDa γ-zein has been successfully cloned by GWAS and linkage mapping analysis (Liu et al., 2016).

We detected a lead SNP responsible for Alb in the Wx gene region (**Figure 5** and **Table 1**), This region is considered a hotspot of major QTLs (6-2; qPC-6; and pro6) for protein content in rice (Aluko et al., 2004; Wang et al., 2008; Lou et al., 2009; Yu et al., 2009). Analysis of the correlation among Alb, amylose content and quantity of 2.3 kb Wx RNA showed that Wx might influence Alb content, that needs transgenic experiment to verify the function of Wx. Additionally, Starchmetabolism-related genes, such as PUL (Fujita et al., 2009), ISA2 (Utsumi et al., 2011), GBSSII (Hirose and Terao, 2004), Flo4 (Kang et al., 2005), and AGPS2a (Akihiro et al., 2005) were associated with SSPs in different populations. Overexpression of albumin gene RAG2 significantly increased total protein content, prolamin, glutelin, and amylose content, but decreased total starch (Zhou et al., 2016). Carbon and nitrogen metabolisms show a cooperative modification, and consequently, validation the function of these associations might help to better understand the genetic relationship between SSP and starch contents in rice grain.

### Application in the Improvement of Grain Quality in Rice

Grain with good quality could be developed by regulating the SSP content. SSP is a typical quantitative trait typically affected by environment (Shewry, 2007). Combination of conventional breeding and molecular techniques, e.g., markerassisted selection (MAS), may provide a more efficient approach for improving the SSP content of the grain than classical breeding alone (Zhang et al., 2008). The Lgc1 mutation has been used to development new low easy-to-digest protein rice varieties with low glutelin content and high prolamin content, which is useful for patients with chronic renal failure (Mochizuki and Hara, 2000; Nishimura et al., 2005; Morita et al., 2009). O2 mutation could increase free lysine and tryptophan levels by reducing the synthesis of zeins in maize, which is useful for QPM breeding (Mertz et al., 1964; Liu et al., 2016). Genes and possible causative SNPs identified in the present study could be used as potential targets for rice grain nutritional quality improvement. Glutelin is the most easily to digest and contains high lysine (Tanaka et al., 1995). The lead SNP sf0709447538 co-located in QTLs (7–4, 7–5, and 7–6) was validate to have an effect on glutelin, which could be targets for map-based cloning of the candidate genes to illuminate the molecular mechanism of glutelin and improve grain nutritional quality by MAS. RM1 and RP6 are adjacent genes on chromosome 7, and with higher levels expression and higher Pro, hap2 of RP6 and hap3 of RM1 are distributed mainly in the indica subpopulations (**Figure 4** and Supplementary Table S6). Therefore, the region covering RM1 and RP6 from japonica subpopulation can be a promising target for reducing Pro content and achieving better nutritional quality in indica cultivars. Wx may be a key gene for regulating the content of Alb and amylose content. The availability of the Wx gene sequence provides the possibility of improving the protein content via Wx gene modification. Different haplotype combinations of candidate genes for SSP would produce grains with different eating and nutritional quality. Genes and possible causative SNPs identified in the present study could be useful for breeding rice cultivars with favorable eating and nutritional quality.

### CONCLUSION

In the present study many associations were identified for five SSP traits using GWAS in two environments. We detected novel associations, known SSP genes, and SNPs adjacent to known starch-metabolism-related genes. We also analyzed haplotypes of known grain-quality-related genes. Our results suggested that

GWAS was an effective way to identify genes for rice SSP traits and the level of 3.3 kb Wx pre-mRNA is positively correlated with Alb content, providing new insights into the genetic basis of rice quality. Overall, we provided useful information that could be used in future gene functional studies and rice quality improvement.

### AUTHOR CONTRIBUTIONS

YH designed and supervised the experiments. PC, ZS, GL, and BW performed all the phenotypic evaluations. LM, YiL, PC, WD, and BP performed analysis and interpretation of the data. PC wrote the paper. GW, YaL, and DZ provided rice germplasm samples. GG, QZ, JX, and XL participated in the field management and logistic work.

### FUNDING

This work was supported by grants from the from the Ministry of Science and Technology of China (Grants

### REFERENCES


2016YFD0100501, the National Program on R&D of Transgenic Plants (2016ZX08009-003-004 and 2016ZX08001002-002), the National 863 Project (2014AA10A604), and the earmarked fund for the China Agriculture Research System (CARS-01-03) of China.

### ACKNOWLEDGMENTS

We thank Xufeng Bai, Peng Yun, Qiuxiang Luo, Haijiao Dong, Hao Zhou, Pingbo Li, Quanxiu Wang, Dujun Wang, Yuanyuan Zheng, Zhongmin Han, Xiaokai Li, Hu Zhao, and Wei Chen for editing, suggestions, and assistance.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018.00612/ full#supplementary-material

(Oryza sativa L.) and the function of PUL on starch biosynthesis in the developing rice endosperm. J. Exp. Bot. 60, 1009–1023. doi: 10.1093/jxb/ern349



amylopectin biosynthesis in rice endosperm. Plant Physiol. 156, 61–77. doi: 10.1104/pp.111.173435


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Chen, Shen, Ming, Li, Dan, Lou, Peng, Wu, Li, Zhao, Gao, Zhang, Xiao, Li, Wang and He. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Banana21: From Gene Discovery to Deregulated Golden Bananas

Jean-Yves Paul<sup>1</sup> \*, Robert Harding<sup>1</sup> , Wilberforce Tushemereirwe<sup>2</sup> and James Dale<sup>1</sup>

<sup>1</sup> Centre for Tropical Crops and Biocommodities, Queensland University of Technology, Brisbane, QLD, Australia, <sup>2</sup> National Agricultural Research Organisation, Kampala, Uganda

Uganda is a tropical country with a population in excess of 30 million, >80% of whom live in rural areas. Bananas (Musa spp.) are the staple food of Uganda with the East African Highland banana, a cooking banana, the primary starch source. Unfortunately, these bananas are low in pro-vitamin A (PVA) and iron and, as a result, banana-based diets are low in these micronutrients which results in very high levels of inadequate nutrition. This inadequate nutrition manifests as high levels of vitamin A deficiency, iron deficiency anemia, and stunting in children. A project known as Banana21 commenced in 2005 to alleviate micronutrient deficiencies in Uganda and surrounding countries through the generation of farmer- and consumer-acceptable edible bananas with significantly increased fruit levels of PVA and iron. A genetic modification approach was adopted since bananas are recalcitrant to conventional breeding. In this review, we focus on the PVA-biofortification component of the Banana21 project and describe the proof-of-concept studies conducted in Australia, the transfer of the technology to our Ugandan collaborators, and the successful implementation of the strategy into the field in Uganda. The many challenges encountered and the potential future obstacles to the practical exploitation of PVA-enhanced bananas in Uganda are discussed.

Keywords: East African highland banana, staple crop, Uganda, micronutrient deficiency, vitamin A deficiency, pro-vitamin A, carotenoids, biofortification

### INTRODUCTION

Vitamin A (VA) or retinol is an important nutrient which supports vital physiological and developmental functions. Since it cannot be synthesized de novo, VA has to be acquired through a diversified diet (Fraser and Bramley, 2004; Kimura et al., 2007; Beyer, 2010; Fitzpatrick et al., 2012). Whereas retinol is derived directly from animal sources, plant-derived pro-vitamin A carotenoids (PVACs) such as α- and β-carotene must first be converted into retinol by the body (van den Berg et al., 2000). The majority of populations living in developing countries depend on starchy food staples such as cassava (Manihot esculenta), maize (Zea mays), potato (Solanum tuberosum), rice (Oryza spp.), plantain, and banana (Musa spp.) which unfortunately are largely deficient in essential micronutrients such as PVACs.

Vitamin A deficiency (VAD) causes a number of VAD disorders including night and total blindness, premature death (Sommer and Vyas, 2012), and reduced immunity leading to increased risk of childhood infections and high infant mortality (Herbers, 2003; Bai et al., 2011). VAD affects an estimated 190 million pre-school children worldwide, most of whom live in developing countries, with a reported 5.17 million registered cases of clinical or severe levels of VAD (WHO, 2009). It is also estimated that 250,000–500,000 children become blind due to VAD each year,

### Edited by:

Felipe Klein Ricachenevsky, Universidade Federal de Santa Maria, Brazil

### Reviewed by:

Michael A. Grusak, Children's Nutrition Research Center, United States Hamid Khazaei, University of Saskatchewan, Canada

> \*Correspondence: Jean-Yves Paul jy.paul@qut.edu.au

### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 13 February 2018 Accepted: 09 April 2018 Published: 26 April 2018

### Citation:

Paul J-Y, Harding R, Tushemereirwe W and Dale J (2018) Banana21: From Gene Discovery to Deregulated Golden Bananas. Front. Plant Sci. 9:558. doi: 10.3389/fpls.2018.00558

**170**

half of whom die within 12 months of losing their sight (WHO, 2009; Barber et al., 2012). In Uganda, 20% of children aged 6 months to 5 years and 19% of women aged 15–49 years suffered from VAD in 2006, predominantly in the low-income demographics within central Ugandan communities, which heavily rely on banana as a staple food (UDHS, 2006).

In the Great Lakes region of East Africa highland bananas (EAHBs) are an important food security crop and the main food staple (Adeniji et al., 2010). In Uganda, it is estimated that 75% of farmers grow bananas contributing to about 7% of global banana and plantain production (Ssebuliba et al., 2006). In addition, Uganda is the largest banana consumer in the world with an estimated per capita consumption of 220–250 kg/year (Tushemereirwe et al., 2006). In rural populations of Uganda where EAHBs form a major and sometime unique part of the diet, 20 µg/g dry weight (dw) β-carotene equivalent (β-CE) is the minimum amount of PVA required in banana fruit to provide 50% of the estimated average requirement (EAR) of VA (Paul et al., 2017). Despite their considerable biodiversity, most EAHB varieties grown in East Africa have low levels of essential micronutrients such as iron (Fe), zinc (Zn), and PVACs (Davey et al., 2009; Fungo et al., 2010). Overreliance on banana in this region, and in Uganda in particular, has contributed to the exacerbation of micronutrient deficiency-related illnesses.

Strategies such as diet diversification, supplementation, and food fortification have been used to help alleviate some of these ailments with varying levels of success (WHO, 2009; Gómez-Galera et al., 2010; Fitzpatrick et al., 2012; Bruins and Kraemer, 2013). However, whereas these interventions are very successful in an urban context where the target population is in the vicinity of service providers, they fail to reach rural communities that are most in need (Dalmiya and Palmer, 2007; Victora et al., 2008; WHO, 2009). As such, in Uganda, VADs in children are relatively low in urban areas but remain elevated in rural communities (UDHS, 2006).

More recently, biofortification has emerged as a complementary, cost-effective, and sustainable approach to deliver micronutrient-dense crops to the poorest and hardestto-reach communities. This can be achieved through either conventional breeding, where the necessary traits are available within the accessible "breeder's gene pool," or through genetic modification. Examples of successful conventional breeding approaches include the biofortification of maize (Egesel et al., 2003; Harjes et al., 2008), sweet potato (Mwanga et al., 2009), and cassava (Ceballos et al., 2012). These biofortified products are already being disseminated in various parts of the world including Uganda (Anderson et al., 2007; Hotz et al., 2012; Talsma et al., 2016). Arguably the most successful example of biofortification through genetic engineering is the development of Golden Rice (GR) (Ye et al., 2000; Paine et al., 2005). Considering the popularity of EAHBs in East Africa, biofortification of this food staple with enhanced levels of PVACs (or other micronutrients) is now believed to be the best long-term, sustainable, and cost-effective strategy to ease the burden of VAD in high risk populations of Uganda. The use of conventional breeding to develop PVA-biofortified EAHBs is constrained by their low fertility and the lack of high PVA EAHB varieties in the known gene pool. Further, any new varieties developed are unlikely to possess the attributes of locally preferred landraces of EAHBs. Genetic modification is therefore the fastest and most reliable approach to improve the existing preferred varieties.

Here, we review the foundations, goals, achievements, future prospects, and challenges of Banana21<sup>1</sup> , a project that undertook the challenges of alleviating micronutrient deficiencies in Uganda by enhancing the nutritional content of its staple food, EAHBs, through genetic modification.

### THE PROMISE OF BANANA21 – MORE NUTRITIOUS BANANAS MADE IN UGANDA, BY UGANDAN SCIENTISTS, FOR THE UGANDAN PEOPLE

Banana21 is one of four original projects funded by the Grand Challenges in Global Health (GCGH) program of the Bill and Melinda Gates Foundation (BMGF) to "Create a Full Range of Optimal, Bioavailable Nutrients in a Single Staple Plant Species." Banana21 is a collaborative research project between the Centre for Tropical Crops and Biocommodities at Queensland University of Technology (QUT) in Australia and the National Banana Research Program of the National Agricultural Research Organisation (NARO) of Uganda. The project aims to help "Alleviate VAD and iron deficiency anemia through the micronutrient enhancement of the staple food of Uganda, bananas."

From the outset, it was recognized that the success of the project depended on its objectives being widely understood and accepted in Uganda. As such, we considered it extremely important to train and empower young Ugandan scientists to deliver the project milestones thus ensuring the biofortified genetically modified (GM) bananas were generated in Uganda, by Ugandan scientists, from Ugandan varieties, and for the benefit of the Ugandan people. As a consequence, in addition to scientific discoveries and their applications, Banana21 has been a capacity building project where technology transfer continues to be at the core of every phase, milestone, and decision-making activity.

An overview of the Banana21 project strategy is shown in **Figure 1**. Initially, the proof-of-concept research was developed at QUT using the locally grown dessert banana cultivars "Cavendish." The QUT components included gene and promoter discovery, tissue culture and transformation, field trials of GM bananas, and downstream fruit sampling and analysis. The development and implementation of a comprehensive stewardship plan was also a key project component. Following proof-of-concept in Australia, the technology and know-how was to be continuously transferred to NARO in Uganda to generate PVA-biofortified local EAHB varieties. Further, it was necessary for the infrastructure to be implemented at NARO to ensure that the GM plants produced would be tested in the laboratory and the field with the consistency and rigor necessary to generate the data required for a GM product deregulation dossier.

<sup>1</sup>http://www.banana21.org/

### THE CHALLENGES AND ACHIEVEMENTS OF BANANA21 – TRANSGENIC BIOFORTIFIED EAHBs WITH SIGNIFICANTLY HIGHER FRUIT PVA CONTENT

### Phase 1 – Early Discovery and Technology Transfer

HDR, higher degree research.

At the commencement of the project in 2005, the only practical demonstration of PVA biofortification in a staple food crop was GR. The initial GR strategy involved the re-engineering of a carotenoid biosynthesis pathway in normally carotenoid-free rice endosperm by the endosperm-specific [glutelin 1 (gt1) promoter] expression of a daffodil-derived phytoene synthase (psy) transgene in combination with the constitutive (CaMV 35S) expression of a bacterial phytoene desaturase (crtI) gene. Although successful, the level of carotenoid accumulation in rice endosperm was still considered to be too low for practical exploitation (Ye et al., 2000). Subsequent research revealed that the origin of the psy transgene and choice of promoter were important factors affecting PVAC accumulation levels (Paine et al., 2005). This led to the development of Golden Rice 2 (GR2) whereby the use of a maize-derived psy transgene and bacterial crtI, both under the control of the gt1 promoter, resulted in high levels of PVAC accumulation (Paine et al., 2005). Based on this success, the GR2 strategy was deemed to be the most logical approach to develop VA-biofortified bananas. One of the major initial challenges was the identification of suitable transgenes and promoters for expression in banana. As such, the early phase of the project focused on designing and testing large numbers of expression constructs containing a suite of different promoters and transgene combinations.

Due to the paucity of information regarding transgene expression in banana fruit, it was necessary to assess a range of different promoters for their spatiotemporal activity in banana, particularly fruit. As such, several constitutive promoters were isolated from different sources in addition to promoters that controlled the expression of genes involved in banana fruit development. These were fused to the β-glucuronidase reporter gene (uidA), transformed into banana embryogenic cell suspensions (ECS) (Khanna et al., 2004) and transgenic plantlets regenerated. A list of all promoters tested, their origin, and specificity is presented in **Table 1**.

In combination/parallel with the promoter characterization study, a range of different transgenes were also assessed. Initially, the strategy used to develop GR2 was adapted to banana and the first generation of expression constructs was made to express the maize phytoene synthase 1 (ZmPsy1B73) transgene alone or in combination with the bacterial (Pantoea ananatis) carotene desaturase transgene (PaCrtI) and controlled by various promoters (**Table 1**). Our efforts also focused on isolating and using PVA-associated cisgenes from banana in the hope of minimizing gene silencing and thus achieving more stable gene expression over several generation. The use of banana-derived cisgenes/s was also considered to be more advantageous from a future deregulation perspective. The biodiversity of banana and plantain is huge especially in Southeast Asia and Papua New Guinea where bananas are believed to have originated (Perrier et al., 2011). As such, the ability of various cultivar to accumulate PVACs in their fruit is also very diverse (Davey et al., 2009; Ekesa et al., 2015). Of particularly interest to us was a small group of bananas originating from the Pacific called the Fe'i bananas (Musa troglodytarum) which accumulate extremely high levels of fruit PVACs (Englberger et al., 2003). One such Fe'i banana variety, "Asupina," became our model cultivar for not only understanding carotenoid metabolism and accumulation in banana fruit but also as the source of a cisgene encoding a phytoene synthase, MtPsy2a (Mlalazi et al., 2012; Buah et al., 2016). Following the cloning and molecular characterization of MtPsy2a, expression cassettes containing this gene with and without PaCrtI and controlled by various promoters were also constructed and used to generate transgenic "Cavendish" banana lines in order to assess the levels of PVA accumulation in fruit.

During Phase 1, an important training program was also initiated whereby several students from Uganda and Kenya commenced their higher degree research (HDR) doctoral studies at QUT. The research projects developed for these students focused primarily around banana tissue culture and transformation technologies so that the knowledge and experience gained in the QUT laboratories would be transferred back to Africa for the benefit of the project. Simultaneously,


TABLE 1 | List of promoters and genes tested in AFT-1 of the Banana21 project.

training of technical staff and capacity building around infrastructure and laboratory equipment began at NARO, Uganda. A major component of this training program was the development of protocols and standard operating procedure (SOP) for generating and transforming ECS of local banana varieties. From the beginning, the Ugandan component of the project focused on establishing ECSs from three popular Ugandan banana cultivars, namely "Nakitembe," "M9," and "Sukali Ndiizi" since these were considered the most appropriate target cultivars for biofortification. True EAHB varieties such as "Nakitembe" are the preferred cultural choice among Ugandans and are usually grown in the highlands where disease pressure is minimal. The disease-resistant EAHB hybrid "M9" is less popular but was chosen as it is more productive in the lowlands where diseases such as black Sigatoka are a serious limitation to banana production. "Sukali Ndiizi" was chosen because it is a small sweet banana which is popular among children, the most vulnerable target population.

### Phase 2 – Proof of Concept, Field Trials in Australia and Uganda

The banana fruit PVAC target levels necessary to deliver 50% of the EAR of VA in vulnerable populations was estimated at 20 µg/g dw β-CE. The PVA-biofortification proof-of-concept research was done in Australia using transgenic "Cavendish" and "Lady Finger" bananas as the models with the aim of obtaining the target fruit PVA levels with no changes in agronomical characteristics. One of the major limitations of the project was the lengthy timeframe from transformation of banana ECS through to fruit harvest which is approximately 30 months. This limitation precluded the serial testing of our expression constructs. Consequently, our strategy was to transform "Cavendish" and "Lady Finger" ECS in Australia with a large number of different expression constructs, regenerate transformants, and field trial all the plants in parallel. At the end of Phase 1, between 10 and 30 independents transgenic "Cavendish" banana lines for each of the expression constructs had been generated. Considering the very large number of lines to be tested, and the fact that bananas are a very large crop, it was not practical to conduct a trial in the glasshouse. With these limitations in mind, we assessed these lines directly in the field without prior glasshouse characterization. The first Australian field trial (AFT-1) of GM "Cavendish" banana lines was subsequently established in 2009. This trial contained all available independent lines for each construct but only a single plant per line. A total of 28 constructs were tested, 14 to test promoter activity (as promoter/uidA reporter gene fusions) and 14 to test the same promoters in combinations with three transgenes (MtPsy2a, ZmPsy1B73, and PaCrtI) encoding proteins involved in the biosynthesis of PVACs (**Table 1**). This was the first GM banana field trial in Australia and was conducted under a license issued by the Australian Office of the Gene Technology Regulator (OGTR)<sup>2</sup> .

Of the 14 promoter/uidA fusion combinations tested in the field, three promoters (Ubi, Exp1, and Aco) consistently conferred the strongest levels of GUS expression. The constitutive maize polyubiquitin (Ubi) promoter showed consistently strong activity from the earliest stages of fruit development through to maturity. In contrast, the banana fruitspecific expansin 1 (Exp1) and ACC oxidase (Aco) promoters were only active in the later stages of fruit development (Paul et al., 2017). When we subsequently investigated the accumulation of PVACs during fruit development, a similar trend was seen whereby constitutive expression of either

<sup>2</sup>http://www.ogtr.gov.au/internet/ogtr/publishing.nsf/Content/dir076rarmp-toc~ dir076rarmp-ch1~dir076rarmp-ch1s5

MtPsy2a or ZmPsy1B73 using the Ubi promoter increased PVA accumulation from the earliest stages of fruit development while PVAC accumulation was restricted to the late stages of fruit development when the same transgenes were tested under the control of either the Exp1 or Aco promoters. Analysis of fruit samples from plants transformed with MtPsy2a, ZmPsy1B73, and/or PaCrtI revealed that constitutive (Ubi) expression of MtPsy2a resulted in the highest fruit β-CE levels which reached almost 19 µg/g dw in the plant crop (**Figure 2A**). Although this was just below the target level, it nonetheless represented a 11-fold increase over wild-type banana PVA baseline levels and was highly encouraging (Paul et al., 2017).

A second Australian field trial (AFT-2) was planted in September 2012 to test the efficacy of new PVA biofortification constructs as well as to monitor transgene stability in multiple plants of selected lines from AFT-1 through three successive generations. The additional PVA genes and promoters tested in this second field trial (**Table 2**) were tested in combination with the MtPsy2a gene under the control of various promoters. Unfortunately, none of the new combinations tested were found to elevate fruit PVA levels as high as those achieved by the constitutive expression of MtPsy2a alone in AFT-1. However, an extremely interesting and important outcome from AFT-2 was the observation that fruit PVA levels increased over successive generations in some lines to levels exceeding the target. For example, the fruit PVA levels of a line containing Ubi-MtPsy2a was found to increase from 11.7 µg/g dw β-CE in the first generation to 75.1 µg/g dw β-CE in the fourth generation. Qualitative and quantitative phenotypic and agronomical data have been recorded for all the lines tested and has revealed some interesting trends (Paul et al., 2017). The origin of the transgene had the biggest impact with, for example, banana lines expressing the banana MtPsy2a appearing normal with no variation in critical agronomical features such as yield and cycle time. In contrast, although expression of the maize Psy1B73 gene resulted in increased fruit PVA levels, many lines with undesirable phenotypes such as stunting and photo bleaching were observed.

During Phase 2, scientists at NARO had established regenerable ECSs, as well as efficient transformation protocols, for two varieties, "M9" and "Sukali Ndiizi." These breakthroughs allowed the generation of transgenic banana plants from both varieties containing either the ZmPsy1B73 or the "Asupina" derived phytoene synthase 2a (MtPsy2a) transgenes. Since the results of AFT-1 were not available when this activity commenced, these gene were placed under the control of the banana-derived Exp1 promoter (Mbabazi, 2015). The transgenic lines generated from both cultivars were subsequently grown in a field trial at NARL-NARO, Kawanda Uganda in 2010 to assess the fruit PVAC levels. This first Ugandan field trial (UFT-1), conducted under authorization from the Ugandan National Biosafety Committee (NBC), was a very important milestone for Banana21 as it represented the first GM crop field trial in sub-Saharan Africa where the events had been created by local scientists from a national laboratory.

At the completion of Phase 2, proof-of-concept for PVA biofortification had been demonstrated in Australia using "Cavendish" and "Lady Finger" bananas and revealed that target fruit PVA levels could be achieved using the constitutive (Ubi) or fruit-specific (Aco) expression of MtPsy2a alone (Paul et al., 2017). Further, very few off-type traits, such as reduced yield and increased cycle time, were observed in the transgenic lines and the trait appeared stable over successive generations. Importantly, the field trial in Uganda (UFT-1) produced multiple lines with fruit PVA levels higher than their respective controls including one Exp1-MtPsy2a line of M9 with 33.1 µg/g dw β-CE (Mbabazi, 2015). With this knowledge, a new generation of plant expression vectors were made at QUT and two constructs containing Ubi-MtPsy2a and Aco-MtPsy2a

FIGURE 2 | (A) PVA-biofortified "Cavendish" banana in Australia and (B) PVA-biofortified (left) and wild-type control (right) "Nakitembe" EAHB in Uganda.


TABLE 2 | List of new promoters and genes tested in AFT-2 of the Banana21 project.

were transferred to NARO for transformation into EAHB varieties.

### Phase 3 – Product Development: Early Events Selection Field Trial in Uganda

This phase of the project (2012–2017) initially involved generating a total of 200 independent transgenic lines each of EAHB cultivars "M9" and "Nakitembe" containing MtPsy2a under the control of either the Ubi or the Aco promoters. Following molecular characterization by the now fully trained technical staff and highly qualified scientists at NARO, the transgenic lines were field planted (UFT-2) at NARL-NARO, Kawanda in August 2014 to identify suitable elite lines for further testing in multi-location field trials (MLTs). During those 5 years, the NARO team showed exceptional professionalism in learning and implementing good practices around generating, handling, and tracking GM products. The results from UFT-2, which will soon be published, clearly demonstrate that fruit PVACs can consistently accumulate at levels above the required target of 20 µg/g dw β-CE in GM-biofortified "M9" and "Nakitembe" without phenotypic alteration of the plants (**Figure 2B**).

From the current transgenic line selection trial at Kawanda, 10 elite lines each of "M9" and "Nakitembe" will be selected to progress through to future MLTs. The initial selection process began in 2017 by selecting lines with fruit PVA levels equal or greater than 20 µg/g dw β-CE at the full green developmental stage (harvesting stage in Uganda) and yield within 20% of non-GM controls plants. From this initial selection, molecular analysis was used to identify lines containing fewer than three copies of the integrated expression cassettes with a preference for single integrations. Although selection of "single copy" events is preferred in seed crops to produce homozygous lines that do not segregate for the transgenic trait in future generations, in a vegetatively propagated crop such as banana, it is only rationalized to increase the likelihood of events with "clean insert."

During this phase, NARO scientists also conducted a preliminary blind sensory panel test of traditionally prepared banana meal (matooke) using fruit from both non-GM and PVA-biofortified "M9" and "Nakitembe" to compare appearance and texture but not taste. Interestingly, 80% of the panelists (n = 15) rated fruit from one of the "M9" PVA-biofortified lines as the most preferred whereas fruit from its non-GM counterpart had the highest dislike proportion (29%).

On the basis of these encouraging results, Banana21 entered its fourth and final phase of funding in October 2017. This final phase is focused on generating all the data required for the deregulation of a GM PVA-biofortified EAHB in Uganda. The prospect of the release of the world's first deregulated GM banana developed from an African laboratory is as much exciting as it is daunting.

### THE FUTURE CHALLENGES OF BANANA21 – THE NOT SO LONG ROAD AHEAD TO DEREGULATION

From a technology perspective, the groundwork in Uganda has been completed and the exceptional results from UFT-2 have provided the ideal platform to further select elite lines to progress through to MLTs and ultimately to farmer release.

Prior to MLTs, critical data necessary for the compilation of a deregulation dossier must be obtained. Whole-genome sequencing of the pre-selected lines is necessary to allow the site(s) of transgene insertion into the host genome to be identified. This is necessary to ensure that (i) the transgene is intact, (ii) there are no new open-reading frames created, (iii) there is no disruption of endogenous open-reading frames, and (iv) no plasmid sequence has been integrated into the host genome. Finally, it is essential that the composition of the fruit from the deregulated lines is very similar to that derived from the wild type. Therefore, compositional analysis for food characteristics such as calories, calories from fat, carbohydrates, protein, ash, and moisture will be done, and lines with values >15% different to controls will be discarded.

The next phase of the project involves the identification of two lines (one lead and one reserve) each of "M9" and "Nakitembe" from the MLTs, and obtaining the necessary agronomic, biochemical, and molecular data to ultimately prepare the dossier for submission to regulators in Uganda for deregulation and general release. At the end of this phase of the project (December 2021), these four lines are expected to meet all the agronomic, biochemical, and molecular analyses and biosafety assessment required for deregulation in Uganda under the proposed Biosafety Bill.

One of the major hurdles for Banana21 is the lack of a regulatory framework for biotechnology in Uganda. Without a regulatory body controlling the safe application

of biotechnologies, Banana21 will not be able to release PVA-biofortified "Golden bananas" to farmers and consumers in need. After ratifying the Cartagena Protocol on Biosafety in 2002, it took 6 years for Uganda to approve a policy on Biotechnology and Biosafety in 2008. Under the current policy, the NBC supervises all GMO activities up to the stage of Confined Field Trial (CFT) under the supervision of the National Council of Science and Technology (UNCST) Act 1990. Therefore in 2012, the National Biotechnology and Biosafety Bill, 2012, was introduced into parliament to provide a regulatory framework and guide the implementation of modern biotechnology in Uganda to minimize any potential risks to the environment, human, and animal health. After 5 years, Uganda's Parliament passed the National Biotechnology and Biosafety Bill into law becoming the Biosafety Act, 2017 on October 4, 2017. The new law will not only benefit Banana21 but also a multitude of other biotechnology products developed in Uganda such as bananas with bacterial wilt resistance, drought tolerant maize, bollworm resistance and herbicide tolerant cotton, and new cassava varieties with resistance to Cassava mosaic and brown streak viruses.

Our initial target of 20 µg/g dw β-CE was calculated using a bioconversion factor of PVACs to retinol of 6:1 based on the results of a study using Mongolian gerbils (Bresnahan et al., 2012). Underestimating this bioconversion factor would raise the target above its current value. For this reason, Banana21 with financial help form HarvestPlus and the BMGF, commissioned a nutrition study at Iowa State University to determine a more accurate bioconversion ratio in humans. Although the results from this study are not yet available, some of the PVA-biofortified lines that have been developed under Banana21 have over four-times the initial target value with no yield penalty and could potentially be substituted if all other deregulation criteria are met.

The development of robust diagnostics for banana viruses was another important component of this project. These diagnostics form the basis of a banana virus indexing protocol that has been rolled out to ensure that the plantlets derived from the Banana21 project are virus tested, thus reducing the potential distribution of infected planting material to farmers.

Since the majority of banana growers in Uganda are subsistence farmers, the distribution strategy adopted in the future will need to minimize the cost of planting material while maximizing the rate of distribution. Initial propagation of the lines to be released will be done at NARO and the plantlets will be sent to small banana micropropagation laboratories and also used to establish small "mother gardens" for the initial production of suckers. A cost-effective and self-sustaining strategy for dissemination will then involve identifying "innovative farmers" that will be given suckers. For every sucker, they will be asked to give away two suckers to neighbors who in turn will be asked to give away two suckers for each one received under the scheme.

### REFERENCES

Adeniji, T. A., Tenkouano, A., Ezurike, J. N., Ariyo, C. O., and Vroh-Bi, I. (2010). Value-adding post harvest processing of cooking bananas

A key component of the next phase of the project will be the implementation of a comprehensive stewardship and communication plan. This includes (i) forming a Technical Advisory Committee (TAC) that meets regularly and provides scientific, strategic, and biosafety expertise, (ii) implementing and regularly updating SOPs, (iii) keeping accurate and safe records of the data with tools such as the BananaTracker software developed by QUT, and (iv) meetings with the Australian OGTR and Food Standards Australia and New Zealand (FSANZ) to seek advice on the requirements for deregulation if these lines were to be deregulated in Australia. The NARO team has also been involved in various communication activities in an attempt to educate the public and de-mystify the use of GMOs. Important stakeholders are targeted through workshops and information sessions, as well as various paper-based and audio-visual communication materials.

### FINAL REMARKS AND CONCLUSION

From the outset in 2005, Banana21 has been on a trajectory to develop lines of EAHBs with levels of fruit PVACs that would provide 50% of the EAR of VA with consumption of only 300 g per person per day. Based on the significant progress thus far, it is highly likely that the transgenic lines developed under Banana21 will be released by 2021 and have a significant impact in alleviating VAD in a sustainable way, especially in rural Uganda where bananas are a fundamental part of the culture. The PVA-enhanced, disease-resistant "M9" line will have the greatest impact in lower elevations of Uganda where the disease pressure is high, while the PVA-enhanced "Nakitembe" line will have greatest impact in the highlands where there is much lower disease pressure. The importance of banana as a food security crop (perennial nature, year-round production, and ability to cope with long periods of drought) associated with a low cost, farmer-driven distribution strategy should ultimately see "Golden bananas" adopted as a widespread and efficient VAD alleviating strategy in the next decade.

### AUTHOR CONTRIBUTIONS

J-YP and RH drafted the initial manuscript while WT and JD reviewed and provided the constructive criticisms.

### FUNDING

The Banana21 team project was, and still is, supported by a Grant from the Bill & Melinda Gates Foundation and the Department for International Development (United Kingdom).

Anderson, P., Kapinga, R., Zhang, D., and Hermann, M. (2007). "Vitamin A for Africa (VITAA): an entry point for promoting orange-fleshed sweetpotato to

<sup>(</sup>Musa spp. AAB and ABB genome groups). Afr. J. Biotechnol. 9, 9135–9141.

combat vitamin A-deficiency in sub-Saharan Africa," in Proceedings of the 13th ISTRC Symposium, Apartado, 711–720.


maize for plant breeding trials. Food Chem. 100, 1734–1746. doi: 10.1016/j. foodchem.2005.10.020


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Paul, Harding, Tushemereirwe and Dale. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fpls-09-00558 April 24, 2018 Time: 17:15 # 8

# Improving the Yield and Nutritional Quality of Forage Crops

Nicola M. Capstaff\* and Anthony J. Miller\*

John Innes Centre, Norwich, United Kingdom

Despite being some of the most important crops globally, there has been limited research on forages when compared with cereals, fruits, and vegetables. This review summarizes the literature highlighting the significance of forage crops, the current improvements and some of future directions for improving yield and nutritional quality. We make the point that the knowledge obtained from model plant and grain crops can be applied to forage crops. The timely development of genomics and bioinformatics together with genome editing techniques offer great scope to improve forage crops. Given the social, environmental and economic importance of forage across the globe and especially in poorer countries, this opportunity has enormous potential to improve food security and political stability.

#### Edited by:

Felipe Klein Ricachenevsky, Universidade Federal de Santa Maria, Brazil

### Reviewed by:

Zhipeng Liu, Lanzhou University, China Giovanna Attene, University of Sassari, Italy

#### \*Correspondence:

Nicola M. Capstaff nicola.capstaff@jic.ac.uk Anthony J. Miller tony.miller@jic.ac.uk

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 06 December 2017 Accepted: 06 April 2018 Published: 24 April 2018

#### Citation:

Capstaff NM and Miller AJ (2018) Improving the Yield and Nutritional Quality of Forage Crops. Front. Plant Sci. 9:535. doi: 10.3389/fpls.2018.00535 Keywords: forage, nutritional enhancement, grass production, legumes, breeding, management

## INTRODUCTION

Forage grasslands are used to feed livestock and globally it has been estimated that they represent 26% of the land area, and 70% of agricultural area (FAO, 2010). Such crops are significant economically, as the European example shows (see **Figure 1**). Forage crops are usually grasses (Poaceae) or herbaceous legumes (Fabaceae). Some tree legumes such as mulga (Acacia aneura) and leadtree (Leucaena leucocephala) are also grown in desert and tropical grasslands (Muir et al., 2011). In the tropics, popular grasses include Napier grass (Pennisetum purpureum), Brachiaria, and Panicum species. In the poorest parts of the world livestock production is critically important for smallholders' livelihoods. Sub-Saharan Africa is an example and frequently women maintain the livestock production systems (Njuki and Sanginga, 2013). In temperate climates, the main grasses include bentgrass (Agrostis spp.), fescue (Festuca spp.), ryegrass (Lolium spp.) and orchard grass (Dactylis spp.) or hybrids of these. For example, Festuca and Lolium hybrids has been developed from 1970s (Ghesquière et al., 2010) giving rise to crops such as Festulolium pabulare which combines the superior forage quality of Lolium multiflorum with the persistence and stress tolerance of Festuca arundinacea. Some maize (Zea mays) cultivars have been specifically bred for forage. The commonly cultivated herbaceous legumes are trefoil (Lotus corniculatus), medics (Medicago spp.), clover (Trifolium spp.) and vetches (Vicia spp.). Brassica forage species include cultivars of oilseed rape (Brassica napus) and kale (Brassica oleracea). Fodder beet (Beta vulgaris) is another temperate forage. The combination of forage crops grown in any country varies depending on climate and livestock needs, however, the perennial legume lucerne or alfalfa (Medicago sativa) is the most widely cultivated as it can be grown with both temperate and tropical grasses, or as a standalone crop. This is a huge topic to review as there are so many species grown across the world, therefore we have chosen to focus on a few examples, the tropical grasses Pennisetum and Brachiaria, and more prominently the temperate crops Lolium and alfalfa.

In an ideal world, we would all eat pulses rather than the animal products generated from them, as grain legumes are the food that offers the most sustainable future (Foyer et al., 2016). There is continued pressure from many groups to lower human consumption of animal products due to livestock efficiency issues and for human health (Cramer et al., 2017). There is a lack of reliable statistics for the proportion of adults adopting a plant-based diet, but it is estimated to be between 1 and 10% of the population in developed western countries such as within the European and United States (Mcevoy and Woodside, 2010) and studies support these diets as healthy and nutritionally adequate (American Dietetic Association and Dietitians of Canada, 2003). However, the consumption of livestock products can be regarded as important to a healthy diet due to their high nutrient density (CAST, 2013) regardless of the numerous efficiency and environment concerns (Di Paola et al., 2017), particularly true in developing countries where undernourishment incidences are estimated as ∼4–22% of the population (Alexandratos et al., 2006). Livestock production can convert non-edible crops such as the forages into human food, with sustainable intensification possible when inputs and outputs of the system are balanced (Derner et al., 2017).

Moreover, the cultural and social significance of livestock cannot be underestimated and the trend of increased global production is set to continue (Thornton, 2010). Livestock feature prominently across all cultures both in cuisine, but also music and literature. Additionally, in many developing countries the rearing of livestock such as cattle and goats are vital in times of hardship; many view animals as living 'piggy-banks,' that can for example pay the family school fees (Herrero et al., 2013). Therefore, in practice livestock production is set to continue throughout the world and forage crops will be grown for coming decades. Plant research has chiefly focussed on grain crops, but here we argue that there is enormous potential for improving forages. Improving the yield and nutritional quality of forage crops can help mitigate the unsustainable negative impacts of livestock production.

### FORAGE CROPS IN LIVESTOCK DIETS

Forage crops can be feed directly to livestock or can be processed by partial drying or pre-digestion. Because of this processing, animal feeds can be categorized as either bulky feeds or concentrates. Bulky feeds are also termed forage and are produced from grass, cereal and legume cropping as described above, such as alfalfa, Lolium or a mixture of the two. This forage can be provided to animals directly through grazing pasture land or in a processed form, such as hay (where water content is >15%) or dried (pelleted) biomass. Concentrates are generally cereal, oilseed and legumes seeds, or bi-products of their preparation for human food, biofuel and textile. They can also include high energy feedstuffs such as sugar-rich crop molasses and fats of animal origin, for example fish by-catch discards. In industrialized countries, production of both these categories of feed can surpass the amount produced for plant-based food for human consumption; in United States over double dry matter per-capita per year (DM cap/yr.) is produced for animal feed than for foodstuffs (Krausmann et al., 2008).

Livestock diet can therefore be exclusively forage or largely forage with concentrate supplementation. Concentrate supplementation is used to compensate nutritional deficiencies in the forage supply, increase animal performance such as milk production or at particularly challenging periods of development, for example calving. Due to most livestock diet being of forage this review focuses on the main crops used worldwide and will not discuss concentrates. The amount plant science has contributed to improvements in concentrates has been underappreciated and undervalued in literature, however, the role these crops have on livestock production has been reviewed previously (Erb et al., 2012).

Forage crops can be grown in mixed species cultivation to provide nutritional and environmental benefits. By offering livestock mixed grazing pastures or blending feeds, nutritional quality can be enhanced. For example, alfalfa is the highest-yielding perennial forage legume and produces more protein per unit area than other forage legumes and so can be grown alone or in combination with a range of different grass species. Well-managed alfalfa is normally grown successively for 3 or more years, but if harvested too late in the season the crop cannot survive the winter (Bélanger et al., 2006).

### FORAGE NUTRITIONAL CONTENT

### Digestibility

The nutritional status of a forage crop depends upon the concentration (and ratios) of carbohydrates, proteins, and lipids. The composition of these organic nutrients determines the digestibility (D-value) of each crop which along with mineral and vitamins provides the amount of energy which can be derived by the animal (ME measured in MJ/kg DM) (Osbourn, 1980). Such calculations are becoming increasingly prevalent when growers are deciding which crop to grow based and particularly dependent on if the animal is non-ruminant or ruminant.

In forage crops 50–80% of DM is carbohydrate; if this percentage is too low then supplements of grains can be added. The primary types of carbohydrate are the insoluble structural saccharides cellulose and hemicellulose, or the storage forms such as starch and water-soluble polymers (e.g., fructans). These are degraded into simple sugars through cleavage of glycosidic bonds, either by the animal itself (non-ruminant and ruminants) or via microbial digestion and subsequent animal absorption (ruminants only). Different ratios of carbohydrates within the forage crop will have altered downstream digestibility for the animal, especially if the cell-wall structure constrains digestion by the microbial population or limits plant cell wall penetration (Weimer, 1996). Although lignin, a polyphenolic compound within forage, is not a carbohydrate, it has a dramatic impact on the digestibility of cellulose hemicellulose; lignin binds with structural carbohydrates and cell wall proteins and reduces nutrient availability. For forages increased lignin concentration in the growing crop will increase the percentage of indigestible DM. Of the major forage crops grown globally grasses, particularly Lolium perenne, have high digestibility due to high soluble sugar content alongside low lignin content (Ruckle et al., 2017).

Animal digestion of simple carbohydrates produces monosaccharides which can be readily metabolized. In ruminants, only microbial digestion of structural carbohydrates produces simple sugars which are subsequently metabolized to pyruvate. Pyruvate is absorbed by the animal and is metabolized further into volatile fatty acids (VFAs) which are a major energy source, (Bergman, 1990). Ruminants absorb VFAs in their rumen, and the rate of this is dependent on the concentration of individual VFAs, rumen pH and the absorptive area in the ruminal lining.

### Protein

Nitrogen (N) availability to animals is predominantly from forage proteins and are estimated using crude total protein Kjeldahl measurements. Protein is usually abundant in the major form of Ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO), although relative amounts vary between species (Wallace et al., 1997). This is especially true when comparing content in grasses with herbaceous legumes, with red clover (Trifolium pratense), white clover (Trifolium repens) and lucerne (Medicago sativa) grown widely due to their high protein value (Ruckle et al., 2017). Again, lignin will severely affect the digestibility of protein. Some micronutrients like proanthocyanidins or condensed tannins also change the digestibility of protein, but they inhibit protein degradation through binding. This can be advantageous as rapid protein degradation is causative of bloat, however, too high a tannin content will mean protein passing through the digestive track is unabsorbed and therefore a loss in nutrition value (Lees, 1992; Piluzza et al., 2014). This means there is a balance between reduced bloat and animal productivity (Mueller-Harvey, 2006). All grasses contain little or no proanthocyanidins, whereas many legumes especially big trefoil (Lotus pedunculatus) and Sericea lespedeza (Lespedeza cuneata) can have levels as high as 18% DM (Barry and Manley, 1984; Mueller-Harvey, 2006). Other N-containing compounds can be found in forage such as nucleic acids, nitrate and ammonia (Wallace et al., 1997).

### Lipids

Lipids in forage crops are mostly found as polyunsaturated fatty acids (PUFAs) in the range of 10 – 30 g kg−<sup>1</sup> (Hatfield et al., 2007) of which the most abundant is α-linolenic acid [62% total lipids (Clapham et al., 2005)], with linolenic and palmitic acid also being present (Harfoot and Hazlewood, 1988). These dietary lipids are important in final animal product quality; forage diets with lower PUFA levels than cereal diets can produce leaner meat (Wood et al., 2004; Van Elswyk and McNeill, 2014). Moreover, fresh forage has been shown through numerous studies to produce milk with lowered PUFA content and increased trans-fatty acids (Elgersma et al., 2006; Chilliard et al., 2007). Studies have been used to profile PUFAs across forage species, with grasses tending to have more α-linolenic acid when compared to legumes and legumes in turn having higher linolenic acid content (Boufaïed et al., 2003). Striking differences in PUFA content can be seen within species through profiling cultivars, and moreover the harvest period and its environment (Elgersma et al., 2003; Clapham et al., 2005). For example Lolium perenne, Festuca pratensis (meadow fescue), and Festulolium hybrids of the two have been shown to vary not only between species at the beginning of their growth season, but more prominently between individual cutting regimes (Dewhurst et al., 2001).

### Trace Elements

Minerals and trace elements from forages are important for maintaining livestock health. As there is a move toward using fewer antibiotics in animal production the nutritional balance of feed takes on additional importance. Zinc is particularly important for the immune system and supplements can be added to animal feed, but addition of too much results in wasteful excretion, reviewed in Brugger and Windisch (2015). Contrastingly, avoiding accumulation of toxic minerals can also be important for forage crops. Getting the balance right is crucial as low levels of selenium can be beneficial for livestock, but high concentrations are toxic (Zhu et al., 2009). Some elements accumulated in plants can make them unpalatable for livestock, but the ability of forage crops to grow fast and quickly recover from cutting makes them ideal crops for phytoremediation [e.g., Napier grass, (Ishii et al., 2015)].

### Biomass Production

Probably the most important trait of any forage crop is rapid biomass production, as crops are either cut or grazed directly, and nutritional quality depends on the rate of biomass production. Intensive production with faster growth often decreases this nutritional, but this depends on the species grown and some cultivars have better recovery from defoliation. Plant height correlates well with biomass for most crops (e.g., maize) and this factor together with ground area cover are the criteria underpinning methods to assess yields (Freeman et al., 2007).

Many plant species can be grown for forage production, but the ability of the shoot meristem to respond with increased growth after cutting is essential. In some forage species, aboveground grazing or cutting has been correlated with increased root exudation (Paterson and Sim, 1999). This flush of

carbon release by roots can stimulate rhizosphere microbes that in turn help to mobilize soil nutrients to sustain aboveground regrowth. Maintaining an optimal nutrient and water supply is very important for forage biomass production. For example, the importance of N supply for re-growth after cutting grass has been demonstrated (Dawson et al., 2004). Furthermore, the previous N status of alfalfa influences its regrowth ability (Meuriot et al., 2004, 2005).

### IMPROVING FORAGE CROPS

### Cultivar Breeding

Due to the relatively recent cultivation of forage crops compared to other agricultural plant species, there were few improvements before 1900. Recently, agricultural trends and the global economic importance of forages, mean new cultivars have been bred. These improvements are helped by many closely related wild populations which can be used in development of new lines (Boller and Green, 2010). The most desirable improvements are increasing dry matter yield (DMY), crop durability and resistance to diseases particularly by pathogenic fungus and pests particularly nematodes, digestibility of DM, and nutritional content of this tissue. Arguably the greatest improvements have been made in breeding of Medicago spp., Trifolium spp., Lolium, and Festuca. Large scale breeding programs include testing of these crops, such as NE1010, a multistate cooperative effort of 15 institutes across 12 North-eastern states of United States and Canada (NIMSS, 2017). Similar tropical grass breeding programs include the Brachiaria partnership between the International Centre for Tropical Agriculture based in Colombia (C.I.A.T.), the Ugandan National Livestock Resources Research Institute (NaLIRI), the Tanzania Livestock Research Institute (TALIRI), the Institute of Agricultural Research of Mozambique (IIAM) and the Brazilian Agricultural Research Corporation (E.M.B.R.A.P.A.) (CIAT and CGIAR, 2015) which is being conducted across Eastern and Southern Africa.

Breeding programs for forage crops are fraught with difficulties. Individual plants have high genotypic and phenotypic heterogeneity with many species being polyploid, a problem which is exacerbated by in-breeding across many grasses, and few agronomic traits being linked to distinct genes (Poehlman, 1987; Vogel and Pedersen, 1993). Studies have focussed on this problem in specific legumes (Jahufer et al., 2002; Riday and Brummer, 2007; Collins et al., 2012; Luo et al., 2016) and grasses (de Araüjo et al., 2002; Piano et al., 2007; Blackmore et al., 2016). Regardless of these problems there have been some major developments in breeding lines for forages, especially in Medicago and Lolium. **Figure 2** shows a brief historical timeline of Lolium cultivation, and includes the current breeding regimes for grasses; future breeding possibilities are also included and discussed in later sections. One of the most interesting breeding developments is the exploitation of closely related species of Lolium and Festuca (Thomas and Humphreys, 1991; Humphreys et al., 2003) to create hybrid Festulolium cultivars. These cultivars have the high quality characteristics of Lolium combined with the stress tolerance and persistence found in Festuca (Ghesquière et al., 2010). Backcrossing of Festulolium have generated novel hybrids with more stable protein content when compared to parental lines (Humphreys et al., 2014). Advances in phenotyping are making it easier to include the quantification of characteristics in the field; such as high level imaging of growing crops to accurately determine later traits like biomass (Walter et al., 2012).

New cultivars are being helped by advances in sequencing methods that can provide more transcriptomic data (Barrett et al., 2009; Pfeifer et al., 2013; Yates et al., 2014), including the identification of SNPs which may be investigated to improve Lolium (Blackmore et al., 2016) and Trifolium (Nagy et al., 2013). Draft genomes for such crops and cultivars are becoming increasingly common (Byrne et al., 2015; De Vega et al., 2015; VanBuren et al., 2015) as well as more evidence that model species like Brachypodium can direct research (Brkljacic et al., 2011; Rancour et al., 2012). Such research is providing clues to candidate genes which could be used for nutritional enhancement.

### Candidate Genes for Nutritional Enhancement

Identification of potential candidate genes is usually through quantitative trait loci (QTL) analysis or marker-assisted selection (MAS) provided from the above completed genomes. Those identified are studied in relation to biomass and growth traits; in M. sativa QTL has been used for lodging resistance and vigor (McCord et al., 2014), plant height and regrowth following harvests in association with MsaciB (Robins et al., 2007), candidate gene analysis for flowering and stem height through CONSTANS-LIKE (Herrmann et al., 2010) and biochemical markers of ROS resistance genes for drought tolerance correlated to DM (Maghsoodi et al., 2017). The expression of other ROS associated genes of the Iron-Superoxide Dismutase family (Myouga et al., 2008) have also been linked to increases in DM in both the legume M. sativa (McKersie et al., 2000) and grass Lolium cultivars (Warnke et al., 2002).

In Lolium, transcriptomics showing differentially expressed genes between wild-type and a dwarf mutant enabled identification of three key genes associated with dwarfism (Li W. et al., 2017), which were subsequently used for forward screens. Markers are used to infer both phenotypic traits and to track inheritance to aid breeding. For instance chloroplast SSRs have been investigated in Lolium through a similar technique as above (Diekmann et al., 2012). A thorough re-annotation of the model forage M. truncatula genome has also identified hundreds of small, secreted peptides coded by both macronutrient-responsive and nodulation-responsive genes, which could aid reverse genetics for improving many forage crops, especially M. sativa (de Bang et al., 2017). Iterative mapping software such as BioMercator (Sosnowski et al., 2012) has been used in Lolium to perform meta-QTL analysis using readily available published data (Shinozuka et al., 2012), consequently providing new candidate genes from previous work including orthologs of rice amino acid biosynthesis genes and a marker for reproductive traits, showing how new algorithms

can exploit old data. Moreover, BioMercator used to decipher flowering time and height in M. truncatula (Julier et al., 2007) directly implicated the above research into CONSTANS-LIKE in M. sativa (Herrmann et al., 2010; Julier et al., 2010). Such potential ease for transferring model plant knowledge to forage crop research is further discussed below.

Despite the need to ensure optimal nutritional content especially in the end-product feed, rapid vegetative biomass accumulation is the most desirable trait of a good forage crop, especially those which undergo extensive cutting throughout the growing season. Due to this phenomenon, candidate genes for improving the crops are associated with either photosynthesis or nitrogen use efficiency (NUE). More generally for resistance to biotic and abiotic stresses there is also a huge opportunity for improving traits in forage crops using our genetic knowledge from model plants (see **Figure 3**). **Figure 3** summarizes some of the traits that can be considered for all forage crops.

Such improvements in traits could be aided by achievements in transformation and genetic marker techniques. Reproducible and high efficiency transformation has been developed for temperate cultivars of Festuca (Wang and Ge, 2005; Zhang et al., 2006) and Lolium (Bajaj et al., 2006; Badenhorst et al., 2016); and more recently for some of the tropical grasses such as Pennisetum (Gondo et al., 2017) and Brachiaria (Cabral et al., 2015). Some examples of gene editing forage crops to confer stress tolerances have been successful in aiding both biomass increases but also nutritional quality. Transformation of M. sativa with the Arabidopsis Enhanced Drought Tolerance1 gene produced plants with not only increases in root length, shoot height and vegetative biomass, but also increases in proline, soluble sugar and chlorophyll content under drought stress when compared to wild-type (Zheng et al., 2017). Importantly these increases were shown both in the laboratory but also in field conditions. This study also identified the increased expression of many interesting genes, including M. sativa Heat Shock Protein23 (HSP23), a gene already shown to enhance abiotic stress tolerance in both Nicotiana tabacum and Festuca (Lee et al., 2012a,b) along with other members of the MsHSP family (Li et al., 2016; Li Z. et al., 2017). Similarly, for the Ethylene Response Factor (ERF) family studies have been shown that introducing the M. sativa gene into other plants can confer enhanced resistance to salinity; MsERF9 and MsERF11 in Nicotiana and Arabidopsis, respectively (Chen et al., 2012a,b).

### Protein and N Budget

Forage NUE is a target for breeding, particularly as protein content of crops is so valuable. Protein accumulation is linked to N status and when the supply is supra-optimal greater storage occurs. When compared with grain crops that have been bred for high seed starch, forage crops often require N in greater amounts due to their increased growth, storage capacity and higher fiber content (Parsons et al., 1991). For forage crops, it is the leaf

tissue biomass that is harvested rather than grains/roots/tubers. Principally NUE for forage crops can be based on N utilization efficiency (NUtE) as we are interested in the highest achievable biomass of the shoot which will form the content to be dried for feed production (Xu et al., 2012). Not only does this include biomass, but also the relative N levels in this tissue; it is not enough to only have a high yield of biomass in the shoots, it must also yield optimal amounts of N. Moreover, when looking at the effect of fertilizer use we are also interested in how both the biomass and N status change on application and thus also N uptake efficiency (NUpE). Forage crops offer challenges for NUE as there is a requirement for optimal yield of shoot biomass with a high N content (NUtE) while also optimizing N fertilizer acquisition (NUpE) throughout the growth season.

refinement method and bootstrapping = 100. The phylogeny was built using Newick format in iTOL v3.4.3 (Letunic and Bork, 2007) and a radial phylogenetic tree produced. The tree was color coded to show model species, human food crops and forage crops for both legumes and grasses, although it should be noted that many can overlap in their uses.

As NUE is an important criterion for biomass improvements, many genes relating to N acquisition or metabolism have been the subject of study in model systems. Additionally, genes important in carbon metabolism have also been the focus, due to the links between C:N ratios for plant growth (Jaradat et al., 2009). Despite the long evolutionary divergence between grasses and legumes, many key candidate genes have high genetic similarity, meaning one can use known genes which effect a trait in a forage crop from one species and investigate it within another. For example, a range of vegetative N storage proteins have been identified and the reviewed for leaves (Muntz, 1998) and roots (Bewley, 2002). To illustrate this further, the phylogeny in **Figure 4** is the known and predicted coding sequences for rbcS including the model species Arabidopsis thaliana, many significant grass and legume crops. Many of the forage crops have high similarity in their coding sequence to more well-studied crop species. For example, Medicago and Trifolium rbcS sit closely to the legume species which have their genomes sequenced [Cajanus cajan, Cicer arietinum, Glycine max, Lotus japonicus, Medicago truncatula, and Phaseolus vulgaris (Jacob et al., 2016)]. Such sequences can provide a wealth of potential genes of interest for breeding programs (Araújo et al., 2015; Rauf et al., 2016). For

example, investigation of Heat Shock Protein in M. truncatula, found a homologous HSP70 in M. sativa and had a substantial role in stress tolerance when conferred to A. thaliana (Li Z. et al., 2017).

In forage crop vegetative biomass, the most important nutrients for livestock are proteins and water-soluble carbohydrates (WSCs), and ideally the post-harvest quality of these should be maintained. There has been considerable interest in developing organ specific proteome reference maps for stems and leaves. The dominant proteins in these tissues are photosynthetic enzymes such as RuBisCO and RuBiCO small unit (rbsS), which for M. truncatula make up ∼28.9% of leaf tissue, or other carbon-fixation genes for example glyceraldehyde 3-phosphate dehydrogenase and triose phosphate isomerase, with structural protein such as lignin biosynthesis being more concentrated in stems (Watson et al., 2003). As the D-value of forage is mostly linked to cell wall concentration and a reduction of this can aid digestibility (Jung and Allen, 1995; Jung et al., 2012), some proteomes have looked even more specifically at such tissues (Gokulakannan and Niehaus, 2010).

Some research has focussed on transgenic approaches to increase and enhance amino acids and proteins. As many forages have low concentrations of the sulfur-containing amino acids of methionine and cysteine, both important in animal and human nutrition (Ball et al., 2006), some studies have specifically aimed at increasing these levels by over-expression. These have included using lupins (Lupinus albus) (Molvig et al., 1997; Tabe et al., 2010) and soybean (Dinkins et al., 2001; Tabe and Droux, 2002), used as forage sources.

Apart from cultivar differences which can be improved with breeding programs or specific transgenic approaches, the most significant changes in nutritional content is due to stresses (Araújo et al., 2015). Consequently, stress proteomes have also been used for vegetative tissue; lupin stem proteins have been analyzed under water stress to show increases in serine protease and cysteine protease required for remobilization of proteins (Pinheiro et al., 2005); in grasspea (Lathyrus sativus) seedlings under either salinity, low temperature or ABA stress gave rise to the identification of 48 stress-responsive proteins (SRPs) which include those important dominant proteins discussed above (Chattopadhyay et al., 2011); in M. sativa drought conditions showed remobilization of RuBisCO-derived N could compensate for the decreases in N assimilation (Aranjuelo et al., 2011). Moreover, through harvesting regimes, forage crops undergo extreme stress which has shown to cause the remobilization of vegetative storage proteins (VSPs) to boost new shoot regrowth in both Medicago and Trifolium as well as being important for cultivars with better cold tolerance (Avice et al., 2003), whereas Lolium has shown how defoliation increases the relative proportions of certain proteins, particularly asparagine and glutamine (Bigot et al., 1991).

Finally, the N consumed by livestock is recycled and increasing ruminant productivity is a major target for as the conversion of plant to microbial protein is inefficient. It was estimated that as much as 70% of the plant N eaten by animals for milk or meat production is excreted as ammonia or urea to the environment (MacRae and Ulyatt, 1974; Kingston-Smith et al., 2008; Kingston-Smith et al., 2010). Furthermore, the process of rumen fermentation is important for the generation of greenhouse gasses like methane (Bannink et al., 2008; Dijkstra et al., 2011).

### Rhizosphere Microbiome

The impact of genomics extends beyond the crop plants to their environmental interface. For example, the rhizosphere microbiome is likely to be a future target for improving the nutritional quality of crops. Epiphytic bacteria living on and in the plant, may be important for crop health and nutrition, and some microorganisms can fix atmospheric N within legume root systems. Bacteria living with plants may be able to assist in digestion and absorption of forage eaten by livestock. These bacteria may improve the uptake of trace elements in the animal gut by the production of specific binding molecules and/or siderophores. In the soil, the rhizosphere microbiome is important for nutrient cycling and uptake, particularly in low input systems like those grown in the tropics. The inoculation of new forage crops with beneficial microorganisms is likely to be a target for research and use in future crops, coupled with rhizosphere microbiome research of root exudate composition.

The root is known to directly modify the rhizosphere population by altering the chemical constituents of root exudates. For example, the roots of the tropical grass Brachiaria specifically produce a chemical shown to inhibit nitrifying bacteria and to specifically block ammonia-oxidizing pathways in soil bacteria, the first step in the process of converting ammonium to nitrate (Byrnes et al., 2017). Soil ammonia-oxidizing bacteria quickly convert urea or NH<sup>4</sup> <sup>+</sup> fertilizer to NO<sup>3</sup> <sup>−</sup>. Soil N form is fundamental for crop acquisition, as NO<sup>3</sup> <sup>−</sup> is mobile and readily leached while NH<sup>4</sup> <sup>+</sup> binds. Nitrification inhibitors have been identified in root exudates from several legumes and grasses including sorghum and rice, but by far the largest activity was detected in the tropical grass Brachiaria humidicola (Subbarao et al., 2009). In rice, the ability of root exudates to inhibit nitrification varied between cultivars from 5 to 50%, but was not significantly higher in three ancestral lines (Tanaka et al., 2010). The biological nitrification inhibitor (BNI) activity of root exudates has been assayed using a recombinant luminescent reporter ammonia-oxidizing bacteria Nitrosomonas europaea (Subbarao et al., 2006). In Brachiaria, roots exudate the cyclic diterpene "brachialactone," (Subbarao et al., 2009); brachialactone has a 5-8-5-membered ring system and a γ-lactone ring and contributed to 60–90% of the BNI activity released from the roots of this tropical grass. This exciting example offers the potential for transferring this trait to other forage crops to improve NUE. In the future synthetic pathways to produce plant nitrification inhibitors will be fully elucidated, providing the opportunity to capture this trait in forages and transfer to other crops to improve yield and nitrogen acquisition.

### Digestibility

As protein digestion and uptake in livestock is directly related to energy availability (ME) (Nocek and Russell, 1988; McCarthy

et al., 1989) it is important to increase WSC in many forage crops, especially grasses (Miller et al., 2001). In Lolium WSCs include fructans which are the most important storage polysaccharide and thus improved metabolism of fructan from sucrose can help improve the D-value (Chalmers et al., 2005). Use of distinct Fructan:Fructan 6G-fructosyltransferase sequence variants has shown to increase fructan levels at warmer temperatures in Lolium, thus hoping to aid development of high sugar-content grasses even at changing climates (Rasmussen et al., 2014). The amounts of WSCs are strongly associated with the N availability to the root (Roche et al., 2017) highlighting the importance of C:N balance in vegetative tissue (Louahlia et al., 2008). Furthermore, the amounts of WSCs varies between varieties as well as within the environment; Lolium cultivars AberMagic, AberDart, and AberElite all had highest growth rates correlated to highest WSC concentration during spring/summer, corresponding to high N availability from the roots alongside optimal photosynthesis conditions (Winters et al., 2010). Recent advances in the identification and manipulation of photosynthesis promoters for both Lolium perenne RBCS and Chlorophyll a/b Binding (CAB) (Panter et al., 2017) has provided transgenic lines for assessing increases in yield, fiber and, more importantly for digestibility, the fructan concentrations in both pseudostem and leaf blades in field trials (Badenhorst et al., 2018). Such work provides a platform for future studies to identify promotors important in other nutritional traits.

The amounts of resistant starch are important for the digestibility and nutritional content of forage crops. Resistant starch (RS) generally has lower digestibility until it reaches the large intestine (Englyst et al., 1999), where in ruminants more digestion can occur (Raigond et al., 2015). Research studies have shown that M. sativa has advantages as a feed source over cereals for enhanced D-value (Giuberti et al., 2018). One major difference between dietary RS is that it is seen to have advantages in the human diet by providing more fiber, but disadvantages in livestock feed for non-ruminants as it remains undigested. In general, lower RS will improve the digestibility of forage crops for both ruminant and non-ruminant livestock. As a crops D-value is closely linked with its starch, protein and lignin content, genomic studies have begun large-scale genome-wide association studies (GWAS) to confirm correlations across a range of traits, such as using three distinct alfalfa cultivars with a high-throughput genotyping-by-sequencing approach (Biazzi et al., 2017). However, this study did highlight that differences in SNPs associated in different tissue types (shoots and leaves) can vary in correlation with traits such as protein content, and so care must be taken when using GWAS to aid crop improvements.

Another substantial nutrient in forage crops is that of proanthocyanidins or condensed tannins (CTs). CTs bind to protein making it unavailable to digestion for ruminants until it reaches the rumen, and thusly an important trait in increasing the D-value of a crop (Min et al., 2003), although too high a CT content can be harmful restricting fermentation, especially in low leaf protein content species. A compromise is therefore desirable, with the moderate CT of 2–4% of the forage biomass giving the optimal D-value (Dixon et al., 2005). Whilst some species of legumes have optimum levels of CTs such as Lotus corniculatus, others such as Onobrychis viciifolia and Trifolium ambiguum are often poor choices for forage in many climates (Min et al., 2003; Baker, 2012); which means there is more scope to increase CTs concentrations in high yielding species where they are low such as M. sativa and Trifolium repens rather than increase growth traits aforementioned (Burggraaf et al., 2006; Salunkhe et al., 2017).

As the CT synthesis pathway has been well-characterized in Arabidopsis with the transcriptional regulators R2R3 MYB, bHLH, and WD40 protein identified as having a central role in final CT content (Lepiniec et al., 2006). Such knowledge can be used to manipulate forage crops. The R2R3 MYB homolog MtPAR in the M. truncatula seed coat has been characterized and hairy root transformation in alfalfa resulted in the accumulation of CTs to the level of ∼20 mg/g shoot biomass (Verdier et al., 2012), although this is still below the desirable concentration. A similar study showed that expression of the TaMYB14 transcription factor from a low-yielding forage activates CT biosynthesis in both Trifolium and Medicago (Hancock et al., 2012). Other approaches have involved characterizing early steps of CT biosynthesis in M. truncatula in the hope to later target crop relatives (Pang et al., 2007), whilst others have looked at how relative amounts of CT differ between leaves and higher concentration containing flowers to see if changing flowering in Trifolium could improve its D-value (Burggraaf et al., 2008). There has been an effort to engineer better digestibility in some forage cultivars (Wang and Brummer, 2012) and microbial pre-digestion after cutting and before feeding, including microbial supplements (Boyd et al., 2011; West and Bernard, 2011; Elghandour et al., 2015), can be used to enhance this.

### Biomass Production

As biomass yield is the main target for forage crop improvement, more rapidly growing cultivars can be targeted for breeding. Studies have consequently focussed on heading date (Fe et al., 2015) and flowering time regulation (Skøt et al., 2011; Shinozuka et al., 2012) in Lolium by developing Genomic Prediction models and QTL mapping as described previously (Nuñez and Yamada, 2017). Manipulating genes involved in delayed senescence has been targeted for increasing biomass yields. The introduction of the 5<sup>0</sup> flanking region of the Zea mays cysteine protease gene SEE1 in Lolium multiflorum has shown this promoter region to increase leaf lifespan by approximately 8–16 days (Li et al., 2004). A similar study using the Arabidopsis Senescence-Associated Gene12 (SAG12) promotor also delayed senescence in M. sativa with notable chlorophyll and yield increases even after 3 months of growth (Calderini et al., 2007). A final example is the expression of the Panicum virgatum NAC1 and NAC2 transcription factors in Arabidopsis atnap lines (mutants with defective senescence) to restore wild-type phenotype, predominantly measured using total chlorophyll concentrations (Yang et al., 2015).

However, fast growth must also be coupled with the ability of the plants to respond to cutting by providing rapid regrowth. Growth rates and recovery from cutting are traits that are relatively easy to select for in breeding trials, and have been

of regular interest to researchers for many decades in both Lolium and Medicago (Wilman et al., 1977; Vance et al., 1979). In addition, cutting experiments using 13C and 15N in both Lolium and Medicago have shown how the soil is affected both for dissolved organic C and N, and microbial biomass, demonstrating that management schemes can be critical to subsequent soil health (Schmitt et al., 2013). An ability to rapidly regrow may increase the susceptibility of the plant to insect and pathogens and this is worthy of further investigation. The relationship between tissue wounding and plant immunity is a topic that is quickly developing and there is now good evidence that tissue growth rate is closely linked with immunity (Huot et al., 2014).

Thusly, management schemes for forage crops are very important for yield. For example, choosing when to cut or graze a crop is crucial for subsequent regrowth of the plant (Karn et al., 2006; Asaadi and Yazdi, 2011; Bumb et al., 2016). To assist in this choice there is scope for the use of molecular markers, with the future possibility of a PCR test for the optimal time harvest based on the expression of candidate genes like storage proteins. Such tissue testing of crops can also be used for decisions on the timing of fertilizer applications as the two evaluations are made at around the same time. There is scope to identify a suite of marker genes that can be used to help decide when these key decisions are made.

Mixed cropping schemes are already widely used for forage crops and there are clear advantages in growing legumes and grasses together. Legumes increase soil N through their N fixation symbiosis with Rhizobium, with their biological nitrogen fixation ranging from 32 to 115 kg ha−<sup>1</sup> (Iannetta et al., 2016). This can in turn decrease subsequent fertilizer use for crops grown thereafter, a reduction between 23 and 31 kg N ha−<sup>1</sup> (Preissel et al., 2015). Numerous intercropping regimes have been tested including modeling of various climatic and soil texture parameters (Bachinger and Reining, 2009). Transfer of N from legume to crop, including in grasslands, has been investigated (Pirhofer-Walzl et al., 2012). However, it is still unknown how this interaction affects N movement and leaching through the soil profile. Such an investigation is required to give evidence of environment changes as well as crop productivity. Mixed species cultivation also has advantages for disease and extreme weather resistance as the susceptibility of the plants to these stresses varies between cultivars and species. Forage breeding has focussed on monoculture selection regimes and there is scope for better mixed species crops that could be included in trials for new varieties. Some advantages and disadvantages of mixed forage crops are summarized below in **Table 1**.

Growing forage crops for improved nutritional quality has not been a target for breeding programs, rather yield and climate tolerance have been the drivers. Future crops must be tolerant of climate changes and weather extremes. Unlike many crops where monocropping is most productive, forage crops have the advantage that they can be easily grown in combination without lowering productivity. Such a trend has been shown across multiple trials as well as increasing biodiversity (Tilman et al., 2001, 2006; Weigelt et al., 2009). As with any system that promotes biodiversity whilst still being productive, this can TABLE 1 | The advantages and disadvantages of growing forage crops in mixed systems.


mean not only lowered costs to manage but also help cultural agriculture acceptability, with consumers becoming more aware of the effect the production of their food has environmentally (Scherr and McNeely, 2008).

### Trace Elements

Plant research has focussed on the goal of biofortifying cereals, but there is also potential to improve the nutritional quality of forage crops. The economic importance of livestock production in the poorest parts of the world offers the opportunity to biofortify animal crops thereby improving the health of these animals and both directly and indirectly their owners. The knowledge base developed for grain biofortification (e.g., candidate plant metal transporters) has yet to be applied to forage crops. For example, transporter proteins for iron and zinc storage have been identified in cereals (Connorton et al., 2017; Menguer et al., 2017) and their equivalents in forages have yet to be identified.

Although very abundant in most soils, silicon is particularly required by grasses (Tubana et al., 2016) and is therefore likely to be important for the optimal growth of many forage crops. Silicon is important for cell wall structure and therefore resistance to pathogens and pests, however, it may have a negative impact on digestibility. The supply of this nutrient may become limiting for forage crops, particularly as the plant biomass is regularly removed from the field and silicon is not yet a routine addition to fertilizer.

Most species of forage crops can form mycorrhizal associations and this type of symbiosis is important for acquisition of trace elements. For natural grazing, these symbiotic associations are particularly important, but when fertilizer is added to cultivated forage crops mycorrhiza are suppressed (MacLean et al., 2017). Enhancing this symbiosis by inoculation of forage crops with mycorrhizal fungi has the potential to improve the mineral element composition of the feed. The fungal symbiosis has additional benefits for the plant by increasing the soil area mined for nutrients and water; this can be crucial during extreme weather events such as drought. Furthermore, a balanced and optimized root rhizosphere microbiome is essential for optimal root function and this applies to all crops including forage (Mommer et al., 2016).

### Environmental Footprint of Forage Crops

As in all agriculture, improving water and nutrient use efficiency is a target for forage crops. The general fertilizer requirement of maize grown for forage and for grain are the same as that for a biomass crop. N requirements differ greatly for forage crops, and legumes and rhizome crops like Miscanthus have low N requirements (Dierking et al., 2017). Improving NUE using transporter marker genes as indicators of the crop status in the field could be valuable (Fan et al., 2017). Targeting particularly the NUpE component of NUE is important for minimizing the wasteful and environmentally damaging losses of excess N fertilizer additions.

As discussed above for protein content, biomass production and cutting/grazing decisions there is the potential to develop gene markers that can indicate the N status of each type of forage crop. Mixed plant communities tend to have better NUE, probably because each species has a different temporal pattern of N uptake, resulting from different growth rates and root architecture (Tilman et al., 2001; Weigelt et al., 2009). In more affluent countries the relatively low chemical fertilizer prices do not encourage more judicious use of fertilizer for forage crops, but the threat of legislation for overuse has provided a new incentive for better fertilizer use efficiency. There is plenty of scope for improving the NUE of forage crops particularly as breeding programs have not focussed on this trail. For water acquisition, the long tap roots of Medicago are ideal for penetrating deep for water and nutrients. Varietal differences in this important trait have long been known (McIntosh and Miller, 1980) and the choice of cultivar depends on the soil type, climate and cropping regime that is required.

### CONCLUSIONS AND FUTURE DIRECTIONS

### Future Performance Improvements Using Genomics

The availability of genomics and bioinformatics has revolutionized all biology and as databases expand to include more species and cultivars this information can assist forage breeders to improve crop performance. The future possibilities for breeding of forage crops using Lolium as an example are shown in **Figure 2**. By comparing cultivar sequence information and using GWAS for traits such as high vegetative tissue concentrations of protein, NUpE or specific trace elements the nutritional quality and yield of forage crops can be improved. Some SNPs in key genes that have been identified in model plants can be the targets for gene editing techniques (Bonhomme et al., 2014; Slavov et al., 2014; Thorogood et al., 2017). TILLING lines are also being used in many forage crops to study gene function (Carelli et al., 2013; Dalmais et al., 2013; Manzanares et al., 2016). Furthermore, as shown with the rbcS example in **Figure 4**, sequence information can be used for the design of PCR primers which can be used for tissue testing. These tests can be used to rapidly identify general health and nutritional status of crops as well as specific pathogens. One bottleneck is likely to be the transfer of the new genetic information into forage crops. For example, GM forage crops may be more acceptable to the public, as if fed to animals their entry into the human food chain is indirect. The use of CRISPR/Cas9 technology may provide an acceptable route for such manipulations, and as with many crops such feasibility studies have begun in forage crops; the mutation of the Medicago sativa Squamosa Promoter Binding Protein Like9 (SPL9) has been attempted and validated (Gao et al., 2018), although poor genome editing efficiency is limiting advances at present. Many candidate genes have been identified which may be quickly transferred into forage crops, but the technology for transformation is limiting development of these improved plants. In the future genome editing may become more accepted, particularly perhaps for animal feed crops.

### Focusing on Roots

As discussed above high-yield, low-input vegetative biomass is desired for forage crop production. This has meant aboveground phenotyping strategies are being widely developed using predominantly imaging and spectral data (Walter et al., 2012), although more research is needed to see how vegetative phenotyping will work across different species, especially in mixed-cropping systems. However, although the need for welldeveloped, established root systems is clearly important (Kell, 2011; Nacry et al., 2013), breeding for belowground traits has been largely disregarded. This is unsurprising as with all crops, root phenotyping is difficult, being hidden in the soil and therefore labor intensive and difficult to sample. Any current root system improvements have been the consequence of vegetative drought and salinity assays discussed previously.

Consequently, there has been a shift of focus toward breeding for underground traits in forage crops; across plant science this has been termed the next green revolution step (Lynch, 2007; Den Herder et al., 2010). Before phenotyping can even begin it is necessary to determine which kind of improvements are necessary, of which 2 main categories are found. The first is to improve root systems for the plant itself. This could include increasing fine root biomass, lateral root initiation, or in the case of legumes nodulation by Sinorhizobium, for increased nutrient uptake (Jackson et al., 1997; Ariel et al., 2010; Downie, 2010; Wang et al., 2010), or instead increasing root density or taproot length for either nutrient and/or water uptake, or resilience to stress such as defoliation (Dawson et al., 2004; Erice et al., 2007; Ghesquière et al., 2010; Kell, 2011).

The second category is the improvement of root systems to aid the environment. This target is to improve agricultural land not just for production but also in terms of the ecosystem services, and this is especially true in the case of forage crops (Marshall et al., 2016). Forages and grasslands can provide ecosystem services that are wide-ranging and highly linked to root function including soil C-sequestration important for climate change (Kell, 2011, 2012), or lowering run-off of land thus helping to lessen flooding and soil erosion (Macleod et al., 2013). The idea of using both non-leguminous and leguminous forage crops as cover crops to mitigate climate change is gaining appreciation, (Kaye and Quemada, 2017). Another point to note is that many perennial grasses including Miscanthus and Panicum can be used

for biofuel production but the characteristics required for a forage crop do not always match with those of a biofuel (Yang and Udvardi, 2018), although efficient root function and structure is likely to be a characteristic desirable for both agricultural sectors.

Whether to improve plant performance or that of the environment, advances in phenotyping root systems will be crucial, including characterizing the plasticity of the system whilst the plant is growing. At present there are a plethora of root analysis software available (Paez-Garcia et al., 2015), but these require imaging roots either grown artificially such as on plates or already taken from the field and therefore evasive. There is therefore an increased interest in developing imaging techniques of plants grown in clear media to chart phenotypic changes throughout growth, or more promisingly the use of X-ray computed tomography (CT) scanning to give high resolution 3D models of the growing root system (Zhu et al., 2011).

### Developing Management Systems

At present forage growers cannot easily and reliably determine the N status of their crops. For maximum biomass production, it is important to maintain the N status of the crop throughout the growing season and this requires an optimized soil N supply (Hofer et al., 2017). Application of too much N fertilizer results in wasteful run-off and sub-optimal supply results in decreased biomass production. Studies have already shown, through <sup>15</sup>N labeling of Lolium, how deficiency caused by low N fertilizer application causes an increase in the protein substrate pool whereas the store pool decreased in size and turnover rate (Lehmeier et al., 2013). This highlights the importance of fertilizer studies for N composition of forage crop vegetative tissue. Maintaining N supply for maximal yield is limited by two factors: (1) unreliable and unreproducible tests for soil N levels (Knight, 2006) and (2) an easy reliable measure of the crop's status.

Presently farmers take limited samples across their growing area in the hope that this is representative of the N in the whole plot through the growing season. Nevertheless, this does not indicate a plant's N status or provide a measure of NUE. Some research has focused on the use of spectral data to evaluate crop efficiency (Foster et al., 2017), but such techniques require further investigation and can give false readings caused by pathogen attack. Sensors for N contents of soil are also being developed, however, these can be a costly solution (Shaw et al., 2016). Due to these problems, it may be better if the farmer could determine the crop N status directly and then make a more informed decision as to how they should subsequently fertilize the plot. This would enable more efficient fertilizer use, thus increasing forage biomass with lowered costs. Furthermore, for forage that includes legumes

### REFERENCES

these N budget problems are complicated by the additional input of gaseous N-fixation. Other strategies of crop testing should be developed to reliably inform the grower of NUE efficiency.

### Final Animal Product Studies

As forages are grown to rear livestock which in turn becomes food products for humans it is also important to view research in plant science from a livestock study prospective, of which has been touched upon above when discussing nutritional composition of crops. At present many countries adopt large-scale, concentratefeeding led livestock production like that of the United States, with many potential human health risks due to bacteria, antibiotic-resistant bacteria, prion, and dioxin presence in end products (Sapkota et al., 2007). Despite a rise in concentratefeeding, forage crops are still used widely as the main source of feed due to its high-yields of DM and energy for low costs (Reynolds, 2000), although usually studies focus on investigating a combination of both especially at various stages of development. For example, studies comparing growth of cattle fed a grass-diet instead of a linseed diet found the end product meat had a healthier fatty acid profile high in beneficial n-3 PUFAs, but the cattle were more slow-growing and thus the meat quality was poorer (Nuernberg et al., 2005). Similar outcomes have also been found for milk production from dairy cows in high-forage systems (Dewhurst et al., 2006). If improvements could be made in forage quality, especially more high-sugar varieties as outlined above, then potentially huge improvements in the animal production can be made.

In conclusion, utilizing the information obtained from the research effort to improve grain crops and the knowledge gathered from model systems like Arabidopsis, offers an excellent future perspective for improving the nutritional quality and yield for forage crops.

### AUTHOR CONTRIBUTIONS

NC and AM wrote the manuscript and conceived the perspective, read, and approved the final manuscript.

### FUNDING

This project was supported by grants BB/J004588/1 and BB/J004561/1 from the Biotechnology and Biological Sciences Research Council (BBSRC) and the John Innes Foundation. NC was supported by an iCASE studentship from the BBSRC, grant BB/M015203/1, with the support of The British Association of Green Crop Driers Ltd.

American Dietetic Association and Dietitians of Canada (2003). Position of the American dietetic association and dietitians of Canada: vegetarian diets. J. Am. Diet. Assoc. 103, 748–765. doi: 10.1053/jada.2003.50142

Aranjuelo, I., Molero, G., Erice, G., Avice, J. C., and Nogués, S. (2011). Plant physiology and proteomics reveals the leaf response to drought in alfalfa (Medicago sativa L.). J. Exp. Bot. 62, 111–123. doi: 10.1093/jxb/erq249

Alexandratos, N., Bruinsma, J., Bödeker, G., Schmidhuber, J., Broca, S., Shetty, P., et al. (2006). "World agriculture: towards 2030/2050. Interim report. Prospects for food, nutrition, agriculture and major commodity groups," in ESA Working Paper No. 12-03, eds J. Bruinsma and N. Alexandratos (Rome: FAO).


tobacco plants enhances resistance to salinity. Mol. Biol. Rep. 39, 6067–6075. doi: 10.1007/s11033-011-1421-y


under contrasting N managements. Front. Plant Sci. 8:1618. doi: 10.3389/fpls. 2017.01618


suppression of plant defense. J. Plant Physiol. 167, 1533–1541. doi: 10.1016/j. jplph.2010.06.023


crop plants. Commun. Soil Sci. Plant Anal. 40, 2688–2711. doi: 10.1080/ 00103620903173756


of gibberellins in a shade-tolerant mutant of perennial ryegrass. Front. Plant Sci. 8:868. doi: 10.3389/fpls.2017.00868




and longevity of leaves, longevity of tillers and dimensions and weights of leaves and 'stems' in Lolium. J. Agric. Sci. 89, 65–79. doi: 10.1017/S0021859600027209


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Capstaff and Miller. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Joint Exploration of Favorable Haplotypes for Mineral Concentrations in Milled Grains of Rice (Oryza sativa L.)

Guo-Min Zhang1†, Tian-Qing Zheng<sup>2</sup> \* † , Zhuo Chen3†, Yong-Li Wang<sup>1</sup> , Ying Wang<sup>1</sup> , Yu-Min Shi <sup>4</sup> , Chun-Chao Wang<sup>2</sup> , Li-Yan Zhang<sup>1</sup> , Jun-Tao Ma<sup>1</sup> , Ling-Wei Deng<sup>1</sup> , Wan Li <sup>1</sup> , Tian-Tian Xu<sup>2</sup> , Cheng-Zhi Liang<sup>3</sup> , Jian-Long Xu2,5 \* and Zhi-Kang Li 2,5

*<sup>1</sup> Northern Japonica Rice Molecular Breeding Joint Research Center, Chinese Academy of Sciences, Haerbin, China, 2 Institute of Crop Sciences, National Key Facility for Crop Gene Resources and Genetic Improvement, Chinese Academy of Agricultural Sciences, Beijing, China, <sup>3</sup> Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing, China, <sup>4</sup> Guangxi Key Laboratory of Rice Genetics and Breeding, Rice Research Institute, Guangxi Academy of Agricultural Sciences, Nanning, China, <sup>5</sup> Shenzhen Institute of Breeding for Innovation, Chinese Academy of Agricultural Sciences, Shenzhen, China*

### Edited by:

*Huixia Shou, Zhejiang University, China*

### Reviewed by:

*Yongzhong Xing, Huazhong Agricultural University, China Qingyao Shu, Zhejiang University, China*

\*Correspondence:

*Tian-Qing Zheng tonyztq@163.com Jian-Long Xu xujlcaas@126.com*

*†These authors have contributed equally to this work.*

#### Specialty section:

*This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science*

Received: *04 December 2017* Accepted: *22 March 2018* Published: *12 April 2018*

### Citation:

*Zhang G-M, Zheng T-Q, Chen Z, Wang Y-L, Wang Y, Shi Y-M, Wang C-C, Zhang L-Y, Ma J-T, Deng L-W, Li W, Xu T-T, Liang C-Z, Xu J-L and Li Z-K (2018) Joint Exploration of Favorable Haplotypes for Mineral Concentrations in Milled Grains of Rice (Oryza sativa L.). Front. Plant Sci. 9:447. doi: 10.3389/fpls.2018.00447* Grain minerals in rice, especially those in milled grains, are important sources of micro-nutrition elements, such as iron (Fe), zinc (Zn), manganese (Mn), copper (Cu), and selenium (Se), and of toxic heavy metal elements, especially cadmium (Cd), for populations consuming a rice diet. To date, the genetic mechanism underlying grain mineral concentrations (GMCs) in milled grain remains largely unknown. In this report, we adopted a set of 698 germplasms consisting of two subsets [*indica*/*Xian* (X-set) and *japonica*/*Geng* (G-set)], to detect quantitative trait loci (QTL) affecting GMC traits of Fe, Zn, Cd, Mn, Cu, and Se in milled grains. A total of 47 QTL regions, including 18 loci and 29 clusters (covering 62 Cd loci), responsible for the GMCs in milled grains were detected throughout the genome. A joint exploration of favorable haplotypes of candidate genes was carried out as follows: (1) By comparative mapping, 10 chromosome regions were found to be consistent with our previously detected QTL from linkage mapping. (2) Within eight of these regions on chromosomes 1, 4, 6, 7, and 8, candidate genes were identified in the genome annotation database. (3) A total of 192 candidate genes were then submitted to further haplotype analysis using million-scale single nucleotide polymorphisms (SNPs) from the X-set and the G-set. (4) Finally, 37 genes (19.3%) were found to be significant in the association between the QTL targeting traits and the haplotype variations by pair-wise comparison. (5) The phenotypic values for the haplotypes of each candidate were plotted. Three zinc finger (like) genes within two candidate QTL regions (*qFe6-2* and *qZn7*), and three major GMC traits (Fe, Zn, and Cd) were picked as sample cases, in addition to non-exhausted cross validations, to elucidate this kind of association by trait value plotting. Taken together, our results, especially the 37 genes with favorable haplotype variations, will be useful for rice biofortification molecular breeding.

Keywords: favorable haplotype joint exploration, grain mineral concentration, GMC, quantitative trait locus, QTL, milled grain, biofortification molecular breeding, rice (Oryza sativa L.)

### INTRODUCTION

Micronutrient malnutrition (or "hidden hunger") is widespread throughout different countries (Kumssa et al., 2015), especially among poor populations whose daily caloric intake is confined to staple cereals (Gregorio and Htut, 2003; Ma et al., 2008; Bhullar and Gruissem, 2013). The development of biofortified cereals, especially mineral-dense rice, remains an efficient way to alleviate malnutrition in developing countries worldwide, including China (Gregorio and Htut, 2003; De Steur et al., 2012). Meanwhile, as a side effect of modernization, heavy metal pollution of arable land has become more and more severe; concentrations of toxic minerals, especially cadmium (Cd), are increasing in cereal grains, which threatens human health (Al-Saleh and Shinwari, 2001; Huang et al., 2007; Fu et al., 2008; Hang et al., 2009). Currently, with the fast expansion of rice cultivation to Northeast China, the grain mineral concentrations (GMCs) in early-matured japonica/Geng type rice have become more and more important in rice production.

In addition to its relatively small genome, rice remains the world's most popular staple food crop (Dawe et al., 2010; GRiSP, 2013); therefore, both the biofortification and the relief of heavy metal pollution in rice have attracted increased research attention. The GMCs belong to complex traits controlled by multiple quantitative trait loci (QTL). Some QTL mapping studies have been carried out with different populations (Tang, 2007; Lu et al., 2008; Shen et al., 2008; Garcia-Oliveira et al., 2009; Zhang et al., 2009, 2011, 2014; Zhong, 2010; Anuradha et al., 2012; Bekele et al., 2013; Du et al., 2013; Kumar et al., 2014; Norton et al., 2014; Huang et al., 2015; Nawaz et al., 2015; Hu et al., 2016), and in-silico mapping (Chandel et al., 2011) for the GMCs in brown rice has been performed. GMC-related QTL tend gather in four regions on chromosomes 2, 3, 4, 6, 7, and 11, respectively. Specifically, there are three regions gathering QTL controlling Cd concentration in rice grains on chromosomes 4, 7, and 11, among which, the one on chromosome 7 is supported by evidence from four different tests. The single causative gene was identified as OsNramp1 (Ueno et al., 2009a,b, 2010; Ishikawa et al., 2010; Tezuka et al., 2010; Abe et al., 2011). However, just as the other cloned genes identified as associated with GMCs, such as, OsVIT (Zhang et al., 2012) and OsNAS (Lee et al., 2009) for Fe, OsLCT1 (Uraguchi et al., 2014) and OsHMA3 (Ueno et al., 2010) for Cd, OsNramp5 (Ishimaru et al., 2012; Liu et al., 2017; Tang et al., 2017) for Mn, and OsHMA4 (Huang et al., 2016) for Cu, it's also mainly responsible for the GMCs in the aleuronic layer rather than the endosperm, which is the major part of the milled grain. Currently, attempts have been made by a few molecular biologists using endosperm-specific promoters to improve the GMCs in milled grains (Zheng et al., 2010; Masuda et al., 2012). However, the genetic mechanism of GMCs in milled grains remains largely unknown.

Previously, we used two sets of backcrossed inbred lines (BILs) derived from the same donor, and two elite new varieties in Southwestern China, Ce258 and Zhongguangxiang1 (ZGX1) as recipients, to assess the genetic background and the genotypic by environment (G × E) effects of GMC traits in rice milled grains using QTL mapping (Xu et al., 2015). Therefore, in the present study, QTL information from that linkage mapping work was used to confirm the results of a genome-wide association study (GWAS) using a set of 698 sequenced germplasms. Favorable haplotype joint exploration for candidate genes within important QTL regions was also carried out.

### MATERIALS AND METHODS

### Plant Materials and Field Experiments

A set of 698 germplasms was adopted in this study. The set comprised two subsets, one was an indica/Xian subset (Xset) including 265 accessions randomly chosen from the 3K genome project (The 3,000 Rice genomes project, 2014), and the other was an early japonica/Geng subset (G-set), which included 433 accessions with sequencing data from similar sequencing pipelines. According to their maturation times, the X-set was planted at Sanya (18.3◦N, 109.3◦E) of Hainan province and the G-set was planted at Haerbin (45.8◦N, 126.65◦E) of Heilongjiang province. A small set of accessions was used as a control panel to check the variances between different environments.

All of the above plant materials were transplanted into the field at a spacing of 13.2 cm between individuals and 25 cm between rows, with a final planting density of approximately 18,000 individuals per 667 m<sup>2</sup> . Field management was carried out according to the local farmers' practice. At the mature stage (about 40 days after flowering), seeds were bulk-harvested for each line. The seeds were air-dried and stored for 3 months in a drying house before being evaluated for the mineral concentrations (GMCs) in the milled grains.

Basic physical and chemical properties of the soil in the paddy field were analyzed using routine analytical methods of agricultural chemistry (Lu, 1999).

### Evaluation of Grain Mineral Concentrations (GMCs)

Dried seeds of each line were de-hulled, polished and then milled into flour, according to the surging and grind-milling method described in our previous report (Xu et al., 2015), to prevent possible mineral contamination, especially by Fe. About 0.3 g of rice flour was digested with 6 ml of HNO<sup>3</sup> and 0.2 ml of H2O<sup>2</sup> using a microwave digestion system (Microwave300, Anton PAAR, Graz, Austria), with the following parameters: 5 min at 700 W, 700–1,200 W for 10 min, and 1,200 W for 20 min. The samples were then transferred to a block heater at 160◦C for further digestion. The remaining 1 ml of digested sample was diluted with 50 ml of Milli-Q water before analysis. The Fe, Zn, Cd, Mn, Cu, and Se concentrations in the digested samples were determined using the methods described in our previous report (Xu et al., 2015). Two standards and two controls were set in each testing batch. Three replications of the tests were performed for each sample.

### Genotyping by Sequencing and Shared SNP Extraction

The X-set germplasms were re-sequenced with an averaged depth of more than 10× (The 3,000 Rice genomes project, 2014). The cleaned reads were then mapped to the reference genome of Nipponbare (IRGSP1.0), and about 14 M high-quality single nucleotide polymorphisms (SNPs) were identified (The 3,000 Rice genomes project, 2014). Based on these 14 M SNP, a set of 2.9 M SNPs related to potential protein-coding areas was carefully selected. To build an SNP set for primary association studies, a subset of about 27,921 SNPs was selected from the 2.9 M SNPs by choosing one SNP per 100 counts, as described in our previous GWAS mapping work (Zhang et al., 2016). For the Gset germplasms, the quantity of the full set of SNPs was about 4 M. Finally, about 13 K SNP markers shared by both sets were extracted and submitted for further analyses, including sample clustering, principal component analysis (PCA), and GWAS mapping. These analyses were also carried out with the X-set and G-set data independently and compared with the pooled data. To perform deeper mining, favorable haplotypes were jointly explored for candidate genes within important QTL regions, based with the original 14 M and 4 M SNPs in the X-set and G-set, respectively.

### Data Analysis, QTL Mapping, and Haplotype Analysis

Basic statistical analysis of the GMC traits, including the analysis of variance (ANOVA) and Duncan's t-test, were conducted using SAS software (S. A. S. I. Inc., 2004). The basic scenario of a compressed mixed linear model (Zhang et al., 2010), implemented in the Genomic Association and Prediction Integrated Tool (GAPIT) Version 2 (Lipka et al., 2012), was adopted for association analysis between QTL-flanking markers and GMC traits for the pooled, the X-set, and the G-set. Parameters for GAPIT were set with reference to our previous report (Zhang et al., 2016). A relatively stringent threshold was adopted to identify significant correlations between the SNPs and GMC traits, comprising a −log10(P)-value of 6.0. To minimize to the possibility of type II errors in QTL detection (Li, 2001), a relatively loose threshold of 3.0 was adopted for the loci with supporting evidence from our previous linkage mapping report (Xu et al., 2015) or other references. The allelic effects were estimated by setting the Major.allele.zero = TRUE in GAPIT Version 2 to identify the donors of favorable alleles and their effects on GMC traits.

Subsequently, a joint exploration of favorable haplotypes was carried out according to the following steps: (1) By comparative mapping, we compared the results from the association mapping with the linkage mapping results from our previous report (Xu et al., 2015). The regions containing the jointly detected QTLs were then subjected to candidate gene analysis. (2) We searched the regions in the annotation dataset with wet-lab supporting evidence from the Rice Annotation Project database (RAP-DB) (Ohyanagi et al., 2006). (3) We then screened the genes by annotation information. If there were any obvious supporting evidence from the functional annotation, representing the relationships between the gene and the QTL targeting trait, then these genes would be highly focused in the next step. (4) Next, we compared all the mean values of the targeting traits for all the haplotypes of each candidate gene using pairwise comparisons with Duncan's t-test to identify significant associations between the variations of haplotypes and the QTL targeting traits. (5) Finally, we plotted the QTL targeting trait values for the haplotypes of each candidate in a straight-forward view. This joint haplotype exploration of the candidate genes was performed with the aid of Perl scripts and the full sets of SNPs in the X-set and G-set, respectively. For graphing and plotting, both Excel and R scripts were used.

## RESULTS

### Performance of the 698 Sequenced Accessions

Among the 698 sequenced accessions, a wide range of variation was found for the GMC traits in the milled grains. As shown in **Figures 1A–C**, the concentrations of three major GMC traits (Fe, Zn, and Cd) ranged from 0.9 to 9.1 ppm, 5.8 to 29.6 ppm, and 0.002 to 0.054 ppm, with mean values of 2.4, 16.4, and 0.009 ppm, respectively. The concentrations of the other three GMC traits (Mn, Cu, and Se) (**Figures 1D–F**) ranged from 3.6 to 22.0 ppm, with a mean value of 9.7 ppm; from 0.8 to 7.5 ppm, with a mean value of 3.2 ppm; and from 0.01 to 0.11 ppm, with a mean value of 0.04, ppm, respectively. All the GMC traits fitted normal or normal-like distributions in the pooled set, except for the Cd concentration, which showed a binomial-like distribution (**Figure 1**). Notably, when we highlighted samples from the X-set and G-set with different colors, a major proportion of G-set samples were found to have higher Zn and Cu, but lower Cd concentrations. For the other three GMC traits, the phenotypic value distributions between the two sets overlapped markedly, especially for the Se concentration. The affects on the GMC trait values were caused by multiple factors, including different environmental conditions, especially the soil (**Supplementary Table 1**), as well as the genetic factors, were much more complex than we expected. Nevertheless, according to the ANOVA results based on the control panel (**Supplementary Table 2**), all the genotypic variances showed higher statistical significances than the environmental variances. Although limited by the diversity of the control panel, the effects of the genotypic variances for most GMC traits were only marginally significant or insignificant, except for the Zn and Mn concentrations.

The Manhattan plots presenting the GWAS mapping results of the six GMC traits were shown in **Figures 2A–F**. Sample clustering and PCA analyses were also carried out based on the 13 K SNPs. The PCA result for the pooled data is shown in **Figure 2G**, and the kinship between the 698 accessions is presented in **Figure 2H**. For comparison, the PCA results obtained from the X-set and G-set independently are also shown in the **Supplementary Figure 1**. The results showed that the segregating pattern of the pooled set was quite similar to that of the X-set, whereas the G-set seemed relatively uniform. Considering that the optimum setting of the PCA value might vary according to different GMC traits, during the GWAS analysis with GAPIT, the Model.selection was set as TRUE for the optimum PCA value setting.

(A–F) Distribution graphs for Fe, Zn, Cd, Mn, Cu, and Se concentrations in milled grains, respectively.

## Identification of Loci Controlling the Six GMC Traits

According to the comparisons between the GWAS results from the pooled set and the two independent sets (X-set and G-set) shown in **Supplementary Figures 2**–**7**, a compensating mode was found between them. This meant that most signals in the pooled set were donated by either the X-set or the G-set, although the significance levels of the signals in the pooled set would be somewhat reduced if they were not significant in both subsets. To focus on the GMC QTL throughout different populations (also termed genetic background independent) and environments (also termed stably expressed), we adopted the results from the

nucleotide polymorphism (SNP) genotyping data; (H) VanRaden map for the Kinship of the 698 germplasms.

analyses based on the pooled set for further joint exploration of favorable haplotypes. A total of 47 QTL regions, including 18 loci and 29 clusters covering 62 Cd loci (**Table 1**, **Figure 2**) were detected by GWAS mapping for the six GMC traits from these 698 sequenced accessions. They included six loci for Fe, four loci for Zn, three loci for Mn, two loci for Cu, three loci for Se, and 62 loci belonging to 29 clusters for the Cd concentration. The average –log<sup>10</sup> value for these loci was 5.2 (range, 3.1–9.9). The –log<sup>10</sup> values varied by different GMC traits: It was 4.0 for Fe (range, 3.1–4.8), 3.5 for Zn (range, 3.4–3.6), 5.5 for Cd (range, 3.2–9.9), 4.3 for Mn (range, 4.0–4.6), 5.0 for Cu (range, 4.8–5.1), and 4.5 for Se (range, 3.8–5.6).

Alleles from the germplasms increased the GMCs at about 38 (47.5%) of the above 80 loci, while they decreased the GMCs at the other 42 (52.5%) loci. Among the 42 loci with GMC decreasing alleles from the germplasms, 35 (83.3%) loci were responsible for the Cd concentration. However, among the 38 loci with GMC increasing alleles from the germplasms, 27 (71.1%) loci were responsible for the Cd concentration. Thus, according to the effects of GMC traits for human health, there were only 46 (57.5%) loci with favorable alleles from our 698 sequenced germplasms in comparison with the reference genome.

According to their physical position, the 62 loci associated with the Cd concentration could be group into 29 QTL clusters


TABLE 1 | Quantitative trait loci (QTL) affecting grain mineral concentrations (GMCs) detected by a genome-wide association study (GWAS) in a panel of 698 germplasms.

*(Continued)*

#### TABLE 1 | Continued


*<sup>a</sup>QTL cluster.*

*<sup>b</sup>Favorable allele effect (FAE) values of the peak markers.*

*<sup>c</sup>A GMCs-related QTL detected by linkage mapping in our previous report (Xu et al., 2015), in which the three parents for the BC populations were also involved in our germplasms for the GWAS mapping.*

*<sup>d</sup>The number in brackets are reference codes as listed in reference section.*

(**Table 1**). Sixteen (55.2%) clusters harbored at least two loci (range, 2–5; mean = 3.6 loci/cluster). The three largest clusters were found on chromosomes 8, 11, and 12. Each of them harbored five loci for Cd concentration. Reverse allelic effects from the germplasms were detected for different loci gathered in one cluster. Among 14 (48.3%) of them, a single locus was found for each cluster.

### Haplotype Analysis of the GMC Candidate Regions

We chose a total of 10 regions with supporting evidence from our linkage mapping for candidate gene scanning. A total of 192 coding genes with wet-lab evidence according to the RAP-DB (Ohyanagi et al., 2006) were identified in eight of the ten candidate regions (**Supplementary Table 3**). No significant relationship was found between the annotation information and the GMC traits; therefore, all 192 genes were submitted for further analysis. Candidate gene haplotype analysis was then carried out for these genes. Statistical comparisons between the mean values of the three major GMC traits (Fe, Zn, and Cd) were then carried out for different haplotypes of the genes in the X-set and G-set, respectively.

Based on the results of Duncan'st-test for the haplotypes of the above candidate genes, 37 genes were found to have significant associations between the haplotype variations and the targeting trait of the QTL region (**Table 2**). There were no obvious GMC trait-related genes based on the annotation information from RAP-DB (**Supplementary Table 3**); therefore, three genes associated with zinc binding domain and/or zinc finger, which have not yet been reported to be related to the GMC traits, were chosen as sample cases in addition to non-exhausted cross validations. The genes were Os06g0489500 (Chr6:16404065- 17615233) for qFe6-2, and Os07g0568300 (Chr7:22841126- 22941126) and Os07g0569700 (Chr7:22841126-22941126) for qZn7. We performed trait value plotting for these samples following the above tests for all 192 candidate genes. We focused on the three major GMC traits: Fe, Zn, and Cd. Most phenotypic values between the different haplotypes for Os07g0569700 were insignificant, except for the Cd concentrations in the G-set. Thus, we only showed the significant results for the other two genes [Os06g0489500 (marginally associated with Fe) and Os07g0568300 (highly associated with Zn)] in **Figures 3**, **4** for the X-set and G-set data, respectively.

From the X-set, among the five haplotypes of Os06g0489500 (**Figures 3a,c,e**), Hap3 seemed to be the most favorable one, which is associated with relatively higher Fe and Zn concentrations, but without significant affects on the Cd concentration, compared with the other haplotypes. Hap1 was the second choice. It was associated with an increased Fe concentration, but a relatively lower Zn concentration, and an insignificantly higher Cd concentration. Among the eight haplotypes of Os07g0568300 (**Figures 3b,d,f**), Hap6 seemed to be the most unfavorable one, being associated with relatively lower Fe and Zn concentrations, and an insignificantly higher Cd concentration. Hap4 and Hap8 from the X-set were only associated with higher Zn concentration and had no significant effect on the Fe or Cd concentrations. Additionally, mild but significant effects of Hap2, Hap5, and Hap7 on Zn concentrations were also detected compared with Hap6.

In the G-set (**Figure 4**), among the nine haplotypes of Os06g0489500, Hap2 increased not only Zn but also Cd concentrations compared with the other haplotypes. Hap7 significantly reduced Zn, but had insignificantly increased the Cd concentration. Os07g0568300 was only associated with the Fe concentrations in the G-set. Among the seven haplotypes, Hap6 was favorable, which significantly increased the Fe, but had no significant effects on the Zn or Cd concentrations.

In addition to these two significant candidate genes shown in sample cases, all 37 genes listed in **Table 2** will become the focus for further functional verification in our future work.

### DISCUSSION

### Comparison of Identified GMC QTL With Reported Genes/QTL

As described in another report for GMC QTL mapping in milled grain of rice (Hu et al., 2016), the statistical significances of QTL for the GMCs in milled grain are much lower compared with the QTL detected for GMCs in brown rice grains. This phenomenon also appeared in our association mapping experiment. Thus, we adopted two thresholds, including a relative loose one to minimize the type II error. Finally, we mapped a total of 80 loci (**Table 1**, **Figure 2**). Ten (12.5%) of them including qFe6-2, qFe7, qZn7, qZn12, qCd1-1/qCd1-2, qCd4-7, qCd6-2, qCd8-1, and qCd11-1 were consistent with the loci from our previous linkage mapping work, including qFe6, qFe7, qZn7, qZn12, qCd1, qCd4, qCd6, qCd8, and qCd11, respectively.

Twenty (25%) of these 80 loci were also supported by loci reported in other works. Some were supported by multiple references. For example, qZn1 covered the region marked by id1005056–id1005058 (Norton et al., 2014), qZn7 was consistent with qZn7 (Huang et al., 2015; Hu et al., 2016) and qZN-7 (Lu et al., 2008), as well as the marker id7003641, which was significantly associated with the Zn concentration (Norton et al., 2014). QTL qCd7-1 was supported by qSCd7/ qGCd7 (Ishikawa et al., 2010) and qCdp7 (Abe et al., 2011). Some QTL were supported by single piece of evidence. The QTL qCd2-2 was consistent with qCd2b (Zhang et al., 2014). The loci qCd3-1, and qCd3-5 were consistent with two different reported qCd3 (Zhang et al., 2014; Huang et al., 2015), while the loci qCd11-8, and qCd11-9 were covered by a same relatively large region of qCd11 (Kashiwagi et al., 2009). The other 12 loci, qCd4-7, qCd5-1, qCd5- 7, qCd6-2, qCd6-3, qCd6-6, qCd7-2, qCd8-1, qCd8-3, qCd11-7, qCd12-9, and qCu5 were consistent with qCd4-2 (Kashiwagi et al., 2009), qCd5 (Zhang et al., 2014), qCd5.1 (Huang et al., 2015), Segment\_on\_Chr6 (Ishikawa et al., 2005), OsLCT1 (Uraguchi et al., 2014), qCd6 (Zhang et al., 2014), qCDCN-7 (Shen et al., 2008), Segment\_on\_Chr8(Ishikawa et al., 2005), qCd8 (Zhang et al., 2014), qCd11 (Tang, 2007), qSCd12 (Ishikawa et al., 2010), and qCu5 (Zhang et al., 2014), respectively.

Notably, five loci (6.3%) including qZn7, qCd4-7, qCd6-2, and qCd8-1 were supported by multiple pieces of evidence from our linkage mapping and other references. Thus, they would be of higher value for breeding application, with characteristics of stable expression and/or genetic background independence.

### Multiple Evidence for QTL Detection for GMCs

Although, the statistical significance for qFe6-2 and qZn7 was only marginal (both with –log<sup>10</sup> = 3.6) in GWAS mapping, they still possessed independent supporting evidence from the linkage mapping work (Xu et al., 2015). Additionally, there were supporting references of qZn7 (Huang et al., 2015; Hu et al., 2016), and qZN-7 (Lu et al., 2008), and the significant marker regions of id1005056–id1005058 (Norton et al., 2014). Thus, the joint application of the GWAS and linkage mapping again showed its power for QTL mapping, even for the traits with relatively low heritability, such as the GMC traits. Sometimes, multiple independent marginal evidences, when taken together, are more powerful than one single strong association signal.

Japonica/Geng and indica/Xian differ markedly in their ability to accumulate Cd (Ueno et al., 2010; Uraguchi et al., 2011), which is much more significant than for the other GMC traits. Thus, when we pooled the two subsets together for the analysis, a population similar to those used for bulk-segregant analysis, with a bi-nominal distribution, was formed. This explained why the



*<sup>a</sup>QTL detected by linkage mapping in our previous report (Xu et al., 2015). <sup>b</sup>Loci detected by a genome-wide association study (GWAS) in this work. <sup>c</sup>The gene highlighted as sample cases in the latter part of haplotype analysis were shown in bold. <sup>d</sup>X-set* = *indica/Xian set, G-set* = *japonica / Geng set,* \**,* \*\**,* \*\*\**, and* \*\*\*\* *represents significant level of 0.05, 0.01, 0.001, and 0.0001, respectively, in the pair-wise comparison using Duncan's t-test for the different haplotypes of each gene.*

Cd QTL gained more statistical power in the GWAS mapping (**Figures 1C**, **2C**). By contrast, the distributions of other GMC traits were not so significantly associated with the population structure, when divided by subsets (**Figures 1A,B,D–F**). In addition, only the locus significant in both sets, or at least highly significantly in one set, would be detected within the pooled data. Those peaks with an average level of significance in a single set would be highly likely to decrease in the analyses using the pooled data. However, this kind of underestimation of the QTL underlying the other five GMC traits would not have a large affect on the exploration of the really important loci that are suitable for practical breeding, especially those with multiple pieces of evidence that support the QTL, such as qZn7 and qFe6-2.

Additionally, many closely linked QTL with reverse allelic effects for the Cd concentration were identified in QTL clusters

along all the chromosomes, except for chromosomes 2 and 7. Thirteen (44.8%) of these clusters were supported by evidence from our previous mapping work or by other reports (**Table 2**). The largest clusters, Clst8C, Clst11b, and Clst12b, were each was found to harbor five loci for Cd concentration. Clst11b was also supported by evidence from multiple references. In our previous report, genetic overlaps were found for QTL controlling different GMC traits. Commonly, chromosomal crossovers in this kind of germplasm panel were thought to occur more frequently than in a bi-parental population. Thus, in this mapping work, with the improvement of mapping resolution compared with SSR linkage mapping in a bi-parental population, the details of the genetic overlap between GMC traits, especially those caused by tight linkage, may be magnified. The exact mechanisms underlying these Cd regions require further investigation.

Finally, according to the joint favorable haplotype exploration, we found that functional annotation could not always offer sufficient useful information during the candidate genes screening. By contrast, the QTL targeting trait comparison

would effectively help to narrow down the candidate genes from 192 to 37 by removing more than 80% unrelated information.

### Implications for Molecular Biofortification Breeding

This work offers at least three useful implications for the biofortification molecular breeding of rice. The first is that the QTL or candidate gene haplotypes underlying the GMC traits detected in this report, as well as those from our previous report (Xu et al., 2015), showed multiple effects on more than one GMC trait. Thus, in biofortification molecular breeding work on crops, especially rice (Oryza sativa L.), a possible trade-off between the improvement of favorable GMCs, such as Fe and Zn, and the accumulation of toxic heavy metal elements, such as Cd, in the milled grain should be taken into consideration. Selection of favorable haplotypes of candidate genes during molecular breeding would decide the final success of the breeding products. For example, if we chose Hap3 from the X-set for Os06g0489500, a relatively higher Fe and Zn concentration in the milled grain would be obtained, together with an insignificantly lower Cd concentration; however, if Hap1 of the X-set was adopted, the improved Fe concentration would be accompanied by a relatively lower Zn concentration and an insignificantly higher Cd concentration (**Figures 3a,c,e**). Thus, when we construct a scheme for backcross (BC) breeding, which is commonly adopted in biofortification breeding, using certain germplasms with higher favorable GMCs, such as Fe and/or Zn, as donors and an elite line as recurrent parents (RPs), at least two important steps should be taken during parental selection. First, the existing haplotypes of the target genes in the RPs and donors should be clarified by genotyping and haplotype analysis. Second, different GMCs, especially nutrient minerals and toxic minerals, should be balanced. For different RPs, different elite donors with suitable haplotypes should be selected for crossing.

The second point is that according to the mean values for the GMCs based on the haplotypes in the X-set and G-set, the Cd concentration is significantly lower in the G-set. This is consistent with known differences in Cd accumulation between indica/Xian and japonica/Geng (Ueno et al., 2010; Uraguchi et al., 2011). Thus, not only could the favorable haplotypes within the subspecies be used, but also those from across the subspecies could be taken into consideration. For example, for hybrid breeding, where most products belong to indica/Xian type, favorable haplotypes to decrease the unfavorable GMCs, such as Cd, could be imported from the japonica /Geng donors.

Finally, by combining the joint exploration of the GWAS mapping results with the results from our previous linkage mapping work, and the reference data from other reports, it was possible to identify the QTL regions for the GMCs in the milled grain more reliably. All the mapped loci, especially those that were jointly detected, as well as their favorable haplotypes, offer an opportunity to enhance the Fe and/or Zn concentrations, but control Cd accumulation, in milled rice grains. Biofortification molecular breeding using the favorable haplotypes jointly explored in this work, involving marker assisted selection and/or gene editing, would be the next step of our on-going studies.

### AUTHOR CONTRIBUTIONS

T-QZ, J-LX, and Z-KL: Conceived and designed the experiments; T-QZ, G-MZ, Y-MS, Y-LW, and YW: Performed the experiments; C-CW and T-QZ: Analyzed the data; ZC, C-ZL, T-TX, L-YZ, J-TM, L-WD, and WL: Contributed reagents, materials, and analysis tools; T-QZ and J-LX: Wrote the paper.

### REFERENCES


### FUNDING

This work was supported by grants from the National Key R&D Program of China (2016YFD0101801), the Agricultural Science and Technology Innovation Program and the Cooperation and Innovation Mission (CAAS-XTCX2016009), the Open Funding Program from the Guangxi Key Laboratory of Rice Genetics and Breeding (160-380-16-3), the Scientific Program Guangxi Province (GuiKe AB16380119), the Chinese Academy of Sciences Strategic Priority Research Program Fund (XDA08020302), Helongjiang Province Science Fund for Distinguished Young Scholars (JC201214), the Shenzhen Peacock Plan, and the Green Super Rice Project Bill and Melinda Gates Foundation (OPPGD1393).

### ACKNOWLEDGMENTS

We would like to thank the native English speaking scientists of Elixigen Company (Huntington Beach, California) for editing our manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018. 00447/full#supplementary-material

Supplementary Figure 1 | Comparison of populations presented by clustering and principal component analysis (PCA) in three sets of data.

Supplementary Figure 2 | Comparison of genome-wide association study (GWAS) mapping results in three sets of data for Fe in the milled grains.

Supplementary Figure 3 | Comparison of genome-wide association study (GWAS) mapping results in three sets of data for Zn in the milled grains.

Supplementary Figure 4 | Comparison of genome-wide association study (GWAS) mapping results in three sets of data for Cd in the milled grains.

Supplementary Figure 5 | Comparison of genome-wide association study (GWAS) mapping results in three sets of data for Mn in the milled grains.

Supplementary Figure 6 | Comparison of genome-wide association study (GWAS) mapping results in three sets of data for Cu in the milled grains.

Supplementary Figure 7 | Comparison of genome-wide association study (GWAS) mapping results in three sets of data for Se in the milled grains.

Supplementary Table 1 | Physical and chemical characteristics of the soil in the experimental fields.

Supplementary Table 2 | Analysis of variance (ANOVA) for the control panel under two environments.

Supplementary Table 3 | Coding genes in the candidate regions used for haplotype analysis.


accumulation in grain and aerial plant parts in rice. BMC Plant Biol. 9:8. doi: 10.1186/1471-2229-9-8


of Cd between roots and shoots in rice. Plant Cell Physiol. 50, 2223–2233. doi: 10.1093/pcp/pcp160


of brown rice grown in Cd-polluted soils. Euphytica 180, 173–179. doi: 10.1007/s10681-011-0346-9


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer QS and handling Editor declared their shared affiliation.

Copyright © 2018 Zhang, Zheng, Chen, Wang, Wang, Shi, Wang, Zhang, Ma, Deng, Li, Xu, Liang, Xu and Li. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Toward Eradication of B-Vitamin Deficiencies: Considerations for Crop Biofortification

Simon Strobbe and Dominique Van Der Straeten\*

Laboratory of Functional Plant Biology, Department of Biology, Ghent University, Ghent, Belgium

'Hidden hunger' involves insufficient intake of micronutrients and is estimated to affect over two billion people on a global scale. Malnutrition of vitamins and minerals is known to cause an alarming number of casualties, even in the developed world. Many staple crops, although serving as the main dietary component for large population groups, deliver inadequate amounts of micronutrients. Biofortification, the augmentation of natural micronutrient levels in crop products through breeding or genetic engineering, is a pivotal tool in the fight against micronutrient malnutrition (MNM). Although these approaches have shown to be successful in several species, a more extensive knowledge of plant metabolism and function of these micronutrients is required to refine and improve biofortification strategies. This review focuses on the relevant B-vitamins (B1, B6, and B9). First, the role of these vitamins in plant physiology is elaborated, as well their biosynthesis. Second, the rationale behind vitamin biofortification is illustrated in view of pathophysiology and epidemiology of the deficiency. Furthermore, advances in biofortification, via metabolic engineering or breeding, are presented. Finally, considerations on B-vitamin multi-biofortified crops are raised, comprising the possible interplay of these vitamins in planta.

### Edited by:

Alexander Arthur Theodore Johnson, University of Melbourne, Australia

### Reviewed by:

Francesco Di Gioia, University of Florida, United States Aymeric Goyer, Oregon State University, United States

### \*Correspondence:

Dominique Van Der Straeten Dominique.VanDerStraeten@ugent.be

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 19 January 2018 Accepted: 21 March 2018 Published: 06 April 2018

#### Citation:

Strobbe S and Van Der Straeten D (2018) Toward Eradication of B-Vitamin Deficiencies: Considerations for Crop Biofortification. Front. Plant Sci. 9:443. doi: 10.3389/fpls.2018.00443

Keywords: micronutrients, biofortification, metabolic engineering, folate, pyridoxine, thiamine, crop improvement, plant development

## INTRODUCTION

In an era of tremendous technological capabilities, insufficient accessibility to nutritious food, a primary human need, still affects over two billion people on a global scale (Bailey et al., 2015; Gupta, 2017). Though approximately 800 million people endure energy deficit due to inadequate amounts of calories in their diet (Haddad et al., 2016; FAOSTAT, 2017), the relative abundance of undernourishment has dropped to almost half in the last 25 years. Unfortunately, the degree of undernourishment witnessed over the last decades has recently passed a minimum, as undernourishment is estimated to have affected 815 million people in 2016, as opposed to 777 million in 2015 (FAOSTAT, 2017). The general trend of a decrease in global undernourishment can be mainly attributed to yield improvement of important staple crops such as rice, maize and wheat, which has more than doubled since 1960 (Long et al., 2015). Unfortunately, caloric malnutrition represents only a portion of the food-related burden of diseases, as micronutrient malnutrition (MNM) is present in over one–fourth of the world's population (Bailey et al., 2015; Blancquaert et al., 2017; De Lepeleire et al., 2017).

Micronutrient malnutrition comprises shortage of dietary micronutrients, including minerals (iron, zinc, iodine, selenium, etc.) as well as vitamins (Bailey et al., 2015). Micronutrient malnourishment, commonly referred to as the "hidden hunger," is known to induce diseases and disorders in many populations, not particularly confined to the developing world. Pregnant women and young children are most vulnerable for MNM, often resulting in death (Bailey et al., 2015; De Steur et al., 2015; Win, 2016; De Lepeleire et al., 2017). MNM can be considered an urgent global concern, persistent in many populations and remaining largely hidden (Ruel-Bergeron et al., 2015). Anemia, a condition of suboptimal hemoglobin level (Scott et al., 2014), illustrates the disastrous impact of combined micronutrient shortage on human physiology, as its occurrence has been linked with deficiency in iron (Camaschella, 2015), pro-vitamin A (Semba et al., 1992; West et al., 2007), thiamin (vitamin B1) (Franzese et al., 2017), pyridoxine (vitamin B6) (Clayton, 2006; Hisano et al., 2010) and folate (vitamin B9) (Moll and Davis, 2017). Anemia is held responsible for almost 2 million deaths of children under 5 years old on a yearly basis (Scott et al., 2014), and is estimated to affect more than 2 billion people globally. It is estimated that half of the cases of anemia could be attributed to deficiency in one or more micronutrients, though primary factors are iron and folate shortage (Scott et al., 2014).

High incidence of micronutrient deficiencies are, in many cases, related to monotonous diets, largely consisting of energy rich, starchy staples (Blancquaert et al., 2017; De Lepeleire et al., 2017). These crops including wheat, rice, potato, cassava, corn and plantain have the tendency to contain inadequate levels of vitamins and therefore expose the population, consuming massive amounts of these staples, to the risk of vitamin deficiencies (**Table 1**). This is a downside of the cheap supply of energy rich staples, which enabled the aforementioned decrease in caloric malnourishment.

Given the observation that MNM have a detrimental effect on global human health, there is a great need to strongly reduce these deficiencies, also stated in the Copenhagen consensus, where micronutrient interventions were ranked as the number one priority, related to Sustainable Development Goal 2 (SDG2), requiring great global investment (Copenhagen Consensus, 2012). Fortunately, there are several means to combat MNM in an effective way, which can be divided in education, supplementation and biofortification. Behavioral interventions, consisting of educational efforts encouraging dietary diversification, are the ideal means to improve the micronutrient status of a population (Reinbott et al., 2016). This strategy, however, requires changes in cultural or religious habits of certain communities, as well as recurrent interventions (Blancquaert et al., 2017). Fortification includes the administration of micronutrient to the population under the form of pills or fortification of food products (such as flour). The latter method, which is mandatory in many countries, has proven to be a rapid medium to ensure optimal micronutrient levels in the troubled populations (Pasricha et al., 2014; Sandjaja et al., 2015; Atta et al., 2016; Rautiainen et al., 2016; Wang et al., 2016). Unfortunately, supplementation depends on specialized infrastructure and appears difficult to implement in poor rural populations who have the highest demand for micronutrient interventions (Blancquaert et al., 2017). Luckily, biofortification, which involves the augmentation of the natural nutritional value of crops, can be addressed as a valuable additional strategy in the battle against MNM (Blancquaert et al., 2017; Garcia-Casal et al., 2017; Martin and Li, 2017; Strobbe and Van Der Straeten, 2017). Biofortification of locally consumed crops does not require changes in consumer behavior and demands only a one-time investment (De Steur et al., 2015, 2017; Bouis and Saltzman, 2017). Biofortification of staple crops, massively consumed in deficient populations, is an excellent way to supply sufficient micronutrients (Blancquaert et al., 2017; De Lepeleire et al., 2017). Biofortification sensu stricto, thereby omitting agricultural interventions (Cakmak and Kutman, 2017; Watanabe et al., 2017), comprises breeding techniques as well as genetic engineering approaches (Blancquaert et al., 2017; Bouis and Saltzman, 2017). Breeding strategies have the advantage to be easily implemented in agriculture, as they do not require exhaustive regulations (Mejia et al., 2017). However, the scope of the breeding approaches is confined to sexual compatibility, thereby lacking the ability to exploit useful animal or prokaryotic derived characteristics (Strobbe and Van Der Straeten, 2017). Biofortification via metabolic engineering, overrules this restriction. Furthermore, the latter approaches enable creation of a biofortification strategy blueprint, applicable to a wide variety of food crops (Strobbe and Van Der Straeten, 2017). Metabolic engineering does, however, require a great knowledge of the specific micronutrient metabolism and its importance in the physiology of the plant.

This review reflects upon the acquired knowledge which enabled successful B-vitamin biofortification in food crops, bundling information on thiamin (B1), pyridoxine (B6), and folates (B9). This evaluation includes vitamin function in plant growth and development as well as importance in human pathophysiology, epidemiology and accomplishments in biofortification. Furthermore, in view of multi-biofortification, the simultaneous biofortification of multiple vitamins and minerals, possible synergistic or adverse effects of micronutrient combinations, are scrutinized. Multi-biofortification endeavors are the step-stone for future eradication of MNM. A list of abbreviations can be found in **Supplementary Table S1**.

### VITAMIN B1 – THIAMIN

Thiamin is a water soluble B-vitamin consisting of a pyrimidine ring, linked to a thiazole moiety by a methylene bridge (Lonsdale, 2006) (**Figure 1**). Vitamin B1 consists of different itamer forms of thiamin, predominantly occurring as thiamin and its different phosphate esters thiamin pyrophosphate (TPP) and thiamin monophosphate (TMP). However, other forms of thiamin do exist, such as thiamin triphosphate, though their contribution to the total pool is rather marginal (Gangolf et al., 2010). In literature, thiamin(e) is sometimes confusingly used to describe the total pool of the different B1-vitamers, here simply referred to

### TABLE 1 | Vitamin content of six major staple crops.


Vitamin content data were retrieved from the USDA database (USDA, 2016). RDA data were retrieved from (Trumbo et al., 2001). Values represent vitamin content of 100 g fresh edible portion of each crop. Fold enhancement needed to reach the RDA of the corresponding vitamin upon consuming the highest average national serving, is indicated between brackets. Data on staple crop consumption are derived from FAOSTAT (FAOSTAT, 2017). <sup>1</sup>Highest recommended daily allowance (RDA). <sup>2</sup> In 2013. <sup>3</sup>Milled equivalent.

FIGURE 1 | Thiamin biosynthesis in plants. Synthesis of pyrimidine and thiazole moieties as well as their condensation occurs in plastids. Biosynthesis pathway is shown in blue, enzymes in black. Transport across membranes is proposed to be carrier-mediated (barrels), of which the identified mitochondrial TPP carrier is indicated (red barrel) (Frelin et al., 2012). The chemical structure of thiamin is depicted, of which the free hydroxyl group can be pyrophosphorylated by action of thiamin pyrophosphokinase (TPK). End product feed-back, performed by TPP on the THIC riboswitch, is depicted in red. Products: Gly, glycine; NAD+, nicotinamide adenine dinucleotide; SAM, S-adenosylmethionine; AIR, 5-aminoimidazole ribonucleotide; HET-P, 4-methyl-5-β-hydroxyethylthiazole phosphate; HMP-P, 4-amino-2-methyl-5-hydroxymethylpyrimidine phosphate; HMP-PP, HMP-pyrophosphate; TMP, thiamin monophosphate; TPP, thiamin pyrophosphate. Enzymes: THIC, HMP-P synthase; THI1, HET-P synthase; TH1, HMP-P kinase/TMP pyrophosphorylase; TPK, thiamin pyrophosphokinase; TH2, TMP phosphatase, PALE GREEN1.

as B1. The bioactive vitamer is TPP, serving as cofactor in multiple enzymatic reactions.

### Biosynthesis

Biosynthesis of TPP in plants takes place in plastids, cytosol and mitochondria (**Figure 1**). Although the TMP vitamer is synthesized in the chloroplast, subsequent enzymatic reactions in mitochondria and cytosol are required to finalize the de novo biosynthesis of TPP, the bioactive vitamer (Mimura et al., 2016; Goyer, 2017; Hsieh et al., 2017). Formation of TMP involves the synthesis of the thiazole and pyrimidine moieties, followed by their condensation, all of which occur in chloroplasts (Goyer, 2017). The thiazole moiety of the thiamin structure is created by the action of 4-methyl-5 β -hydroxyethylthiazole phosphate (HET-P) synthase (THI1), requiring NAD<sup>+</sup> and glycine as a substrate and yielding HET-P (Godoi et al., 2006). In this reaction the THI1 enzymes is consumed, as it supplies a sulfide from a cysteine residue, making THI1 a 'suicidal' enzyme. 4-amino-2-methyl-5 hydroxymethylpyrimidine phosphate (HMP-P) synthase (THIC) is an iron-sulfur cluster containing enzyme, catalyzing the formation of HMP-P, the pyrimidine intermediate of thiamin biosynthesis (Raschke et al., 2007; Kong et al., 2008; Goyer, 2017). This reaction requires 5-aminoimidazole ribonucleotide (AIR) (derived from purine metabolism) and S-adenosylmethionine (SAM) as substrates (Chatterjee et al., 2008). This is the key regulatory step of thiamin biosynthesis, as the promotor is under control of the circadian clock, and the terminator is feedback inhibited by the end product TPP (Bocobza et al., 2013). This terminator contains a riboswitch, rarely seen in eukaryotes (Wachter et al., 2007). The riboswitch acts by binding TPP in its pre-mRNA state, resulting in the formation of an unstable splice variant of the THIC gene, thereby causing a lowered THIC activity (Wachter et al., 2007; Bocobza et al., 2013). Next, the bifunctional enzyme harboring HMP-P kinase and TMP pyrophosphorylase activities (TH1), catalyzes the HMP-P phosphorylation and subsequent condensation of HMP-PP and HET-P to form TMP (Ajjawi et al., 2007b). TMP forms the end product of plastidial thiamin biosynthesis and is further processed by the PALE GREEN1/TH2 enzyme, after exiting the plastids (Mimura et al., 2016; Hsieh et al., 2017). The subcellular localization of TH2 action, being cytosolic, mitochondrial, or both, has been debated. First, the enzyme was discovered in the cytosolic fraction of Arabidopsis (Ito et al., 2011; Mimura et al., 2016). Later experimental evidence, utilizing translational fusions with green fluorescent protein (GFP), confirmed the cytosolic location of the TH2 enzyme, realized from a (preferred) native secondary translational initiation site (Mimura et al., 2016). This secondary translational initiation site yields a protein which lacks the functional N-terminal mitochondrial targeting peptide of TH2. Therefore, TH2 was considered to be predominantly residing in the cytosol though likely also present in mitochondria. More recently, TH2 was found to be almost exclusively localized in the mitochondria of th2/pale green1 rescued with a GFP-fused TH2, controlled by the cauliflower mosaic virus (CaMV) 35S promoter (Hsieh et al., 2017). These seemingly contradicting findings can be explained by the fact that a strong constitutive promoter (35S) preceding the full coding sequence of TH2, favors the first translational start site and incorporates the mitochondrial targeting peptide into the arising protein (Hsieh et al., 2017). However, these findings highlight the ability of sole mitochondrial TH2 activity to complement the th2/pale green1 mutant, indirectly hinting at the existence/necessity of a mitochondrial TMP importer as well as a thiamin exporter. Combining these results, it can be concluded that TMP dephosphorylation, executed by TH2, likely occurs primarily in the cytosol and to a lesser extent in mitochondria. However, it cannot be excluded that this subcellular localization of TH2 action might change in different conditions/tissues/species. The reaction mediated by TH2 yields thiamin. In turn, thiamin, is the substrate for thiamin pyrophosphokinase (TPK), producing the active vitamer TPP in the cytosol (Ajjawi et al., 2007a). TPP subsequently travels to the different subcellular locations via carriers. Two mitochondrial TPP carriers have been partially characterized in plants (Frelin et al., 2012).

### Role in Plant Physiology

Vitamin B1, sometimes called the 'energy vitamin,' plays a crucial role in plant energy homeostasis, mostly by the role of TPP as a cofactor (Goyer, 2010). TPP is a cofactor for three enzymes which are central to the energy metabolism. First, the pyruvate dehydrogenase (PDH) complex, catalyzing pyruvate decarboxylation, yielding acetyl CoA and NADH, necessary for the tricarboxylic acid (TCA or Krebs) cycle and biosynthetic processes, respectively (Bocobza et al., 2013). In addition, in the tricarboxylic acid cycle, α-keto-glutarate dehydrogenase (2-oxoglutarate dehydrogenase E1 component, OGDH) functioning also requires TPP, further accentuating its critical role in central metabolism (Rapala-Kozik et al., 2012). Third, TPP is an essential cofactor of the transketolase (TK) enzyme, playing a key role in the Calvin cycle as well as the pentose phosphate pathway, which renders pentose sugars as well as NADPH to the cell (Goyer, 2010).

Hence B1, in its form of TPP, controls a few key steps in central aerobic energy metabolism. In this regard, TPP has been suggested to influence the flux through these pathways, as it is required in their rate-limiting steps (Bocobza et al., 2013). Indeed, modestly increasing TPP concentration, through introduction of a non-functional riboswitch in Arabidopsis, induced enlarged activity of TPP-dependent enzymes (PDH, OGDH, and TK) (Bocobza et al., 2013). Moreover, these plants emitted larger amounts of CO2, suggesting overactive oxidative metabolism. This is further confirmed by the observation of depleted starch reserves at the beginning of the light period in high TPP lines (Bocobza et al., 2013). This presents a clear rationale behind the strict circadian regulation on B1 biosynthesis. As a consequence of its influence on central metabolism, metabolite composition, particularly that of amino acids, is severely altered in plants with aberrant B1 composition (Bocobza et al., 2013).

B1 metabolism was shown to have a clear function in enabling plants to cope with biotic as well as abiotic stresses. Thiamin biosynthesis as well as B1 levels were observed to increase upon application of abiotic stresses such as high light, drought, salt

and oxidative stress, conferring tolerance (Kaya et al., 2015; Yee et al., 2016). Remarkably, this B1-induced tolerance to oxidative stress was concomitant with decreased production of reactive oxygen species (ROS) (Tunc-Ozdemir et al., 2009). The exact molecular basis for this role of B1 in stress adaptation of plants remains partly unknown. Given the enhanced expression of TPP-dependent enzymes in plants exposed to drought stress, the influence of B1 on abiotic stress control seems to be via its end product TPP (Rapala-Kozik et al., 2012). On the other hand, the B1 biosynthesis enzyme THI1, responsible for plastidial thiazole biosynthesis, appears to be able to directly regulate stomatal closure (Li et al., 2016). The potential of B1 to enhance abiotic stress tolerance was, however, not observed in engineered Arabidopsis lines (Dong et al., 2015). Considering biotic stresses, B1 has been shown to confer systemic acquired resistance (SAR) (Ahn et al., 2007; Bahuguna et al., 2012; Boubakri et al., 2012). Thiamin-treated plants depicted enhanced ROS (hydrogen peroxide, produced upon up-regulation of superoxide dismutase) accumulation upon infection (Bahuguna et al., 2012), contrasting their role of decreasing ROS in abiotic stresses (Tunc-Ozdemir et al., 2009; Kaya et al., 2015). By doing so, thiamin provokes priming, a state in which the plant has the ability to react more rapidly upon infection (Conrath et al., 2006; Ahn et al., 2007). This priming effect, described for B1, was confirmed in high B1 engineered Arabidopsis (Dong et al., 2015), but not seen in rice (Dong et al., 2016). Moreover, thiamin treatment of plants induced higher accumulation of phenolic compounds, salicylic acid (SA) (through higher phenylalanine ammonia lyase activity) and nitrogen assimilation (via increased nitrate reductase activity) (Bahuguna et al., 2012).

### Pathophysiology and Epidemiology

The central role of B1 in central (oxidative) metabolism in humans is reflected in its pathophysiology upon vitamin deficiency. TPP plays an indispensable role in energy metabolism as a cofactor in cleavage of α-keto acids (Adeva-Andany et al., 2017), as well as general oxidative metabolism, identical to its role in planta. While having lost the ability to synthesize thiamin during their evolution, humans possess the potential to interconvert the different thiamin phosphate-esters (Zhao et al., 2001; Banka et al., 2014).

B1 was the first vitamin for which deficiency was characterized, as it was considered a "vital amine' (hence 'vitamine'), defined as a substance inducing the disease beriberi upon insufficient consumption (Lonsdale, 2006). Beriberi is a disease occurring upon severe B1 deficiency, divided in wet and dry beriberi, depending on whether it is manifested in the cardiovascular system or in the peripheral nervous system, respectively (Abdou and Hazell, 2015). B1 deficiency can cause heart problems, and even lead to heart failure (Roman-Campos and Cruz, 2014). Different symptoms, such as enlarged heart and increased venous pressure, have been reported. B1 deficiency has also been linked to Sudden Infant Death Syndrome (SIDS), due to brainstem malfunctioning related to hypo-oxidative metabolism (Lonsdale, 2015). An insufficient supply of the B1-vitamin can induce severe alternation of the nervous system, which leads to a disorder called Wernicke's encephalopathy (WE) (Jung et al., 2012). WE involves the arising of selective brain lesions, the first symptoms of which include confusion, apathy and impaired awareness, eventually ending in coma and death. B1 deficiency-induced disorders are in many cases easily reverted with thiamin application, and often witnessed in patients suffering from chronic alcoholism (Butterworth, 1993). The detrimental effect of B1-deficiency on brain functioning can be explained by the strong dependency of the brain on the oxidative metabolism (Butterworth, 1993; Gibson et al., 2005).

One of the greatest risks of B1-deficiency, along with the lethal consequences of untreated WE, is the difficulty of diagnosis, leaving many illnesses untreated (Harper, 2006). Although cases of severe beriberi have become rare, outbreaks of B1 deficiency-induced beriberi have been reported on a global scale, causing many deaths, even upon sufficient access to healthcare (Luxemburger et al., 2003; Ahoua et al., 2007). Hence, in developing countries, B1-deficiency often is not linked to the observed casualties (Barennes et al., 2015). Moreover, infantile exposure to B1-deficiency was recently shown to have longterm effects on motor functions and balance of the child (Harel et al., 2017). Furthermore, elderly people have been shown to be highly susceptible to B1-deficiency, even in the developed world (Hoffman, 2016). Indeed, an investigation in New York state (United States) identified 14% of elderly as being B1-deficient (Lee et al., 2000). B1-deficiency is likely to be exacerbating Alzheimer's disease, and could therefore be considered a serious threat, definitely not confined to the developing world (Gibson et al., 2013).

Good sources of vitamin B1 are, besides animal-derived products (meats, liver, eggs, and dairy products), beans and peas, nuts and whole grains (Lonsdale, 2006; USDA, 2016). Different massively consumed crops, such as rice, cassava, potato and plantain contain inadequate amounts of B1 (**Table 1**). In the case of rice, polishing, which removes the aleurone layer to avoid rancidification, eliminates many nutritionally valuable substances, including B1 (Goyer, 2017). This is illustrated by the original observation of B1-deficiency induced paralysis and death in fowls fed with polished rice, reversible by administration of the rice polishings (Lonsdale, 2006). Therefore, overconsumption of such staples in a monotonous diet, imposes a serious threat to human health. Furthermore, high carbohydrate intake increases the need of dietary B1, which is explained by its role in carbohydrate catabolism (Elmadfa et al., 2001). This emphasizes the need for increased B1 levels in these popular starchy crops.

### Biofortification

Engineering of the thiamin biosynthesis pathway to augment thiamin content in plants has been attempted recently (Dong et al., 2015, 2016). The key step in thiamin -and therefore B1 engineering is the first committed step in plastidial pyrimidine biosynthesis, THIC (Raschke et al., 2007). Activity of the THIC enzyme seems to be a major determinant of B1 biosynthesis, as indicated by the oscillations of the corresponding mRNA transcript with TMP levels (Bocobza et al., 2013). Moreover, this gene harbors a TPP-binding riboswitch in its 3<sup>0</sup> UTR, which enables it to destabilize its mRNA upon high TPP prevalence (Wachter et al., 2007). This feedback mechanism,

rather unique in eukaryotes, further highlights THIC as a regulatory point in B1 biosynthesis and therefore the THIC gene as an ideal candidate in metabolic engineering approaches (Pourcel et al., 2013). Indeed, eliminating this riboswitch, thereby removing the feedback inhibition on THIC, elevates thiamin level 1.6-fold in Arabidopsis (Bocobza et al., 2013). Enhancing the flux toward biosynthesis of the pyrimidine intermediate is likely insufficient for accumulation of B1. Indeed, feeding Arabidopsis seedlings with the intermediates pyrimidine and thiazole indicates that both are necessary to achieve higher levels of B1 (Pourcel et al., 2013). Combined overexpression of THIC and THI1, the plastidial thiazole biosynthetic enzyme (Godoi et al., 2006), further enhanced B1 levels of Arabidopsis over threefold compared to wild type (Dong et al., 2015). Similarly, implementation of this combined engineering strategy in rice resulted in B1 increase of 2.5-fold in leaves and 5-fold in unpolished grains (Dong et al., 2016). However, B1 levels remained barely affected in polished rice seeds. Future engineering strategies in B1-biofortification will tackle additional bottlenecks in B1-accumulation as well as applying engineering strategies to specific tissues (Goyer, 2017). Taken into account the detrimental effects of THIC-riboswitch elimination, resulting in chlorotic plants with enhanced carbohydrate oxidation (Bocobza et al., 2013), B1 biofortification should be approached with caution.

Besides metabolic engineering, there are opportunities to enhance B1 content in crops via breeding techniques. Indeed, several (wild) potato varieties were identified which harbor over 2-fold difference in B1 content compared to popular agricultural potato cultivars (Goyer and Sweek, 2011). Similarly, up to 2.7-fold variation was found in different cassava accessions (Mangel et al., 2017). Previously, over 10-fold B1 variation has been measured in rice (Kennedy and Burlingame, 2003). Recently, a genome wide association study (GWAS) identified multiple quantitative trait loci (QTL), underlying B1 content in common wheat (Li et al., 2017). These results imply that breeding strategies could help in acquiring higher B1 levels in popular/regional crop varieties. On the other hand, elevating of B1 levels through exposure to certain biological stresses has been suggested, as this proves to augment B1 biosynthesis by significantly increasing the expression of the biosynthesis genes (Kamarudin et al., 2017).

### VITAMIN B6

Vitamin B6 represents a group of water-soluble molecules with similar biochemical properties, consisting of pyridoxine (PN), pyridoxal (PL), pyridoxamine (PM), and their phosphorylated esters (Fudge et al., 2017). PN, PL and PM differ by carrying a hydroxymethyl, an aldehyde or an aminomethyl substituent, respectively (**Figure 2**) (Hellmann and Mooney, 2010). Considering these six vitamers, the phosphorylated pyridoxal (PLP, **Figure 2D**) is the most bioactive, functioning as a cofactor in over a hundred reactions (Fudge et al., 2017). B6 can be considered a powerful antioxidant, comparable to carotenes (vitamin A) and tocopherols (vitamin E), as they are able to quench ROS (Bilski et al., 2000).

pyridoxine-phosphate (PNP), (F) pyridoxamine-phosphate (PMP).

### Biosynthesis

De novo biosynthesis of vitamin B6 takes place in the cytosol and comprises only two enzymes (**Figure 3**). Pyridoxal phosphate synthase protein (PDX1) generates pyridoxal 5<sup>0</sup> -phosphate (PLP) utilizing ammonia, glyceraldehyde 3-phosphate (G3P) and ribose 5<sup>0</sup> -phosphate (R5P) as substrates (Titiz et al., 2006). This ammonia originates from the reaction catalyzed by the PDX2 glutaminase, which releases ammonia from glutamine to yield glutamate (Tambasco-Studart et al., 2007). Furthermore, PMP/PNP oxidase (PDX3) is considered a crucial step in PLP salvage, ensuring its retrieval from the PMP and PNP vitamers (Sang et al., 2007). The non-phosphorylated vitamers PL, PM, and PN, can be converted to their corresponding phosphorylated vitamers by the action of the SALT OVERLY SENSITIVE 4 kinase (SOS4) (Shi et al., 2002). Finally, a pyridoxal reductase (PLR1) was identified, mediating a NADPH-requiring conversion of PL to PN (Herrero et al., 2011). Through these reactions plants are capable of balancing the different vitamer forms of B6, which is required to ensure controlled growth and development (Colinas et al., 2016).

### Role in Plant Physiology

Vitamin B6 is involved in a plethora of metabolic reactions, serving as cofactor or required as an antioxidant (Tambasco-Studart et al., 2005; Mooney and Hellmann, 2010). PLP is considered to function as a cofactor for about 200 enzymatic reactions in Arabidopsis (Fudge et al., 2017). These PLPdependent enzymes, covering oxidoreductases, transferases, hydrolases, lyases, and isomerases, can be explored using the B6 database tool (Percudani and Peracchi, 2009). These reactions roughly cover the whole spectrum of plant metabolism. In doing so, B6 is required in amino acid synthesis as well as catabolism (Mooney and Hellmann, 2010). This is illustrated by the Arabidopsis mutant reduced sugar response (rsr4-1), harboring

a mutated B6 biosynthesis gene (PDX1), exhibiting a decreased content of shikimate, altered levels of different amino acids, and higher levels of TCA constituents (malate, citrate and fumarate) (Wagner et al., 2006). Similarly, Arabidopsis mutants for the PLP salvage enzyme PDX3 (involved in B6 vitamer interconversions) contained aberrant amino acid profiles. The initial step in starch breakdown, α-glucan phosphorylase, requires PLP as a cofactor (Mooney and Hellmann, 2010). Furthermore, PLP-dependent enzymes play a role in synthesis of glucosinolates (Mikkelsen et al., 2004). Remarkably, biosynthesis of the plant hormones auxin (Zhao, 2010) and ethylene (Van de Poel and Van Der Straeten, 2014; Vanderstraeten and Van Der Straeten, 2017) as well as ethylene breakdown (Nascimento et al., 2014) involve PLP-requiring enzymes. B6 levels have also been linked to nitrogen metabolism as pdx3 lines were shown to be ammonium dependent (Colinas et al., 2016). This link is further strengthened by the observation that the ammonium transporter mutant amt1 has altered B6 levels (Pastor et al., 2014).

On top of its vast influence on plant metabolism via PLPdepending enzymes, B6 plays a crucial role as an antioxidant (Vanderschuren et al., 2013; Fudge et al., 2017). Arabidopsis mutants with a lowered B6 status, exhibit distinct phenotypes including poor seed development, delayed flowering and reduced plant growth, while complete knock-outs are lethal (Vanderschuren et al., 2013). The lowered tolerance of these mutants to salt, high light, ultraviolet light, and oxidative stress illustrate the importance of B6 as a stress protector (Vanderschuren et al., 2013). Upon heat stress, a non-catalytic pyridoxine biosynthesis protein (PDX1.2), ensures sufficient B6 production by aiding its paralogs (PDX1.1 and PDX1.3), resulting in an increase of B6 content (Moccand et al., 2014; Dell'Aglio et al., 2017). Conversely, Arabidopsis lines, engineered for enhanced B6 content, display enhanced tolerance to abiotic stresses (Raschke et al., 2011). Furthermore, these plants exhibit enlarged cells, leading to larger organs. Interestingly, their amino acid and sugar composition is severely altered, reflecting the broad influence of B6 on plant metabolism.

### Pathophysiology and Epidemiology

Vitamin B6, especially PLP, is crucial for correct human functioning, as it is required as a cofactor for around 4% of all enzyme activities (Ueland et al., 2017). Most of these reactions involve amino acid synthesis and catabolism, in which PLP serves as a cofactor in transaminations, aldol cleavages and carboxylations. Furthermore, PLP plays a role in energy metabolism as it is involved in gluconeogenesis and lipid

metabolism. Moreover, B6 is necessary in the biosynthesis of heme as well as neurotransmitters (Ueland et al., 2017). In addition, B6 plays an important role as an antioxidant (Justiniano et al., 2017) and is even known to aid in enzyme folding (Cellini et al., 2014).

In parallel with its functions in human metabolism, B6 deficiency is manifested in a broad spectrum of disorders. Most notably, B6 deficiency is known to provoke neurological disorders, such as peripheral neuropathy (Ghavanini and Kimpinski, 2014) and epileptic seizures (Skodda and Muller, 2013). Moreover, B6 deficiency might be linked to anemia, given the ability of B6 intake to cure some cases of the disease (Hisano et al., 2010). Furthermore, B6 deficiency has been associated with cardiovascular diseases, stroke, rheumatoid arthritis, diabetes and different types of cancer including colorectal, lung, breast, and kidney (Ueland et al., 2017).

Although investigation on vitamin B6 deficiency on a global scale is lacking, there is evidence supporting the existence of persistent deficiency in several populations (Fudge et al., 2017). Indeed, studies in the United States and South Korea concluded that around one-in-four people have sub-optimal B6 status (Pfeiffer et al., 2013; Kim and Cho, 2014). Furthermore, half of the elderly in nursing homes in Norway were considered B6 deficient (Kjeldby et al., 2013). The situation in developing countries is estimated to be even worse, given the observation that over half of the population of Uganda and Sudan remain B6 deficient (Fudge et al., 2017). Knowing the detrimental effect this deficiency, remaining undiagnosed, could exert on human health, there is a strong need to supply these people with satisfactory amounts of B6.

Humans, unable to synthesize B6 de novo, predominantly depend on their diet for sufficient B6 acquisition, as gut bacteria can be considered as suppliers of marginal amounts of different vitamins (LeBlanc et al., 2013; Fudge et al., 2017). Good sources of dietary B6, besides animal-derived products such as fish and meat, are fresh vegetables including carrots and onions (USDA, 2016; Fudge et al., 2017). However, bioavailability should be considered, given the observations that up to half of the B6 pool could be lost as a result of incomplete digestibility, which is found to be more problematic in plant-based food sources compared to animal products (Roth-Maier et al., 2002). Furthermore, the most consumed staple crops in the world are considered poor sources of dietary B6 (Fudge et al., 2017) (**Table 1**).

### Biofortification

Metabolic engineering approaches rely on the knowledge acquired of the relatively simple plant B6 biosynthesis pathway, mainly involving PDX2 (Tambasco-Studart et al., 2007) and the pyridoxal phosphate synthase protein (PDX1) (Titiz et al., 2006). In a metabolic engineering strategy, overexpression of both PDX1 and PDX2 genes yielded up to fourfold increase in B6 levels, while overexpression of the single genes only generated marginal effects (Raschke et al., 2011). Interestingly, enhanced plant biomass in aerial organs with similar overall morphology as well as tolerance to oxidative stress were observed in two-gene engineered plants with increased B6 content. When targeted to roots, the two-gene approach, enabled almost sixfold augmentation of B6 in cassava, without any severe alteration in yield (Li et al., 2015). The success of this two-gene engineering strategy therefore supports assessment in different crops, as well as investigation of possible influences on crop physiology and yield.

So far, analysis of crop germplasm has revealed limited variation (<2-fold) in B6 composition of potato (Mooney et al., 2013) and wheat (Shewry et al., 2011). However, screening of vast accessions of a particular crop could identify interesting lines and thereby also pinpoint novel important QTLs and maybe novel genes influencing B6 homeostasis (Fudge et al., 2017).

### VITAMIN B9

Folate is a collective term for a group of water soluble B9 vitamins. Folates can be considered tri-partite structures, consisting of a pterin ring linked to the para-aminobenzoate (p-ABA) moiety carrying a γ-linked glutamate tail (Scott et al., 2000; Rebeille et al., 2006) (**Figure 4**). The different folate species, called vitamers, are chemically different on three levels, being the oxidation state, the glutamate tail length and the nature of C1-substituents (Blancquaert et al., 2010; Strobbe and Van Der Straeten, 2017). These properties all exert an influence on folate stability. First, oxidized folates are considered more stable, given the susceptibility of the pterin – p-ABA linkage to (photo-) oxidative cleavage (Blancquaert et al., 2010). Tetrahydrofolates (THF), the most reduced folate forms, harboring a fully reduced B-ring in the pterin moiety, are the active cofactors. Conversely, folic acid, containing an aromatic pterin B-ring, is more stable, though exhibiting marginal natural occurrence (Blancquaert et al., 2010; Gorelova et al., 2017b). In this respect, the term 'folic acid' is used to indicate the synthetic folate analog. Second, folate entities greatly differ in their glutamate tail length, as they carry one to eight glutamates (Garratt et al., 2005; Strobbe and Van Der Straeten, 2017). Polyglutamylated folates are thought to possess enhanced in vivo stability as their polyglutamate tail ensures cellular retention as well as augmented association with folate dependent enzymes (Blancquaert et al., 2014). Third, folates species can differ in their attached C1- units, giving rise to an array of folate entities, affecting their stability and biological role (**Figure 4**).

### Biosynthesis

In plants, folate biosynthesis is executed in different subcellular localizations (**Figure 5**). The pterin 'branch' resides in the cytosol (Strobbe and Van Der Straeten, 2017). Here, the first committed step is executed by GTP cyclohydrolase I (GTPCHI), utilizing GTP as a substrate and yielding 6-hydroxymethyldihydropterin (HMDHP) (Basset et al., 2002). An alleged pterin mitochondrial importer is considered to ensure translocation of HMDHP to the mitochondrion (Hanson and Gregory, 2011; Strobbe and Van Der Straeten, 2017). The plastidial p-ABA branch supplies the p-ABA moiety of the folate molecule (**Figure 5**). Here, the first committed step is performed by aminodeoxychorismate synthase (ADCS), using chorismate, originating from the shikimate pathway (Herrmann and Weaver, 1999), as a substrate (Sahr et al., 2006). Given the hydrophobic nature of p-ABA,

it is thought to reach the mitochondria by diffusion through membranes (Hanson and Gregory, 2011; Strobbe and Van Der Straeten, 2017). Upon entering the mitochondria, HMDHP is pyrophosphorylated and coupled with p-ABA to form dihydropteroate. These enzymatic reactions are executed by the bifunctional HMDHP pyrophosphokinase/dihydropteroate synthase (HPPK/DHPS) (Gorelova et al., 2017a). Subsequently, dihydropteroate is converted to dihydrofolate (DHF) by the action of dihydrofolate synthetase (DHFS) (Ravanel et al., 2001), followed by a reduction catalyzed by dihydrofolate reductase as part of a bifunctional enzyme dihydrofolate reductase/thymidylate synthase (DHFR-TS) (Gorelova et al., 2017b), yielding THF. Folate biosynthesis is finalized upon polyglutamylation of THF, by the action of folylpolyglutamate synthetase (FPGS) (Ravanel et al., 2001; Mehrshahi et al., 2010).

### Role in Plant Physiology

The chemical structure of folates makes them ideal carriers of C1-substituents, conferring a central role in carbon metabolism of nearly all living organisms (Blancquaert et al., 2010), with the exception of some Archaea (Gorelova et al., 2017b). Thereby, folates are both needed for proper anabolism as well as catabolism of cellular compounds. They play an essential role in the synthesis of purines as well as thymidylate and are therefore indispensable in DNA synthesis and growth (Stover, 2004). Furthermore, folates are required in biosynthesis of many plant metabolites including pantothenate (vitamin B5) and formyl methionyl tRNA as well as serine and glycine interconversion and catabolism of histidine (Blancquaert et al., 2010). Furthermore, folates are needed in production of lignin, ensuring cell wall rigidity (Srivastava et al., 2015). In addition, iron-sulfur cluster enzymes depend on folates for their assembly (Waller et al., 2010). Given their role as C1 donors and acceptors, folates play a key role in the methyl cycle (Blancquaert et al., 2010). 5-methyl-THF, is required as a methyl-donor in the conversion of homocysteine to methionine, which is necessary for replenishing of the SAM-pool (Blancquaert et al., 2010). SAM in its turn, functions as methyl storage in supplying this C1-unit to a wide range of methyltransferases, including DNA methyltransferases. Therefore, insufficient folate can alter the methyl-cycle homeostasis and evoke epigenetic changes by alteration in the DNA methylation pattern (Zhou et al., 2013). A disequilibrated folate homeostasis greatly influences epigenetic functioning through genome-wide hypomethylation, lowered histone methylation and transposon derepression, as witnessed in Arabidopsis methyleneTHF dehydrogenase/methenylTHF cyclohydrolase (MTHFD1) mutants (Groth et al., 2016). Similarly, aberrant functioning of FPGS, the enzyme responsible for extension of the glutamate tail, evoked upregulation of transposable elements (typically repressed by methylation), which could be reverted via administration of 5-methyl-THF (Zhou et al., 2013).

Additional to their requirement in catabolism and anabolism of essential plant metabolites, folates appear to have a profound influence on plant growth and development. In non-photosynthetic plastids, the plastidial pool of folates influences plant energy metabolism by inhibiting starch formation (Hayashi et al., 2017). The mechanism is thought to operate via depletion of the ATP pool -required in starch assembly from sucrose- upon folate shortage, regulated by the folate-dependent DHFR-TS (Hayashi et al., 2017). Remarkably, the interplay of folate and sugar metabolism was shown to modulate auxin signaling, hence controlling plant development (Stokes et al., 2013). Moreover, folates possess the ability

FIGURE 5 | Folate biosynthesis is plants. Folate (vitamin B9) biosynthesis in plants occurs in three subcellular compartments: the cytosol, the plastids (green) and the mitochondrion (red). Biosynthesis pathway is shown in blue, enzymes in black. Polyglutamylated folates are considered the end product of folate biosynthesis. Products: ADC, aminodeoxychorismate; p-ABA, para-aminobenzoate; DHN-P3, dihydroneopterin triphosphate; DHN-P, dihydroneopterin monophosphate; DHN, dihydroneopterin; HMDHP, 6-hydroxymethyldihydropterin; HMDHP-P2, HMDHP pyrophosphate; DHP, dihydropteroate; DHF, dihydrofolate; Glu, glutamate; THF, tetrahydrofolate. Enzymes: ADCS, ADC synthase; ADCL, ADC lyase; GTPCHI, GTP cyclohydrolase I; DHNTPPH, dihydroneopterin triphosphate pyrophophohydrolase; NSP, non-specific phosphatase; DHNA, DHN aldolase; HPPK, HMDHP pyrophosphokinase; DHPS, DHP synthase; DHFS, DHF synthetase; DHFR, DHF reductase; FPGS, folylpolyglutamate synthetase.

to influence seed composition, demonstrated by the high N-content of Arabidopsis plastidial FPGS (atdfb-3) loss-offunction mutant seeds (Meng et al., 2014). This reveals an interaction between folate metabolism and N-metabolism in darkness. Folate metabolism was also shown to maintain root development in the indeterminate state, via FPGS functioning (Reyes-Hernandez et al., 2014). Folate synthesis and therefore accumulation is high during germination and in meristematic tissues, coherent with their demand upon cell division and concomitant DNA synthesis (Rebeille et al., 2006). Moreover, folate biosynthesis is stimulated upon light exposure, indicating a higher folate requirement (Rebeille et al., 2006). Indeed, the production of chlorophyll is dependent on folate (Van Wilder et al., 2009). Moreover, folates are able to ensure sufficient NADPH production, thereby controlling cellular redox state by a balanced functioning of DHFR-TS genes, needed in detoxification of ROS originating from photosynthesis or photorespiration (Gorelova et al., 2017b). In photorespiration, folate is directly required as a cofactor for the serine hydroxymethyltransferase in the glycine decarboxylase complex (Collakova et al., 2008; Maurino and Peterhansel, 2010). Finally, folate biosynthesis enzymes are known to influence plant stress responses, possibly through generation of folate biosynthesis intermediates (Storozhenko et al., 2007b; Navarrete et al., 2012).

Given the influence of folates on plant development, their homeostasis and accumulation is considered to be tightly regulated, depending on their tissue specific requirement

(Rebeille et al., 2006). Indeed, recent insights in folate metabolism of Arabidopsis confirm fine-tuning of folate accumulation by feed-back inhibition of a regulatory DHFR-TS homolog (DHFR-TS3) (Gorelova et al., 2017b). Together, these findings raise caution toward possible implications upon folate biofortification, as an increased folate pool might influence different aspects of plant physiology (Van Wilder et al., 2009).

### Pathophysiology and Epidemiology

Humans lack the ability to synthetize folates de novo. However, they possess DHFR and FPGS enzymes, thereby allowing conversion of DHF to THF and polyglutamylated folates, respectively (Masters and Attardi, 1983; Garrow et al., 1992). Hence, humans are almost completely reliant on their diet for adequate folate supply, given that the gut microbiome has a marginal contribution to the folate pool (Camilo et al., 1996; LeBlanc et al., 2013). As the usage of folates as C1 donors and acceptors originated early in evolution, being implemented by prokaryotes and all eukaryotes, their basic functioning in plants is very similar to that in humans. Thus, folates are important in DNA synthesis and in supplying methyl groups to proteins, lipids, and DNA, through their necessity in SAM replenishment (Saini et al., 2016). Similar to plants, changes in folates levels have the potency to change the human epigenome (Bistulfi et al., 2010). Folates are required in methylation of myelin basic protein, which is pivotal for the compaction of myelin around the neuron sheath, thereby ensuring sufficient nerve conduction (Ramaekers and Blau, 2004; Bottiglieri, 2005).

Upon inadequate dietary folate intake, folate status can drop, a condition known as folate deficiency, which has a broad pathophysiology. Folate deficiency results in decreased erythrocyte development, causing megaloblastic anemia (Lanzkowsky, 2016). The elevated levels of homocysteine, resulting from low folate status, can induce vascular diseases, such as coronary artery disease and strokes (Antoniades et al., 2009; Guo et al., 2009; Zeng et al., 2015). The most notable consequence of folate deficiency is its detrimental impact on neurulation. This is revealed by the occurrence of neural tube defects (NTDs) such as spina bifida, encephalocele and anencephaly, caused by folate deficiency (Geisel, 2003; Youngblood et al., 2013; Greene and Copp, 2014). Last but not least, different forms of cancer have been linked to inadequate folate status, including colorectal (Feng et al., 2017), prostate (Price et al., 2016), and pancreatic tumors (Yallew et al., 2017).

Folate deficiency is still a global problem, predominantly present in the developing world, yet persisting in many populations of the developed world as well (Blancquaert et al., 2014; Zaganjor et al., 2015). Moreover, even populations blessed by the availability and opportunity of a diverse and folaterich diet, remain susceptible to deficiency, as illustrated by the low folate status measured in the Swedish population (Eussen et al., 2013; Gylling et al., 2014) and the observed sub-optimal folate levels in 39% of Belgian first trimester pregnancies (Vandevijvere et al., 2012). Worldwide, 300,000 pregnancies are estimated to be affected by NTDs annually, half of which are considered to be caused by insufficient maternal folate status

(Flores et al., 2014). China, inhabited by almost 1.4 billion people, recorded a countrywide prevalence of NTDs as high as 0.24% (Blancquaert et al., 2014). More strikingly, Shanxi province, located in Northern China, has amongst the highest incidence rates of NTDs in the world, as high as 1.39% (Li et al., 2006).

Fortunately, noteworthy advances have been made in the fight against folate malnutrition. Educational efforts, advocating a diverse diet containing folate rich foods such as green leafy vegetables and fermented products, is the primary strategy to diminish folate deficiency (Strobbe and Van Der Straeten, 2017). Folic acid, the synthetic form of folate as administered in pills, has been implemented in fortification strategies, which have ensured a significant reduction of neural tube defects (Williams et al., 2015; Wang et al., 2016). Unfortunately, high folic acid intake can also impose unwanted side effects, since excessive accumulation of unmetabolized folic acid has been linked to colorectal cancer and impaired immunity (Cho et al., 2015; Selhub and Rosenberg, 2016). Moreover, both folic acid fortification and supplementation are costly interventions, which are difficult to implement in poor rural regions in need (Blancquaert et al., 2014). Therefore, biofortification, via metabolic engineering or breeding is advised to ensure a stable cost-effective means to fight folate deficiency (De Steur et al., 2012, 2015; Blancquaert et al., 2014; Strobbe and Van Der Straeten, 2017).

### Biofortification

Over the last decades, many successful folate biofortification approaches have been conducted, thereby additionally acquiring new insights in folate metabolism in certain crops and tissues (De Lepeleire et al., 2017; Strobbe and Van Der Straeten, 2017). The most widely attempted folate metabolic engineering approach is the enhancement of GTPCHI activity, proven to be a fruitful strategy in prokaryotes (Sybesma et al., 2003). This approach has been confirmed to be functional in plants by the engineering of cis-genic Arabidopsis lines, over-expressing GTPCHI (Hossain et al., 2004). This single gene approach, introducing GTPCHI, referred to as G-engineering, has been implemented in rice (Storozhenko et al., 2007a), tomato (de la Garza et al., 2004), maize (Naqvi et al., 2009), lettuce (Nunes et al., 2009), potato (Blancquaert et al., 2013a), and Mexican common bean (Ramírez Rivera et al., 2016). The highest fold enhancement, reached in the edible portions of these crops is a ninefold folate increase in lettuce. This could possibly be due to a difference in regulation in leafy tissue. However, single gene approaches have hitherto not resulted in over 10-fold increase in folate content. A bigenic approach was substantially more successful, adding ectopic expression of aminodeoxychorismate synthase (ADCS) (GA-strategy). In tomato (de la Garza et al., 2007) and rice (Storozhenko et al., 2007a) this led to 25- and 100-fold folate enhancement, respectively. Unfortunately, this approach, able to reach the desired levels in tomato and rice, does not promise to be universally applicable, as it only resulted in limited enhancements in Arabidopsis and potato (Blancquaert et al., 2013a). In rice seeds, ADCS has been indicated as the most important limiting factor in folate biosynthesis, additional to

GTPCHI (Dong et al., 2014a). Building further on these findings, novel biofortification approaches aimed at further gene stacking, using mitochondrial folate biosynthesis genes (Strobbe and Van Der Straeten, 2017). Indeed, additional introduction of FPGS in GA-engineered plants did not only result in elevated folate levels in rice endosperm (100-fold) and potato tubers (12-fold) respectively, but also in enhanced folate stability upon storage (Blancquaert et al., 2015; De Lepeleire et al., 2017). Increasing storage stability has also been addressed by introduction of mammalian folate binding proteins (Blancquaert et al., 2015). This strategy is promising, as it could limit the aforementioned undesired effects of folate increase on plant physiology, via sequestration of the active folate pool. Moreover, recent discovery of plant folate binding proteins creates novel opportunities in folate biofortification via metabolic engineering (Puthusseri et al., 2018).

Breeding endeavors, aimed at acquiring elite crop variants with augmented folate content in the edible portion, though not implemented so far, have shown to be feasible (Andersson et al., 2017; Bouis and Saltzman, 2017). Upon availability of high throughput folate quantification in the food matrix, screening of vast germplasm collections could lead to identification of high folate varieties (De Brouwer et al., 2010; Strobbe and Van Der Straeten, 2017). In this respect, over sevenfold variation in milled rice folate content was described by examination of 78 accessions (Dong et al., 2011). More recently, unpolished brown rice folate content was found to vary up to threefold in 150 examined accessions (Aiyswaraya et al., 2017). Similar screening has been employed in barley (Andersson et al., 2008), red beet (Wang and Goldman, 1996), potato (Goyer and Sweek, 2011; Robinson et al., 2015), tomato (Iniesta et al., 2009), muskmelon (Lester and Crosby, 2002), common bean (Khanal et al., 2011; Jha et al., 2015), lentil (Jha et al., 2015), (chick)pea (Jha et al., 2015), spinach (Shohag et al., 2011), and strawberry (Mezzetti et al., 2016). Furthermore, these variations could be utilized to identify interesting QTLs, underlying folate content, in GWAS (Khanal et al., 2011; Dong et al., 2014b). These techniques, though limited in their potential folate enhancement, are promising, as they might face lower regulatory restrictions, hence allow more rapid implementation in agriculture, reaching the populations in need (Mejia et al., 2017; Potrykus, 2017).

### B-VITAMIN INTERPLAY

Multi-biofortification is considered an important goal in the fight against MNM (Blancquaert et al., 2014; Strobbe and Van Der Straeten, 2017). However, possible effects of altered micronutrient levels upon each other as well as on basic plant growth and development, should be taken into consideration. Examination of the role of B-vitamins in plant metabolism evidently reveals that inducing their accumulation could alter plant physiology. This has been conspicuously observed in metabolic engineering approaches of B1 (Bocobza et al., 2013; Dong et al., 2015) and B6 (Raschke et al., 2011). Furthermore, B9 enhancement, though not depicting any severe effect on

plant growth, has shown to alter the rice seed metabolism (Blancquaert et al., 2013b). The influence of B-vitamins on plant metabolism is, however, at least partly intertwined, indicating the importance of detailed investigation of the effect of their combined biofortification (**Figure 6**).

In central energy metabolism, both folate and B1 appear to negatively influence the plant's ability to accumulate starch (Bocobza et al., 2013; Hayashi et al., 2017). PLP (B6) is also involved in starch breakdown, though there are no indications to suspect increased starch breakdown upon elevation of PLP levels (Zeeman et al., 2004; Mooney and Hellmann, 2010). In the biosynthesis of B6, G3P, an intermediate in central energy metabolism (glycolysis), serves as a substrate (Fudge et al., 2017), the steady state concentration of which might be altered in B1 engineered lines (Bocobza et al., 2013). In altering this central metabolism equilibrium, B1 augmentation might influence the flux through the shikimate pathway (Bocobza et al., 2013), the activity of which is required in the plastidial part of folate biosynthesis (Strobbe and Van Der Straeten, 2017). In this shikimate pathway, PLP (B6) is required as cofactor. Folates are able to generate NADPH (Gorelova et al., 2017b), replenish the SAM pool (Blancquaert et al., 2010), and are needed in the biosynthesis of iron sulfur cluster enzymes (Waller et al., 2010). Interestingly, THIC, pinpointed as the rate limiting step in B1 biosynthesis, contains an iron-sulfur cluster and requires SAM for its catalytic activity (Pourcel et al., 2013). Strongly increased THIC activity would therefore require enhanced SAM turnover (Palmer and Downs, 2013), for which enhanced folate levels might have a beneficial effect. NADPH is on the other hand required for pyridoxal reductase activity in B6 homeostasis. Ribose 5<sup>0</sup> -phosphate, an important substrate in B6 biosynthesis (Fudge et al., 2017), is a product of the pentose phosphate pathway, the flux of which might be controlled by B1 (Bocobza et al., 2013). Similarly, AIR, an important substrate in B1 biosynthesis (Pourcel et al., 2013), is derived from purine metabolism, the synthesis of which is dependent on folate (Strobbe and Van Der Straeten, 2017). Moreover, B1, B6, and B9 have been linked to nitrogen metabolism. First, thiamin application is known to stimulate nitrogen assimilation (Bahuguna et al., 2012). Second, B6 content was observed to be altered in the ammonium transporter mutant amt1 (Pastor et al., 2014). Moreover, PLP (B6) salvage mutant pdx3 is depending on ammonium (Colinas et al., 2016). Third, folate biosynthesis mutants (atdfb-3, plastidial FPGS) harbored enhanced nitrogen content of seeds (Meng et al., 2014). Remarkably, given the labile nature of folate, increasing in planta stabilization of folates has been the subject of biofortification strategies (Blancquaert et al., 2015). Therefore, enhancing levels of antioxidants, such as B6, has been proposed as an additional biofortification strategy, protecting the folate pool from oxidative cleavage (Blancquaert et al., 2014).

### REFERENCES

Abdou, E., and Hazell, A. S. (2015). Thiamine deficiency: an update of pathophysiologic mechanisms and future therapeutic

### FINAL REMARKS

Creation and evaluation of multi-biofortified crops would not only offer a sustainable solution to eradicate MNM, but also help to elucidate the interplay of different micronutrients. The availability of novel tools, allowing facilitated cloning of multiple genes paved the way toward such multi-biofortification (Engler et al., 2014). Furthermore, a prerequisite in biofortification strategies is to consider stability upon storage of the crop product, as well as after food processing and bioavailability upon human consumption (Blancquaert et al., 2015; Diaz-Gomez et al., 2017). Different agronomical techniques could be employed, alone or in combination, to augment vitamin content of crops. Metabolic engineering of the complete pathway, or symbiosis with bacteria, might be appropriate ways to tackle vitamin B12 deficiency (DeMell and Holland, 2016). Metabolic engineering strategies could be developed in a precise way, enabling the creation of food crops which harbor an ideal balance of energy supply and micronutrient delivery, while exhibiting marginal effects on plant physiology. These novel crop varieties could, in combination with fortification and dietary interventions eradicate MNM, alleviating a great global burden.

### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

### FUNDING

SS is indebted to the Agency for Innovation by Science and Technology in Flanders (IWT) for a predoctoral fellowship. DVDS acknowledges support from Ghent University (Bijzonder Onderzoeksfonds, BOF2009/G0A/004), and the Research Foundation—Flanders (FWO, project 3G012609).

### ACKNOWLEDGMENTS

The authors thank Jolien De Lepeleire for the helpful suggestions and critical comments on the manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018.00443/ full#supplementary-material

TABLE S1 | An overview of the abbreviations.

considerations. Neurochem. Res. 40, 353–361. doi: 10.1007/s11064-014- 1430-z

Adeva-Andany, M. M., Lopez-Maside, L., Donapetry-Garcia, C., Fernandez-Fernandez, C., and Sixto-Leal, C. (2017). Enzymes involved in branched-chain

amino acid metabolism in humans. Amino Acids 49, 1005–1028. doi: 10.1007/ s00726-017-2412-7


toward sustainable health. Crit. Rev. Plant Sci. 29, 14–35. doi: 10.1080/ 07352680903436283


folates in rice. J. Chromatogr. B. 878, 509–513. doi: 10.1016/j.jchromb.2009. 12.032


invasive tumour stage-specific malignancy of human colorectal cancers. Int. J. Cancer 141, 2537–2550. doi: 10.1002/ijc.31008



in human primary keratinocytes and reconstructed epidermis. Photochem. Photobiol. 93, 990–998. doi: 10.1111/php.12720



a systematic review and meta-analysis. J. Nutr. 144, 906–914. doi: 10.3945/jn. 113.189589


pathway in Arabidopsis thaliana. New Phytol. 202, 1223–1236. doi: 10.1111/nph. 12757


clusters in all domains of life. Proc. Natl. Acad. Sci. U.S.A. 107, 10412–10417. doi: 10.1073/pnas.0911586107


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Strobbe and Van Der Straeten. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# You Shall Not Pass: Root Vacuoles as a Symplastic Checkpoint for Metal Translocation to Shoots and Possible Application to Grain Nutritional Quality

Felipe K. Ricachenevsky1,2 \*, Artur T. de Araújo Junior<sup>2</sup> , Janette P. Fett2,3 and Raul A. Sperotto<sup>4</sup>

<sup>1</sup> Departamento de Biologia, Programa de Pós-Graduação em Agrobiologia, Universidade Federal de Santa Maria, Santa Maria, Brazil, <sup>2</sup> Programa de Pós-Graduação em Biologia Celular e Molecular, Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil, <sup>3</sup> Departamento de Botânica, Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil, <sup>4</sup> Centro de Ciências Biológicas e da Saúde, Programa de Pós-Graduação em Biotecnologia, Universidade do Vale do Taquari – UNIVATES, Lajeado, Brazil

### Edited by:

Manuel González-Guerrero, Universidad Politécnica de Madrid (UPM), Spain

### Reviewed by:

Seçkin Eroglu, ˘ ˙ Izmir University of Economics, Turkey Gian Pietro Di Sansebastiano, University of Salento, Italy

> \*Correspondence: Felipe K. Ricachenevsky felipecruzalta@gmail.com

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 08 February 2018 Accepted: 14 March 2018 Published: 03 April 2018

### Citation:

Ricachenevsky FK, de Araújo Junior AT, Fett JP and Sperotto RA (2018) You Shall Not Pass: Root Vacuoles as a Symplastic Checkpoint for Metal Translocation to Shoots and Possible Application to Grain Nutritional Quality. Front. Plant Sci. 9:412. doi: 10.3389/fpls.2018.00412 Plant nutrient uptake is performed mostly by roots, which have to acquire nutrients while avoiding excessive amounts of essential and toxic elements. Apoplastic barriers such as the casparian strip and suberin deposition block free diffusion from the rhizosphere into the xylem, making selective plasma membrane transporters able to control elemental influx into the root symplast, efflux into the xylem and therefore shoot translocation. Additionally, transporters localized to the tonoplast of root cells have been demonstrated to regulate the shoot ionome, and may be important for seed elemental translocation. Here we review the role of vacuolar transporters in the detoxification of elements such as zinc (Zn), manganese (Mn), cadmium (Cd), cobalt (Co) and nickel (Ni) that are cotransported with iron (Fe) during the Fe deficiency response in Arabidopsis thaliana, and the possible conservation of this mechanism in rice (Oryza sativa). We also discuss the evidence that vacuolar transporters are linked to natural variation in shoot ionome in Arabidopsis and rice, indicating that vacuolar storage might be amenable to genetic engineering without strong phenotypical changes. Finally, we discuss the possible use of root's vacuolar transporters to increase the nutritional quality of crop grains.

Keywords: biofortification, nutritional quality, vacuolar transport, metals, ionome

### ROOT APOPLASTIC AND SYMPLASTIC CONTROL OF METAL UPTAKE

Roots are the primary sites of nutrient absorption and as such they must carefully control elemental uptake. This is accomplished via selective transporters at the plasma membrane of root cells at the epidermal and cortical cell layers. Root cells have their cytoplasm connected by plasmodesmata, membrane-lined channels that cross cell walls and allow diffusion of solutes between adjacent cells. The continuum cytoplasm-plasmodesmata of several cells make up the symplast (Rutschow et al., 2011). Once a molecule crosses an epidermal or cortical cell plasma membrane, it can move radially from the external layers into the internal stele and reach the pericycle by diffusion.

Ricachenevsky et al. Root Vacuolar Transporters and the Ionome

The next step is xylem loading, which is also dependent on membrane selective transporters that efflux nutrients out of the symplast. Thus, the symplastic route depends on the coordination between influx transporters at the external root cell layers, efflux transporters at the internal cell layers, and diffusion between cells that are symplastically connected (Miwa and Fujiwara, 2010). Indeed, influx and efflux transport systems characterized to date were shown to be important for the control of elemental concentrations in the xylem sap and consequently in the shoots (**Figure 1A**; Miwa and Fujiwara, 2010; Sasaki et al., 2016).

Solutes from the rhizosphere can also move radially into root tissues penetrating extracellular spaces between cells and in cell walls, comprising the apoplast. However, at the endodermal cells, a lignin band deposited in anticlinal cell walls, the Casparian Strip, provides an extracellular barrier to prevent free diffusion into the stele through the apoplast (**Figure 1A**; Naseer et al., 2012). This cell wall modification interrupts the apoplastic communication from the external cell layers into the stele, thus making selective nutrient transport into the symplast necessary for ions to reach the xylem. Since endodermal cells contact both the external (connected to the rhizosphere) and internal (connected to the xylem) apoplast compartments, they are crucial to root nutrient uptake (Geldner, 2013; Barberon et al., 2016).

The composition of the Casparian Strip bands and the mechanisms and genes involved in its formation during endodermal development are being dissected in detail in the model species Arabidopsis thaliana (Geldner, 2013). It was shown that Casparian Strips are actually made of lignin, not suberin (Naseer et al., 2012). Recently, it was demonstrated that suberin deposition on endodermal cell surfaces in response to nutritional stress could block access of apoplast solutes to plasma membranes, therefore making absorption by the epidermis and cortical cells necessary (Barberon et al., 2016). Thus, it is expected that changes in Casparian Strip bands and suberin lamellae deposition would result in altered radial nutrient movement in roots and modified access to the xylem and shoot translocation. Indeed, changes in Casparian Strip porosity results in leakage of nutrients from one apoplastic compartment to another, which changes xylem sap concentrations and consequently perturbs the shoot ionome (Hosmani et al., 2013; Kamiya et al., 2015; Huang and Salt, 2016). Thus, maintenance of diffusional barriers in the root apoplast is important for controlling root-to-shoot translocation of nutrients.

### ROOT CELL VACUOLES AS CHECKPOINTS FOR METAL DIFFUSION IN THE SYMPLAST AND ROOT-TO-SHOOT TRANSLOCATION IN ARABIDOPSIS: THE Fe DEFICIENCY EXAMPLE

With proper apoplast diffusional barriers and the consequent symplastic control of absorption, the rate of uptake from

FIGURE 1 | Root vacuolar compartmentalization regulating the ionome. (A) Checkpoints of ion radial movement within roots. (1) Endodermal diffusion barriers block ions entry in the apoplast connected to the xylem. (2) Influx and (3) efflux transporters control ion concentration in the root symplast. Influx and efflux transporters may also be present in plasma membranes of other cells, but are shown in epidermis and pericycle for clarity. (4) Root vacuoles can restrict symplastic movement of ions, and therefore decrease or increase their availability for xylem loading and shoot/seed translocation. Different cell layers may have distinct vacuolar repertoire for storage. Gray arrows indicate diffusion within the symplast through plasmodesmata. Epi, epidermis; Cor, cortex; En, endodermis; Per, pericycle. Red band – Casparian Strip; yellow = suberin deposition. (B) Data from Genevestigator showing regulation under Fe deficiency of Arabidopsis genes AtIRT1 (AT4G19690), AtHMA3 (AT4G30120), AtFPN2 (At5G03570), AtMTP3 (AT3G58810), and AtMTP8 (AT3G58060). (C) Graphical visualization of the coexpressed gene network of the same genes as in (B) using ATTED-II (http://atted.jp/). Nodes (hexagons and circles) represent genes, while straight lines represent coexpression. Nodes in hexagonal shape and orange color represent genes of interest. Nodes in circle shape and purple color represent genes within the network which are directly connected to the genes of interest. Nodes in circle shape and light blue color represent genes which are connected to the purple circle genes. (D) Data from Genevestigator showing regulation under Fe deficiency of rice genes LOC\_Os03g46470 (OsIRT1), LOC\_Os02g43410 (OsYSL15), LOC\_Os07g12900 (OsHMA3), LOC\_Os05g03780 (OsMTP1), LOC\_Os03g12530 (OsMTP8.1), LOC\_Os02g53490 (OsMTP8.2), and LOC\_Os06g36450 (not yet characterized, but most similar gene to AtFPN1/AtFPN2; named OsFPN1).

the soil and xylem loading presumably determines the concentration of a given element and the amount of rootto-shoot translocation. However, root vacuoles also control nutrients and trace elements concentrations in the root symplast (**Figure 1A**). Studies in Arabidopsis have shown that specific vacuolar transporters expressed in roots perform vacuolar compartmentalization, which can impact xylem loading and root-to-shoot translocation. Loss-of-function of these transporters result in higher translocation of respective elements to shoots, presumably due to increased element availability in the root symplast for efflux into the xylem (Arrivault et al., 2006; Morrissey et al., 2009).

A striking example where vacuolar compartmentalization for multiple elements is part of a coordinated response, in which vacuoles detoxify elements that increase their concentrations due to excessive uptake, is observed during Fe deficiency response in Arabidopsis (**Figures 1B,C**). The classical Fe acquisition mechanism (reduction strategy, or strategy I) includes rhizosphere acidification by an H+-ATPase, Fe3<sup>+</sup> reduction to Fe2<sup>+</sup> by a membrane-bound, extracellular-facing reductase protein, and Fe2<sup>+</sup> uptake by the high affinity transporter AtIRT1 (Brumbarova et al., 2015). AtIRT1 has broad specificity, being able to transport other divalent metals, such as Zn2+, Mn2+, Co2+, Cd2+, and Ni2<sup>+</sup> (Korshunova et al., 1999; Barberon et al., 2014), which are potentially harmful. Indeed, increased concentrations of Zn, Mn, Co, and Cd in Arabidopsis shoots are part of the ionomics profile associated with physiologically Fe deficient plants, even if Fe concentration is not affected (Baxter et al., 2008). Recent work showed that non-Fe metals regulate AtIRT1 localization at the plasma membrane, which suggests that plants must balance Fe and other metal uptake through AtIRT1 under low Fe for optimal nutrition (Barberon et al., 2014).

This observation indicates that AtIRT1 is the main route of entry for these metals, which transiently accumulate in roots of Fe deficient plants. The vacuolar transporters AtMTP3, AtMTP8, AtFPN2, and AtHMA3, which are, respectively, Zn, Mn, Co/Ni, and Cd/Zn transporters (Arrivault et al., 2006; Schaaf et al., 2006; Morel et al., 2009; Morrissey et al., 2009; Eroglu et al., 2016), are coordinately up regulated upon Fe deficiency, presumably in order to decrease local high concentrations in the root symplast (**Figures 1B,C**; Buckhout et al., 2009; Thomine and Vert, 2013). Consequently, their activity can reduce metal accumulation in shoot tissues. Therefore, the action of vacuolar transporters in compartmentalization of metals into root vacuoles indirectly control the shoot ionome, indicating that root vacuoles are a checkpoint for metal movement into the xylem and can fine-tune the accumulation of essential but/or potentially toxic elements in shoots.

Interestingly, both AtFPN2 and AtHMA3 were shown to be involved in natural variation of Co and Cd shoot concentrations, respectively (Morrissey et al., 2009; Chao et al., 2012). Accessions harboring an insertion in the coding sequence of AtFPN2, which results in a truncated version of the protein, were hypersensitive to Co and Ni, and had increased concentrations of Co in shoots. The increased Co accumulation was more pronounced in conditions of low Fe availability, indicating that AtIRT1 uptake and AtFPN2 vacuolar compartmentalization work in concert to control Co movement in the symplastic xylem loading and root-to-shoot translocation (Morrissey et al., 2009). Regarding AtHMA3, a non-functional allele is present in several accessions of Arabidopsis, resulting in higher Cd concentration in shoots (Chao et al., 2012). AtHMA3 allele variation was considered the primary determinant of Cd concentration variation in shoots of Arabidopsis multiple accessions (Chao et al., 2012). These data suggest that vacuolar sequestration in roots might be important not only to general metal detoxification, but that fine tuning of detoxification could be involved in local adaptation of distinct genotypes within a species. Therefore, allele diversity of root vacuolar transporters might help to explain natural variation in the shoot ionome.

### ROOT VACUOLAR COMPARTMENTALIZATION UNDER Fe DEFICIENCY IN RICE

There is little evidence for a conserved mechanism during Fe deficiency response in other species than Arabidopsis, although some of the orthologous genes are also up-regulated by Fe deficiency (**Figure 1D**). In rice, the best model species for monocots, Fe deficiency induces the combined strategy (with elements from both classical strategies I and II), which upregulates OsIRT1 (Ricachenevsky and Sperotto, 2014). However, evidence that OsIRT1 has broad specificity is still lacking, although over-expression of OsIRT1 leads to increased Fe, Zn, and Cd concentrations in rice plants (Lee and An, 2009). In rice, the MTP group 1 clade has only one member, named OsMTP1 (Ricachenevsky et al., 2013). OsMTP1 has been suggested to detoxify Zn into vacuoles as part of basal Zn tolerance, resembling the AtMTP1 function (Menguer et al., 2013; Ricachenevsky et al., 2015). Thus, an AtMTP3-like gene (i.e., with a function to detoxify high Zn under Fe deficiency) might be lacking in rice. Still, OsMTP1 may be somewhat involved in the Fe deficiency response, since expression data indicates it might be up-regulated in rice roots under low Fe conditions (**Figure 1D**).

The rice ortholog of AtHMA3, named OsHMA3, has a role in Cd vacuolar detoxification in roots. Positional cloning and natural accession screening has shown that OsHMA3 is the causative gene of variation in Cd shoot concentrations (Miyadate et al., 2011; Yan et al., 2016). Conversely, overexpression of OsHMA3 resulted in increased Cd tolerance, with Cd concentrations increasing in roots and decreasing in shoots (Sasaki et al., 2014). These results show that OsHMA3 performs a similar role as AtHMA3, and that both are targets for variation in Cd concentrations within each species.

There are two AtMTP8 orthologous genes in rice: the duplicated gene pair OsMTP8.1 and OsMTP8.2. Both proteins are localized at on the vacuole and are involved in Mn tolerance, similar to what is described for AtMTP8. However, decreased expression or loss-of-function of both transporters results in lower Mn concentrations in roots, but not in

shoots, indicating that they may work differently than AtMTP8 regarding its role in controlling shoot translocation (Chen et al., 2013; Takemoto et al., 2017). Interestingly, OsMTP8.1 seems to be up-regulated by Fe deficiency, similar to AtMTP8 (**Figure 1D**).

### ROOT VACUOLAR COMPARTMENTALIZATION IMPACTS THE GRAIN IONOME

Recent work has clearly shown that root vacuolar transporters can also affect mineral accumulation in grain. In rice, OsHMA4 is the causative gene of high grain Cu phenotype found in some accessions. Loss-of-function or natural variants with decreased OsHMA4 function result in increased Cu concentration in shoots and grains, as well as decreased concentration in roots and in root cell sap. Thus, OsHMA4 detoxifies Cu into root vacuoles, which decreases Cu translocation to shoots and grains (Huang et al., 2016). Similarly, the rice tonoplast-localized OsABCC1 transporter was shown to detoxify arsenic (As) by transporting As(III)-phytochelatin into vacuoles. The osabcc1 mutants have increased As sensitivity and As accumulation in grains (Song et al., 2014). In Arabidopsis, the duplicated pair AtABCC1/AtABCC2 also has a similar role (Song et al., 2010), again indicating that there is conservation of function in distantly related species. Moreover, OsHMA3 natural variation was clearly linked to high/low Cd in rice grains (Yan et al., 2016).

The vacuolar iron transporter (VIT) family also deserves attention. In Arabidopsis, the AtVIT1 gene is key for correct distribution of Fe within seeds (Kim et al., 2006). Interestingly, rice has two genes, OsVIT1 and OsVIT2, which are involved in Fe storage in flag leaves. Intriguingly, high Fe (a common condition in lowland rice) up regulates OsVIT2 in roots, indicating that rice plants might have a mechanism to avoid Fe toxicity using root vacuolar compartmentalization. Moreover, OsVIT1/OsVIT2 also seem to regulate Fe distribution in seeds (Zhang et al., 2012), and endosperm-specific over-expression of the orthologous wheat gene TaVIT2 results in increased Fe content in wheat grains (Connorton et al., 2017).

Based on that, we expect that Arabidopsis mutants and/or natural variants with weak alleles for AtMTP3, AtMTP8, AtFPN2, and AtHMA3 would have higher concentrations of their respective substrates in seeds. Indeed, AtMTP3-RNAi plants show increased Zn concentrations in whole inflorescences and siliques, and marginal increase in seeds (Arrivault et al., 2006). Under Mn sufficient conditions, loss-of-function mtp8 mutants and WT showed similar or slightly increased Mn concentration in seeds, whereas under Mn deficient conditions the mutants had decreased concentration compared to WT (Chu et al., 2017; Eroglu et al., 2017). These results suggest that in the presence of Mn, the lack of vacuolar root transporter allow more Mn to be translocated to seeds, compensating for decreased sink strength for Mn in the mtp8 mutant, which is apparent under Mn deficient conditions (Eroglu et al., 2017). It would be interesting to have data on Mn and Zn concentration in seeds of mtp8 mutants and AtMTP3-RNAi plants grown under high concentration of each element. Moreover, mutants or natural variants for AtHMA3 and AtFPN2 did not have their seed metal concentration evaluated (Morel et al., 2009; Morrissey et al., 2009; Chao et al., 2012), and no data for seeds of these plants is available in the Ionomics Hub database<sup>1</sup> .

It is important to note that available evidence indicates that vacuolar compartmentalization is one of the checkpoints controlling the seed ionome, but by no means the only one. Depending on the element and its chemical speciation, other transporters expressed throughout the plant could contribute to the regulation of translocation to developing seeds (Sperotto et al., 2012; Punshon et al., 2018). An interesting example is highlighted by natural variation in AtHMA3, which is linked to species-wide Cd concentration variation in leaves but not to Zn and cobalt (Co) variation to the same extent, despite being able to transport both (Morel et al., 2009; Chao et al., 2012). One possibility is that elements such as Zn might be more tightly regulated, and thus variation in vacuolar transporter activity in roots might be compensated by other transporters (Chao et al., 2012). Thus, it remains to be tested the extent to which vacuolar compartmentalization is important to seed accumulation, which elements are most impacted, the transporters involved in this regulation, and how that varies in different species.

## CONCLUSION

Biofortification of grains has been a long sought goal on the plant nutrition field, especially for Fe and Zn, the two most commonly lacking minerals in the human diet (Sperotto et al., 2012; Ricachenevsky et al., 2015). Arsenic is also a problem in rice, since it can accumulate in grains to harmful levels for human consumption (Punshon et al., 2017). Root vacuolar compartmentalization works in concert with other checkpoints to control elemental translocation to the shoots and, consequently, to the grains. Therefore, we should expect that mutants and/or accessions with loss-of-function alleles coding root vacuolar transporters have increased concentrations of the respective elements in the xylem sap, and potentially increased concentrations in seeds.

We conclude that (1) tonoplast-localized root transporters can fine-tune symplastic concentrations of ions, together with apoplastic barriers and influx/efflux transporters; (2) plants are likely to tolerate changes in vacuolar storage capacity without strong changes in phenotype, since natural variation harbors loss-of-function alleles; and (3) orthologous genes in distantly related species might be hotspots of genetic variation (Morrissey et al., 2009; Chao et al., 2012; Huang et al., 2016; Yan et al., 2016). Thus, vacuolar transporters in roots are good candidates to search for interesting alleles and for engineering both shoot and seeds' ionome for biofortification and nutritional quality.

<sup>1</sup>www.ionomicshub.org

### AUTHOR CONTRIBUTIONS

fpls-09-00412 March 29, 2018 Time: 16:46 # 5

FR, JF, and RS wrote the manuscript. FR and AdAJ drew figures and presentation of previously published public data. All authors approved the final manuscript.

### REFERENCES


### FUNDING

Authors were supported by Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES).

Proc. Natl. Acad. Sci. U.S.A. 112, 10533–10538. doi: 10.1073/pnas.15076 91112


in iron-dependent nickel detoxification in Arabidopsis thaliana roots. J. Biol. Chem. 281, 25532–25540. doi: 10.1074/jbc.M601062200


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Ricachenevsky, de Araújo Junior, Fett and Sperotto. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Nicotianamine Synthase Gene Is a Useful Candidate for Improving the Nutritional Qualities and Fe-Deficiency Tolerance of Various Crops

#### Tomoko Nozoye1,2 \*

<sup>1</sup> Center for Liberal Arts, Meiji Gakuin University, Kanagawa, Japan, <sup>2</sup> Department of Global Agricultural Sciences, Graduate School of Agricultural and Life Sciences, The University of Tokyo, Tokyo, Japan

With the global population predicted to grow by at least 25% by the year 2050, the sustainable production of nutritious foods will be necessary for human health and the environment. Iron (Fe) is an essential nutrient for both plants and humans. Fe is poorly soluble, especially at high pH levels, at which it is difficult for living organisms to accumulate sufficient Fe. In plants, Fe deficiency leads to low yield and poor nutritional quality, as it significantly affects chlorophyll synthesis. Fe deficiency is a worldwide agricultural problem that is especially serious in soils with a high pH, such as calcareous soils, which comprise approximately 30% of cultivated soils worldwide. Genetic improvements in crops that can tolerate Fe deficiency will be required to meet the demands for crop production and could ultimately contribute to the amelioration of global warming. Nicotianamine (NA) is an Fe chelator in plants that is involved in metal translocation in the plant body. In mammals, NA inhibits angiotensin I-converting enzyme, which plays a key role in blood pressure control. It was recently shown that the enhancement of NA production using nicotianamine synthase is useful for increasing not only NA but also Fe and Zn levels in crops such as rice, soybean, and sweet potato. Additionally, these plants showed Fe-deficiency tolerance in calcareous soil. These results suggested that NAS overexpression simultaneously improves food quality and increases plant production. This review summarizes progress in generating crops overexpressing NAS.

Keywords: calcareous soil, iron (Fe), zinc (Zn), Fe deficiency, nicotianamine (NA)

### INCREASING Fe DEFICIENCY TOLERANCE COULD CONTRIBUTE TO FOOD SECURITY AND AMELIORATE GLOBAL WARMING

Iron (Fe) is an essential nutrient for virtually all living organisms. Under aerobic conditions, Fe is oxidized to Fe(III) compounds, and their solubility in water is poor. Therefore, most Fe is not available to plants, although mineral soils contain 6% Fe by weight. Plants suffering Fe deficiency show leaf chlorosis, and their yield and nutritional quality are impaired dramatically

### Edited by:

Alexander Arthur Theodore Johnson, University of Melbourne, Australia

#### Reviewed by:

Giacomo Cocetta, Università degli Studi di Milano, Italy Wricha Tyagi, Central Agricultural University, India

\*Correspondence: Tomoko Nozoye atom1210@mail.ecc.u-tokyo.ac.jp

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 03 December 2017 Accepted: 28 February 2018 Published: 27 March 2018

#### Citation:

Nozoye T (2018) The Nicotianamine Synthase Gene Is a Useful Candidate for Improving the Nutritional Qualities and Fe-Deficiency Tolerance of Various Crops. Front. Plant Sci. 9:340. doi: 10.3389/fpls.2018.00340

**235**

(Marschner, 1995). This problem is exacerbated in soils with a high pH, such as calcareous soils, and is one of the major problems for crop production. Calcareous soils comprise approximately 30% of the cultivated soils worldwide (Chen and Barak, 1982). As the world population continues to increase, it is predicted that we will need 1.5 times more food in 2050 (High Level Expert Forum - How to Feed the World in 2050, 2009). It will be necessary to increase food production to meet this demand. However, the recent increase in atmospheric CO<sup>2</sup> levels is causing climate change, and it will be difficult to expand the area occupied by cultivated land by removing forests, which already contribute 2% of the CO<sup>2</sup> emissions (Intergovenmental Panel on Climate Change [IPCC], 2007). Problem soils, including calcareous soil, comprise 67% of the land globally (Food and Agriculture Organization of the United Nations). Improvements in plant growth in calcareous soils have great potential to increase the production of plant biomass and reduce atmospheric CO<sup>2</sup> levels, which will ultimately contribute to ameliorating global warming (Conway, 2012; Schroeder et al., 2013). In addition, Fe is necessary for human health, and its deficiency causes anemia, an easily identified disease that is a serious health problem, especially in developing countries (Welch and Graham, 2004). Ultimately, Fe in the human diet comes from plant uptake from the soil. Therefore, biofortification, i.e., increasing the Fe level in food plants, would improve human health. Appropriate target levels of Fe might differ according to the crop and target countries, as food cultures differ. For a rice-based diet, the target concentration of Fe was estimated to be 14.5 µg/g dry weight (DW) in polished rice grains, which is more than twice the present amount in rice (6 µg/g DW) (Hotz and McClafferty, 2007; Johnson et al., 2011). Therefore, genetically modified crops that can tolerate Fe deficiency while taking up sufficient Fe from calcareous soils would have a great impact on food security and contribute to ameliorating global warming.

### Fe ACQUISITION STRATEGIES IN PLANTS

To acquire sparingly insoluble Fe from soil, plants have evolved two main strategies to acquire soil Fe (Marschner et al., 1986). Higher plants, not including graminaceous plants, which include soybean and sweet potato, are categorized as Strategy I plants, which reduce Fe(III) to Fe(II) by ferric-chelate reductases, and then take up Fe(II) via ferrous iron transporters IRT1 (Eide et al., 1996; Robinson et al., 1999; Vert et al., 2002). By contrast, graminaceous plants, including important staple crops such as rice, barley, and maize, are categorized as Strategy II plants, which produce and secrete Fe(III) chelators called mugineic acid family phytosiderophores (MAs) from their roots via the TOM1 transporter (Nozoye et al., 2011, 2013) and solubilize sparingly soluble Fe(III) in the rhizosphere (Takagi, 1976).

Nicotianamine (NA) is a non-proteinogenic amino acid that was first found in tobacco (Noma et al., 1971). NA chelates many metal cations, including Fe, zinc (Zn), copper (Cu), and manganese (Mn) (Beneš et al., 1983; Murakami et al., 1989; von Wirén et al., 1999). NA exists in all plants examined so far, including Strategy I and II plants (Hell and Stephan, 2003; Takahashi et al., 2003; Schuler et al., 2012), and plays an important role in the internal transport of metal nutrients (Mori et al., 1991; Kawai et al., 2001; Hell and Stephan, 2003; Takahashi et al., 2003; Suzuki et al., 2006; Schuler et al., 2012). In graminaceous plants, NA also serves as an intermediate for the biosynthesis of MAs (Takagi, 1976; Mori and Nishizawa, 1987; Shojima et al., 1990). NA synthase (NAS) converts three molecules of S-adenosyl methionine into NA (Shojima et al., 1989, 1990; Higuchi et al., 1995). NAS genes were first isolated from barley and have subsequently been cloned from several plants species, including Arabidopsis, barley, rice, and maize (Herbik et al., 1999; Higuchi et al., 1999, 2001; Suzuki et al., 1999; Mizuno et al., 2003). Rice possesses three members: OsNAS1-3. OsNAS1, and OsNAS2 are mainly expressed in Fe-deficient roots and shoots, whereas OsNAS3 is also expressed in Fe-sufficient shoots (Inoue et al., 2003). It was suggested that all three have important roles in NA production under Fe-deficient conditions, although their roles might differ slightly.

### NICOTIANAMINE IS ALSO AN ATTRACTIVE FUNCTIONAL COMPONENT IN HUMAN HEALTH

In mammals, NA inhibits angiotensin I-converting enzyme (ACE), which plays a key role in blood pressure control (Kinoshita et al., 1993). ACE plays a role in the renin–angiotensin system in the maintenance of blood pressure and fluids, as well as electrolyte homeostasis (Re, 2004). ACE inhibitors are widely used as antihypertensive agents (Chirumamilla et al., 2001; Re, 2004) The oral administration of NA causes ACE inhibitory activity in vitro and antihypertensive effects in spontaneously hypertensive rats; moreover, the strength of ACE inhibition is correlated with NA content (Izawa et al., 2008). The inhibitory activity of NA against ACE is very strong (Kinoshita et al., 1993; Kataoka, 2005). Almost all vegetables contain more than 44 µg/g DW NA, which has the ability to inhibit ACE activity by more than 60–70% (Izawa and Aoyagi, 2012). In addition, NA from pumpkin not only improves hypertension, but also long-term memory function (Takada, 2011). In fact, brain-penetrating ACE inhibitors such as captopril reduce the incidence of Alzheimer's disease in elderly hypertensive patients (Ohrui et al., 2004). Therefore, increased intake of NA through the diet could be effective for primary prophylaxis of hypertension and Alzheimer's disease.

### TRANSGENIC APPROACH TO INCREASING NA IN PLANTS

Several reports have described transgenic plants generated by introducing the NAS gene (**Table 1**). The concentration of endogenous NA differs among crops (Izawa et al., 2008;


Izawa and Aoyagi, 2012). The antihypertensive effect of NA was first identified in soybean (Kinoshita et al., 1993), which contains the highest amount of NA among the crops examined thus far (Izawa et al., 2008). In agreement with the endogenous NA level, the NA concentration in transgenic soybean was highest among the HvNAS1-overexpressing plants. The NA concentration in the transgenic soybean was increased to 768.1 µg/g DW in the seeds under the control of the cauliflower mosaic virus (CaMV) 35S promoter, which was four times higher than in non-transgenic (NT) seeds (Nozoye et al., 2014a). In sweet potato, overexpression of HvNAS1 by the CaMV 35S promoter increased the NA concentration to 339.5 µg/g fresh weight (FW) in the leaves and 225.9 µg/g FW in the storage roots, which were 9.1 and 4.6 times higher, respectively, than in NT plants (Nozoye et al., 2017). In comparison, the NA concentration in HvNAS1-overexpressing tobacco (a dicot) by the CaMV 35S promoter was 78.9 µg/g FW in the leaves, which was 8.7 times higher than in NT plants (Kim et al., 2005). It was suggested that NAS genes are separated into two clusters between Gramineae and dicots and that the NASs in soybean and sweet potato were most similar in the dicot cluster (Nozoye et al., 2017). The NASs in soybean and sweet potato might have high enzymatic activity and produce more NA. The endogenous NA concentrations also differed among the tissues in rice (**Table 1**). The NA concentrations in leaves tended to be higher than that in the seeds. Consistent with the endogenous NA levels, the NA concentrations in leaves were also higher than those in seeds in NAS-overexpressing rice plants. Overexpression of HvNAS1 by the CaMV 35S promoter increased the NA concentration to 75.8 µg/g DW in the polished seeds, which was 10.6 times higher than in NT seeds (Masuda et al., 2009). Overexpression of HvNAS1 by the OsActin1 promoter increased the NA concentration to 30.3 µg/g DW in the polished seeds, which were 16 and 5.1 times higher, respectively, than in NT plants (Masuda et al., 2009). By overexpressing rice NAS genes (OsNAS1–3) in rice under the control of an enhanced CaMV 35S promoter, the NA concentration in rice seeds increased to 210 µg/g DW, which was 9.3 times higher than in NT seeds (Johnson et al., 2011). In comparison, by overexpressing OsNAS1 in rice under the control of the maize ubiquitin promoter, the NA concentration in rice leaves increased to 400 µg/g DW, which was 6.7 times higher than in NT leaves (Zheng et al., 2010). Using seed-specific expression of OsNAS1 under the control of the rice glutelin promoter, the NA concentration in rice seeds increased to 65 µg/g DW, which was 5.2 times higher than in NT seeds (Zheng et al., 2010). This concentration was slightly lower than that in NAS-overexpressing rice seeds under the control of ubiquitous promoters. Additionally, in these plants, the NA concentration in shoots was not different from that in NT plants. These results suggest that it is possible to achieve a greater increase in NA in seeds by enhancing NA mobilization and translocation from leaves (and roots) to seeds. The average increase (fold change) in NA concentration was 7.6 and did not differ significantly among the crops, suggesting that the amount of endogenous NA is not a factor that limits the NA concentration. The combined enhancement of NAS and NA transporters could further elevate the NA level in the edible parts of the plant.

fpls-09-00340 March 24, 2018 Time: 13:56 # 3

TABLE 1


concentrations

 in

NAS-overexpressing

 plants.

### ENHANCEMENT OF NA INCREASED THE Fe AND Zn CONCENTRATIONS IN PLANTS

Nicotianamine plays an important role in metal transport in the plant body. It was suggested that NA is involved in the translocation of Fe and Zn into seeds in rice, Arabidopsis, tomato, and tobacco (Higuchi et al., 1996; Takahashi et al., 2003; Kim et al., 2005; Masuda et al., 2008, 2009; Schuler et al., 2012). Fe is readily oxidized and precipitated in the apoplasm of both roots and shoots. Therefore, Fe uptake from the apoplasm is important for plant growth. In the Arabidopsis double IRT1 and Nramp1 mutant, Fe was precipitated and accumulated in the apoplast of the roots, while the Fe concentration in shoots was dramatically reduced compared with NT (Castaings et al., 2016). In NASoverexpressing plants, the Fe and Zn concentrations were also increased (**Table 1**). NA might be involved in the mobilization of Fe and Zn in the apoplasm. In the seeds of HvNAS1 overexpressing soybean plants, the Fe and Zn concentrations increased to 110 and 65 µg/g DW, which were 2 and 1.45 times higher, respectively, than in NT plants (Nozoye et al., 2014a). In HvNAS1-overexpressing sweet potato, the Fe and Zn concentrations increased to 52.9 and 17 µg/g FW in the leaves and 15.1 and 3.5 µg/g DW in the storage roots, which were 3 and 3, and 2.1 and 3.5 times higher, respectively, than in NT plants (Nozoye et al., 2017). In HvNAS1-overexpressing tobacco, the Fe and Zn concentrations increased to 5.3 and 9.6 µg/g FW in the leaves, which were 5 and 2.3 times higher, respectively, than in NT plants (Kim et al., 2005). In HvNAS1-overexpressing rice via the CaMV 35S promoter, the Fe and Zn concentrations increased to 9 and 45 µg/g DW, respectively, in the polished seeds, which were 2 and 1.5 times higher than in NT seeds (Masuda et al., 2009). In contrast, in HvNAS1-overexpressing rice via the OsActin1 promoter, the Fe and Zn concentrations were 170 and 25 µg/g DW in the leaves and 5 and 40 µg/g DW in the polished seeds, respectively, which did not differ significantly from those in NT plants (Masuda et al., 2009). By overexpressing rice NAS genes (OsNAS1–3), the Fe and Zn concentrations in rice seeds increased to 81 and 91 µg/g DW, respectively, which were 3.5 and 2.2 times higher than in NT seeds (Johnson et al., 2011). In OsNAS1-overexpressing rice via the maize ubiquitin promoter, the Fe and Zn concentrations in rice seeds were increased to 28 and 120 µg/g DW, respectively, which were 2.3 and 5.5 times higher than in NT seeds (Zheng et al., 2010). Seed-specific expression of OsNAS1 under control of the rice glutelin promoter, increased the Zn concentration in rice polished seeds to 29.07 µg/g DW, which was 2.3 times higher than in NT seeds (Zheng et al., 2010); however, the Fe concentration was not altered in the polished seeds. In these plants, the Fe and Zn concentrations in leaves were increased to 18 and 30 µg/g DW, which were 1.8 and 1.9 times higher, respectively, than in NT leaves. As with the NA concentration, the Fe and Zn concentrations tended to be higher in soybean seeds; however, this difference was not significant compared to the NA concentration. In agreement, the increases in Fe and Zn concentrations were lower than those of the NA concentration (**Table 1**). The average increases in Fe and Zn were 2.2 and 2.3, respectively, whereas that of NA was 7.6. There might be a factor limiting the increases in the metals compared with NA. Because Fe and Zn are taken up from the soil via roots, modification of the uptake system might further increase Fe and Zn. It is also possible that the increased NA in NAS-overexpressing plants was not translocated in the plant body efficiently. There might be potential to increase Fe and Zn by changing the flow of NA in the plant body.

### NAS-OVEREXPRESSING PLANTS SHOWED TOLERANCE TO Fe DEFICIENCY

Several NAS-overexpressing plants have been confirmed to tolerate Fe deficiency compared to NT plants (Lee et al., 2009; Nozoye et al., 2014a, 2017). The plant growth is dramatically reduced under Fe-deficient conditions. The plant heights and soil and plant analyzer development (SPAD) values (which represent the chlorophyll content) of NAS-overexpressing soybean and sweet potato plants were higher than those of NT plants when grown in calcareous soil with low Fe availability, suggesting that these transgenic plants were conferred tolerance to Fe deficiency. Under normal soil conditions, their growth did not differ. In rice plants, the Fe-deficiency tolerance of NAS-overexpressing rice plants in calcareous soil was not determined. HvNAS1-overexpressing rice exhibits enhanced NAS activity in Fe-deficient roots (Higuchi et al., 2001) and contains a higher amount of NA and deoxymugineic acid than NT plants in both roots and shoots (Masuda et al., 2009). Transgenic rice lines expressing barley NAS genes exhibit increased tolerance to low Fe availability in calcareous soil (Suzuki et al., 2008). Rice plants overexpressing OsIRO2, a transcription factor that enhances expression of Fe deficiency-inducible genes including OsNAS1 and OsNAS2, showed improved tolerance to low Fe availability in calcareous soil (Ogo et al., 2011). These results suggest that overexpression of the NAS gene in rice also enhances tolerance to Fe deficiency.

It was recently suggested that NA may be involved in Fe homeostasis; enhanced NA production induced Fe deficiency signaling and mobilization of Fe in the plant body (Nozoye et al., 2014b,c). Since NA has the ability to chelate Fe, NA may enable the de-repression of Fe deficiency-inducible genes by drawing Fe from an unknown Fe-sensing mechanism, and further increase the NA and deoxymugineic acid (a primary MAs) levels. NA has long been considered a candidate long-distance Fe signaling molecule in both gramineous and dicot plants (Curie and Briat, 2003); however, this has not yet been proven. In rice, NAS overexpression positively modulates Fe homeostasisrelated genes (Wang et al., 2013). NA accumulation in Osnaat1 mutants triggers a constitutive Fe deficiency response (Cheng et al., 2007). In Arabidopsis, NA-over-accumulating plants showed an Fe-deficient phenotype and expressed Fe-inducible genes at higher levels than did NT plants; however, they also contained more Fe than did NT plants, suggesting that an increase in the NA apoplastic pool sequestered Fe,

which controls plant Fe homeostasis (Cassin et al., 2009). The overexpression of ZINC-INDUCED FACILITATOR 1 (ZIF1) in Arabidopsis increased the amount of NA in the roots and shoots and led to Fe deficiency (Haydon et al., 2012). ZIF1 is a vacuolar membrane-localized putative transporter required for Zn tolerance that is hypothesized to transport NA from the cytoplasm into the vacuoles. Perturbing the subcellular distribution of NA may have profound effects on Fe with respect to subcellular distribution and inter-organ partitioning. In agreement with this phenomenon, it was revealed that AtYSL1 and AtYSL3, Fe-NA transporters, are required for proper longdistance Fe signaling (Kumar et al., 2017). A ysl1ysl3 doublemutant did not up- or down-regulate Fe deficiency-induced or -repressed genes, while it contained markedly low tissue levels of Fe compared to NT plants. These results suggest that NA may be involved in long-distance signaling to maintain Fe homeostasis. In NAS-overexpressing plants, the increased NA might induce Fedeficiency-inducible genes that contribute to conferring tolerance to Fe deficiency.

### CONCLUSION

Overexpression of the NAS gene enhances NA levels in several crops, including crops in which endogenous NA is already high, such as soybean and sweet potato. Additionally, NAS overexpression enhances the Fe and Zn concentrations and confers tolerance to Fe deficiency in calcareous soil. The increase

### REFERENCES


in NA tended to be higher in leaves than in seeds, while the increases in Fe and Zn were lower than those of NA. These results suggest the potential to increase NA, Fe, and Zn concentrations further in the edible parts of crops. Further analysis of NA translocation in the plant body will allow for improved engineering strategies not only to accumulate bioavailable Fe in edible parts, but also to increase the tolerance of plants to low Fe availability to meet the demands of plant production and to solve problems such as inadequate diet, food shortages, and global warming in the near future.

### AUTHOR CONTRIBUTIONS

TN designed and wrote the manuscript.

## FUNDING

This publication was supported by a grant-in-aid for Young Scientists (B) (Grant No. 15K18658) from JSPS KAKENHI (to TN) and by a grant from Uragami-zaidan (to TN).

### ACKNOWLEDGMENTS

I thank Prof. Naoko K. Nishizawa for reading and commenting on the manuscript.


Soil Sci. Plant Nutr. 22, 423–433. doi: 10.1080/00380768.1976.1043 3004


**Conflict of Interest Statement:** The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Nozoye. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Zinc and Iron Concentration as Affected by Nitrogen Fertilization and Their Localization in Wheat Grain

Bal R. Singh<sup>1</sup> \*, Yadu N. Timsina<sup>1</sup> , Ole C. Lind<sup>2</sup> , Simone Cagno2,3 and Koen Janssens<sup>3</sup>

<sup>1</sup> Faculty of Environmental Sciences and Natural Resource Management, Norwegian University of Life Sciences, Ås, Norway, <sup>2</sup> Centre of Environmental Radioactivity, Faculty of Environmental Sciences and Natural Resource Management, Norwegian University of Life Sciences, Ås, Norway, <sup>3</sup> Department of Chemistry, University of Antwerp, Antwerp, Belgium

Nearly half of the world cereal production comes from soils low or marginal in plant available zinc, leading to unsustainable and poor quality grain production. Therefore, the effects of nitrogen (N) rate and application time on zinc (Zn) and iron (Fe) concentration in wheat grain were investigated. Wheat (Triticum aestivum var. Krabat) was grown in a growth chamber with 8 and 16 h of day and night periods, respectively. The N rates were 29, 43, and 57 mg N kg−<sup>1</sup> soil, equivalent to 80, 120, and 160 kg N ha−<sup>1</sup> . Zinc and Fe were applied at 10 mg kg−<sup>1</sup> growth media. In one of the N treatments, additional Zn and Fe through foliar spray (6 mg of Zn or Fe in 10 ml water /pot) was applied. Micro-analytical localization of Zn and Fe within grain was performed using scanning macro-X-ray fluorescence (MA-XRF) and laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS). The following data were obtained: grain and straw yield pot−<sup>1</sup> , 1000 grains weight, number of grains pot−<sup>1</sup> , whole grain protein content, concentration of Zn and Fe in the grains. Grain yield increased from 80 to 120 kg N ha−<sup>1</sup> rates only and decreased at 160 kg N ha−<sup>1</sup> g. Relatively higher protein content and Zn and Fe concentration in the grain were recorded with the split N application of 160 kg N ha−<sup>1</sup> . Soil and foliar supply of Zn and Fe (Zn + Fes+<sup>f</sup> ), with a single application of 120 kg N ha−<sup>1</sup> N at sowing, increased the concentration of Zn by 46% and of Fe by 35%, as compared to their growth media application only. Line scans of freshly cut areas of sliced grains showed co-localization of Zn and Fe within germ, crease and aleurone. We thus conclude that split application of N at 160 kg ha−<sup>1</sup> at sowing and stem elongation, in combination with soil and foliar application of Zn and Fe, can be a good agricultural practice to enhance protein content and the Zn and Fe concentration in grain.

Keywords: nitrogen application, zinc and iron uptake, zinc and iron distribution in grain, wheat, LA-ICP-MS, MA-XRF

### INTRODUCTION

Cereals are genetically low in Zn and Fe concentration, with reduced bioavailability (Graham et al., 2001; Cakmak, 2002). About half of the world cereal production come from soils low in plant available Zn (Cakmak, 2002), leading to poor quality of cereal grain with respect to Zn content. The situation is similar, concerning Fe deficiency in cereals. About one third of the developing countries' population and 10% of Americans and Canadians experience Zn deficiency or are at risk

### Edited by:

Raul Antonio Sperotto, University of Taquari Valley, Brazil

### Reviewed by:

Ümit Barı ¸s Kutman, Gebze Technical University, Turkey Lourdes Hernandez-Apaolaza, Universidad Autonoma de Madrid, Spain

> \*Correspondence: Bal R. Singh balram.singh@nmbu.no

### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 01 November 2017 Accepted: 22 February 2018 Published: 09 March 2018

#### Citation:

Singh BR, Timsina YN, Lind OC, Cagno S and Janssens K (2018) Zinc and Iron Concentration as Affected by Nitrogen Fertilization and Their Localization in Wheat Grain. Front. Plant Sci. 9:307. doi: 10.3389/fpls.2018.00307

**242**

of it (Hotz and Brown, 2004), erasing the geographical and political boundaries. Every year Fe and Zn deficiency causes deaths of about 800000 children and 2.4% of global disabilityadjusted life years worldwide (DALYs), while the corresponding value of DALYs for Zn is 1.9% (Mutangadura, 2004). DALYs are calculated as the sum of years of life lost (YLLs) and the years lived with disability (YLDs) based on 291 causes and 20 age groups of both sexes. The consumption of white flour made predominantly from endosperm of wheat grain discarding bran in the milling process has even worsened the degree of Fe and Zn malnutrition. This is because Fe and Zn accumulate in higher concentrations in the embryo and aleurone layer than in endosperm of a wheat grain (Šramková et al., 2009; Cakmak et al., 2010). Hence, the consumption of whole grain wheat rather than white wheat flour has been advocated to increase the daily Fe and Zn intake.

Nitrogen fertilization is known not only to increase wheat grain yield but also to facilitate the uptake of Fe and Zn in wheat grain (Cakmak et al., 2010; Shi et al., 2010). The uptake and transport of Fe and Zn to grain is probably facilitated by metal chelating compounds (Kutman et al., 2010), such as 2-deoxymugineic acid (DMA) mainly for the translocation of Fe and Zn from flag leaves to grain in wheat (Barunawati et al., 2013). Kutman et al. (2011) reported that N nutrition is critical in both the uptake and translocation of Zn and Fe to wheat grain and they showed that at high N rate, nearly 80% and 60% of total shoot Zn and Fe, respectively, were harvested with grain. Improving N status of plants from low to sufficient resulted in threefold increase in shoot Fe content of wheat plants (Aciksoz et al., 2011) Similarly, Erenoglu et al. (2011) demonstrated that N is a critical player in the uptake and accumulation of Zn in plants and thus deserves special attention in biofortification of food crops with Zn. Depending on N supply, Zn remobilization from pre-anthesis sources provided almost all grain Zn, when the Zn supply was withheld at anthesis (Kutman et al., 2012). Cakmak et al. (2010) found co-localization of protein, Fe and Zn in embryo and aleurone layer of wheat grain, indicating that the protein rich grains accumulate higher amount of Zn and Fe in wheat grain. Increasing Zn and N supply had a major impact on Zn accumulation in the endosperm, reaching concentrations higher than the current breeding targets (Persson et al., 2016).

Cakmak et al. (2010) suggested the positive role of soil and foliar applied Zn and Fe in increasing respective metal concentrations in durum wheat grain and also claimed that increased activity of Zn and Fe in the source (flag leaf and stem) during grain filling could be increased by additional Zn and Fe application through soil or by foliar application. Habib (2012) showed that joint Zn-Fe application could increase in grain concentration more than with their separate application. However, the concentrations of Zn and Fe depend on the size of wheat grains (Velu et al., 2011) and number of grains per spike (Nowack et al., 2008). Haslett et al. (2001) and Timsina (2014) demonstrated the role of phloem transport of Zn in wheat plants by performing stem girdling, and they showed that <sup>65</sup>Zn supplied on upper leaf was transported to lower leaves and root tip.

The concentration of minerals vary within a grain, depending on its portions. For example, wheat endosperm contains about 15 mg kg−<sup>1</sup> Zn, while germ and aleurone holds about 150 mg kg−<sup>1</sup> Zn (Šramková et al., 2009). By using laser ablationinductively coupled plasma-mass spectrometry (LA-ICP-MS), Wang et al. (2011) depicted higher concentration of Zn in the aleurone layer and crease vascular tissue with decreasing gradient of Zn from crease vascular tissue to endosperm, suggesting that translocation of Zn toward the endosperm occurred through the crease vascular tissue. Moreover, protein rich grain accumulated higher amount of Zn and Fe in wheat than low protein grain (i.e., Fe = 71 mg kg−<sup>1</sup> and Zn = 57 mg kg−<sup>1</sup> vs. Fe = 36 mg kg−<sup>1</sup> and Zn = 30 mg kg−<sup>1</sup> ) (Ozturk et al., 2009). This showed that higher protein or nitrogen content favors the accumulation of Zn and Fe in wheat grain (Peleg et al., 2008; Ozturk et al., 2009; Kutman et al., 2010).

In spite of the available literature on the role of N on Zn and Fe uptake by plants, information on the optimum rate and time of N application, and its effect on Zn and Fe uptake under varying levels of micronutrients in the soil are scanty. Similarly, the localization of these micronutrients in grains is not fully understood. We hypothesized that (i) N fertilization increases protein yield components of wheat and the concentration of Fe and Zn in grain, (ii) foliar Zn and Fe spraying increases their concentration in wheat grain, and (iii) micro-analytical techniques can provide information on the location of Zn and Fe in wheat grain. To test these hypotheses, we investigated the interactive effect of N, Zn, and Fe on grain yield, protein content and nutrient concentration in a pot experiment conducted in an environmentally controlled growth chamber. In addition, we investigated the distribution of Fe and Zn in selected wheat grains by using scanning macro-X-ray fluorescence (MA-XRF) and laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS).

### MATERIALS AND METHODS

### Growth Media

Artificially prepared growth media, according to the OECD 207 guideline (OECD, 1984), were used for growing plants. They consisted of 80% sand (<2 mm), 10% peat (<4 mm), and 10% kaolin on a dry weight basis. In the absence of sphagnum peat, unfertilized natural peat, produced by Econova Garden AB, Sweden, was used. Air dried peat was sieved through 4 mm wire mesh and average moisture content was determined by drying nine representative samples in an oven at 105◦C for 24 h for the correction of moisture content in peat. Average moisture content varied from 41 to 48% depending on peat delivery bags. To maintain the pH of growth media at 6.5 ± 0.2, CaCO<sup>3</sup> was mixed at rates from 0.5 to 5 g per 100 g growth media and the amount of lime required was calculated from the liming curve obtained. Hand mixed growth media and lime were prepared. The homogeneous growth media mixture was filled in 3-l plastic pots containing 2015 g mixture (dry weight) in each pot.

### Experimental Design and Fertilizer Rates

The experiment was set up as a complete randomized factorial design (**Table 1**). and It consisted of two major treatment factors: N treatments and Zn-Fe treatments. The experiment was further



Where, N-treatments: Single soil application of N at sowing: N1N80, single application of N at sowing equivalent to 80 kg N ha−<sup>1</sup> mixed wit growth media. N1N120, single application of N at sowing equivalent to 120 kg N ha−<sup>1</sup> mixed with growth media. N1N160, single application of N equivalent to 160 kg N ha−<sup>1</sup> mixed with growth media. Split application of N: N2N120, Split application of N equivalent to 120 kg N ha−<sup>1</sup> . 70% (equivalent to 84 kg N ha−<sup>1</sup> ) of allocated N was applied at sowing time and 30% (equivalent to 36 kg N ha−<sup>1</sup> ) at the beginning of stem elongation. N2N160, Split application of N equivalent to 160 kg N ha−<sup>1</sup> . 70% (equivalent to 112 kg N ha−<sup>1</sup> ) of allocated N was applied at sowing time and 30% (equivalent to 48 kg N ha−<sup>1</sup> ) at the beginning of stem elongation. Zn–Fe treatments: Soil application of Fe and Zn at sowing: Zn, Zn mixed with soil at sowing. Zn + Fe, Zn and Fe mixed with growth media at sowing. Fe, Fe mixed with soil at sowing. Soil plus foliar application of Fe and Zn at booting stage: Zn<sup>s</sup> <sup>+</sup> <sup>f</sup> , application of Zn at sowing plus 30% of growth media applied Zn as foliar spray. (Zn + Fe)<sup>s</sup> <sup>+</sup> <sup>f</sup> , application of Zn and Fe at sowing plus 30% of growth media applied zinc and iron as foliar spray. Fe<sup>s</sup> <sup>+</sup> <sup>f</sup> , application of Fe at sowing plus 30% of growth media applied iron as foliar spray.

divided into two groups: (i) experiments with growth media application of all treatment factors (5 N treatments × 3 Zn–Fe treatments = 15 growth media treatments) and (ii) experiments with growth media l plus foliar spray of Zn and Fe (2 N treatments × 3 Zn–Fe treatments = 6 growth media plus foliar treatments) (**Table 1**). Both treatment factors were incorporated into the same experiment to see the combined effect of them. Among the five N-treatments, three were single N application to growth media before sowing at the rates equivalent to 80, 120, and 160 kg N ha−<sup>1</sup> , and two were split N applications at the rates equivalent to 120 kg N ha−<sup>1</sup> and 160 kg N ha−<sup>1</sup> . In split application treatments (N120 and N160 kg N ha−<sup>1</sup> ), 70% N was applied at sowing and 30% at the stem elongation phase. Similarly, the three Zn–Fe treatments included Zn, Fe, and Zn + Fe. The rate of Zn and Fe application at sowing was 10 mg kg−<sup>1</sup> of growth media. Foliar spray of Zn and Fe (equal to 6 mg pot−<sup>1</sup> ) was made at booting stage in the two N treatments (single application and split application of 120 kg N ha−<sup>1</sup> . This was done to assess the effect of foliar spray of Zn and Fe on wheat grain yield and Zn and Fe concentration. Since the artificially made growth media used supplied all nutrients required for plant growth, the need of having a control pot was not felt.

All basic nutrients and the treatment factors (N, Zn, and Fe) were applied in deionized water solution. Powdered calcium carbonate was mixed with growth media to maintain soil pH at 6.5 ± 0.2. The treatment combinations and rates and sources of N fertilizers and micronutrients are presented in **Tables 1**, **2**, respectively. The solution volume of all nutrients was fixed to 25 ml, which was later taken into account while watering the growth media after sowing. All added nutrients and lime were mixed manually to the growth media to get a homogeneous distribution. The second dose of nitrogen in split nitrogen treatments (N2N120 and N2N160), amounting to 30% of the total N, was added at the beginning of stem elongation and watered immediately, so that N could spread properly. A handheld sprayer was used. The sprayed solution of 10 ml water per pot contained 6.0 mg of Zn as zinc sulfate and 6.0 mg of Fe Fe-EDTA mixed with DP-Klebemiddel surfactant with a concentration of 0.5 ml per one liter solution. Spraying was done after complete emergence of flag leaf at booting stage and 10 mL solution was sprayed several times to ensure that the whole solution was effectively sprayed on plant leaves.

### Plant Growth and Harvesting

Wheat plants were grown in a control growth chamber at about 21◦C. The duration of day and night length was 8 and 16 h, respectively. The source of light was halogen metal halide lamps by POWERSTAR HQI-BT 400W/D. The test plant was a hard red winter wheat variety "Krabat" used by farmers since 2011 which is claimed to be medium early in growth period, high yielding with good agronomical characteristics, medium protein, relatively good disease resistance and baking quality. Twenty seeds were sowed in each pot, which after 1 week were thinned to eight plants. While watering for the first time, the amounts of water contributed by peat and other liquid nutrients were taken into consideration to maintain moisture at 60% of field capacity throughout the growth period. It was achieved by weighing the pots with soil mixture and plants regularly and adding water to compensate the weight loss.

Plant were harvested at maturity by cutting each spike separately, and these were kept in bags for each pot. After removal of spikes, straw was cut just at the base of first node. Grain and straw yields were recorded after oven drying at 75◦C for 48 h. Wheat grain and straw were ground in a ball mill (Retsch MM301), with ball and container walls of zirconium to avoid sample contamination. However, for the sake of brevity, only grain yield, protein, Zn, and Fe concentration in grain are reported in this paper.

### Chemical Analysis

Total N in grain nitrogen was analyzed by dry combustion as described by Bremner and Mulvaney (1982). The whole grain protein (WGP) was obtained by multiplying the total N by a factor of 5.70 (ISO, 2009).


All solutions were added to growth media before sowing at volume 25 ml. Rows with bold letters represent for treatment factors.

About 0.2 g of ground wheat flour was digested in 5 ml conc. HNO<sup>3</sup> for about 2 h in ultra clave microwave reactor (MLS-MILESTONE, ultra-CLAVE III) at 250◦C and at 160 bar pressure. The digested samples were diluted to 50 ml by adding double de-ionized water (B-pure, Barnstead). Three Standard Reference Materials (SRM) (SRM1567a wheat flour) and 5 method blanks (5 ml HNO<sup>3</sup> solution) were also digested along with grain samples. Concentrations of Fe and Zn were analyzed by an inductively coupled plasma optical emission spectrometer (ICP-OES, Perkin-Elmer Optima 5300 DV) in wheat samples, SRM, and method blanks.

The lower detection limits (LOD's) and lower quantification limits (LOQ's) were determined for the concentration of Fe and Zn in the method. Measured concentrations of Fe and Zn in all samples were higher than LOD's (Average of blanks plus 3 times the SD) and LOQs (Average of blanks plus 10 times the SD). The accuracy of analytical method was determined by the analysis of three replicates of standard reference materials (SRMs 1567a wheat flour). The measured concentrations of Fe and Zn in SRMs were in accordance with the certified concentration limits and the RSD was <5%.

### Localization of Fe and Zn in Wheat Grain

Six selected wheat grains from different treatments containing relatively high Zn (42–99 mg/kg) and Fe (44–115 mg/kg) concentration were used. These grain samples were collected from different pots showing higher Zn and Fe concentration and mostly from foliar experiment and thus here the effect of N on Zn and Fe location was not clearly investigated. At first, the Environmental scanning electron microscope with dispersive X-ray spectrometry (ESEM-EDS) technique, available at the Norwegian University of Life Sciences (NMBU), Norway, was used for observation of grain morphology and element distribution. However, the limit of detection of ESEM-EDS (0.1% w/w) was not sufficiently low to detect and quantify Zn and Fe in the expected concentrations. For trace element 2D distribution analysis, Scanning macro- X-ray fluorescence (MA-XRF) and laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS) were performed at the University of Antwerp, Belgium.

Each seed was sliced in two parts. Slicing of grains was done with a razor blade that was cleaned with ethanol prior to use. The half grains were measured with no additional sample preparation. In the case of MA-XRF, they were mounted by means of adhesive tape on a diapositive frame positioned vertically on sample holder. The measurement, after verifying the sample-sourcedetector distance, was performed in an entirely non-invasive way.

In the case of LA-ICP-MS, they were introduced mounted horizontally on plasticine on the floor of the sample chamber, which was closed and purged. Detailed photographs of the surface were acquired by means of the instrument software, and the line profiles were drawn and ablated on the basis of those.

MA-XRF was performed using a non-commercial selfassembled Scanning macro- X-ray fluorescence (XRF), with the setup named instrument C (Alfeld et al., 2011). The elemental maps were recorded with a step size of 25-µm and dwell time of 400 ms per point with tube settings of 50 kV and 1.0 mA (35 W). The beam size at the focal point was approximately 50 µm. XRF maps were obtained for each element detectable in the grains. The most relevant elements were Fe, Zn, and K, whose distribution is shown in **Figure 6**. For a more detailed analysis of these elements and their co-occurrence in the grain, the following technique was used.

LA-ICP-MS was performed with a New Wave NWR193 ArF excimer laser and a Varian 7700 quadrupole ICP-MS. The ablation of sample was performed in helium gas (He) and transported to the plasma in argon gas (Ar). The flow rate was set to 0.4 l/min for carrier gas and 0.7 l/min for makeup gas. The forward power was set to 1350 watt. The line scan was executed at the speed of 10 µm/s with dimension more than 3 mm length and 100 µm beam diameter. The repetition rate of scan was 20 Hz at 90% energy capacity. The fluence was maintained at approximately 8 J/cm<sup>2</sup> . The laser warm up at the beginning of scan lasted until 20 s and washout begin after about 290 s and lasting until 350 s. This generated gross

element counts, which were further refined in the following way. Background counts were collected along the scan line before and after the wheat grain (which was located in the middle of a scan line). Net counts were determined by subtracting from the gross counts the average background counts (after removal of outliers) for each element (K, Fe, Zn). Finally, the normalized counts of Zn and Fe were determined by dividing their respective net counts with net counts for K. Potassium, (39K) was used as the normalizing element since it occurs in a more evenly distributed way throughout the grain, and particularly in the crease, as made visible by MA-XRF in **Figure 6D**. The use of K-normalized counts for Fe and Zn, helps evidencing any real increase of either element in spots/areas of the grain, independently from total ion count and surface/positioning effects.

### Statistical Analysis

The analysis of variance (ANOVA) was performed by twoway ANOVA and the relations between variables were analyzed by regression model using Minitab 16. During the regression analysis, data for independent variables were centered on their average value when necessary. For centering of data, each observed value was subtracted from the average of the respective variable. The comparison between all treatments, considering the interaction of treatment factors and main effects of N- and Zn– Fe- treatments, was carried out by Tukey comparison. In all cases, data were analyzed considering 5% level of significance (p = 0.05).

### RESULTS

### Grain Yield

The grain yield pot−<sup>1</sup> increased while increasing N rate from 80 to 120 kg N ha−<sup>1</sup> , but decreased when N was increased to 160 kg N ha−<sup>1</sup> . A similar trend was also observed for split application of N from 120 to 160 kg N ha−<sup>1</sup> . The single application of 120 kg N ha−<sup>1</sup> at sowing (N1N120) and split application of 160 kg N ha−<sup>1</sup> at sowing and stem elongation (N2N160) resulted in the highest yield (**Figure 1**). Likewise, the growth media application of Zn produced higher yield than growth media applied Fe and Zn + Fe, particularly at the application rate of 160 kg N ha−<sup>1</sup> at sowing (**Figure 1**).

The growth media plus foliar application (Zns+<sup>f</sup> or Fes+<sup>f</sup> ) for Zn and Fe, single or together, increased the grain yield in comparison to their growth media application at sowing. The single application (N1N120) of 120 kg N ha−<sup>1</sup> resulted in significantly higher yield (3.02 ± 0.04 g pot−<sup>1</sup> ) than the split application (N2N120) of 120 kg N ha−<sup>1</sup> (2.87 ± 0.04 g pot−<sup>1</sup> ) (**Table 3**) in all combinations of Zn and Fe, except for growth meida plus foliar spray of Zn. The Zn-Fe treatment (Zn + Fe) did not show a significant difference in mean grain yield for experiments with growth media l plus foliar application (Zn + Fe<sup>s</sup> + <sup>f</sup> ) of Zn and Fe (p = 0.107).

### Whole Grain Protein (WGP)

The growth media and foliar application (Zns+<sup>f</sup> or Fes+<sup>f</sup> ) resulted in significantly higher wheat grain protein (WGP) as compared to growth media application (Zn or Fe) (p = 0.021) at single as well as split N equivalent to 120 kg N ha−<sup>1</sup> (**Table 4**). Irrespective of N treatments, average increase of WGP for growth media l plus foliar application of Zn + Fe<sup>s</sup> <sup>+</sup> <sup>f</sup> , Zn<sup>s</sup> <sup>+</sup> <sup>f</sup> and Fs + fe was 8.4, 6.5, and 7% as compared to their growth media applied rate (Zn + Fe, Zn, Fe). Among all treatments, the growth media plus foliar application of Zn + Fe at split application of 120 kg N ha−<sup>1</sup> (Zn + Fe<sup>s</sup> <sup>+</sup> <sup>f</sup> at N2N120) produced the highest protein rich grains (8.96 ± 0.133%), suggesting that split N application is a better method to achieve higher protein in wheat grain. Zinc in combination with single growth media application of 120 kg N ha−<sup>1</sup> (Zn at N1N120) gave the lowest protein content (7.95 ± 0.06%) (**Table 4**).

### Iron Concentration in Wheat Grain

Iron concentration in grain was significantly affected by growth media supplied N (p < 0.001). The increasing rate of N from N1N80 to N1N160) resulted in significantly higher Fe concentration in grain when Fe was applied alone (**Figure 2**).

The growth media and foliar application (Fe<sup>s</sup> <sup>+</sup> <sup>f</sup> and Zn + Fe<sup>s</sup> <sup>+</sup> <sup>f</sup> ) of Fe increased the Fe− concentration in wheat grain significantly (p < 0.001) as compared to their growth media applied rate (Fe and Zn + Fe). Both growth media (Zn + Fe) or growth media plus foliar application (Zn + Fe<sup>s</sup> <sup>+</sup> <sup>f</sup> ) of Zn + Fe showed higher Fe- concentration in wheat grain, but a significantly higher Fe-concentration was achieved only with Zn + Fe<sup>s</sup> <sup>+</sup> <sup>f</sup> at N1N120 (**Figure 3**).

### Zinc Concentration in Wheat Grain

Grain Zn-concentration responded significantly to the main effect of N-treatment (p < 0.001). When N rate at sowing increases from 80 to 120 kg N ha−<sup>1</sup> (N1N80 to N1N120), the concentration of Zn in grain tends to decrease (**Figure 4**). For instance, single application of 120 kg N ha−<sup>1</sup> (N1N120) and split application of 160 kg N ha−<sup>1</sup> (N2N120) showed about 10% less Zn as compared to single application of 80 and 160 kg N ha−<sup>1</sup> (N1N80 or N1N160). All other combinations of treatments showed generally the same Zn concentration.

In the experiments with growth media plus foliar spray of Zn and Fe (Zn<sup>s</sup> <sup>+</sup> <sup>f</sup> and Zn + Fe<sup>s</sup> <sup>+</sup> <sup>f</sup> ), Zn- concentration in wheat grain increased significantly (p < 0.01), but was not affected by N- treatments and the interaction between N- and Zn–Fetreatments. The increase in the grain Zn-concentration in foliar sprayed treatments (Zn<sup>s</sup> <sup>+</sup> <sup>f</sup> and Zn + Fe<sup>s</sup> <sup>+</sup> <sup>f</sup> ) was higher than with Zn or Zn + Fe treatments at N1N120 or N2N120, but a significantly higher Zn- concentration was achieved only with Zn + Fe<sup>s</sup> <sup>+</sup> <sup>f</sup> at N1N120 (**Figure 5**).

### Relationship Between Fe- and Zn-Concentration and Grain Yield Parameters

A multiple linear regression including all measurements for 84 growth pots provided a valid relationship (p < 0.001) between Feconcentration in grain and the total grain weight (TGW), number

FIGURE 1 | Grain yield at N when Zn, Zn + Fe, and Fe applied to growth media at sowing. N1 and N2 for single and split N. 80, 120, and 160 stand for rate of N in kg ha−<sup>1</sup> . Bars with same alphabet at head are not significantly different at 5% level of significance.



∗∗Significant at p = 0.01; #Not significant at p = 0.05; Avg., Average; Treatment means ± 1 SE (n = 4) followed by same upper case alphabet are not significantly different for N- × Zn–Fe-treatments. Average means followed by same lower case alphabet are not significantly different for respective N- or Zn–Fe- treatments. Tukey comparison was performed at 5% level of significance. N1 and N2 stand for single and split application of N. N120 stands for growth media application of 120 kg N ha−<sup>1</sup> . Zn, Zn + Fe and Fe without suffix for growth media application of Zn–Fe- treatments and with suffix's + f' for growth media plus foliar application.

TABLE 4 | Mean ± 1 SE (n = 4) whole grain protein (%) in wheat grains at experiment with growth media plus foliar application of Zn and Fe.


∗∗∗Significant at p = 0.001; <sup>∗</sup> significant at p = 0.05; Avg., Average. The significance of letters after means and treatments are explained in Table 3.

of grains pot−<sup>1</sup> and grain Fe-uptake (Equation 1) rather than with a single variable. The coefficient of TGW and number of grains pot−<sup>1</sup> were negative, but positive for grain Fe-uptake. It indicated that grain Fe-concentration had a tendency to increase when total Fe-uptake in the grain increased but tend to lower with increase in grain yield parameters: TGW and number of grains pot−<sup>1</sup> .

Grain Fe − concentration (mg kg−<sup>1</sup> ) = 69.89 − 0.96 TGW (g) − 0.321 No. of grains pot−<sup>1</sup> + 261.46 Grain Fe − uptake grain (mg pot−<sup>1</sup> ) (1)

p < 0.001 (for regression model, No. of grains pot−<sup>1</sup> , Grain Fe- uptake); p > 0.05 for TGW;

Similarly, a regression analysis of Zn concentration in grain with TGW, number of grains pot−<sup>1</sup> and grain Feuptake in together showed a significant relation (p < 0.01) (Equation 2). The regression model defined about 72% of the variability in the grain Zn- concentration indicating the role of other variables in its determination. Positive coefficients of TGW and total Zn- uptake in grain indicated that grain Znconcentration tend to increase with these factors and negative coefficient for number of grains pot−<sup>1</sup> hint for decrease in

FIGURE 2 | Mean ± 1 SE (n = 4) bar plot for the responses of grain Fe- concentration at N-treatments when Zn + Fe and Fe applied to growth media at sowing. N1 and N2 for single and split N. 80, 120, and 160 stand for rate of N in kg ha−<sup>1</sup> . Bars with same alphabet at head are not significantly different at 5% level of significance.

alphabet at head are not significantly different at 5% level of significance.

grain Zn- concentration when number of grains pot−<sup>1</sup> tended to increase.

Grain Zn − concentration (mg kg−<sup>1</sup> ) = 45.625 − 0.646 TGW (g) − 0.211 No. of grains pot−<sup>1</sup> + 120.98 Grain Zn − uptake (content) in grain (mg pot−<sup>1</sup> ) (2)

p < 0.001 (for regression model, Grain Zn- uptake; No. of grains pot−<sup>1</sup> ); p > 0.05 (for TGW).

### Localization of Zn and Fe in Wheat Grain

Element distribution map of half wheat grains generated by the MA-XRF are shown in **Figure 6**. Relatively bright spots in the maps represent higher X- ray signal from the respective elements. This signal is influenced by the element concentration in the sample, among other factors. However, the analyzed surfaces of grains were slightly irregular and this gives rise to somewhat hazy depiction of the element distribution.

In the MA-XRF maps, it is obvious that Fe and Zn concentrations are rather variable inside the wheat grain.

FIGURE 4 | Mean ± 1 SE (n = 4) bar plot for the responses of grain Zn- concentration at N-treatments when Zn and Zn + Fe applied to growth media at sowing. N1 and N2 for single and split N. 80, 120, and 160 stand for rate of N in kg ha−<sup>1</sup> . Bars with same alphabet at head are not significantly different at 5% level of significance.

However, Fe and Zn seem to have similar patterns throughout all grains. Iron concentrated mainly in embryo and to some extent along the aleurone layer in the crease area (**Figure 6B**). Zn appears to be present in the embryo and along crease just outside the endosperm (**Figure 6C**).

LA-ICP-MS was performed on three out of the six samples that were analyzed by MA-XRF. LA-ICP-MS was used to obtain normalized signals for Fe and Zn along a well-defined line profile transversally across the grains (**Figure 7**). Generally, the profiles for both elements appear to vary in a similar fashion, confirming their co-localization. The normalized Zn and Fe counts clearly peaked at well-defined points. Higher signals of both elements were found at both ends of the grain and, for Zn only, at the embryo-aleurone interface in the middle of the grain. The

(A) Position of wheat grains subjected to MA-XRF. (B–D) Element distribution maps (1400 × 400 pixels) of wheat grains. (B) Iron, (C) Zinc and (D) Potassium.

lowest signals for both elements were encountered within the endosperm. Higher normalized signals can be seen for Fe than for Zn at both ends of the grain, generally associated with bran.

## DISCUSSION

### Grain Yield

Among N treatments, grain yield increased from the N rate of 80 kg N ha−<sup>1</sup> (N1N80) to the N rate of 120 kg N ha−<sup>1</sup> (Ntreatments: N1N120 and N2N120) but it decreased at the highest N rate of 160 kg N ha−<sup>1</sup> . This suggested positive yield response to increasing N application rate up to a definite rate only (Marino et al., 2009; Abedi et al., 2011) and decline beyond this level. Split application of 160 kg N ha−<sup>1</sup> (70% at sowing and 30% at stem elongation) produced higher grain yield than at split application of 120 kg N ha−<sup>1</sup> . Reduction in grain yield at higher N rate may be associated to dilution of Zn and Fe, thus limiting their supply. As pointed out by Kutman et al. (2011), significant reduction in grain yield was caused by the high N treatment under the discontinued Zn regime. Dilution of Zn may have affected grain yield by impairing the reproductive development (Cakmak and Engels, 1999). It is likely that the reproductive development in later spikes was negatively affected by poor supply of Zn (Kutman et al., 2011).

Grain yield was significantly higher for growth media application of Zn than growth media applied Fe or Fe + Zn. The results are in line with Silspour (2007) and Nadim et al. (2012). Nadim et al. (2012) recorded significant increase in grain yield with soil applied Zn (in the form of ZnSO4) in comparison with soil applied Fe. The effect of N on grain yield and the increase in grain yield for growth media applied Zn was also associated with the increased number of grain pot−<sup>1</sup> (data not

FIGURE 7 | LA-ICP-MS Normalized counts along scan profiles. Normalized counts represent the relative count intensity of Zn and Fe with respect to K. The signals for Zn and Fe for two scan lines are shown. The arrows indicate approximate location for elevated Zn and Fe signals in the grain fall on scan line. The white part in the picture of the wheat grain is the endosperm and bran is on either sides of endosperm. The embryo is at the top the grain. The bran at right hand side of endosperm is the crease.

shown). Kutman et al. (2012) showed that the increase in the grain yield due to improved N and Zn supply was parallel to the increase in spike number and eventually grain amount. Nadim et al. (2012) and Jiang et al. (2013) in their respective studies pointed out higher leaf area index and photosynthetic rate in connection with soil applied Zn at sowing time. Ekiz et al. (1998) noticed significant increase in wheat and other cereals grain yield when Zn (7 kg ha−<sup>1</sup> ) was applied in Zn- deficient soil.

### Whole Grain Protein

fpls-09-00307 March 8, 2018 Time: 11:2 # 10

In the experiment with soil application of nutrients, N-treatments N2N160 and N1N80, resulted in comparatively higher WGP than other N treatments. Each of these N- treatments represented different grain yield groups in this study. At N2N160, the grain yield was highest but N1N80 treatment produced lowest yield. At N2N160, the increase in protein with increasing grain yield was supported by increased available N, when 48 kg N ha−<sup>1</sup> was supplied at stem elongation since change in protein with change in yield mainly depends on the available N (Brown et al., 2005). In addition, Brown et al. (2005) illustrated that both increasing and decreasing grain protein with higher grain yield may be due to N surplus firstly and N limitation secondly. Similarly, for N1N80, higher protein was associated with lower yield, suggesting the concentration effect (Marschner, 1995). The higher protein concentration may not be the result of sufficient N but it could be due to the reduction in grain yield by limited available N and environmental limitation (Fowler, 2003) leading to less dilution.

Late N application at stem elongation in split N- treatments enhanced grain protein in comparison with single N application at sowing (Elhanis et al., 2000; Abedi et al., 2011). When initial N rate at sowing was sufficient, the split N applied at stem elongation period assured the increase in both yield and protein. For example, split application of 160 kg N ha−<sup>1</sup> (70% at sowing and 30% at stem elongation) increased both protein and grain yield. However, split application of 120 kg N ha−<sup>1</sup> increased protein concentration but not the yield in comparison to single application of 120 kg N ha−<sup>1</sup> at sowing, possibly because of limited availability of initial N needed for an increase of number of grains per spike (Li et al., 2001).

### Iron and Zinc Concentrations in Wheat Grain

In general, higher concentrations of Fe and Zn in grain were recorded for the treatments with lower grain yield and lower concentrations when higher grain yields were achieved. Studies in the past have mentioned that dilution of Zn and Fe in wheat grain occurs at increased grain yields (Liu et al., 2006; Gomez-Becerra et al., 2010). Multiple linear regression analysis (Equation 1 and Equation 2) presented a decreasing tendency of Fe- and Zn- concentrations in grain with increase in grain yield parameters: TGW and number of grains pot−<sup>1</sup> . Similarly, Zhao et al. (2009) reported that Zn- concentration of wheat grain correlated negatively with grain yield, but the correlation with grain weight was weak. A positive correlation of Zn- and Fe- uptake (i.e., total uptake in grain) suggested that higher concentrations of Zn and Fe in wheat grain were due to the increased uptake from soil or translocation of Zn and Fe from vegetative parts to the grain (Cakmak et al., 2010; Kutman et al., 2011).

A multiple linear regression insinuated a dynamic relation among grain Zn- and Fe- concentrations, their uptake in to grain and grain yield parameters (Marschner, 1995) suggesting that the process was governed by sink-source relation. The negative coefficients for grain yield components suggested that the dilution of Zn and Fe in grain was due to combined effect of grain size and the number of grains pot−<sup>1</sup> (sink size) as indicated by Sperotto et al. (2013) in rice plant, pointing involvement of factors other than grain (sink strength) only. Other factors could be the availability of metals (Marschner, 1995), for instance, Zn and Fe during grain filling (Cakmak et al., 2004, 2010; Kutman et al., 2011) or factors contributing dry weight (starch) in grain, which increases the size, and weight of grain (Marschner, 1995; Pleijel et al., 1999).

The application of 30% higher Zn and Fe, either separately (Zn<sup>s</sup> <sup>+</sup> <sup>f</sup> and Fe<sup>s</sup> <sup>+</sup> <sup>f</sup> ) or together (Zn + Fe<sup>s</sup> <sup>+</sup> <sup>f</sup> ), as foliar spray in addition to soil application caused positively significant increase in Zn- (Kutman et al., 2010) and Fe-concentrations in grain (Cakmak et al., 2010; Habib, 2012). For instance, foliar applied Fe (F<sup>s</sup> <sup>+</sup> <sup>f</sup> ) and Zn + Fe (Zn + Fe<sup>s</sup> <sup>+</sup> <sup>f</sup> ) increased the Fe- concentration in grain by 34 and 64% in comparison to growth media applied Fe and Zn + Fe, respectively. The respective increases for Zn were 17 and 46% for foliar applied Zn (Zn<sup>s</sup> <sup>+</sup> <sup>f</sup> ) and Zn + Fe (Zn + Fe<sup>s</sup> <sup>+</sup> <sup>f</sup> ). This could be explained by the increased activity of Zn and Fe in sources (flag leaf and stem) during grain filling (Cakmak et al., 2010) when additional Zn and Fe was supplied at booting. The increase was notably higher for the application of Zn + Fe together, similar to the finding of Habib (2012), where Fe and Zn concentration in wheat grain increased by applying Zn and Fe together as foliar spray.

### Localization of Zn and Fe in Wheat Grain

In this study, concentration map of Zn and Fe, obtained by MA-XRF and normalized count plots provided by LA-ICP-MS, evidenced the co-existence of both elements, especially at embryo stage and just outside the endosperm and the aleurone layer. This is in accordance with the results obtained using staining technique developed by Cakmak et al. (2010), where the colocalization of protein, Zn, and Fe in embryo was claimed to be due to the co-segregation. Similarly, Kutman et al. (2010) also showed the co- existence of Zn and protein in a durum wheat grain. Tsuji et al. (2006) used a µ-XRF technique for the elemental mapping of biological materials and found µ-XRF was useful for the analysis of element distribution in grain samples. In elemental map of black wheat and buck or soba wheat by µ-XRF technique, Zn and Fe were found to be located at either embryo and/or coat of grains (Tsuji et al., 2006). In this study, LA-ICP-MS revealed higher signals for both Zn and Fe in the embryo and bran portions, including the aleurone and crease area, with low signals

in the endosperm (**Figure 7**). These results are similar to those reported by Cakmak et al. (2010) and Wang et al. (2011), where the distribution of Zn in wheat grain and its translocation to the endosperm were shown. Based on the decreasing concentration gradient of Zn from crease area toward endosperm, Cakmak et al. (2010) suggested that Zn and Fe are translocated and distributed through the crease and then pass in to the endosperm.

To clearly define the location of Zn and Fe and their gradients, from bran to endosperm, crease area to endosperm and embryo to endosperm, for instance, the identification of the direction of element supply is essential. For this, the spatial resolution should be higher than in this study. Besides, higher sample numbers and improved sample preparation should ensure improved results, allowing for instance to avoid possible topography/surface effects on element signals.

### CONCLUSION

The rate of N application at sowing caused an increase in grain and straw yield up to the N rate of 120 kg N ha−<sup>1</sup> and a decrease at higher rate of N. The increase in grain yield was primarily determined by the increase in the number of grains pot−<sup>1</sup> or number of grains spike−<sup>1</sup> . The split application of 160 kg N/ha increased the grain and straw yield more than split application of 120 kg N/ha. The growth media application of Fe and Zn interacted with N to increase protein, Zn and Fe concentration in wheat grain. The foliar sprayed Zn and Fe at booting stage of wheat significantly increased the whole grain protein, total uptake and concentration of Fe and Zn in grain.

MA-XRF and LA-ICP-MS results indicated the co-localization of Zn and Fe in grain especially in the embryo and the aleurone. LA-ICP-MS also indicated higher concentration of Zn and Fe in the crease area and lower in the endosperm, indicating that Zn and Fe could translocate into the endosperm (the common source of flour in daily food) via crease tissue.

### REFERENCES


### AUTHOR CONTRIBUTIONS

BS: planning of experiment, supervision of student, and writing of the manuscript. YT: conduction of the whole experiment, preparation of samples, and thesis writing for his master degree. OL: assitance in the planning, sample preparation, localization studies, and reading of the manuscript. SC: running of LA-ICPMS and MA-XRF studies and reading of the manuscript. KJ: assistance in LA-ICPMS and MA-XRF studies.

### FUNDING

The research part of this master study was financed by the project "Mineral Improved Food and Feed Crops for Human and Animal Health" (Project No. 332160UA) and by a grant from the Norwegian Ministry of Foreign Affairs under the Program for Higher Education, Research and Development (HERD) in Western Balkan. The financial assistance for conducting this study is gratefully acknowledged. We also acknowledge the assistance by CERAD: this study has been funded by the Norwegian Research Council through its Centre of Excellence (CoE) funding scheme (Project No. 223268/F50). This research was supported by the Hercules Foundation (Brussels, Belgium) under grant AUHA09004 and FWO (Brussels, Belgium) Project Nos. G.0C12.13 and G.01769.09.

### ACKNOWLEDGMENTS

We wish to thank Matthias Alfeld and Kevin Hellemans for their precious support in data interpretation and measurements by means of MA-XRF and LA-ICPMS, respectively. This manuscript is based on Mr. Timsina's thesis and that this thesis is the only medium this content has appeared in and that the publication of this content is in line with the policy of the Norwegian University of Life Sciences.

Brown, B., Westcott, M., Christensen, N., Pan, B., and Starck, J. (2005). Nitrogen Management for Hard Wheat Protein Enhancement. Oregon, DC: PNW, 578.



Ethylenediaminetetraacetic acid. J. Agric. Food Chem. 56, 4643–4649. doi: 10.1021/jf800041b


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Singh, Timsina, Lind, Cagno and Janssens. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Analysis of Yellow Striped Mutants of Zea mays Reveals Novel Loci Contributing to Iron Deficiency Chlorosis

#### David Chan-Rodriguez1,2 and Elsbeth L. Walker<sup>2</sup> \*

<sup>1</sup> Plant Biology Graduate Program, University of Massachusetts Amherst, Amherst, MA, United States, <sup>2</sup> Department of Biology, University of Massachusetts Amherst, Amherst, MA, United States

The micronutrient iron (Fe) is essential for photosynthesis, respiration, and many other processes, but it is only sparingly soluble in aqueous solution, making adequate acquisition by plants a serious challenge. Fe is a limiting factor for plant growth on approximately 30% of the world's arable lands. Moreover, Fe deficiency in humans is a global health issue, affecting 1.62 billion people, or about 25% of the world's population. It is imperative that we gain a better understanding of the mechanisms that plants use to regulate iron homeostasis, since these will be important targets for future biofortification and crop improvement strategies. Grasses and non-grasses have evolved independent mechanisms for primary iron uptake from the soil. The grasses, which include most of the world's staple grains, have evolved a distinct 'chelation' mechanism to acquire iron from the soil. Strong iron chelators called phytosiderophores (PSs) are synthesized by grasses and secreted into the rhizosphere where they bind and solubilize Fe(III). The Fe(III)-PS complex is then taken up into root cells via transporters specific for the Fe(III)- PS complex. In this study, 31 novel, uncharacterized striped maize mutants available through the Maize Genetics Cooperation Stock Center (MGCSC) were analyzed to determine whether their mutant phenotypes are caused by decreased iron. Many of these proved to be either pale yellow or white striped mutants. Complementation tests were performed by crossing the MGCSC mutants to ys1 and ys3 reference mutants. This allowed assignment of 10 ys1 alleles and 4 ys3 alleles among the novel mutants. In addition, four ys<sup>∗</sup> mutant lines were identified that are not allelic to either ys1 or ys3. Three of these were characterized as being non-allelic to each other and as having low iron in leaves. These represent new genes involved in iron acquisition by maize, and future cloning of these genes may reveal novel aspects of the grass iron acquisition mechanism.

Keywords: iron, phytosiderophores, yellow stripe, maize, mutants

### INTRODUCTION

The global demand for crops with high concentrations of nutrients in edible tissues is increasing due to current trends in population growth, global climate change, and decreasing arable land resources (Eckardt et al., 2009). Iron (Fe) deficiency in humans is a global health issue, affecting 1.62 billion people, or about 25% of the world's population, and it is imperative that we gain a better

### Edited by:

Felipe Klein Ricachenevsky, Universidade Federal de Santa Maria, Brazil

### Reviewed by:

Yoshiko Murata, Suntory Foundation for Life Sciences, Japan Sebastien Thomine, Centre National de la Recherche Scientifique (CNRS), France

> \*Correspondence: Elsbeth L. Walker ewalker@bio.umass.edu

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 01 December 2017 Accepted: 29 January 2018 Published: 20 February 2018

#### Citation:

Chan-Rodriguez D and Walker EL (2018) Analysis of Yellow Striped Mutants of Zea mays Reveals Novel Loci Contributing to Iron Deficiency Chlorosis. Front. Plant Sci. 9:157. doi: 10.3389/fpls.2018.00157

understanding of the mechanisms that plants use to regulate iron homeostasis, since these will be important targets for future biofortification strategies (McLean et al., 2009; Murgia et al., 2012). Quantitative trait loci (QTL) have been identified for maize grain iron accumulation (Zhang et al., 2017) and identification of additional components of the maize iron homeostatic apparatus may help to elucidate the genes underlying such QTL. Although we have learned a great deal through the study of model organisms such as Arabidopsis, it is important to note that the grasses, which include most of the world's staple grains, use phytosiderophores (PSs) that are secreted into the rhizosphere where they bind and solubilize Fe(III) (Tagaki, 1976; Tagaki et al., 1984, 1988). PSs are not made or used by non-grass species.

Biofortification of crops has been restricted by our limited knowledge of the molecular mechanisms controlling iron uptake, translocation, accumulation, and deposition in the grain. Attempts to increase iron content have been promising, but these efforts have been focused on the relatively small set of known genes that are involved in iron homeostasis. The iron-storage protein, ferritin (Briat and Lobreaux, 1998; Briat et al., 1999), has been expressed in rice endosperm, to increase iron and zinc content (Goto et al., 1999; Drakakaki et al., 2005). The iron uptake machinery has been a target for biofortification by engineering key enzymes involved in PS synthesis (Higuchi et al., 1999; Takahashi et al., 1999, 2001; Suzuki et al., 2008). These efforts have been only partially successful, suggesting that identifying additional genes involved in mobilization and translocation within the plant could be helpful to develop additional strategies for the production of biofortified crops.

In plants, iron is essential for photosynthesis, respiration, and many other processes, but is only sparingly soluble in aqueous solution, making adequate acquisition by plants a serious challenge (Marschner, 1995). Furthermore, iron is highly reactive and if over-accumulated can cause cellular damage. As a response to these key properties of iron, plants have evolved highly regulated iron mechanisms to ensure efficient and tightly controlled acquisition from the soil. Most plants use a combination of rhizosphere acidification, iron reduction, and uptake via the ZIP (ZRT, IRT-like protein) family transporter, IRT1 (iron-regulated transporter). In this strategy, iron is first solubilized and then taken up from the soil, as reviewed in Walker and Connolly (2008), Jeong and Guerinot (2009), and Morrissey and Guerinot (2009). In contrast, the grasses, which include most of the world's staple grains, have evolved a distinct 'chelation' mechanism to acquire iron from the soil. PSs are synthesized by grasses and secreted into the rhizosphere where they bind and solubilize Fe(III) (Tagaki, 1976; Tagaki et al., 1984, 1988). The Fe(III)-PS complex is then taken up into root cells via transporters specific for the Fe(III)-PS complex (Romheld and Marschner, 1986; von Wiren et al., 1994). This mechanism is also known as 'Strategy II.' The Fe(III)-PS uptake transporter Yellow Stripe1 (YS1) has been studied extensively (Curie et al., 2001; Yen et al., 2001; Roberts et al., 2004; Schaaf et al., 2004; Murata et al., 2006; Harada et al., 2007; Inoue et al., 2009; Lee et al., 2009), and is a proton-coupled symporter of Fe(III)-PS complexes (Schaaf et al., 2004).

Phytosiderophores are chemically quite distinct from bacterial and fungal siderophores (Miethke and Marahiel, 2007) and belong to a class of compounds called mugineic acids (Ma and Nomoto, 1996), with a well-worked out biosynthesis (Mori and Nishizawa, 1987; Kawai et al., 1988; Shojima et al., 1990; Ma et al., 1995; Takahashi et al., 1999; Kobayashi et al., 2001). In contrast to the details established for PS biosynthesis and Fe-PS uptake, the molecular details of PS secretion have not been as well-characterized. In several grass species, PSs are secreted according to a diurnal cycle, with release occurring several hours after sunrise (Zhang et al., 1991; Walter et al., 1995; Ma et al., 2003; Reichman and Parker, 2007; Ueno et al., 2007; Nagasaka et al., 2009; Bernards et al., 2014). Large numbers of vesicles have been observed in barley roots just prior to the daily release of PS, suggesting that PSs are secreted by exocytosis (Nishizawa and Mori, 1987; Sakaguchi et al., 1999; Negishi et al., 2002). Furthermore, microarray analysis of barley roots indicated that expression of genes associated with polar vesicle transport increases in the early morning (Negishi et al., 2002). The anion channel blockers anthracene-9-carboxylic acid and phenylglyoxal were shown to inhibit PS secretion by barley roots (Sakaguchi et al., 1999), potentially indicating that anion channels are involved in loading PS to secretory vesicles. Alternatively, anion channels in the plasma membrane (PM) could be responsible for PS transport across the PM. Major facilitator superfamily transporters with PS efflux activity were recently identified in rice and barley, and have been called transporter of mugineic acid (TOM1) (Nozoye et al., 2011).

A classically known mutation in maize called yellow stripe3 (ys3; Wright, 1961) renders plants unable to secrete PSs, even though PSs are synthesized in normal amounts (Basso et al., 1994; Lanfranchi et al., 2002). The Ys3 gene in maize is located between 85,618,053 and 114,789,459 on chromosome 3 based on two genetic markers (IDP3861 and IDP4688) on the IBM2 2008 Neighbors map. A partial gene with similarity to TOM1 (GRMZM2G063306 also called ZEAMMB73\_058478) is located within this interval in the maize reference sequence version 3 but the sequence contained two sequence gaps in the region occupied by GRMZM2G063306. Based on sequence similarity and strong expression during iron deficiency, this gene was suggested as a candidate for the locus affected in ys3 mutants (Nozoye et al., 2013; Li et al., 2014), but genetic evidence for this assignment has not been presented.

In spite of this progress in understanding the process of PS synthesis, release, and uptake of Fe-PS complexes, there are many gaps in our understanding of what makes a particular grass species or cultivar 'iron efficient.' In Kentucky bluegrass, for example, the amount of PS release does not correlate well with resistance to iron deficiency (Buxton et al., 2012). Because of this complexity, we sought to understand the genes in Zea mays (maize) that contribute to iron efficiency, by examining the set of maize mutants available through the Maize Genetics Cooperation Stock Center (MGCSC) that have been described as 'yellow striped' or 'green striped.' Both these descriptions may refer to iron deficiency chlorosis that is typical in both

ys1 and ys3 maize mutants, and is characterized by yellow interveinal regions and green veins. By performing allelism tests with ys1 and ys3 reference mutant plants, we have identified novel yellow striped mutants (that we designate as ys<sup>∗</sup> , pending gene identification and assignment of new nomenclature) that may shed light on additional components contributing to iron efficiency in maize. We further characterized the sequence of GRMZM2G063306 in the WT B73 genome and the ys3 reference mutants. We have identified four new alleles of ys3. Based on the evidence from our sequencing of GRMZM2G063306 in multiple independent ys3 mutants, we present strong genetic evidence that GRMZM2G063306 (ZmTOM1) corresponds to the Ys3 gene of maize.

### MATERIALS AND METHODS

### Plant Material and Growth Conditions

Maize (Zea mays) plants of B73 and W22 inbred lines were used as WT reference in our experiments, as indicated in the text. Uncharacterized yellow striped mutants were obtained from the MGCSC<sup>1</sup> .

For all experiments involving genetic crosses, stocks were grown at the University of Massachusetts Crop and Animal Research and Education Center, South Deerfield, MA, United States, during the summer season between May and September. Mutant plants were supplemented with foliar iron (Fe-EDDHA) through growing season to alleviate chlorosis. For the purposes of initial phenotyping, plants were grown in the greenhouse in a 4:1 v/v mix of potting soil and Turface. All phenotyping was also repeated under field conditions. Supplemental light was supplied with high-pressure sodium lamps to give a 20 h light period each day. For quantitative polymerase-chain-reaction (PCR) analysis, plants were grown in a sand:Turface mix (9:1 v/v) irrigated with water until germination and then irrigated with modified Hoagland's nutrient solutions, with 1 mM KH2PO4, 3.75 mM KOAc, 5 mM Ca(NO3)2, 1.25 mM KNO3, 2 mM MgSO4, 3.75 mM NH4OAc, 46 uM H3BO3, 9.1 uM MnCl2, 0.77 uM ZnSO4, 0.32 uM CuSO4, and 0.83 uM H2MoO<sup>4</sup> (Yordem et al., 2011) containing 100 µM FeSO4-EDTA every 48 h. Plants were grown for 10 days after germination before the root tissue was collected.

### PCR and Sequencing of ZmTOM1 in ys3 Mutant Lines

Genomic DNA was extracted from leaves of ys3 mutant plants (ys3:04HI-A632GN-144, ys3:67-2403, ys3:04HI-Oh43xA632GN-187, and ys3:07IL-B73GN-279) and the exons of ZmTOM1 (GRMZM2G063306/Zm00001d041111) were amplified using primers listed in Supplementary Table 1. Amplifications were performed using ExTaq polymerase (Takara, Madison, WI, United States), with cycling conditions of 95◦C, 60 s followed by 35 cycles of 95◦C, 15 s, 55◦C, 30 s, and 72◦C, 60 s, with a final extension step at 72◦C for 5 min. PCR products were gel purified before sequencing.

### Real-Time PCR Analysis (qRT-PCR)

The root tissue was flash frozen in liquid nitrogen after harvesting. The frozen root tissue was ground using a Tissuelyser (QIAGEN, Valencia, CA, United States) in 2 ml tubes containing 3.2 mm chrome steel beads (BioSpect Products, Bartlesville, OK, United States). Total RNA was extracted using QIAGENR <sup>R</sup> Neasy Plant Mini Kit (QIAGEN, Valencia, CA, United States), and on-column DNAse treatment step was included for all samples. cDNA was synthesized from 750 ng of total RNA using SuperScript IV VILO (Life Technologies, Carlsbad, CA, United States). For real-time PCR (RT-PCR) analysis, Quantprime primer design webtool (Arvidsson et al., 2008) was used to design ZmTOM1 primers. The primer efficiency of each set of primers (Supplementary Table 2) was evaluated empirically by serial dilution curve of cDNA. PowerUPTM SYBRTMGreen Master Mix (Life Technologies, Carlsbad, CA, United States) was used in quantitative RT-PCR experiments. A two-step PCR protocol was used with the following conditions: initial cycle of 50◦C, 120 s, and 95◦C, 120 s, and 40 cycles of 95◦C, 15 s, and 60◦C, 60 s. After two-step cycling was completed, melting curve was performed to ensure that single amplicon was obtained from each reaction. To determine transcript levels, the threshold cycle (Ct) values from target gene was normalized to ZmGAPDH reference gene for each sample and by the 11C<sup>t</sup> method, we calculated fold change compared to B73 WT. Data represent three biological replicates.

### Inverse PCR

Genomic DNA (∼1 µg) was digested with AciI and NlaIII (New England Biolabs) for 2.5 h at 37◦C and reaction was stopped by incubating for 20 min at 65◦C. The DNA was then diluted 25 fold, and ligation was performed using 20 units of Epicenter <sup>R</sup> T4 DNA ligase (Illumina, Inc., Madison, WI, United States) overnight at either 20◦C for blunt ends or 4◦C for sticky ends. The resulting ligation was purified using phenol/chloroform (1:1, v/v) and ethanol precipitation in the presence of 40 µg of glycogen. Then, 1/6 of the purified ligation was used as template for the 1st round of PCR, with primers oZmTOM1\_4456 and oZmTOM1\_4504 for AciI restriction digest, or primers oZmTOM1\_3641 and oZmTOM1\_5012 for NlaIII restriction digestions. The 2nd round of PCR was performed using 1 ul of a 1:100 dilution of the PCR product from the 1st round as template using nested primers oZmTOM1\_4338 and oZmTOM1\_4573for AciI digested DNA or oZmTOM1 4774 and oZmTom1\_5154, for NlaIII digested DNA. Amplifications were performed using ExTaq polymerase (Takara, Madison, WI, United States), with cycling conditions of 95◦C, 2 min, and 25 cycles of 95◦C, 15 s, 57◦C, 30 s, 72◦C for 2 min, and a final elongation step for 10 min. Primer sequences for inverse PCR (iPCR) are listed in Supplementary Table 3.

### Metal Measurement

Leaves of at least 10 individual plants were collected from 19 day-old plants grown in the greenhouse and samples were dried at 65◦C for 72 h. In every experiment, all controls and mutants were grown simultaneously and using the same soil batch. Metal

<sup>1</sup>http://maizecoop.cropsci.uiuc.edu/

concentrations were determined by inductively couple plasma mass spectrometry (ICP-MS) at the Donald Danforth Plant Research Institute.

### RESULTS

### Complementation Testing of Yellow Striped Mutants from MGCSC

We obtained 31 mutants classified as having a yellow striped phenotype from the MGCSC (**Table 1**). These were planted in the field and phenotypic analysis indicated that 21 of the lines showed the phenotype typical of iron deficiency chlorosis. In the other 10 lines, we either did not observed stripes at all or else observed a solid yellow or white striped phenotype (**Table 1**). To identify new genes involved in iron uptake in maize, and to identify new alleles for ys3, we performed complementation tests between the uncharacterized yellow striped mutants and the reference maize mutants ys1:ref and ys3:ref. Due to stunting or sterility of some mutant stocks, not all crosses were obtained. From these crosses, we identified 10 new alleles for ys1 and 4 new alleles for ys3. Moreover, we found four novel yellow stripe mutants, ys<sup>∗</sup> -PI262172, ys<sup>∗</sup> :N2398, ys<sup>∗</sup> :PI228180, and ys<sup>∗</sup> :04HI-A632XOh43GN-18, that are not allelic to ys1 or ys3, and thus may represent new maize genes involved in iron uptake or homeostasis.

TABLE 1 | Yellow or green striped mutants from the Maize Genetics Cooperation Stock Center (MGCSC) and results of complementation tests with ys3:ref and ys1:ref.


NT indicates that allelism was not tested.

sequence is shown for comparison. For ys3:04HI-Oh43XA632-GN-187, genomic sequence of the site of insertion is shown, with the 8 bp direct repeat flanking the

### Analysis of ZmTOM1 Coding Sequence in ys3 Alleles

Because a gene with similarity to TOM1 (GRMZM2G063306 also called ZEAMMB73\_058478, here designated ZmTOM1) is located within the genomic interval containing Ys3, this gene has been suggested as a candidate for the locus affected in ys3 mutants (Nozoye et al., 2013; Li et al., 2014). To determine whether the suggested candidate gene, ZmTOM1, underlies the long-known ys3 mutant, we sequenced the exons of ZmTOM1 in all five alleles of ys3 (reference allele and the four novel alleles identified through complementation tests; **Figure 1**) to identify causative mutations. The MGCSC holds three stocks designated as ys3:ref mutants (304A, 311F, and 311G). In all three lines, ZmTOM1 sequences were identical, and contained a 4 bp insertion in exon 9. This insertion causes a frame shift followed by 11 novel amino acids before introducing a premature stop codon

(**Figure 1**). We found a different 4 bp insertion in exon 11 of ZmTOM1 in the ys3:67-2403 allele (**Figure 1**). This 4 bp insertion causes a frame shift followed by 107 new amino acids before a stop codon occurs to terminate the protein prematurely. For the ys3:04HI-Oh43XA632-GN-187 allele, we could not amplify fragments containing exon 10 and 11, but partial sequences from both exons could be obtained. We hypothesized that an insertion could be present between these two exons causing failure to amplify that region. Using iPCR, we identified both left and right borders of an insertion containing the characteristic elements of a transposon. The inserted sequences were flanked by 8 bp direct repeats and contained 130 bp terminal inverted repeats (TIRs). We aligned the TIR sequences with the maize reference sequence and identified two regions in chromosome 7 annotated as Far1-related sequence 5, which corresponds to a mutator-like transposable element (MULE). MULE transposons

insertion underlined.

generate 8–10 bp target sequence duplications and have TIRs of >100 bp. Thus, the insertion has all the elements expected for a MULE transposon inserted in the ys3:04HI-Oh43XA632- GN-187 allele (**Figure 1**). A one nucleotide change at the exon– intron border for exon 5 of ys3:04HI-A632GN-144 was observed. Likewise, a one nucleotide change near the 3<sup>0</sup> end of exon 9 was observed in ys3:07IL-B73GN-279. We hypothesized that splicing could be affected these two alleles, and so investigated ZmTOM1 gene expression in the roots of these plants using Q-RT-PCR. Expression of ZmTOM1 was observed in both ys3:04HI-A632GN-144 and ys3:07IL-B73GN-279 (data not shown). Since the ZmTOM1 transcript was observed, we speculated that altered splicing due to the mutations might be leading to aberrant ZmTOM1 mRNA, so we sequenced the full-length cDNA from each mutant line to test this. We confirmed that the one nucleotide change at the ys3:04HI-A632GN-144 exon–intron junction altered the splice donor site. The mutation causes splicing to occur at a new donor site 3 nucleotides into the adjacent intron (**Figure 2**). As a result, one additional amino acid is inserted without affecting the reading frame. The amino acid is inserted in a strongly conserved region that could lead to a nonfunctional protein. In the ys3:07IL-B73GN-279 allele, the single nucleotide change occurred at the first nucleotide of intron 9, changing the splice donor site from GT to AT. In the mRNA produced by this allele, a new splice donor site is recognized in exon 9, 21 nucleotides upstream from the original donor site (**Figure 2**). The resulting amino acid sequence is thus missing seven residues in a strongly conserved region. Our results show clear genetic evidence that the Ys3 gene is ZmTOM1.

### Analysis of Novel Yellow Striped Maize Mutants

To evaluate whether the yellow striped phenotype in ys<sup>∗</sup> mutants is due to low iron, we analyzed metal levels in leaves of three of the mutants. The ys<sup>∗</sup> -PI262172 mutant was not included in this analysis, because its stunted growth prevented our obtaining sufficient material for this experiment. Visual inspection of the leaves of 12-day-old WT and mutant plants indicates differences in the severity of the observed chlorosis, with ys<sup>∗</sup> :PI228180 having very mild chlorosis and ys∗ :04HI-A632xOh43GN-18, ys1:ref, and ys3:ref having the most marked chlorosis (**Figure 3**). In all three ys<sup>∗</sup> mutants tested, the levels of iron were significantly lower than WT (**Figure 3**) indicating that the plants are iron-deficient. Control ys1 and ys3 plants are also low in iron, as expected. In a segregating population of ys<sup>∗</sup> :04-04HI-A632xOh43GN-18 mutants, the iron concentration in yellow striped siblings was less than half (42%) that of WT siblings. The iron concentration in ys<sup>∗</sup> :04HI-A632xOh43GN-18 was significantly lower even than ys1 and ys3, indicating a very substantial alteration in iron homeostasis in these plants. For ys<sup>∗</sup> :PI228180 and ys<sup>∗</sup> :N2398, iron levels were higher than either ys1 or ys3,

TABLE 2 | Complementation test results among ys<sup>∗</sup> mutants.


but were still significantly lower than the amount in WT control plants.

We also measured the Zn and Mn concentration in the leaves of the mutant plants (**Figure 3**). We note that altered iron homeostasis often causes alterations to multiple metals. For example, ys1 and ys3 mutants, which are clearly impaired in iron uptake, have higher Zn and Mn than WT control plants (**Figure 3**). It is possible that this occurs because PS secretion or uptake directly affects Mn and Zn uptake or translocation, but it is also possible that the mechanism is indirect. Like ys1 and ys3 mutants, ys∗ :04HI-A632xOh43GN-18 and ys<sup>∗</sup> :N2398 plants have higher Mn and Zn than WT control plants. For ys<sup>∗</sup> :PI228180, the Zn concentration in leaves is not significantly different from WT control plants, and the Mn concentration is slightly but significantly lower than that of the WT controls, and much lower that the Mn concentration in the ys1 and ys3 mutants. We note that the soil batch used for growth of the ys<sup>∗</sup> :04HI-A632xOh43GN-18 and ys<sup>∗</sup> :N2398 plants and their controls was different from the batch used to grow ys<sup>∗</sup> :PI228180 and its controls.

### Complementation Tests among ys<sup>∗</sup> Mutants

We performed crosses among three of the four identified ys<sup>∗</sup> mutants to determine how many loci are represented by these three mutants. The ys<sup>∗</sup> -PI262172 mutant was not included in this analysis, because its stunted growth prevented our obtaining the appropriate crosses. F1 seeds were grown in the greenhouse and the phenotypes were recorded. We found complementation among all crosses performed in ys<sup>∗</sup> mutants, indicating that they do not represent alleles. These results show that we have identified three novel genes involved in iron homeostasis (**Table 2**).

## DISCUSSION

### The Rationale for Gene Discovery in Zea mays

Much of the molecular work on iron uptake and homeostasis in grasses has been performed using rice, both because of its properties as a model organism and also because of the fundamental importance of this species as a crop. Still, the Fe(III)-PS uptake transporter, YS1, was first identified in maize by making use of the excellent genetic resources available in this species (Curie et al., 2001). The YS1 gene has been used directly as a strategy for engineering biofortification with mixed results. In an early study using constitutive expression of barley YS1 in rice, plants showed superior growth in alkaline soil conditions but did not contain significantly more iron in grains (Gomez-Galera et al., 2012). Later, barley YS1 expressed in rice was shown to promote the preferential mobilization and loading of Fe in seeds while displacing Cd and Cu (Banakar et al., 2017). At present, two key uptake genes, Ys1 and Ys3 (TOM1), for the grass specific mechanism are understood, as are the genes involved in PS synthesis, but it is unclear whether additional grass specific components exist. If they do, they will need to be discovered directly in grass species such as maize.

### Identification of the Ys3 Gene

Maize ys3 mutants lack the ability to secrete PS (Lanfranchi et al., 2002). ZmTOM1 has been proposed as candidate gene for Ys3 because of its function as PS effluxer (Nozoye et al., 2011) and its location within the same map interval as the genetically identified ys3 mutant allele. Previous reports analyzing the ys3 transcriptome during iron deficiency suggested reduced expression and alternative splicing of ZmTOM1 (Nozoye et al., 2013). However, this approach could not definitively assign ZmTOM1 as the Ys3 gene, since other genes (ZmMATE3/ZmPEZ1) had reduced expression in ys3 mutants, and mutations in ZmTOM1 were not identified. Here, we were able to show that multiple alleles of ys3 could be found among the yellow striped mutants held at the MGCSC, and that each of these carries a unique mutation that is expected to abolish the function of the ZmTOM1 protein.

### Three Novel Yellow Striped Maize Mutants

In this study, we identified three novel yellow striped mutants whose phenotype is apparently caused by low iron content. These mutants represent three different loci involved in iron

### REFERENCES


homeostasis. Genetic mapping to identify the underlying genes responsible for the yellow striped phenotype in these mutants will reveal unknown elements of the iron homeostasis machinery and may provide new options for biofortification. Initially, it appeared as though our screening of the MGCSC mutant collection had reached saturation since multiple alleles for both ys1 and ys3 were obtained. However, three novel loci contributing to iron content in leaves were identified, indicating that saturation mutagenesis has likely not been reached and additional genes causing an iron deficiency induced yellow striped phenotypes in maize could be uncovered. Genetic mapping of the three ys<sup>∗</sup> mutants is underway to discover the genes responsible for these interesting metal homeostasis phenotypes. Future work will also include tests to indicate whether additional iron supply or direct iron supply to the leaves can alleviate the ys<sup>∗</sup> phenotypes, and tests of the iron concentration in grains of the mutant plants to see whether the grain concentration of iron is altered in the mutants.

### AUTHOR CONTRIBUTIONS

EW conceived the project, was responsible for the experimental design, and also performed some of the genetics crosses and phenotyping in the field. DC-R conducted screening and genetics crosses and performed all of the molecular work on the project.

### FUNDING

This work was supported by grants to EW (USDA AFRI Grant Nos. 2009-02268 and NSF IOS-153980).

## ACKNOWLEDGMENTS

We would like to express our sincere gratitude to the lab members, past and present who helped in the cornfield, especially Rakesh K. Kumar and Harry Klein, to Mary Sachs, of the MGCSC, who answered numerous questions and provided us with details pertaining to the stocks used in these studies, and to Dan Jones and Chris Phillips for their expert assistance in the greenhouse.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018.00157/ full#supplementary-material

metals by selective Fe transport. Plant Biotechnol. J. 15, 423–432. doi: 10.1111/ pbi.12637




Wright, J. (1961). A new yellow stripe on chromosome 3. Maize Newslett. 35:111.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Chan-Rodriguez and Walker. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Contribution of NtZIP1-Like to the Regulation of Zn Homeostasis

Anna Papierniak<sup>1</sup> , Katarzyna Kozak<sup>1</sup> , Maria Kendziorek<sup>1</sup> , Anna Barabasz<sup>1</sup> , Małgorzata Palusinska ´ 1 , Jerzy Tiuryn<sup>2</sup> , Bohdan Paterczyk<sup>3</sup> , Lorraine E. Williams<sup>4</sup> and Danuta M. Antosiewicz<sup>1</sup> \*

1 Institute of Experimental Plant Biology and Biotechnology, Faculty of Biology, University of Warsaw, Warsaw, Poland, <sup>2</sup> Faculty of Mathematics, Informatics, and Mechanics, University of Warsaw, Warsaw, Poland, <sup>3</sup> Laboratory of Electron and Confocal Microscopy, Faculty of Biology, University of Warsaw, Warsaw, Poland, <sup>4</sup> Biological Sciences, University of Southampton, Southampton, United Kingdom

Tobacco has frequently been suggested as a candidate plant species for use in phytoremediation of metal contaminated soil but knowledge on the regulation of its metal-homeostasis is still in the infancy. To identify new tobacco metal transport genes that are involved in Zn homeostasis a bioinformatics study using the tobacco genome information together with expression analysis was performed. Ten new tobacco metal transport genes from the ZIP, NRAMP, MTP, and MRP/ABCC families were identified with expression levels in leaves that were modified by exposure to Zn excess. Following exposure to high Zn there was upregulation of NtZIP11-like, NtNRAMP3, three isoforms of NtMTP2, three MRP/ABCC genes (NtMRP5-like, NtMRP10-like, and NtMRP14 like) and downregulation of NtZIP1-like and NtZIP4. This suggests their involvement in several processes governing the response to Zn-related stress and in the efficiency of Zn accumulation (uptake, sequestration, and redistribution). Further detailed analysis of NtZIP1-like provided evidence that it is localized at the plasma membrane and is involved in Zn but not Fe and Cd transport. NtZIP1-like is expressed in the roots and shoots, and is regulated developmentally and in a tissue-specific manner. It is highly upregulated by Zn deficiency in the leaves and the root basal region but not in the root apical zone (region of maturation and absorption containing root hairs). Thus NtZIP1-like is unlikely to be responsible for Zn uptake by the root apical region but rather in the uptake by root cells within the already mature basal zone. It is downregulated by Zn excess suggesting it is involved in a mechanism to protect the root and leaf cells from accumulating excess Zn.

Keywords: zinc, tobacco, ZIP, NtZIP1-like, yeast complementation

### INTRODUCTION

Tobacco (Nicotiana tabacum L cv. Xanthi) has frequently been considered for phytoremediation purposes because of its high biomass and ability to take up and accumulate in leaves high amounts of metals, including zinc (Zn) (Vangronsveld et al., 2009; Herzig et al., 2014; Vera-Estrella et al., 2017). To improve its capacity to take up and store metals in shoots, it has been transformed with a number of metal homeostasis genes, but with limited success (Gisbert et al., 2003; Martínez et al., 2006; Gorinova et al., 2007; Wojas et al., 2008, 2009; Korenkov et al., 2009; Siemianowski et al., 2011; Barabasz et al., 2013; Wang et al., 2015). Recently, it was shown that when expressing metal transporters to engineer new metal-related traits, a major

#### Edited by:

Raul Antonio Sperotto, University of Taquari Valley, Brazil

### Reviewed by:

Manish Kumar Patel, National Institute of Plant Genome Research (NIPGR), India Marc Hanikenne, University of Liège, Belgium

> \*Correspondence: Danuta M. Antosiewicz dma@biol.uw.edu.pl

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 01 December 2017 Accepted: 31 January 2018 Published: 16 February 2018

#### Citation:

Papierniak A, Kozak K, Kendziorek M, Barabasz A, Palusinska M, Tiuryn J, Paterczyk B, ´ Williams LE and Antosiewicz DM (2018) Contribution of NtZIP1-Like to the Regulation of Zn Homeostasis. Front. Plant Sci. 9:185. doi: 10.3389/fpls.2018.00185

**264**

part of the resulting phenotype was due to the modulation of endogenous gene expression (Barabasz et al., 2016; Kendziorek et al., 2016). Therefore, a greater understanding of Zn-homeostasis mechanisms is required to successfully genetically modify the efficiency of Zn accumulation in shoots. Maintaining high Zn in the above ground organs depends on three major processes operating efficiently: Zn uptake from the soil, root-to-shoot translocation and storage in leaves without detrimental toxic effects.

Zn uptake is thought to be mediated primarily by ZIP (ZRT\IRT related Protein) metal transporters. In Arabidopsis thaliana AtZIP2, AtIRT1 and AtIRT3 residing in the plasma membrane have been identified as key players in Zn acquisition by roots (Vert et al., 2002; Lin et al., 2009; Palmer and Guerinot, 2009; Milner et al., 2013). The root-to-shoot translocation of Zn (and other metals) depends on two main factors: the ability to store the metal in the roots; and the efficiency of its loading into xylem vessels. It has been shown that HMAs (Heavy-Metal ATPases) which belong to the P1B-ATPase family (Williams et al., 2000; Williams and Mills, 2005) are involved in both processes. HMA3, identified in A. thaliana and rice, localized in the tonoplast of root cortical cells, limits translocation of Cd from the roots to the shoots by sequestrating the metal into the root vacuoles. There is a suggestion that it could also transport Zn into the vacuoles and control the amount of Zn available for xylem loading and thus the efficiency of its translocation to the shoot (Morel et al., 2009; Ueno et al., 2010; Miyadate et al., 2011). The efficiency of the next step in Zn translocation to shoots - loading of a metal into the xylem vessels, is under the control of two genes with overlapping function, HMA2 and HMA4 (Hussain et al., 2004; Verret et al., 2004; Wong and Cobbett, 2009). Encoded proteins are localized in the roots at the plasma membrane of xylem parenchyma cells where they are responsible for Zn (and also Cd) efflux to the xylem. Decreased translocation of Zn to shoots in the athma2athma4 double mutant led to severe Zn deficiency (Hussain et al., 2004; Wong and Cobbett, 2009; Mills et al., 2012).

Zn transported to the shoots is stored primarily in the mesophyll cells of leaves. Its level of accumulation depends on the ability of the mesophyll cells to store the metal without toxicity. This complex process involves efficient metal import and its loading into the vacuoles, but also regulated redistribution from this compartment. Currently we are far from having a clear picture of all the elements involved. The potential players include members of several transport families. The ZIP genes play a diverse roles, and those present in the plasma membrane are responsible for Zn uptake, while others localized in the tonoplast could contribute to control of Zn release from vacuoles (Guerinot, 2000; Milner et al., 2013; Ricachenevsky et al., 2015). Accumulation of metal/s in the vacuoles also depends on the NRAMP (Natural Resistance-Associated Macrophage Protein) family. Members of this family transport Fe and Mn, while Cd, Zn or Ni can also serve as substrates for some (Nevo and Nelson, 2006; Ricachenevsky et al., 2015). AtNRAMP1 is a plasma membrane Mn uptake system in roots of A. thaliana; Cailliatte et al. (2010), while NRAMP3 and NRAMP4 are involved in metal release (Mn and Fe) from vacuoles in leaves and seeds (Lanquar et al., 2005, 2010). High expression of NRAMP3 and NRAMP4 genes was noted in the leaves of Zn/Cd hyperaccumulating A. halleri (Weber et al., 2004) and Thlaspi caerulescens. Both TcNRAMP3 and TcNRAMP4 were implicated in metal hypertolerance, but the precise role is yet to be determined (Oomen et al., 2009). Loading of metals into vacuoles is provided by the members of the MTP (Metal Tolerance Proteins) family. Residing in the tonoplast, they are involved in sequestration primarily Zn in the vacuoles, but other metals such as Fe, Mn, Cd, Ni or Co can also be substrates for some family members (Gustin et al., 2011; Menguer et al., 2013; Ricachenevsky et al., 2013; Farthing et al., 2017). However, some MTPs are localized in the plasma membrane, and they remove cations from the cytoplasm to the cell wall (Menguer et al., 2013; Migocka et al., 2015). In the leaves, a key protein for Zn sequestration and detoxification is the vacuolar protein MTP1. AtMTP1 from A thaliana contributes to Zn accumulation in leaves and to basal Zn tolerance by sequestering Zn in vacuoles (Kobae et al., 2004; Desbrosses-Fonrouge et al., 2005; Ricachenevsky et al., 2013). MTP1 also has a function in Zn accumulation in shoots of Zn hyperaccumulators such as A. halleri (Dräger et al., 2004) or Thlaspi goesingense (Gustin et al., 2009).

Despite a broad interest in the use of tobacco to remove metals from contaminated soil, knowledge of the metal homeostatic processes in this species is still in its infancy. Only a few metal transport genes have been cloned and characterized so far. NtPDR3 (pleiotropic drug resistance) from Nicotiana tabacum was shown to be highly expressed under Fe-deficiency conditions suggesting its involvement in iron homeostasis (Ducos et al., 2005). MTP family members involved in Zn and Co metabolism were cloned from Nicotiana tabacum (NtMTP1a, NtMTP1b) and Nicotiana glauca (NgMTP1) (Shingu et al., 2005). Also, two orthologs of the Arabidopsis thaliana HMA2 and HMA4 were identified in tobacco, NtHMAα and NtHMAβ. Similar to Arabidopsis genes, NtHMAα and NtHMAβ are responsible for Zn and Cd root-to-shoot translocation (Hermand et al., 2014; Liedschulte et al., 2017). Furthermore, studies performed on tobacco BY-2 cells identified two genes encoding Fe uptake proteins; NtNRAMP3 and NtZIP1 (Sano et al., 2012). A second ZIP family member from tobacco, NtIRT1, was also shown to transport Fe, and its expression depended on the level of Fe and Cd in the medium (Yoshihara et al., 2006; Hodoshima et al., 2007).

To learn more about the molecular mechanism regulating Zn accumulation in tobacco leaves, the aim of this study was to identify the members of the following key metal transport families that could be involved in regulating Zn levels in the leaf blades: ZIP, NRAMP, and MTP. Moreover, taking into account very limited knowledge on the possible contribution of MRPs (multidrug resistance-associated proteins) family members to detoxification of metals, they were also included. MRP/ABCC (Klein et al., 2006; Verrier et al., 2008) transporters are a ubiquitous subfamily of ABC (ATP Binding Casette) transporters which catalyze the export of substrates out of the cytosol in an ATP-dependent manner. Their involvement in Zn and Cd hypertolerance in N. caerulescens was shown by Halimaa et al.

(2014) and also in the detoxification of Cd (Bovet et al., 2003, 2005; Wojas et al., 2007; Gaillard et al., 2008).

The major focus in this study was on proteins mediating Zn import into the tobacco leaf cells from the ZIP family. They were identified and initially characterized in several organisms, for example in Arabidopsis (15 ZIPs; Grotz and Guerinot, 2006), rice (16 ZIPs; Chen et al., 2008), bean (23 ZIPs; Astudillo et al., 2013) and more recently, wheat (Evens et al., 2017). In addition to Zn, ZIPs mediate transport of Mn, Fe, Ni, or Cu. Detailed analysis of the role of ZIP genes is still lacking for many of those identified. Their function has been anticipated primarily based on metalspecific (Zn, Fe, Mn, Cd, and Cu) and concentration-dependent (deficit/sufficient/excess) regulation of ZIP expression in organs. (Bashir et al., 2012; Sinclair and Krämer, 2012; Milner et al., 2013; Evens et al., 2017; Nazri et al., 2017).

Here, bioinformatics analysis of tobacco genome data was performed to identify sequences homologous to chosen Arabidopsis thaliana metal transport genes, and subsequent expression analysis led to the identification of the NtZIP1-like. It was cloned and characterized indicating its specific function in the regulation of Zn homeostasis in tobacco leaves.

### MATERIALS AND METHODS

### Plant Material and Growth Conditions

All experiments were performed on tobacco plants (Nicotiana tabacum var. Xanthi). Surface sterilized seeds (8% sodium hypochloride w/v for 2 min) were germinated on Petri dishes positioned vertically containing quarter-strength Knop's medium, 2% sucrose (w/v) and 1% agar (w/v) (Barabasz et al., 2013). Three weeks following germination, seedlings were transferred to hydroponic conditions. They were cultivated in 2-L pots (5 plants per pot) on aerated quarter-strength Knop's medium for 2 weeks to allow them to adjust to hydroponic conditions. The nutrient solution was renewed every 3–4 days (unless indicated otherwise). Five-week-old plants (3 weeks on plates and 2 weeks on hydroponics) were further used for experiments. They were exposed to chosen Zn (as ZnSO4) concentrations added to quarter-strength Knop's medium. Details are given in the subsections 2.3 and 2.9 below. At the end of each experiment, the plant samples were collected always at the same time of the day (between 10–12 AM). The quarter-strength Knop's medium (containing 0.5 µM Zn) was used as a reference (control) medium in parallel to applied Zn treatments.

Plants were cultivated in a growth chamber at temperature 23/16◦C day/night, 40–50% humidity, 16 h photoperiod, and quantum flux density [photosynthetically active radiation (PAR)] 250 mmol m−<sup>2</sup> s −1 , fluorescent Flora tubes.

### Database Search for Putative Tobacco Metal Transport Family Members

The goal was to identify potential tobacco metal transporters involved in the accumulation of Zn in leaves. There are two sources of tobacco sequences to be used for gene mining. First, the complete genomic tobacco sequence has recently been made available to the public in GenBank with accession code AWOK00000000 (Sierro et al., 2013, 2014). Second, there is the NCBI database which provides already annotated genes from a range of species including tobacco. In tobacco, only several metal transporters have been already identified, cloned and characterized, some more were annotated and their sequences could be found in the NCBI database. Thus, NCBI database likely does not contain all tobacco genes. Therefore, the search for putative tobacco Zn transporters was performed with the use of both AWOK and NCBI databases. Tobacco metal transporters were identified based on homology to the previously annotated sequences of Arabidopsis thaliana genes belonging to the following major metal transport families: (i) ZIPs: ZRT, IRT-like proteins; (ii) NRAMPs: natural resistance-associated macrophage proteins; (iii) MTP: metal tolerance proteins; (iv) MRP/ABCC: multidrug resistance proteins.

The genome of Nicotiana tabacum, Basma Xanthi has been downloaded from http://www.ncbi.nlm.nih.gov/Traces/wgs/?& val=AWOK01 (BX), and a search for tobacco sequences homologous to sequences of metal transporters gene from A. thaliana was performed. For this we used program BLASTn (NCBI Resource Coordinators, 2016) which was run using an amino acid sequence from Arabidopsis against the Basma Xanthi genome. We retained alignments with e-value not exceeding 1e−05. Next we used AAT package (Huang et al., 1997) for analyzing and annotating large genomic sequences containing introns. The predicted exons were further filtered in order to avoid spurious predictions (minimal length at least 10 amino acids, plus a threshold on confidence levels for both boundaries that were returned by AAT package).

In parallel, the NCBI database was used for BLASTn searches of Nicotiana sequences with homology to the already annotated A. thaliana sequences of metal transporters. FGENESH and FGENESH+ tools (Softberry, Mount Kisco, NY, United States<sup>1</sup> ) were used to identify the untranslated regions (UTRs), exons, and introns within the scaffold containing sequences of chosen tobacco genes, and to predict putative proteins encoded by these genes. Protein sequence alignments were performed using ClustalW and the phylogenetic trees were constructed with MEGA7.0 software (Tamura et al., 2013) using the maximum likelihood method with 1000 bootstrap replicates. The prediction of membrane-spanning regions and orientation was performed using Phobius software (Käll et al., 2004).

### Identification of Metal Transport Genes Differentially Regulated in Leaves by Exposure to Zn Excess

The 5-week old tobacco plants (obtained as described in the section "Plant Material and Growth Conditions") were grown for the next 4 days in the control medium, then they were exposed to 200 µM Zn (added to the quarter-strength Knop's medium) for up to 3 days. Quarter-strength Knop's medium (contains 0.5 µM Zn) served as a reference condition. On the 1st, 2nd, and 3rd day of the Zn treatment blades from the 2nd

<sup>1</sup>http://www.softberry.com

leaf (counting from the base of a plant) were collected. Leaves were cut out from each plant, petioles and the major midribs were excised, and the fragments of the blades were immediately frozen in liquid nitrogen. Three independent biological replicate experiments were performed. For each repetition, the leaf blade fragments were collected from a total of 40 plants.

Quantitative Real-Time PCR (RT-qPCR) was used to determine which putative metal transport genes out of those identified by bioinformatics analysis (see section "Database Search for Putative Tobacco Metal Transport Family Members") are differentially regulated in the leaf blades by 200 µM Zn (as compared with the control conditions). Specific primers were designed for the sequences of identified metal transport genes from the ZIP, NRAMP, MTP, and MRP/ABCC families identified in the tobacco genome databases (Supplementary Table S1).

### Cloning of NtZIP1-Like and Bioinformatic Analysis

The whole sequence of ZIP1-like was determined by 5<sup>0</sup> - and 3 0 rapid amplification of cDNA ends (RACE) using SMARTer RACE 5<sup>0</sup> /3<sup>0</sup> Kit (Clontech Laboratories, Inc. and A Takara Bio Company, Mountain View, CA, United States) according to the manufacturer's manual. Briefly, the partial sequence of ZIP1-like (previously identified in the tobacco genome database at AWOK01S302253.1) was used to design genespecific primers (GSPs) for the 5<sup>0</sup> - and 3<sup>0</sup> -RACE reactions [2253- GSP1-1-UPM (5<sup>0</sup> ) 2253-GSP2-1-UPM (3<sup>0</sup> )] (Supplementary Table S1). Amplification of the 5<sup>0</sup> - and 3<sup>0</sup> -end was performed in 50 µl reactions with the use of the Phusion HF polymerase (Thermo Scientific). The PCR product of an expected size was electrophoresed on an 1% agarose/EtBr gel and excised DNA fragment was cleaned with the Macherey-Nagel PCR clean-up Gel extraction (Germany, VWR MANB740609.50) according to the manufacturer's instruction. It was cloned into the pRACE vector (provided with the SMARTer RACE kit) and subsequently the reaction mixture was used to transform Escherichia coli Stellar Competent Cells. The plasmids were isolated from individual colonies, and the presence of the expected insert was confirmed by PCR screening (starters M13/For and M13/Rev), then by sequencing (Genomed, Poland). Nucleic and amino acid sequence alignments between obtained sequence and sequence predicted by Fgenesh program was performed using ClustalW.

The full length NtZIP1-like cDNA sequence was amplified by PCR (Supplementary Table S1), subcloned to pENTRTM/D-TOPO <sup>R</sup> and used for E. coli One ShotTM TOP10 (Invitrogen) transformation. The insert was sequenced to confirm the correct sequence. The sequence of the NtZIP1-like cDNA was deposited to the NCBI database (2015) under the accession number XM\_016652513.

### RNA Extraction

Total RNA was extracted from samples stored in −80◦C with the use of an RNeasy Plant Kit (Syngen, #SY341010) according to the manufacturer's recommendations, followed by DNase I digestion (Qiagen, #79254). The samples of RNA were quantified at 260 nm using a Nanodrop spectrophotometer ND 100 (Nanodrop, Wilmington, DE, United States). RNA concentration and purity was determined before and after DNA digestion using a NanoDrop spectrophotometer ND-1000 (Nanodrop, Wilmington, DE, United States) and the 260/280-nm ratio showed expected values between 1.8 and 2.0. The RNA integrity of samples was also confirmed by electrophoresis in agarose gel.

### Quantitative Real-Time PCR

The cDNA used as a template for the RT-qPCR reaction was synthesized using RevertAidTM First Strand cDNA Synthesis Kits (Fermentas) in a 20 µl reaction volume containing 1–3 µg of aRNA and oligo d(T)18 primers following the manufacturer's protocol. The RT-qPCR reaction was performed according to procedures described in Kendziorek et al. (2016) with minor modifications. It was performed in a Roche mastercycler (LightCycler <sup>R</sup> 480 System, Roche) using Light Cycler480 SYBR Green (Master 0488735001) according to the manufacturer's recommendations. The primers (Supplementary Table S1) were designed using IDT OligoAnalyzer 3.1<sup>2</sup> and OligoCalc: Oligonucleotide Properties Calculator<sup>3</sup> . The tobacco NtPP2A (protein phosphatase 2A; AJ007496) gene was used as the reference gene/internal control and was amplified in parallel with the target gene allowing gene expression normalization and providing quantification. Their stability in the plant samples collected for expression analysis was measured and shown in Supplementary Figure S1. Expression analysis was performed with at least three independent biological replicates. For each sample, reactions were set up in triplicate and means were calculated. Quantification of the relative transcript levels was performed using the comparative dCt (threshold cycle) method. Validation experiments were performed to test the efficiency of the target amplification and the efficiency of the reference amplification. The general quality assessment of the qPCR results was based on the amplification and melting curve profile of the samples in relation to the assay controls (non-template controls).

### Functional Analysis of NtZIP1-Like in Saccharomyces cerevisiae Strains

The full cDNA of NtZIP1-like was amplified using Phusion polymerase with the primers introducing XbaI and BamHI restriction sites for amplification (Supplementary Table S1). Obtained sequence was restriction ligated into the pUG35 yeast expression vector (kindly provided by Dr. M. Migocka, The University of Wrocław). The open reading frame (ORF) of NtZIP1-like was inserted in frame C-terminal to the ORF of EGFP (construct pUG35-NtZIP1-like-EGFP) and with the STOP codon (construct pUG35-NtZIP1-like), and fused with the methionine-repressible MET25 promoter (Kurat et al., 2006; Petschnigg et al., 2009). The same cloning strategy has been used to fuse the EGFP coding region to the N-terminal end of the NtZIP1-like in the pUG36 vector (construct pUG36-EGFP-NtZIP1-like). The resulting constructs and empty vectors were transformed to

<sup>2</sup>http://eu.idtdna.com/calc/analyzer

<sup>3</sup>http://www.basic.northwestern.edu/biotools/oligocalc.html

yeasts using the lithium acetate method (Gietz and Schiestl, 2007).

The yeast strains used in this study were DY1457 (MATa, ade1 can1 his3 leu2 trp1 ura3), the mutant ZHY3 - 1zrt1/zrt2 (DY1457 + zrt1::LEU2, zrt2::HIS3) defective in high and low affinity zinc uptake, and 1fet3fet4 (MATa trp1 ura3 Dfet3::LEU2 Dfet4::HIS3), defective in high and low affinity iron uptake system. Yeast strains were grown on liquid synthetic complete medium (SC-URA-MET/Glu) of the following composition: yeast nitrogen base supplemented with amino acids (without uracil and methionine), 2% (w/v) glucose, pH 5.3 (containing 0.2 mM Zn) overnight at 30◦C with shaking. On the next day the OD<sup>600</sup> was measured, adjusted to OD<sup>600</sup> of approximately 0.2 and yeasts were grown for another 2–5 h. The OD<sup>600</sup> was measured again, adjusted to OD = 0.3, series of dilutions were made (1.0, 0.1, 0.01, 0.001, and 0.0001) and 3 µl aliquots of each yeast culture were spotted onto plates containing (SC-URA-MET/Glu) medium solidified with 2% (w/v) agar supplemented with components depending on needs.

To determine whether Zn is a substrate for the NtZIP1 like, the 1zrt1/zrt2 yeast strain with the expression of pUG35, pUG35-NtZIP1-like, pUG35-NtZIP1-like-EGFP, pUG36 or pUG36-EGFP-NtZIP1-like and WT (DY1457) with the expression of pUG35 or pUG36 (empty vector) were grown on liquid SC-URA-MET/Glu medium (containing 0.2 mM Zn) and spotted onto the agar-solidified SC-URA-MET/Glu medium containing series of EGTA (ethylene glycol-bis(β-aminoethyl ether)-N,N,N<sup>0</sup> ,N0 -tetraacetic acid) concentrations: 2.5, 5.0. 7.5, and 10.0 mM. Yeast growth was monitored for the next 5 days.

To examine if Cd is a substrate for the NtZIP1-like, yeast WT (DY1457) was transformed with the pUG35, pUG35-ZIP1-like and pUG35-ZIP1-like-EGFP, and spotted onto the agar-solidified SC-URA -MET/Glu medium containing range of Cd (as CdCl2) concentrations (5, 10, 20, 50, and 75 µM). The sensitivity to cadmium was monitored.

To determine whether Fe is a substrate for the NtZIP1 like, complementation of the growth defect of 1fet3fet4 mutant line by expression of NtZIP1-like (the same constructs as above were used for expression) was tested on plates containing agar-solidified SC-URA-MET/Glu control medium. Moreover, modification of the sensitivity to high Fe due to expression of NtZIP1-like was examined on medium supplemented with 50 or 100 µM FeCl3.

### Subcellular Localization of NtZIP1-Like Protein

The entire cDNA sequence of the ORF of NtZIP1-like were obtained using Phusion polymerase and primers introducing CACC at the 5<sup>0</sup> end of the amplicon (underlined): forward 5 <sup>0</sup> CACCATGAATAACCACAATGTCCAAGT 3<sup>0</sup> and reverse 5 0 -AGCCCATTTAGCCATCACAGA -3<sup>0</sup> . The CACC overhand in the forward primer is required for directional cloning in the pENTR/D TOPO <sup>R</sup> vector (add provider). Following amplification, the cDNA was ligated into a Gateway entry vector pENTR/D-TOPO (Invitrogen). Fusion proteins with GFP were produced by the recombination (LR reaction) of entry vectors pENTR/D-TOPO-NtZIP1-like with destination vector pMDC43 (N-terminal GFP) (Curtis and Grossniklaus, 2003) using the Gateway system (Invitrogen).

Resulting construct pMDC43-GFP-ZIP1-like was sequenced (Genomed, Poland), then used for determination of the subcellular localization of the NtZIP1-like in tobacco cells. The pMDC43-GFP-ZIP1-like fusion protein was transiently expressed in tobacco leaves as described by Siemianowski et al. (2013). Leaves of 6-week-old WT tobacco grown on control medium were infiltrated with Agrobacterium tumefaciens carrying the pMDC43-GFP-ZIP1-like construct. Three days from the infiltration, leaves were analyzed using a Nikon A1 confocal laser scanning microscope (Melville, NY, United States). GFP signals were detected by excitation with the 488 nm line of the argon laser and emission was recorded between 500 and 560 nm. To confirm plasma membrane localization of NtZIP1-like, cell walls at the plasma membrane border of examined tobacco epidermal cells were visualized by staining with the 50 µM water solution propidum iodide (20 min), a membrane-impermeant red fluorescent dye (Suh et al., 2007; McFarlane et al., 2010). Imaging was detected by 543 nm excitation and 617 nm emission. In parallel, chlorophyll autofluorescence was monitored using a HeNe (543 nm) laser for excitation.

### Hydroponic Experiments

Developmental Regulation of NtZIP1-Like Expression To study the organ-specific expression of NtZIP1-like which depends on a developmental stage of the vegetative phase of growth the 3-week old tobacco plants were transferred from the agar plates to the control liquid medium (see section "Plant Material and Growth Conditions") and cultivated for up to 6 weeks. For the first 3 weeks on hydroponics the nutrient solution was changed every 3–4 days (plants were grown in 2- L plastic pots, five plants per pot). Next they were transferred to 1.2-L pots (two plants per pot) for the consecutive 3 weeks, and the medium was changed every 2nd day. The plant samples were collected at three stages of vegetative development: (Stage 1) small seedlings (3 weeks at the plates and 1 week on hydroponics); (Stage 2) young plants with rosette leaves (3 weeks on plates and 3 weeks on hydroponics); (Stage 3) adult plants with formed stem (3 weeks on plates and 6 weeks on hydroponics). At the Stage 1 and Stage 2 all leaf blades and all roots were collected separately. At the Stage 3 the following organs were collected: (i) from the aerial part of each plant – (a) two young leaves counting from the top (length of the blade of the smallest one was 0.5 cm); (b) two oldest leaves (counting from the base); (c) stem −3 cm of the middle part; (ii) rootstwo segments of the roots which grew out directly from the hypocotyl (adventitious roots were not included into analysis): (a) apical segment: 3–4 cm measured from the tip of the root; (b) basal segment: 3–4 cm measured from the base of the root. Plant samples were immediately frozen in the liquid nitrogen and stored in −80◦C until expression analysis. Three independent biological replicate experiments were performed. For each repetition samples were collected from a total of 30 plants (for Stage 1), 15 plants (for Stage 2) and 10 plants (for Stage 3).

### Regulation of the Expression of NtZIP1-Like by Zinc

To determine if the expression of NtZIP1-like depends on Zn availability, the 5-week old plants (see section "Plant material and growth conditions") were grown for the next 4 days in the control medium, then they were subjected to the following treatments: (i) to Zn deficit (Zn was omitted from the medium) for 4 days; (ii) to Zn deficit for 4 days followed by re-supply with control conditions for 2 days; (iii) to Zn excess (50 µM Zn present in the control medium) for 1 day; (iv) control medium in parallel to all treatments. At the end of each experiment plant material was collected, frozen in liquid nitrogen, and stored in −80◦C for expression analysis. The following organs were collected: (i) the blades of the 2nd and 3rd leaf (counting from the base) without petioles and the major midribs; (ii) two sectors of the roots which grew out directly from the hypocotyl (adventitious roots were not included into analysis): (a) 3–4 cm of the apical region; (b) 3–4 cm of the basal region. Three independent biological replicate experiments were performed. Samples were collected from a total of 10 plants for each repetition.

### Statistical Analysis

All presented data are from one experiment that is representative of three to four independent replicate experiments. Statistical significance was evaluated at the 0.05 probability level using Student's t-test.

### RESULTS

### Bioinformatic Analysis of Transporter Families in Tobacco

Arabidopsis thaliana cDNAs from major metal transport families were used as the query sequences to identify genes encoding Zn transporters in tobacco. By screening the tobacco genome scaffolds, sequences of tobacco genes that significantly matched with the query cDNAs (query coverage > 80%) were selected. The following protein families from A. thaliana were included in the search: (i) ZIPs: ZRT, IRT-like proteins; (ii) NRAMPs: natural resistance-associated macrophage proteins; (iii) MTPs: metal tolerance proteins; (iv) MRP/ABCC: multidrug resistance proteins. For each gene from A. thaliana used as a query, several tobacco homologous sequences were identified on different scaffolds. These sequences were screened for exon orientation, start and end positions, and confidence scores for the boundaries (Supplementary Table S2). Further the selected tobacco scaffolds were screened to identify full putative genomic sequences of NtZIP, NtNRAMP, NtMTP, and NtMRP (ABCC) genes, including transcription start sites, exons, introns, and polyadenylation sites using the FGENESH tool (Salamov and Solovyev, 2000), whereas Phobius system based on a hidden Markov model (HMM) approach, was applied to predict membrane topology of NtZIP1-like protein. The list of genomic sequences comprises twenty-one newly identified tobacco putative metal transporters. Their names were given according to the NCBI terminology of the genes from A. thaliana, which were used as a query (Supplementary Figure S2). Identified sequences (Supplementary Figure S2) were used to design primer pairs (Supplementary Table S1) for determination of their transcript level in the leaf blades of tobacco plants exposed to high Zn.

### Response of Genes from the ZIP, NRAMP, MTP, and MRP/ABCC Families in Tobacco Leaves to High Zn

To determine which of the identified metal transporters could be potentially involved in the regulation of Zn in tobacco leaves, their expression in the blades of plants grown in the presence of 200 µM Zn for up to 3 days was compared to the control conditions (**Figure 1**). From the ZIP family, three genes were identified with a several-fold difference in the transcript level between the Zn-exposed plants relative to those grown at the control medium. The most significant change was noted for NtZIP1-like and NtZIP4 (downregulation) and NtZIP11 (upregulation). Within the NtNRAMPs elevated expression was detected for a putative transporter NtNRAMP3-like. Moreover, modified expression was noted for three isoforms of NtMTP2. Out of identified six putative MRP/ABCC transporters which were subjected to analysis, the expression of three of them (NtMRP10-like and at a lower level NtMRP5-like and NtMRP14 like) were modified by high Zn.

### Phylogenetic Relationship of NtZIP1-Like from Tobacco

In this study, the focus was on finding genes potentially involved in the accumulation of Zn in tobacco leaves, which respond to high Zn. The ZIP family proteins are considered as major Zn uptake transporters (Ricachenevsky et al., 2015). Based on downregulation of NtZIP1-like by high Zn in leaves (**Figure 1**), the assumption was made that it plays a role in Zn influx into the cytosol. Therefore, the NtZIP1-like was chosen for cloning and characterization.

The ORF of the new tobacco ZIP family member – NtZIP1 like, consists of 1104 bp (**Table 1**) with 3 exons (**Figure 2**), and according to the prediction made by the program Fgenesh encodes 367 amino acids. To define the evolutionary relationship between ZIP1-like and the ZIP1 proteins from other organisms, as well as the other ZIPs, a phylogenetic tree was constructed (**Figure 3**). It included ZIP proteins from three species of tobacco (NtmZIP1-like, NsZIP1-like and NaZIP1-like), from A. thaliana, M. truncatula, and V. vinifera. It shows that NtZIP1-like is most closely related to ZIP1 proteins from other organisms including tobacco (NtmZIP1-like, NsZIP1-like and NaZIP1-like), Medicago truncatula (MtZIP1), Vitis vinifera (VvZIP1), and A. thaliana (AtZIP1). Within all ZIP1 sequences under comparison, NtZIP1 (Sano et al., 2012) formed a distinct clade with MtZIP3 and MtZIP4 from M. truncatula. The alignment of protein sequences defined at the phylogenetic tree as the closest homologs showed that the structure of NtZIP1-like is in agreement with the structure of other ZIP family members (Grotz and Guerinot, 2006). It contains eight transmembrane domains (TMs), a longer N-terminal region, a very short C tail, and a cytosolic variable region between TM domains III and IV (**Figure 4**). Histidine residues in the TMs II, IV, and V are highly conserved throughout the entire ZIP family. Our sequence analysis shows

Expression under control conditions was set to 1 as the frame of reference within each experiment. Values correspond to means ± SD (n = 3); those significantly different from the control (Student's t-test) are indicated by an asterisk (P ≤ 0.05).



that NtZIP1-like exhibits high amino acid sequence similarity with AtZIP1 and other known ZIP family members within these three mentioned TM domains (**Figure 4**). Among them, amino acid sequence conservation within the signature region in the fourth TM domain was found. It contains consensus sequences (including a fully conserved histidine residue). On the other hand, a potential metal-binding motif containing multiple histidine residues present in the variable region between TM III and IV, differs between examined proteins primarily in the number of his residues and their localization. In this region, eight histidine residues were found in NtZIP1-like compared to nine present for example in AtZIP1, and only three in NtZIP1 (**Figure 4**). The NtZIP1 protein (Sano et al., 2012) formed also a separate clade containing AtZIP3 and MtZIP3, MtZIP4 and AtZIP5.

The NtZIP1-like shares 56 and 54% identity at the amino acid level with AtZIP1 and NtZIP1, respectively, whereas 58 and 62% at the nucleotide level. The highest homology was found between the NtZIP1-like and other three tested tobacco ZIP1 proteins such as NtnZIP1-like (100%), NaZIP1-like (94 and 96%) and NsZIP1-like (95 and 97%, respectively (**Table 1**).

### NtZIP1-Like Localizes to the Plasma Membrane

To gain insight into the functioning of NtZIP1-like, its subcellular localization was determined by transient expression of the NtZIP1-like protein fused to the N terminus of green fluorescent protein (GFP) under the control of the cauliflower mosaic virus (CaMV) 35S promoter in tobacco leaves.

FIGURE 3 | Phylogenetic analysis of ZIP1 transporters from selected species. The unrooted tree was constructed based on amino acid sequences identified in the Aramemnon (Arabidopsis thaliana) and NCBI database (Nicotiana species, Medicago truncatula, Vitis vinifera), using the MEGA 7.0 software. The lengths of branches are proportional to the degree of divergence. Numbers in the figure represent bootstrap values (1000 replicates). The accession numbers are as follows: Arabidopsis thaliana, AtZIP1 - At3g12750.1, AtZIP2 - At5g59520.1, AtZIP3 - At2g32270.1, AtZIP4 - At1g10970.1, AtZIP5 - At1g05300.1, AtZIP6 - At2g30080.1, AtZIP7 - At2g04032.1, AtZIP8 - At5g45105.1, AtZIP9 - At4g33020.1, AtZIP10 - At1g31260.1, AtZIP11 - At1g55910.1, AtZIP12 - At5g62160.1; Nicotiana attenuata, NaZIP1-like XP\_019256554; Nicotiana tabacum, NtZIP1-like XP\_016507999, NtZIP1- NP\_001312674; Nicotiana sylvestris: NsZIP1-like XP\_009772024; Nicotiana tomentosiformis, NtomZIP1-like XP\_009608181; Medicago truncatula: MtZIP1 - AAR08412.1, MtZIP2 - AAG09635, MtZIP3 - AY339055, MtZIP4 - AY339056, MtZIP5 - AY339057, MtZIP6 - AY339058, MtZIP7 - AY339059; Vitis vinifera, VvZIP1 - XP\_002264603.2.

Three days after the infiltration of the leaves with Agrobacterium expressing pMDC43-GFP-ZIP1-like, the GFP signal (green fluorescence) was detected in tobacco epidermal cells along the cell walls indicating localization of NtZIP1-like protein at the plasma membrane (**Figure 5**). Cell walls were stained with propidium iodide and the green fluorescence (**Figure 5A**) coincided with the red signal, derived from propidium iodide (**Figures 5B,C**). The co-localization of the GFP-derived signal and propidium iodide staining of the cell walls indicates localization of GFP-fused NtZIP1-like protein at the plasma membrane (Pighin et al., 2004; Lee et al., 2010; Siemianowski et al., 2013). Moreover, at higher magnification the signal from the cell wall (red) and from the GFP-labeled plasma membrane (green) were separated (indicated by arrows; **Figures 5E–G**). Altogether, results support the conclusion that NtZIP1-like is a plasma membrane protein.

### Yeast Complementation Supports a Role for NtZIP1-Like as a Zn Transporter

Yeast functional complementation was used to determine the capacity of NtZIP1-like to transport Zn. The yeast zrt1zrt2 double mutant (ZHY3) defective in high and low affinity Zn uptake was used (Eide, 2003). The expression of fulllength cDNA of NtZIP1-like gene fused with the eGFP at its C-terminal end (construct pUG35EGFP-NtZIP1-like-EGFP), as well as at its N-terminal end (construct pUG36-EGFP-NtZIP1 like) did not complement the growth defect of the 1zrt1zrt2 yeast mutant (**Figure 6A**). In contrast, the expression of the construct pUG35-NtZIP1-like (with the STOP codon) fully restored growth under Zn-limited conditions (**Figure 6A**). These result indicates that NtZIP1-like is a plasma membrane protein mediating Zn uptake. The lack of rescue by constructs containing EGFP both at the C- and N-terminal end suggests that the presence of the eGFP protein makes the NtZIP1-like protein dysfunctional.

Some ZIP proteins mediate transport of Cd (Ramesh et al., 2003; Nakanishi et al., 2006; Stephens et al., 2011). To determine if NtZIP1-like is a Cd uptake protein, the wild-type yeast line DY1457 was transformed with pUG35-NtZIP1-like (with the STOP codon), and pUG35-NtZIP1-like-EGFP. If the NtZIP1-like is involved in Cd influx, the yeast transformants should be more sensitive to this metal. As shown in **Figure 6B** the growth of the wild-type transformed with the empty vector or with both tested constructs was limited by a range of Cd present in the medium to the same extent indicating no Cd transport capacity by NtZIP1-like.

Finally, to study the capacity of the NtZIP1-like to transport Fe, a yeast mutant 1fet3fet4 defective in both high- and low affinity Fe uptake systems was transformed with the NtZIP1-like cDNA to examine if it complements the defect in Fe transport. As shown in **Figure 6C**, expression of NtZIP1-like in the 1fet3fet4 did not restore the growth of mutants at control conditions, and it did not modify the sensitivity of yeast to Fe excess. To conclude, the results indicate that the NtZIP1-like does not transport Fe.

### Developmental Regulation of NtZIP1-Like

Our study showed that NtZIP1-like is expressed both in the roots, leaves and stems, but the level depends on the developmental stage (**Figure 7**). The transcript level in the leaves was very low in young 4-week old plants compared to a 6-fold increase in 6 week old tobacco. In the adult 9-week old plants its expression in young leaves was 3-times higher than in the old ones. NtZIP1-like was expressed at a moderate level in stems. In the roots of young 4-week old plants the transcript level was very low, and a 6-fold increase was detected in 6-week old tobacco. It remained at that level in adult 9-week old plants and did not differ in the apical and basal part of the root.

## Expression of NtZIP1-Like Is Zn Regulated

It has been shown that Zn is a substrate for NtZIP1-like. To know more about the possible physiological role of NtZIP1-like

in tobacco, its expression was analyzed in the roots, leaves and stems of plants exposed to Zn excess (50 µM for 1 day), and to Zn-deficiency (no Zn for 4 days) subsequently followed by a replete conditions (4-day Zn deficit followed by 2 days of control conditions). In agreement with downregulation in the leaf blades by 200 µM Zn (**Figure 1**), downregulation by 1-day exposure to 50 µM Zn was detected in leaves, and in the roots (both in the apical and basal segment) (**Figure 8**). Interestingly, its expression was highly upregulated by Zn-deficiency in the leaves and in the basal segment of the roots. No response to low Zn in the medium

was noted in the apical part of the root considered as primarily responsible for Zn uptake (**Figure 8**).

signal from the cell wall (stained red) clearly visible between adjacent GFP-labeled plasma membranes (arrows).

### DISCUSSION

Although tobacco, as a plant with a high capacity to accumulate large amount of metals (including Zn and Cd) in leaves, is used for phytoremediation of metal contaminated soil (Herzig et al., 2003, 2014; Lugon-Moulin et al., 2004; Dguimi et al., 2009; Vangronsveld et al., 2009), metal transporters involved in uptake and storage of metals in leaf tissues remain unknown. Here, based on bioinformatics searches for tobacco metal transporter sequences (Supplementary Figure S2 and Supplementary Table S2), and subsequent analysis of the regulation of the candidate genes by Zn excess (200 µM Zn) in leaves, ten genes (out of twenty-one tested) with significantly modified expression were identified (**Figure 1**). They represent putative metal transport genes that likely contribute to the storage of Zn excess in tobacco leaves, and include transporters involved in sequestration, redistribution and uptake of metals.

In sequestration of Zn in tobacco leaves exposed to 200 µM Zn three isoforms of NtMTP2 may play a role. Elevated expression of two isoforms of NtMTP2-X1 and NtMTP2-X2 (**Figure 1C**) suggests a likely involvement in loading of Zn into vacuoles, which are the major storage compartments within cells. The MTP2 proteins are not fully characterized so far in plants. It is known that the MTP2 belongs to the Group 1 of MTP vacuolar Zn transporters. Phylogenetic analysis showed that MTP1, MTP2, and MTP3 originate from a common MTP1/2/3 ancestors (Gustin et al., 2011).

The concentration of a metal in vacuoles depends not only on efficient loading, but also on the rate of mobilizing vacuolar pool back to the cytosol, which, among others, is under control of NRAMP proteins. In tobacco leaves expression of NtNRAMP3 like was significantly induced by high Zn supply (**Figure 1B**). Oomen et al. (2009) showed that TcNRAMP3/4 (and to a lesser extent AtNRAMP3/4) expression is regulated by Zn supply (lowefficient-excess), though the pattern of the regulation has not been fully established. However, until now Zn has not been shown to be a substrate for NRAMP3. The AtNRAMP3 from A. thaliana and TcNRAMP3 from T. caerulescens mediate efflux of Fe, Cd, and Mn from vacuoles to the cytoplasm (Thomine et al., 2003; Oomen et al., 2009). Ability to transport not only Fe, Mn, Cd but also Zn was shown for AtNRAMP4 and TcNRAMP4 only (Oomen et al., 2009). Thus, future studies will show whether the NtNRAMP3-like is localized to the tonoplast (like AtNRAMP3 or TcNRAMP3) or to the plasma membrane (like e.g., OsNRAMP3; Yamaji et al., 2013), what the substrates are, and what its role in the accumulation of high amounts of Zn in tobacco leaves.

The next identified new putative metal transporters regulated upon high Zn concentration in tobacco leaves are from the MRP/ABCC family. The major changes were found for the NtMRP10-like and NtMRP14-like, whereas to a lesser extent for NtMRP2-like, NtMRP3-like and NtMRP5-like (**Figure 1D**). The MRP/ABCC proteins carry various xenobiotics including metal complexes. Until now, there are only a few studies on plants indicating involvement of MRP/ABCC proteins in the transport of metals as conjugates to various substrates (Klein et al., 2006). Heterologous expression of AtMRP7 in tobacco has suggested a role in Cd transport into the root vacuoles (Wojas et al., 2009). AtABCC1 and AtABCC2 were shown to be targeted to the tonoplast and mediated the vacuolar sequestration of phytochelatin (PC) complexes with Cd(II) and Hg(II) (Park et al., 2012). The MRP/ABCC genes have been shown to be regulated by metals. For example, the expression of AtMRP3 is induced by Cd, Ni, As, Co, and Pb, but not Zn or Fe (Bovet et al., 2003; Zientara et al., 2009). Upregulation by Cd was also

confirmed for AtMRP6 (Gaillard et al., 2008) and AtMRP7 (Bovet et al., 2003), and by high Zn for TcMRP10 in the roots and shoots of Zn hyperaccumulator T. caerulescens (Hassinen et al., 2007). Identification in tobacco leaves of such Zn-responsive MRP/ABCC genes is important for future study on the regulation of Zn homeostasis upon treatment with high Zn.

The emphasis in this study was to shed more light on the regulation of Zn acquisition by cells in the leaves. The ZIP proteins constitute a major Zn uptake system (Sinclair and Krämer, 2012). Here, the NtZIP1-like was cloned and characterized to better understand its function in tobacco.

NtZIP1-like contains an ORF of 1104 bp, encoding a predicted protein of 367 amino acids (**Table 1**). Phylogenetic analysis of the ZIP family proteins shows that the NtZIP1-like forms a distinct clade with other ZIP1 proteins from three tobacco species (NatmZIP1-like, NsZIP1-like, and NaZIP1-like), A. thaliana, M. truncatula and V. vinifera (**Figure 3**). Sequence comparisons (**Figure 4**) showed that the deduced NtZIP1-like protein shares

three independent experiments.

FIGURE 8 | Expression pattern of NtZIP1-like in N. tabacum under various Zn conditions. Plants were grown in standard nutrient solution (control) and then transferred into modified control media: supplemented with 50 µM Zn for 1 day (1d); without Zn for 4 days (4d - Zn deficiency); plants grown at Zn-deficiency conditions for 4 days were transferred to the control medium for 2 days (6d - Zn replete). RT-qPCR analyses was performed on cDNA prepared from leaves (L), apical part of roots (AR) and basal part of roots (BR) of N. tabacum. Gene expression was normalized to the PP2A level. Values correspond to means ± SD (n = 3); those significantly different are indicated by an asterisk (P ≤ 0.05).

all the basic characteristic features of members of the ZIP family of metal transporters. It has eight TM domains, a long N-terminal end, a very short C tail, and a cytoplasmic variable region between TM III and IV (**Figure 4**). The variable region contains a histidine rich domain (HRD) with the motif (HX)<sup>n</sup> (n = 2, 3, and 4) which has been proposed as a metal binding site. The characteristic feature of ZIP proteins is the presence of a signature motif within the TM IV, and highly conserved histidine residues in TM domains II, IV, and V (Eide et al., 1996; Eng et al., 1998; Grotz et al., 1998; Guerinot, 2000; Rogers et al., 2000; Gaither and Eide, 2001). They all are present in the NtZIP1-like (**Figure 4**).

Analysis showed that two tobacco ZIP1 proteins – newly cloned NtZIP1-like and NtZIP1 (Sano et al., 2012), do not cluster together (**Figure 3**). They share 54% identity at the amino acid level (**Table 1**). Comparison of NtZIP1-like and NtZIP1 amino acids sequences (**Figure 4**) showed that an important difference between them lies within a variable cytoplasmic HRD region between the TM III and IV. Although this region displays low sequence conservation among ZIP metal transporters, most of them contain the motif (HX)<sup>n</sup> (n = 2, 3, and 4), which has been proposed as a metal binding site (Eide et al., 1996; Eng et al., 1998; Grotz et al., 1998). Only three (HX) repetitions were detected in NtZIP1, whereas eight in the newly cloned NtZIP1-like. To compare, other ZIP1 proteins contain eight or nine (HX) repetitions. The exact function of the loop between TM III-IV is yet to be determined, however, a study by Nishida et al. (2008) on the TjZNT1 ZIP transporter from Thlaspi japonica showed that deletion of a part of the HRD region containing his residues localized closer to the TM IV (HRD, position 207–217 aa) abolished Zn transport ability. Hence a difference in the structure at the amino acid level might contribute to a different substrate specificity. NtZIP1 and

NtZIP1-like do seem to have different substrate specificities. The expression of NtZIP1 in yeast significantly enhanced Fe accumulation suggesting Fe uptake activity (Sano et al., 2012). In contrast, the NtZIP1-like failed to alter the Fe-limited growth defect of fet3fet4 yeast mutant indicating it may not transport Fe (**Figure 6C**). NtZIP1-like is also unlikely to mediate Cd uptake since its expression in WT yeast did not modify the sensitivity to Cd (**Figure 6B**). Functional complementation of the zrt1zrt2 mutant, defective in Zn uptake supports its potential role as a Zn transporter (**Figure 6A**). NtZIP1-like localizes to the plasma membrane when transiently expressed in tobacco (**Figure 5**). Therefore, our results indicate that NtZIP1-like is a tobacco ZIP1 uptake protein for Zn, but not for Cd or Fe. In general, an ability of ZIP1 proteins to transport Fe was shown for PtZIP1 (Fu et al., 2017). More Fe transporters were identified among other ZIP proteins for example ZmZIP2-8, OsZIP5 and OsZIP8 (Li et al., 2013), MtZIP3, 5, 6 (López-Millán et al., 2004), PtZIP7 (Fu et al., 2017) and NtZIP1 (Sano et al., 2012).

Expression of the NtZIP1-like was detected in all plant organs, which suggests rather a universal role in maintaining Zn homeostasis (**Figure 7**). Its role seems to be more pronounced at later developmental stages. The transcript level is greater in older plants, especially in younger leaves (as compared with the older ones) suggesting a contribution to supplying cells in developing organs with Zn. Analysis of the regulation of the NtZIP1-like expression by Zn availability showed upregulation by Zn-deficiency in the roots and leaves (**Figure 8**), which was similar to AtZIP1 and OsZIP1 (Ramesh et al., 2003; Milner et al., 2013). Interestingly, in the roots upregulation of NtZIP1-like was limited to the basal segment of the root only, and was not detected in the apical part. It is known that the young, apical segment of the root is responsible for acquisition of nutrients, however, not much is known about the role of the older, basal region. Our studies clearly indicate that NtZIP1-like is a Zn-deficiency inducible Zn uptake transporter in leaves and in roots (though in roots the induction takes place only in the basal part; **Figure 8**). Further research is needed to demonstrate the NtZIP1-like tissue-specific expression and regulation, as it is not clear if it is involved in Zn acquisition from the medium, or in internal uptake. Distinct regulation of ZIP genes in a different root sectors has been shown also in rice (Ishimaru et al., 2005). Expression of OsZIP4 was detected in the meristematic region of the Zn-deficient roots. Similarly, AtZIP2p::GUS expression analysis revealed higher induction in the younger region of the roots grown under nutrient-replete conditions, as compared to a lower induction nearer the mature part at the root-shoot junction (Weber et al., 2004). In general, it is known that upregulation upon Zn deficiency conditions and downregulation in replete medium is ascribed to genes involved in the acquisition of micronutrients (Sinclair and Krämer, 2012), and these two features are characteristic for NtZIP1-like (**Figure 8**).

It is known that in the leaves of tobacco plants exposed to Zn excess, the metal is not distributed equally throughout the mesophyll cells. Instead, high concentrations were found in clusters of adjacent cells (Zn-accumulating cells) in contrast to its low level in neighboring non-accumulating ones (Siemianowski et al., 2013). Distinct expression patterns of Zn transport genes must underlie such different Zn uptake and accumulation capacity. Knowing this, we searched for genes differentially regulated in the leaves by high Zn. The NtZIP1-like was identified initially as downregulated by 200 µM Zn (**Figure 1**), and confirmed later as downregulated by 50 µM Zn (**Figure 8**). We hypothesize that the downregulation observed in leaves upon Zn excess could be a part of the molecular mechanism occurring in the low Zn-accumulating cells that prevents them from excessive uptake of Zn. Further comparative studies on the regulation of NtZIP1-like expression in leaves at low and high Zn at the cellular and tissue level are necessary in the future to investigate this.

### CONCLUSION

The bioinformatics analysis using information from the tobacco genome and the detailed expression study has led to the identification of ten new tobacco putative transporters involved in the regulation of Zn accumulation in tobacco leaves. They belong to different major families of metal transporters (ZIP, NRAMP, MTP, and MRP/ABCC), and undergo differential regulation in the leaves of tobacco plants exposed to 200 µM Zn. The upregulation of NtZIP11-like, NtNRAMP3, three isoforms of NtMTP2, three MRP/ABCC genes such as NtMRP10-like, NtMRP5-like and NtMRP14 like, and downregulation of NtZIP1-like and NtZIP4, indicate their contribution to a range of processes underlying uptake, sequestration and redistribution of metals in the cells and tissues. These data provide an important input for further research on metal homeostasis mechanisms in tobacco, the species used for phytoremediation of metal contaminated soil.

The detailed study on the newly cloned NtZIP1-like showed that the encoded protein is localized to the plasma membrane and mediates uptake of Zn, but not Fe or Cd. It is expressed in the roots and leaves – but the level of the transcript depends on the developmental stage. It is also regulated by the availability of Zn, being highly up-regulated by Zn-deficiency specifically in the leaves and in the basal part of the root but not in the apical zone. We have shown previously that tobacco mesophyll cells have a distinct capacity to store Zn in the "Zn-accumulating cells" which are next to non-accumulating ones in the leaf blade (Siemianowski et al., 2013). Our detailed studies on the NtZIP1-like suggest that it might be a candidate gene involved in the restriction of Zn uptake by the mesophyll cells with low capacity to accumulate Zn.

### AUTHOR CONTRIBUTIONS

AP carried out all experiments. KK was involved in yeast study, cloning and expression analysis. MK was involved in cloning, expression analysis and hydroponic experiments.

AB contributed to expression analysis. MP was involved in bioinformatics analysis and hydroponic experiments. JT performed bioinformatics analysis. BP contributed to confocal analysis. LW supervised yeast complementation assays. DA designed the study concept, coordinated the research and supervised experiments, performed data analysis, and wrote the manuscript. All authors read and approved the final manuscript.

### FUNDING

This work was financially supported by National Science Center, Poland (Grant HARMONIA No. NZ3/00527).

### REFERENCES


### ACKNOWLEDGMENTS

We would like to thank Dr. Rafał Milanowski (Department of Molecular Phylogenetics and Evolution, Faculty of Biology, University of Warsaw Biological and Chemical Research Centre) for advice and comments on the construction of the phylogenetic tree.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018.00185/ full#supplementary-material




legume Medicago truncatula. Biometals 24, 51–58. doi: 10.1007/s10534-010- 9373-6


in rice. Nat. Commun. 4, 2442–2453. doi: 10.1038/ncomms 3442


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Papierniak, Kozak, Kendziorek, Barabasz, Palusinska, Tiuryn, ´ Paterczyk, Williams and Antosiewicz. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Biofortified Crops Generated by Breeding, Agronomy, and Transgenic Approaches Are improving Lives of Millions of People around the world

*Monika Garg\*, Natasha Sharma, Saloni Sharma, Payal Kapoor, Aman Kumar, Venkatesh Chunduri and Priya Arora*

*National Agri-Food Biotechnology Institute, Mohali, Punjab, India*

#### *Edited by:*

*Felipe Klein Ricachenevsky, Universidade Federal de Santa Maria, Brazil*

#### *Reviewed by:*

*Hannetz Roschzttardtz, Pontificia Universidad Católica de Chile, Chile Michael La Frano, California Polytechnic State University, United States Ümit Barıs¸ Kutman, Gebze Technical University, Turkey*

#### *\*Correspondence:*

*Monika Garg monikagarg@nabi.res.in, mgarg100@yahoo.com*

#### *Specialty section:*

*This article was submitted to Plant Nutrition, a section of the journal Frontiers in Nutrition*

*Received: 09 August 2017 Accepted: 29 January 2018 Published: 14 February 2018*

#### *Citation:*

*Garg M, Sharma N, Sharma S, Kapoor P, Kumar A, Chunduri V and Arora P (2018) Biofortified Crops Generated by Breeding, Agronomy, and Transgenic Approaches Are Improving Lives of Millions of People around the World. Front. Nutr. 5:12. doi: 10.3389/fnut.2018.00012*

Biofortification is an upcoming, promising, cost-effective, and sustainable technique of delivering micronutrients to a population that has limited access to diverse diets and other micronutrient interventions. Unfortunately, major food crops are poor sources of micronutrients required for normal human growth. The manuscript deals in all aspects of crop biofortification which includes—breeding, agronomy, and genetic modification. It tries to summarize all the biofortification research that has been conducted on different crops. Success stories of biofortification include lysine and tryptophan rich quality protein maize (World food prize 2000), Vitamin A rich orange sweet potato (World food prize 2016); generated by crop breeding, oleic acid, and stearidonic acid soybean enrichment; through genetic transformation and selenium, iodine, and zinc supplementation. The biofortified food crops, especially cereals, legumes, vegetables, and fruits, are providing sufficient levels of micronutrients to targeted populations. Although a greater emphasis is being laid on transgenic research, the success rate and acceptability of breeding is much higher. Besides the challenges biofortified crops hold a bright future to address the malnutrition challenge.

Keywords: malnutrition, biofortification, transgenic, agronomic, breeding

### INTRODUCTION

"Biofortification" or "biological fortification" refers to nutritionally enhanced food crops with increased bioavailability to the human population that are developed and grown using modern biotechnology techniques, conventional plant breeding, and agronomic practices. The United Nations Food and Agriculture Organization has estimated that around 792.5 million people across the world are malnourished, out of which 780 million people live in developing countries (1). Apart from this, around two billion people across the world suffer from another type of hunger known as "hidden hunger," which is caused by an inadequate intake of essential micronutrients in the daily diet (2, 3) despite increased food crop production (4). Besides this overnutrition is growing matter of concern.

So far, our agricultural system has not been designed to promote human health; instead, it only focuses on increasing grain yield and crop productivity. This approach has resulted in a rapid rise in micronutrient deficiency in food grains, thereby increasing micronutrient malnutrition among consumers. Now agriculture is undergoing a shift from producing more quantity of food crops to producing nutrient-rich food crops in sufficient quantities. This will help in fighting "hidden hunger" or "micronutrient malnutrition" especially in poor and developing countries, where diets are dominated by micronutrient-poor staple food crops (5).

**282**

Traditionally, vitamins and minerals have been provided to the masses through nutrient supplementation programs, but it falls short of the goals set by the international health organizations as the supplementation programs rely on external funding that is not guaranteed to be available from year to year. Other limitations are purchasing power of poor people, their access to markets and health-care systems, and lack of awareness regarding the longterm health benefits of these nutrient supplements (6, 7). Hence, biofortification of different crop varieties offers a sustainable and long-term solution in providing micronutrients-rich crops to people. Furthermore, biofortified crops with increased bioavailable concentrations of essential micronutrients are deployed to consumers through traditional practices used by agriculture and food trade which therefore provides a feasible way of reaching undernourished and low income group families with limited access to diverse diets, supplements, and fortified foods. From an economic viewpoint, biofortification is a one-time investment and offers a cost-effective, long-term, and sustainable approach in fighting hidden hunger because once the biofortified crops are developed; there are no costs of buying the fortificants and adding them to the food supply during processing (8–14). Furthermore, in the next few decades, a major population increase might take place in the developing world and with the changing climatic conditions; achieving food security will pose a greater challenge (15, 16). Thus, organizations such as the World Health Organization and the Consultative Group on International Agricultural Research (CGIAR) have included the development of nutritionally enhanced high-yielding biofortified crops as one of their main goals (17).

### NECESSITY AND SOCIOECONOMIC DEVELOPMENT DERIVE BIOFORTIFICATION RESEARCH

Humans require around 40 known nutrients in adequate amounts to live healthy and productive lives (**Table 1**). The mineral elements—sodium, potassium, calcium, magnesium, phosphorous, chlorine, and sulfur—are classified as essential nutrients that

Table 1 | Essential micro- and macronutrients required for good human health.


are required in small amounts in the body. The other class of essential nutrients required in very small amounts in the human body are termed as micronutrients—namely iron, zinc, copper, manganese, iodine, selenium, molybdenum, cobalt, nickel, and vitamin A (18). Collectively, these nutrients play crucial roles in humans and dictate our physical and mental development (19). Many micronutrients act as cofactors for the functioning of various enzymes in the human body and thereby regulate important functions and metabolic processes in our body (20). For humans, agricultural products are the primary source of nutrients, especially for those living in developing countries (21–23). However, the diet of the population based on cereals such as rice, wheat, cassava, and maize contain insufficient amounts of several nutrients such as vitamin A, iron, zinc, calcium, manganese, copper, iodine, or selenium with respect to meeting daily requirements. These nutrient deficient agricultural products cannot support healthy lives and can result in poor health, sickness, increased morbidity and disability, impaired development, stunted mental and physical growth, diminished livelihoods, and reduced national socioeconomic development (24–29). Childhood stunting prevalent in many developing countries is associated with micronutrient malnutrition in children starting from fetal development to 4 years of age (25). Micronutrient deficiencies affect about 38% of pregnant women and 43% of pre-school children worldwide. More than 30% of the world's population has been reported to be anemic (30) and suffering from hidden hunger. The prevalence of anemia is more in developing countries compared with developed countries. Africa and South-East Asian countries are most affected (**Figures 1A,B**). Estimates have indicated that approximately half of this is attributed to iron deficiency (31). Hence, micronutrient malnutrition is the major challenge in many developing countries. Another important point of consideration is uneven distribution of the nutrients among different plant parts (32). For example, the iron content is high in rice leaves, but low in polished rice grain. Apart from under nutrition, growing incidence of overnutrition leading to problems of overweight and in particular, high rate of diabetes is a matter of concern. Consequently, biofortification is also directed toward enhancing the contents of desired micronutrients in the edible portion of crop plants. Nutritional targets for biofortification include elevated mineral content, improved vitamin content, increased essential amino acid levels, better fatty acid composition, and heightened antioxidant levels in crops (12). Biofortification of crop plants can provide enough calories to meet the energy needs along with providing all the essential nutrients needed for sound health. Furthermore, biofortifying the crops which are consumed by the poor population of the world can significantly improve the amount of nutrients consumed by this target population (33).

### BIOFORTIFICATION PATHWAY INCLUDES SEVERAL APPROACHES

Producing nutritious and safe foods, sufficiently and sustainably, is the ultimate goal of biofortification (34). Biofortification of essential micronutrients into crop plants can be achieved through three main approaches, namely transgenic, conventional, and

agronomic, involving the use of biotechnology, crop breeding, and fertilization strategies, respectively. Most of the crops targeted by transgenic, conventional breeding, and agronomical approaches include staple crops like rice, wheat, maize, sorghum, lupine, common bean, potato, sweet potato, and tomato (**Figure 2**). Cassava, cauliflower, and banana have been biofortified by both transgenic and breeding approaches while barley, soybean, lettuce, carrot, canola, and mustard have been biofortified with transgene and agronomic approaches. Higher numbers of crops have been targeted by transgenic means, while the practical utilization of biofortification is higher by breeding methods (**Figures 3A,B**). Cereals being staple crop have been targeted by all three approaches. Same is the case of legumes and vegetables. Interestingly, oil seed biofortification has been achieved through transgenic means, because limited availability of genetic diversity for the targeted component, low heritability, and linkage drag in the targeted crop (**Figure 3B**). Biofortification by breeding has been achieved in crops and specified components when genetic

diversity is available in the utilizable form in the primary, secondary, or tertiary gene pool of the targeted crop. When genetic diversity is unavailable, genetic transformation is the better option. Transgenic-based approach has advantages that a useful gene once discovered, can be utilized for targeting multiple crops (**Figure 4**). Some important genes like phytoene synthase (*PSY*), carotene desaturase, nicotinamide synthase, and ferritin have been utilized in multiple events including multiple crops. In this manuscript, we have compiled the data from research to release on different food crops that are being targeted by the different approaches of biofortification.

### BIOFORTIFICATION THROUGH TRANSGENIC MEANS—MAXIMUM RESEARCHED AND MINIMUM UTILIZED

Transgenic approach can be a valid alternative for the development of biofortified crops when there is a limited or no genetic variation in nutrient content among plant varieties (32, 35). It relies on the access to the unlimited genetic pool for the transfer and expression of desirable genes from one plant species to another which is independent of their evolutionary and taxonomic status. Furthermore, when a particular micronutrient does not naturally exist in crops, transgenic approaches remain the only feasible option to fortify these crop with the particular nutrient (7). The ability to identify and characterize gene function and then utilize these genes to engineer plant metabolism has been a key for the development of transgenic crops (36). Furthermore, pathways from bacteria and other organisms can also be introduced into crops to exploit alternative pathways for metabolic engineering (37).

Transgenic approaches can also be used for the simultaneous incorporation of genes involved in the enhancement of micronutrient concentration, their bioavailability, and reduction in the concentration of antinutrients which limit the bioavailability of nutrients in plants. In addition, genetic modifications can be targeted to redistribute micronutrients between tissues, enhance the micronutrient concentration in the edible portions of commercial crops, increasing the efficiency of biochemical pathways in edible tissues, or even the reconstruction of selected pathways (38–40). Development of transgenically biofortified crops initially involves substantial amount of time, efforts, and investment during research and development stage, but in a long run, it is a cost-effective and sustainable approach, unlike nutrition-based organizational and agronomic biofortification programs (14, 19). Furthermore, genetic engineering has no taxonomic constraints and even synthetic genes can be constructed and used. Transgenic crops with enhanced micronutrient contents hold a potential to reduce micronutrient malnutrition among its consumers, especially poor people in developing countries (12). Numerous crops have been genetically modified to enhance their micronutrient

Figure 3 | Representation of reported biofortified crops by transgenic, agronomic, and breeding means. (A) Comparison of transgenic and breeding approaches of biofortification in terms of relative research and release of commercial crops. While higher emphasis is being laid on transgenic-based biofortification, success rate in terms of cultivar release is higher for breeding-based approach. (B) Percentage of different crops biofortified by different approaches. Cereals have been biofortified in largest number by all three biofortification approaches. Legumes and vegetables have also been targeted by all the approaches in almost equal percentage. Transgenic approach covers highest number of crops. Oilseed crops have been mainly targeted by transgenic approaches due to limited genetic variability.

contents. Among micronutrients, vitamins, minerals, essential amino acids, and essential fatty acids have been targeted by the use of various genes from different sources to enhance the food crop nutritional level (**Table 2**). It has been found that *PSY*, carotene desaturase, and lycopene β-cyclase for vitamins, ferritin and nicotinamine synthase for minerals, albumin for essential amino acids, and Δ<sup>6</sup> desaturase for essential fatty acids have been widely reported as targets for biofortification (**Figure 4**). Successful examples of transgenic method are high lysine maize, high unsaturated fatty acid soybean, high provitamin A and iron

carotene desaturase, nicotinamide synthase, and ferritin have been utilized in multiple events including multiple crops.

rich cassava, and high provitamin A Golden rice. Reports are available for biofortified cereals, legumes, vegetables, oilseeds, fruits, and fodder crops.

### TRANSGENIC CEREALS

### Transgenic Rice (*Oryza sativa*)

Rice has been targeted to address the global challenge of undernutrition. Vitamin deficiency is one of the major challenges

#### Table 2 | Tabulation of crops, nutrients, research status, and concerned publications on biofortification by transgenic means.



(*Continued*)

#### TABLE 2 | Continued


(*Continued*)

#### TABLE 2 | Continued


*Significant amount of information have been generated that hold a bright future to address the malnutrition challenge.*

that affect underprivileged population due to poor affordability. Golden Rice was an important breakthrough in this direction as an effective source of provitamin A (beta-carotene) with a significant potential to reduce disease burden by expressing genes encoding *PSY* and carotene desaturase (41–45). The level of beta-carotene precursor, i.e., phytoene, has been enhanced up to 23-fold by targeting gene encoding carotene desaturase (45). Folic acid (vitamin B9) is important for normal pregnancy and anemia (190). Rice has been genetically modified to increase folate content (up to 150-fold) by overexpressing genes encoding *Arabidopsis* GTP-cyclohydrolase I (GTPCHI) and aminodeoxychorismate synthase [ADCS (46, 47)]. The 100 g of modified rice was found to be sufficient to meet daily folate requirements of an adult individual.

Rice has also been targeted to address the global challenge of iron deficiency anemia. Multiple reports have indicated an increase in iron content in rice by expressing genes encoding, nicotianamine aminotransferase (48), iron transporter *OsIRT1*

(49), nicotianamine synthase 1 (*OsNAS1*) and 2 (*OsNAS2*) (50–52, 191), soybean ferritin (52–54), and common bean ferritin (55). Iron biofortified rice was also synthesized by introducing multiple genes involved in iron nutrition (56–58). In addition to enhanced iron content, improvement in iron bioavailability was also achieved by reducing antinutrient compounds in rice such as phytic acid (59). Similarly, zinc content was also elevated in GM rice by overexpressing *OsIRT1* (49) and mugineic acid synthesis genes from barley [*HvNAS1*, *HvNAS1*, *HvNAAT-A*, *HvNAAT-B*, *IDS3* (60)].

Improvement in quality protein has been addressed by targeting essential amino acid content in rice by expressing seed-specific genes of bean β-phaseolin (61), pea legumin (62); Sesame 2S Albumin (63); soybean glycinin (64); bacterial aspartate kinase, dihydrodipicolinate synthase (DHPS) (65); maize DHPS (66); rice anthranilate synthase α-subunit (67); and *E. coli* aspartate aminotransferase (68). Rice has also been targeted for seed oil quality by increasing amount of polyunsaturated fatty acid that can help in the reduction of bad cholesterol levels in the body and improve human nutrition (192). An essential fatty acid α-linolenic acid has been enhanced in rice by expressing soybean omega-3 fatty acid desaturase (FAD3) gene [*GmFAD3* (69)]. Flavonoids are associated with antioxidant activity and its content in rice has been enhanced by expressing maize C1 and R-S regulatory genes [Myb-type and basic helix-loop-helix-type transcription factors (70)]; and phenylalanine ammonia lyase and chalcone synthase (*CHS*) genes (71). To address the challenge of overnutrition and obesity, the content of less digestible and resistant amylose starch has been enhanced by expression of antisense waxy genes (72, 73) and antisense RNA inhibition of starch-branching enzymes (SBE) (74). Besides introducing micronutrients, expression of functional human milk protein (lactoferrin) in rice grains has opened the possibility for creating a value-added cereal-based ingredients that can be introduced into infant formula and baby food (75, 193).

### Transgenic Wheat (*Triticum aestivum*)

Wheat is one of the most widely grown staple food crops in the world. Researchers have tried to address the challenges of most deficient nutrients like vitamin A, iron, and quality proteins through wheat. The provitamin A content of wheat has been enhanced by expressing bacterial *PSY* and carotene desaturase genes [*CrtB*, *CrtI* (76, 77)]. The iron content in wheat has been enhanced by expression of ferritin gene from soybean (78) and wheat [*TaFer1-A* (79)]. To increase iron bioavailability phytase activity was increased by the expression of the phytochrome gene [*phyA* (80)] and phytic acid content has been decreased by silencing of wheat ABCC13 transporter (81). Protein content, especially essential amino acids lysine, methionine, cysteine, and tyrosine contents of wheat grains were enhanced using Amaranthus albumin gene [*ama1* (82)]. Wheat has also been targeted to improve the antioxidant activity by expressing maize regulatory genes (*C1*, *B-peru*) involved in anthocyanin production (83). To address the challenge of overnutrition and obesity, the content of less digestible and resistant amylose starch has been enhanced by silencing gene encoding SBE [*SBEIIa* (84)].

### Transgenic Maize (*Zea mays*)

Maize is one of the important staple crops in developing countries, and it has been addressed for vitamins, minerals, quality protein, and antinutrient components by means of genetic engineering. Maize endosperm has been enriched with provitamin A (carotenoids) by expressing bacterial *crtB* (85) and multiple (5) carotenogenic genes (86, 194). Vitamin E and its analog are potent antioxidants with implications over human health and many research groups are emphasizing on biofortification of these components in maize crop. Tocotrienol and tocopherol content in maize has been increased by overexpression of homogentisic acid geranylgeranyl transferase [HGGT (87)]. Vitamin C (l-ascorbic acid) a water-soluble antioxidant play roles in cardiovascular function, immune cell development, and iron utilization (88). Its level in corn has been enhanced nearly 100-fold times by recycling oxidized ascorbic acid to reduced form by the expression of dehydroascorbate reductase [DHAR (89)]. On the other hand, Naqvi et al. (90) developed multivitamin corn containing 169-fold the normal amount of beta-carotene, double the normal amount of folate and 6-fold the normal amount of ascorbate by engineering three distinct metabolic pathways.

Bioavailability of micronutrients is hindered by antinutrient components. Bioavailability of iron has been increased by expressing soybean ferritin and *Aspergillus* phytase (91), soybean ferritin (92), *Aspergillus niger phyA2* (93), and silencing the expression of ATP-binding cassette transporter and multidrug resistanceassociated protein (94). As a practical example, BVLA4 30101 variety released by Origin Agritech in China has been biofortified for phytate degradation.

The major maize seed storage proteins, zeins have poor nutritional quality due to lower content of essential amino acids lysine and tryptophan. In maize essential amino acid content has been targeted with significant achievement. Lysine content in maize has been increased by expression of *sb401* from potato (95, 96), single bifunctional expression/silencing transgene cassette (97). Both lysine and tryptophan content have been increased in maize by antisense dsRNA targeting alpha-zeins [both 19- and 22-kDa (98)]. Importance of lysine content in maize is evident from maize varieties rich in lysine *viz.,* Mavrea™YieldGard Maize that has been released by Monsanto in Japan and Mexico; Mavera™ Maize (LY038) by Renessen LLC (Netherland) in Australia, Columbia, Canada, Japan, Mexico, New Zealand, Taiwan, USA The amino acid methionine is a common protein building block that is also important in other cellular processes. Its content has been increased in maize by modifying *cis*-acting site for *Dzs10* (99). Amino acid balance of maize has also been improved by expressing milk protein α-lactalbumin (40).

### Transgenic Barley (*Hordeum vulgare*)

Barley being a model cereal crop has been targeted to improve its micronutrient content. Its zinc content has been improved by overexpression of zinc transporters (100). To increase the bioavailability of iron and zinc, phytase activity has been increased in barely seeds by expression of phytase gene [*HvPAPhy*\_a (101)]. Essential amino acid lysine has been enhanced in barley by expressing DHPS gene [*dapA* (102)]. β glucans are dietary fibers and are believed to dramatically reduce the risk of contracting serious human diseases such as cardiovascular disease and type II diabetes (103). Its content has been increased in barley by overexpression of cellulose synthase-like gene [*HvCslF* (104)]. Resistant starch (amylose only) barley has been produced by the RNAi approach by suppressing all genes coding for SBE [*SBE I, SBE IIa, SBE IIb* (105)]. Content of health promoting polyunsaturated fatty acids, γ-linolenic acid, and stearidonic acid (STA) has been improved in barley by expressing Δ<sup>6</sup> -desaturase [*D6D* (106)]. Barley has been targeted to express human lactoferrin gene [*HLF* (107)]. Apart from this several medicinally and industrially important bioactives including enzymes and antibiotics have been expressed in barley.

### Transgenic Sorghum (*Sorghum bicolor*)

Sorghum is one of the most important staple foods for millions of poor rural people. It has an ability to grow well in harsh environments. It has been targeted to improve provitamin A (beta-carotene) by expressing *Homo188-A* (108). Content of essential amino acid lysine has been improved in sorghum by the introduction of a high lysine protein [HT12 (109)]. One of the issues with sorghum consumption is that its grains are less digestible than the other major staple crops. Its seed storage proteins, γ-kafirin, is resistant to protease digestion. Digestibility index of transgenic sorghum has been increased by RNAi silencing of the *γ-kafirin* (110) and combined suppression involving three genes [*γ-kafirin-1*, *γ-kafirin-2*, and α*-kafirin A1* (111)].

### TRANSGENIC LEGUMES AND PULSES

### Transgenic Soybean (*Glycine max*)

Soybean is a global source of vegetable oil and high-quality protein. The soybean has been targeted to increase provitamin A (beta-carotene), a monounsaturated ω-9 fatty acid (oleic acid) and seed protein contents by expressing bacterial *PSY* gene (112). In a different approach provitamin A (Canthaxanthin) was enhanced by expressing bacterial *PSY* [*crtB*, *crtW*, *bkt1* (113)]. Kim et al. (114) has demonstrated the production of a high provitamin A (beta-carotene) soybean through overexpression of *PSY* and carotene desaturase. Another important nutrient vitamin E activity in barley has been enhanced with increased content of δ-tocopherol and decreased γ-tocopherol by coexpressing 2-methyl-6-phytyl benzoquinol methyltransferase genes [*At-VTE3*; *At-VTE4* (115)]. Soybeans contain approximately 40% protein, but they are deficient in one or more of the essential amino acids, especially the sulfur-containing amino acids, cysteine and methionine. The cysteine content of soybean seeds has been increased through overexpression of the sulfur assimilatory enzyme, O-acetylserine sulfhydrylase (116). Similarly, Dinkins et al. (117) increased methionine and cysteine content in soybean by overexpressing the maize zein protein. The methionine content of soybean has been increased by expressing cystathionine γ-synthase (118, 119). Soybean is rich in healthy oil and has approximately 20% oil content. But 7–10% of the oil contains unstable fatty acid α-linolenic acids that contribute to reduced soybean seed oil quality. It results in the formation of undesirable *trans*-fatty acid as a result of hydrogenation (195). To enhance the agronomic value of soybean seed oil by reducing the levels of α-linolenic acids (18:3), siRNA-mediated gene silencing-based approach has been utilized for silencing of ω-3 FAD3 (120). In another experiment γ-linolenic acid (GLA) and STA (ω-3 fatty acids) content in soybean oil has been increased by expression of Δ<sup>6</sup> -desaturase gene that is responsible for the conversion of linoleic acid and α-linolenic acid to GLA and STA (121). Similarly, STA content has been increased by simultaneous expression of Δ<sup>6</sup> desaturase and Δ15 desaturase (122). Antisense RNA technology has been used to reduce the amount of linoleic acid and palmitic acid and increase the amount of oleic acid by inhibition of expression of Δ12 oleate desaturase [*GmFAD2-1b* (123)] that converts oleic acid into linoleic acid. Soybean seeds are low in isoflavone content. Consumption of isoflavone is associated with human health benefits such as decreased risk of heart disease, reduced menopausal symptoms, and reduced risk of some hormone-related cancers (196). Isoflavone content has been enhanced in soybean seeds by the combination of maize C1 and R transcription factor-driven gene activation and suppression of a competing pathway (124).

Importance of improvement in ω-3 fatty acid content in soybean is evident from the fact that a large number of cultivars with improved oleic, linoleic, and STA have been released by private companies. Transgenic soybean varieties rich in oleic acid *viz*., G94-1, G94-19, G168 have been released in Australia, Canada, Japan, New Zealand, USA; and Treus™, Plenish™ (DP305423) in Australia, Canada, China, European Union, Japan, Mexico, New Zealand, Philippines, Singapore, South Africa, South Korea, Taiwan, USA; and Treus™ (DP 305423 × GTS 40-3-2) in Argentina, Canada, China, Japan, Mexico, Philippines, South Africa, South Korea, Taiwan by Dupont. The transgenic varieties of soybean rich in oleic acid were released by Monsanto, *viz*., Vistive Gold™ (MON87705) in Australia, Columbia, Canada, European Union, Indonesia, Japan, Mexico, New Zealand, Philippines, Singapore, South Korea, Taiwan, USA, Vietnam; MON87705 × MON87708 × MON89788 and MON 87705 × MON 87708 × MON 89788 in Canada. The soybean variety rich in oleic acid and linoleic acid was released in the European Union, Mexico, South Korea, and Taiwan. The other varieties rich in STA *viz*., MON 87769 × MON 89788 were released in Mexico, South Korea, Taiwan and MON87769 released in Australia, Columbia, Canada, European Union, Indonesia, Japan, Mexico, New Zealand, Philippines, South Korea, Taiwan, USA, Vietnam by Monsanto company.

### Transgenic Common Beans (*Phaseolus vulgaris*)

The common bean is among the most important grain legumes used for human consumption. However, although beans are rich in some essential amino acids, e.g., lysine, threonine, valine, isoleucine, and leucine, their nutritional value is limited because of the small amounts of the essential amino acid methionine and cysteine. Common bean methionine content has been increased by the expression of methionine-rich storage albumin from Brazil nut (125).

### Transgenic Lupines (*Lupinus angustifolius*)

Lupine is the major grain legume. The lupine seed protein, in common with the protein of most other grain legumes, is deficient in the sulfur-containing amino acids methionine and cysteine. Its methionine content has been increased by the expression of sunflower seed albumin gene (126).

## TRANSGENIC VEGETABLES

### Transgenic Potato (*Solanum tuberosum*)

Potato is the world's fourth most important source of calories, and it's any nutritional enhancement is of great significance. In potato tuber, provitamin A (carotenoid forms) have been increased by incorporating *PSY* gene (127) and by simultaneous incorporation of three genes: *PSY*, phytoene desaturase, and lycopene β-cyclase (128). Beta-carotene content in tubers has been also enhanced by using RNAi to silence the beta-carotene hydroxylase gene (bch), which converts beta-carotene to zeaxanthin (129) and by regulation of beta-carotene synthesis through expression of lycopene β-cyclase [*StLCYb* (130)]. In another experiment, it has been observed that incorporation of *Or* gene from orange cauliflower mutant leads to increase in carotenoids along with three additional metabolite intermediates phytoene, phytofluene, and z-carotene (131). Zeaxanthin which is another form of carotenoid has been also increased by expressing zeaxanthin epoxidase genes in transgenic potato tuber (132).

The potato has been also targeted for enhancement of vitamin C (ascorbic acid) by overexpressing strawberry *GalUR* (133). Potato tubers are very poor in essential amino acid, methionine, which has been targeted for its enhancement by coexpressing cystathionine γ-synthase (*CgSΔ*90) and methionine-rich storage protein (134). Similarly, silencing of *StMGL1* (135) and antisense inhibition of threonine synthase (136) led to increase in methionine to isoleucine ratio and methionine content (up to 239-folds) in potato tubers. Methionine content has been also enhanced by overexpressing the gene encoding the seed storage protein from *Perilla* [PrLeg polypeptide (137)] and cystathionine γ-synthase (*CgS*) genes (138). Transgenic potatoes expressing Amaranth albumin (*ama1*) result in an increase in total protein content in tubers along with the significant increase in the concentration of several essential amino acids including methionine (139).

High value carbohydrate rich potato tubers has been synthesized by expressing cyclodextrin glycosyltransferases (*CGT*) gene, which results in the production of multipurpose dietary fiber cyclodextrins from starch (140). Potato tubers have been also focused upon to increase the phenolic acid, and anthocyanins contents by the single-gene overexpression or by simultaneous expression of *CHS*, chalcone isomerase (*CHI*), and dihydroflavonol reductase (141). It has been also targeted to improve the content of dietary fiber fructan and inulin (142, 143). Transgenic potato varieties engineered for starch quality, which has reduced amylose and increased amylopectin in starch granules were released by BASF *viz*., Starch Potato (AM 04—1020) in the USA and Amflora™ (EH 92-527-1) in the European Union. Transgenic potato varieties that limit formation of the reducing sugars through starch degradation have been released in Canada and USA by J. R. Simplot Co.

### Transgenic Sweet Potato (*Ipomea batatas*)

Sweet potato is an alternative source of bioenergy and natural antioxidants. It is rich in various phytochemicals, anthocyanins, vitamin C, carbohydrates, potassium, and dietary fiber (197). Its nutrition properties have been further enhanced by increasing the contents of carotene, lutein, and total carotenoids by overexpressing orange *IbOr-Ins* gene in white fleshed sweet potato (144). The antioxidant capacity of orange-fleshed sweet potato cultivar has been increased by overexpression of *IbMYB1* a key regulator of anthocyanin biosynthesis in the storage roots (145).

### Transgenic Cassava (*Manihot esculenta*)

Cassava is an important staple food crop for millions of poor people worldwide as it is tolerant to different stresses. However, cassava is deficient in several important nutrients like provitamin A, vitamin E, iron, and zinc. Cassava biofortification of provitamin A, iron, and zinc has been carried out to reduce their deficiency among the undernourished communities. Telengech et al. (146) as a part of the BioCassava Plus project developed transgenic cassava that expresses beta-carotene in roots using *npt*II, *crtB*, and *DXS*. Similarly, Welsch et al. (147) showed that the cassava plants overexpressing a *PSY* transgene produced yellow-fleshed, high-carotenoid roots. Different transgenic cassava varieties biofortified for enhanced levels of iron, beta-carotene, and zinc are under development and field trials in the Biocassava Plus Program targeted at African countries.

### Transgenic Carrot (*Daucus carota* subsp. *sativus*)

Carrots are one of the most popular vegetables and contain high levels of beta-carotene and vitamins and minerals; however, like many vegetables, these are poor in calcium content (198). Bioavailable calcium content in transgenic carrot has been increased by expressing the *Arabidopsis* H<sup>+</sup>/Ca2<sup>+</sup> transporter [CAX1 (148, 149)].

### Transgenic Lettuce (*Lactuca sativa*)

Lettuce is one of the most popular leafy vegetables all around the world. Compared to spinach, the iron content of lettuce is low. The lettuce has been improved for iron content, yield, and growth rate by expressing a soybean ferritin gene (150).

### Transgenic Cauliflower (*Brassica oleracea*)

Cauliflower is a popular vegetable in several parts of the world. It is rich in antioxidant phytonutrients. Its nutritional value has been further enhanced by increasing beta-carotene content in mutant orange cauliflower by the insertion of a copia-like LTR retrotransponson in the *Or* (151).

## TRANSGENIC OILSEEDS

### Transgenic Linseed (*Linum usitatissimum*)

Linseed edible oil is in demand as a nutritional supplement. Linseed or flax seeds are the richest source of polyunsaturated fatty acids, but linseed oil is highly susceptible to auto-oxidation, which generates toxic derivatives. Genetically modified flax plants with increased antioxidant potential, stable, and healthy oil production has been generated by suppressing *CHS* gene that resulted in hydrolyzable tannin accumulation (152). Very longchain unsaturated fatty acids (VLCPUFA) are important fatty acids with limited supply due to decrease in marine resources such as fish oils. It can be compensated by implementation of VLCPUFA biosynthesis into oilseed crops (153). VLCPUFA such as arachidonic acid (C20:4 n-6), eicosapentenoic acid (EPA C20:5 n-3), and docosahexenoic acid (DHA C22:5 n-3) are considered to be nutritionally beneficial because of their function as cholesterol-lowering agents (199). Researchers have intended to enhance the accumulation of Δ<sup>6</sup> desaturated C18 fatty acids and C20 polyunsaturated fatty acids, including arachidonic and eicosapentaenoic acid by seed-specific expression of cDNAs encoding fatty acyl-desaturases and elongases in linseed (154). Enrichment of carotenoids in flaxseed has been done by the introduction of *PSY* gene [*crtB* (155)]. Transgenic linseed rich in essential amino acids *viz*., CDC Triffid Flax (FP967) has been released by University of Saskatchewan, in Colombia, USA, and Canada.

### Transgenic Canola (*Brassica napus*)

Canola is an important oilseed crop for millions of people around the world. Canola produces edible oil lower in saturated fat and higher in omega-3 fatty acids. To further enhance its health benefits its carotenoid content (mainly alpha and beta-carotenes) has been increased by overexpressing bacterial *PSY* [crtB (37)]. Higher β-carotenoid content has been achieved by simultaneous expression of *PSY*, phytoene desaturase, and lycopene cyclase genes (155) and simultaneous expression of seven bacterial genes; *idi*, *crtE*, *crtB*, *crtI*, *crtY*, *crtW*, and *crtZ* (157). Higher beta-carotene content along with high xanthophylls and lutein contents have been achieved by RNAi silencing of lycopene ε-cyclase [ε-*CYC* (158)] and DET1 (159). Essential amino acid lysine has been increased in canola by expression of aspartokinase (AK) and dihydrodipicolinic acid synthase (DHDPS) genes (160). Increase in level of two fatty acids *viz*., caprylate (8:0) and caprate (10:0) in canola seed oil accompanied by a preferential decrease in the levels of linoleate (18:2) and linolenate (18:3) has been achieved by overexpression of thioesterase gene [Ch FatB2 (161)]. Canola normally does not have any Δ<sup>6</sup> desaturase activity and thus lack GLA. In order to produce GLA more economically and to make it more readily available transgenic lines rich in GLA has been developed by expression of Δ12 or Δ<sup>6</sup> desaturases genes (162, 163). Phytic acid is known as a food inhibitor, which chelates micronutrient and prevents its bioavailability, as human and other monogastic animals lack the phytase enzyme in their digestive track. Transgenic canola varieties *viz*., Phytaseed™ Canola (MPS 961-965) engineered for phytase degradation to enhance the availability of phosphorus in canola has been produced and released by BASF in USA.

### Transgenic Mustard (*Brassica juncea*)

Mustard is an economically significant crop and extensively cultivated for oil throughout the world. It has been targeted for improving the nutritionally important unsaturated fatty acids. This has been achieved by the expression of the enzyme Δ<sup>6</sup> FAD3 that led to the production of gamma linoleic acid in the transgenic mustard (164).

## TRANSGENIC FRUITS

### Transgenic Tomato (*Solanum lycopersicum*)

Tomato is one of the most popular fruits, consumed by billions around the world and is an important source of vitamin C, micronutrients, and other phytonutrients. It derives its color from isopernoid lycopene. Isoprenoids are one of the largest classes of natural products with several thousand compounds. In higher plants, isoprenoids have essential roles in membrane structure (sterols), free radical scavenging (carotenoids and tocopherols), redox chemistry (plastoquinone, ubiquinone), defense mechanisms (phytoalexins), and growth regulation (gibberellins, cytokinins, brassinosteroids, and abscisic acid) (200). Several attempts have been made to increase the isoprenoid content in tomato. The sterol content was elevated in tomato by expression of 3-hydroxymethylglutaryl CoA [*hmgr-1* (165)]. Tomato phytoene and beta-carotene content has been enhanced by expression of 1-deoxy-d-xylulose-5-phosphate synthase [*dxs* (165)]. Higher contents of lycopene, betacarotene, and lutein have also been achieved in tomato by the expression of *PSY* gene [*crtB* (166)]. Double biofortification of carotenoid and flavonoid contents have also been achieved by RNAi technology by suppressing photomorphogenesis regulatory gene [*DET1* (172)]. The beta-carotene content has also been increased by overexpression of lycopene beta-cyclase gene [*beta-Lcy* (167–169)]. Higher contents beta-carotene as well as its hydroxylation product xanthophylls (beta-cryptoxanthin and zeaxanthin) has been obtained by simultaneous expression of *beta-Lcy* and beta-carotene hydroxylase [b-Chy (171)]. Total carotenoid and high value astaxanthin content (hydroxylation product of a beta-carotene) have been enhanced in tomato by expression of beta-carotene ketolase and hydroxylase (170). The tomato has been targeted to improve its vitamin C (ascorbic acid) content by overexpressing GDP-mannose 3′,5′-epimerase [*SlGME1*, *SlGME2* (173)], DHAR (174), and coexpression of three genes GDP-mannose pyrophosphorylase, arabinono-1,4-lactone oxidase, and myo-inositol oxygenase 2 (88, 175). Another important nutrient folic acid has been targeted by overexpression of GTPCHI (176) and aminodeoxychorismate synthase (177).

Tomato has also been selected to increase antioxidant anthocyanins by expression of *CHI* (178), transcriptional activators *AtMYB75* (179), and expression of two transcription factors, *Delila* and *Rosea1* (201). Other antioxidants like chlorogenic acid have been targeted by gene silencing of HQT (180), transresveratrol by expression of stilbene synthase (181), polyphenolic antioxidants by expression of AtMYB12 (182), and genistin by overexpression of isoflavone synthase (IFS) gene (183). Anthocynin rich blue transgenic tomato has been developed by Norfolk plant sciences.

### Transgenic Apple (*Malus domestica*)

Apple has long been recognized as a great source of antioxidants. Apple has been bioengineered with a stilbene synthase gene from the grapevine (*Vitis vinifera* L.) thereby leading to synthesis of resveratrol in transgenic apple, thereby, expanding the antioxidant capacity (184).

### Transgenic Banana (*Musa acuminata*)

The banana, a fourth most important food crop of the developing countries, has been predominantly targeted for beta-carotene. This has been achieved by developing transgenic banana (Super Banana) by expressing *PSY* gene (*PSY2a*) of Asupina banana, which is naturally high in beta-carotene (185).

### TRANSGENIC FODDER

### Transgenic alfalfa (*Medicago sativa*)

Alfalfa is as an important feed legume crop in many countries. Attempts have been made to improve its nutritional status through enhancement of isoflavonoids, essential amino acids, and improve its digestibility. Isoflavonoids are a predominantly legume-specific subclass of flavonoid secondary metabolites. Transgenic alfalfa has been generated by constitutively expressing IFS that is correlated with its increased isoflavonid composition (186). Alfalfa suffers from a limited level of the sulfur-containing amino acids, methionine, and cysteine. Its methionine content has been increased by the expression of cystathionine γ-synthase [*AtCgS* (187)]. Improvement in the digestibility of forages has also been an area of interest as it correlates with animal performance. By targeting three specific cytochrome P450 enzymes for antisense downregulation, transgenic alfalfa lines have been generated with low lignin content (188). Alfalfa has also been engineered to increase phytase activity, and thereby enabling its use in animal feeds, including livestock, poultry, and fish feed (189).

### BIOFORTIFICATION THROUGH AGRONOMIC APPROACHES

Biofortification through agronomic methods requires physical application of nutrients to temporarily improve the nutritional and health status of crops and consumption of such crops improves the human nutritional status (202). In comparison with inorganic forms of minerals, the organic ones are more available for a man, as they can be absorbed more easily; and are less excreted (203) and their toxicity symptoms are less intensive (DRI 2000). It generally relies on the application of mineral fertilizers and/or increase in their solubilization and/or mobilization from the soil in the edible parts of plants. Macrominerals like nitrogen, phosphorus, and potassium (NPK) make an important contribution to the attainment of higher crop yields (204). Through the application of NPK-containing fertilizers, agricultural productivity increased in many countries of the world in the late 1960s and resulted in Green Revolution and saved them from starvation. In the current scenario, these fertilizers are important and necessary to improve crop yield and save the human population from starvation as low-input agriculture cannot feed the current seven billion world population (205). Microminerals iron, zinc, copper, manganese, I, Se, Mo, Co, and Ni are found in varying degrees in the edible portion of certain plants and are usually absorbed from the soil. Improvement of the soil micronutrient status by their application as fertilizers can contribute to decrease in micronutrient deficiency in humans (206). When crops are grown in soils, where mineral elements become immediately unavailable in the soil and/or not readily translocated to edible tissues targeted application of soluble inorganic fertilizers to the roots or to the leaves are practiced. Agronomic biofortification is simple and inexpensive, but needs special attention in terms of source of nutrient, application method and effects on the environment. These should be applied regularly in every crop season and thus are less cost-effective in some cases. Use of mineral fertilizers is evidently feasible in the developed world, as exemplified by the success of Se fertilization of crops in Finland (207), zinc fertilization in Turkey (208), and I fertilization in irrigation water in China (209).

In addition to fertilizers, plant growth-promoting soil microorganisms can be used to enhance the nutrient mobility from soil to edible parts of plants and improve their nutritional status. Soil microorganisms like different species of genera *Bacillus*, *Pseudomonas*, *Rhizobium*, *Azotobacter*, etc. can also be utilized to increase the phytoavailability of mineral elements (210, 211). The N2-fixing bacteria play important role in increasing crop productivity in nitrogen limited conditions (212). Many crops are associated with mycorrhizal fungi that can release organic acids, siderophores, and enzymes capable of degrading organic compounds and increasing mineral concentrations in edible produce (210, 213). Different crops have been targeted through agronomical biofortification to improve the human nutritional status (**Table 3**).

### CEREALS

### Rice Agronomic Biofortification

Micronutrient biofortification through agronomical practices is an alternative strategy to reduce the iron and zinc deficiency in rice grain. Biofortification of rice plants by foliar spray of iron was an effective way to promote iron concentration in rice grains (214–216). Similarly, fortifying germinating rice plantlets with ferrous sulfate lead to increase iron concentration in germinated brown rice [up to 15.6 times the control (215)]. Foliar application of zinc has been reported as an effective agronomic practice to promote rice grain zinc concentration and zinc bioavailability (216, 218–223). On the other hand, application of zinc to soil as fertilizer in addition to a foliar spray proves to be an important strategy to increase the grain zinc content of rice grown in soils with low background levels of zinc (224). Selenium, which is an essential trace element for human health and proved to be a potent antioxidant, has been also increased by the application of selenate as a foliar spray or as fertilizer in rice (216, 225–230).

### Wheat Agronomic Biofortification

Agronomic biofortification has been very efficiently utilized in wheat grain quality improvement. Inclusion of iron in foliar urea fertilizers has been positively correlated with high iron accumulation (231). Application of foliar zinc has reduced human zinc deficiency in regions with potentially zinc-deficient soil and also improved its bioavailability by reducing antinutrient factors like phytic acid (233). Due to significant effects of zinc fertilizers on grain yield, the total amount of zinc-containing NPK fertilizers increased from 0 in 1994 to a record level of 400,000 t per annum in 10–15 years in Turkey. Use of zinc-containing fertilizers increased zinc concentration in grain, and obviously contributed to human nutrition and health in Turkey, especially in rural areas, where wheat provided more than 50% of the daily calorie intake (206). Agronomic biofortification of Se in wheat has been adopted with success in Finland (207). Compound fertilizers supplemented Table 3 | Tabulation of crops, nutrients, research status, and concerned publications on biofortification through agronomic approaches.


(*Continued*)

#### TABLE 3 | Continued


*Physical application of nutrients, growth-promoting soil microorganisms, N2-fixing bacteria and mycorrhizal fungi are utilized to increase the mineral concentration in edible produce.*

with Se were utilized since 1984, and it resulted in an increase in human serum selenium. Apart from chemical and organic fertilizers, researchers have also investigated the role of biofertilizers in promoting the yield of grains. Mycorrhizal fungi along with fertilizers are extensively being used for biofortification (234). Iron biofortification of wheat grains has been accomplished through integrated use of organic and chemical fertilizers and zinc biofortification by using *Bacillus aryabhattai* (235, 236).

### Maize Agronomic Biofortification

Among micronutrients, zinc is required for obtaining nutrientenriched grain and optimum yield in maize. For achieving this, various zinc fertilizer treatments and foliar applications have been carried out in maize crop (237, 239–241). Plant growthpromoting rhizobacteria have led to nutrient enrichment in the plants and have been included in agronomic approaches to develop effective biofortification strategies for the staple crops. One of the effective examples is the maize crop with increased zinc content (242). The Selenium (Se) importance in human and animal health has been known worldwide, and it has also been increased by applying fertilization as an effective agronomic biofortification strategy (226).

### Barley Agronomic Biofortification

The micronutrient profile of barley has been improved by the application of various organic and inorganic biofertilizers. The concentration of zinc and iron in grains has been enhanced by the application of biofertilizers along with inorganic fertilizers and vermicompost (243).

### Sorghum Agronomic Biofortification

Sorghum is cultivated worldwide for grain and fodder. This crop often suffers from the challenge of growing in nutrient poor and contaminated soil. Its nutrient profile has been promoted by the application of fertilizers (both organic and inorganic) that have an additive effect on the yield. Researchers have intended to improve the nutrient uptake and alter the metabolic profile of sorghum by using the combination of plant growth-promoting bacteria and arbuscular mycorrhizal fungi (AMF) (244, 245). Also, the inoculation of *Azospirillum* alone and in combination with phosphate-solubilizing bacteria increased sorghum grain yield and protein content by improving the status of phosphorous and nitrogen in the soil (246).

## LEGUMES

### Soybean Agronomic Biofortification

Selenium-enriched soybean has been produced by the foliar application of selenium complex salts as fertilizers (247).

### Chickpea Agronomic Biofortification

Chickpea has been targeted for the mineral deficiencies, especially the mineral iron, zinc, calcium, copper, manganese, and Mg by using plant growth-promoting actinobacteria (248). Chickpea biofortification for iron and zinc has been addressed by using AMF (249). Similarly, zinc and Se have been fortified in chickpea by foliar spray of respective minerals (250, 251).

### Pea Agronomic Biofortification

Field peas are the second largest legume crop worldwide, also known for their high protein content and its enrichment for zinc has been obtained with foliar zinc applications alone or in combination with soil zinc applications (252).

### Common Bean Agronomic Biofortification

A common bean is an herbaceous annual plant grown for edible dry grain. Beans are a good vehicle for zinc biofortification and have been enriched with zinc by the application of foliar zinc fertilizer (223, 253). Furthermore, it has been studied that administration of organic and chemical fertilizers stimulated the uptake of N, P, K, copper, manganese, and zinc in common bean (254).

## OILSEEDS

### Canola Agronomic Biofortification

Canola supplemented with plant growth-promoting rhizobacteria *viz*. *Azospirillum brasilense*, *Azotobacter vinelandii* along with chemical fertilizers resulted in increased protein, oleic acid, and linoleic acid content in the seed which indicated that rhizobacteria are highly effective in improving yield and nutritive value of canola oil (263).

### Mustard Agronomic Biofortification

Mustard has been targeted for Se enhancement. Plant uptake of Se as selenate has been enhanced by rhizosphere bacteria from a seleniferous area (255).

## VEGETABLES

### Potato Agronomic Biofortification

Field experiments were undertaken to increase zinc concentrations in potato tubers (both flesh and skin of tubers) using foliar zinc fertilizers, which significantly increased tuber zinc concentrations. It was also found that zinc oxide and zinc sulfate were more effective than zinc nitrate as foliar fertilizers for increasing tuber zinc concentrations while maintaining yields (264). Increase in Se content of potato tubers has been reported after foliar application of selenium, selenite, and selenate to potato (256, 257). Foliar application of selenium with humic acids was proven to be a good way to increase the selenium content of potatoes (256).

### Sweet Potato Agronomic Biofortification

Increase in beta-carotene in orange-fleshed sweet potato has been observed with irrigation and chemical fertilizer treatments (258).

## Carrot Agronomic Biofortification

Carrot leaves and storage roots have been supplemented with I and Se by application of both as fertilizers. It has been reported that consumption of 100 g fresh weight of carrots fertilized with I and Se (KICNa2SeO3, KIO3CNa2SeO3) can supply 100% of the recommended daily allowance (259).

## Lettuce Agronomic Biofortification

Lettuce I and Se biofortification have been achieved by the application of KIO3 and Na2SeO4 as foliar spray and nutrient medium (260). Lettuce Se biofortification in the leaves has been carried out with good results after soil agronomic biofortification with an inorganic form of selenium (261).

## FRUIT

### Tomato Agronomic Biofortification

Studies have concluded that a tomato is an excellent crop for iodine biofortification programs when treated with iron fertilizers (262).

### Biofortification through Conventional Breeding—Most Trusted Approach

Biofortification through conventional breeding in the most accepted method of biofortification. It offers a sustainable, cost-effective alternative to transgenic- and agronomic-based strategies. Sufficient genotypic variation in the trait of interest is necessary for conventional breeding to be feasible. Breeding programs can utilize this variation to improve the levels of minerals and vitamins in crops. In conventional plant breeding, parent lines with high nutrients are crossed with recipient line with desirable agronomic traits over several generations to produce plants with desired nutrient and agronomic traits. However, breeding strategies have to sometimes rely on the limited genetic variation present in the gene pool. In some cases, this can be overcome by crossing to distant relatives and thus moving the trait slowly into the commercial cultivars. Alternatively, new traits can be introduced directly into commercial varieties by mutagenesis.

Because this approach is likely to be the most expedient method to improve plants, several international organizations have initiated programs to improve the nutritional content of crops through breeding programs. The Health grain Project (2005–2010) involving 44 partners from 15 countries and over £10 million was carried out in the European Union to develop health promoting and safe cereal foods and ingredients of high eating quality. It has since developed into the Healthgrain forum with a wide range of participants from academia and industry. More than 100 publications have reported bioactive compounds in whole-grain cereals, genetic variation, heritability, and effect on reducing risks of many lifestyle-related diseases (265–267). The CGIAR along with the International Center for Tropical Agriculture (CIAT) and the International Food Policy Research Institute have launched the HarvestPlus program to breed biofortified staple food crops. HarvestPlus is investing heavily to boost three key nutrients-vitamin A, iron, and zinc and is targeting the staple crops, wheat, rice, maize, cassava, pearl millet, beans, and sweet potato in Asia and Africa (268). It is directed to produce staple food crops with enhanced levels of bioavailable essential minerals and vitamins that will have measurable impact on improving the micronutrient status of target populations, primarily resource-poor people in the developing world. The Biocassava Plus program had been initiated to improve the nutrition status of cassava crop. Due to better acceptability, large numbers of crops have been targeted for biofortification through crop breeding (**Table 4**).

## CEREALS

### Rice Breeding

Rice is greatly emphasized for micronutrient enhancement. It is one of the most consumed staple food crop and its biofortification can have a significant effect on malnutrition challenge. The milled rice is poor source of minerals. Different old rice varieties with high iron and zinc content in grain have been screened and the higher mineral trait has been combined with improved agronomic traits by breeding methods. The world's first zinc enriched rice varieties developed by HarvestPlus were released in 2013 by Table 4 | Tabulation of crops, nutrients, research status, and concerned publications on biofortification through breeding.


(*Continued*)


(*Continued*)

#### TABLE 4 | Continued


*Breeding is so far the best method for crop biofortification. Large number of biofortified cultivars have been released by this approach that are helping in addressing the challenge of micronutrient malnutrition prevalent in the developing countries.*

*Released varieties and their country of release have been bold faced.*

the Bangladesh Rice Research Institute (BRRIdhan 62, BRRIdhan 72, and BRRIdhan 64), which is claimed to contain 20–22 ppm zinc in brown rice. In India and Philippines, an improved line (IR68144-3B-2-2-3) was identified in a cross between a highyielding variety (IR72) and a tall, traditional variety (Zawa Bonday) with a high concentration of grain iron [about 21 ppm in brown rice (269)]. Similarly, Jalmagna, a traditional variety which had almost double the iron concentration of common rice variety and zinc concentration, nearly 40% more than that of common rice variety has been identified for further breeding programs to improve iron and zinc concentration (269).

### Wheat Breeding

Wheat as a staple crop is the first and foremost target for biofortification. Wide variation in grain iron and zinc concentrations in wheat and its closely related wild species has been observed that it can be exploited for improvement of modern elite cultivars (270, 272, 297). Utilizing this variation HarvestPlus has released several varieties of wheat with 4–10 ppm higher zinc content. Six varieties of high zinc wheat (BHU 1, BHU 3, BHU 5, BHU 6, BHU 7, and BHU 18) were released in India in 2014 followed by the release of four varieties in Pakistan in 2015 (NR 419, 42, 421, and Zincol). Two varieties BHU 1 and BHU 6 have high yield, disease resistance in addition to high zinc. Recently, variety with high zinc (PBW1Zn) has been released by Punjab Agricultural University, India. Another variety with high zinc and iron content (WB2) has been developed and released by Indian Institute of Wheat and Barley Research, India. Apart from releasing cultivars, several researchers have reported an increase in the zinc and iron content of wheat by plant breeding (208, 270–272). Provitamin A has been another important nutrient targeted for biofortification through breeding. High provitamin A durum wheat variety (HI 8627) has been released by the Indian Agricultural Research Institute (IARI), India in 2005. Several new cultivars have been released after that with the improved beta-carotene content. Yellow pigment content (YPC; carotenoids mainly xanthophyll lutein) in durum wheat is an important quality trait and an antioxidant. A large number of recent durum wheat varieties released in different countries in the past decade show significantly higher YPC than the old varieties released before the 1970s [(273, 274) and others]. Improvement of antioxidant properties contributed by anthocyanins had also been an area of significant research in wheat. Colored wheat (black, blue, and purple) trait has been used in several breeding programs in different countries. Blackgrained wheat cultivar has been released in China after more than 20 years running effort in breeding and has been reported to be high in protein content and selenium (298). The purple wheat cultivar Indigo has been released in Austria in 2006 (299). The purple wheat cultivar PS Karkulka has been registered in Slovakia in 2014. Purple, blue, and black white lines have been developed and registered in India in 2017 (275). The importance of colored wheat can be adjudged from the patent on functional foods from colored wheat in China (CN102217664 B). Apart from this several researchers have worked on different aspects of colored wheat [reviewed in Ref. (276, 277)].

### Maize Breeding

Maize is a cash crop grown for animal feed, industrial purposes (source of sugar, oil, starch, and ethanol) and for use for human consumption. The vast genetic diversity of maize has been the basis for the breeding programs that have generated much of the higher yielding maize used worldwide. Scientists have discovered varieties that have naturally high levels of provitamin A. HarvestPlus is using these lines to breed high-yielding varieties of biofortified maize with higher levels of provitamin A to combat vitamin A deficiency. The provitamin A maize is one of the significant achievements in the field of biofortification. Biofortified orange maize varieties have been grown commercially in Zambia (GV662A, GV664A, and GV665A), Nigeria {Ife maizehyb-3, Ife maizehyb-4, Sammaz 38 (OPV), Sammaz 39 (OPV)} and Ghana {CSIR-CRI Honampa (OPV)} since 2013 (300). Malawi, Zimbabwe (ZS242) and Tanzania have also released biofortified orange maize recently (301). As a positive effect an increase in pupillary response was observed among Zambian children consuming vitamin A biofortified maize (301). Breeders have evaluated antioxidants like tocochromanols, oryzanol, and phenolic compounds in proVA biofortified maize (279). Another significant achievement in the field of maize biofortification is quality protein maize (QPM). Maize breeders have developed QPM with high essential amino acids lysine and tryptophan by incorporating opaque-2 (o2) mutant gene from naturally occurring maize into the maize cultivars. International Maize and Wheat Improvement Center (CIMMYT) has released such hybrid varieties in India (CML176, CML176 × CML186, HQPM4, HQPM-7, VivekQPM-9, HQPM-5, HQPM-1, FQH-4567), China (CML140, CML194, P70), Vietnam (CML161 × CML165), Mexico (CML142 × CML176, CML142 × CML150, CML176, CML170, CML186 × CML149, CML176 × CML186), South Africa (QS-7705), Ghana (GH-132-28), Guinea (Obatampa), Uganda (Obangaina), Benin (Obangaina), Mozambique (Susuma), Brazil (BR-451, BR-473), Venezuela (FONAIAP), Peru (INIA), Colombia (ICA), Honduras (HQ-31), El Salvador (HQ-61), Guatemala (HB-Proticta), and Nicaragua (NB-Nutrinta, HQ INTA-993). For QPM maize breeders, Surinder Vasal and Evangelina Villegas won 2000 world food prize. Maize has also been inbred by recurrent selection scheme, to increase the carotenoids (278) alone or in combination of vitamin E and phenolics (279) and antioxidant power (280). Attempts have been made to increase its vitamin E content (281).

## Sorghum Breeding

The prospects of breeding for micronutrients and beta-carotene rich sorghums have been discussed by Reddy et al. (282). Sorghum varieties have been screened for high minerals, protein (302), lutein, zeaxanthin, and beta-carotene contents (303). Sorghum germplasm has shown large variability and genetic heritability for iron and zinc content (304). Biofortified iron rich sorghum lines (ICSR 14001, ICSH 14002) and hybrids (ICSA 661 × ICSR 196, ICSA 318 × ICSR 94, ICSA 336 × IS 3760) have been bred by ICRISAT and released in India.

New nutritionally high (Fe) sorghum varieties (12KNICSV-22 and 12KNICSV-188) have been released in Nigeria that may boost the malnourished populations, especially children in Nigeria. One of the new varieties (12KNICSV-188) has iron content three times higher than typically grown sorghum. These new varieties involved crossing local Nigerian germplasm with improved lines from ICRISAT (Mali).

### Millets Breeding

Pearl millet is the cheapest source of iron and zinc (305) and large variation has been seen in its germplasm for these micronutrients (283). In India, biofortified (iron and zinc) pearl millet variety "Dhanashakti" and a hybrid ICMH 1201 (Shakti-1201) has been released by ICRISAT, HarvestPlus in 2014. Besides that, two varieties, ICMH 1202 (Nirmal-7) and ICMH 1301, are currently undergoing advanced farm trials. Various well-adapted commercial varieties, their progenies, and hybrids containing high content of iron and zinc in grain have been reported (283, 284).

### LEGUMES AND PULSES

### Lentil Breeding

Lentil, a key pulse in many dryland countries and has easy to cook properties. It has been directed by ICARDA, HarvestPlus for biofortification of iron and zinc with the help of breeding process using genetic diversity stored in gene banks. Research findings have shown that there is a positive correlation of iron and zinc synthesis with protein synthesis, therefore lentil varieties with higher iron, zinc, and protein content can be developed together [ICARDA, HarvestPlus (306)]. High iron and zinc lentil varieties, five in Bangladesh (Barimasur-4, Barimasur-5, Barimasur-6, Barimasur-7, and Barimasur-8), seven in Nepal (ILL 7723, Khajurah-1, Khajurah-2, Shital, Sisir Shekhar, Simal), two in India (L4704, Pusa Vaibhav), one in Ethiopia (Alemaya), and two in Syria (Idlib-2, Idlib-3) has been released by ICARDA, HarvestPlus biofortification program till date. Lentil varieties have been screened for variation in Se content (307).

## Cow Pea Breeding

Cow pea which is also known as poor man meat, rich in protein content has been biofortified for iron content by means of breeding methods. Pant Lobia-1 (2008), Pant Lobia-2 (2010), Pant Lobia-3 (2013), and Pant Lobia-4 (2014) varieties with increased iron content have been released by GB Pant University, Pantnagar, India in collaboration to HarvestPlus.

## Bean Breeding

Studies till date suggest that the iron content of the common bean (*P. vulgaris*) could be increased by 60–80%, while zinc content would be more modest, perhaps around 50%. High heritability has been observed in iron and zinc content in common bean (285, 287, 308). Genes associated with zinc accumulation have been identified in navy bean (286). HarvestPlus is working in this direction and promoting iron biofortified beans in several developing countries. They have released 10 Fe-biofortified common bean varieties in Rwanda (RWR 2245, RWR 2154, MAC 42, MAC 44, CAB 2, RWV 1129, RWV 3006, RWV 3316, RWV 3317, and RWV 2887). HarvestPlus also released ten biofortified iron bean varieties in the Democratic Republic of Congo, i.e., COD MLB 001, COD MLB 032, HM 21-7, RWR 2245, PVA 1438, COD MLV 059, VCB 81013, Nain de Kyondo, Cuarentino, Namulenga.

## VEGETABLES

### Potato Breeding

Potato tubers are the richest sources of antioxidants in human diet. The natural variation of cultivated potato germplasm containing red and purple pigment could possibly represent the contribution of the potatoes to the portion of antioxidants in human nutrition. Therefore, effort of breeders focuses on the breeding of such variants (288). Furthermore, vast genetic variation for micronutrients (291) exists in potato that can be exploited for breeding to further increase iron and zinc levels in human diets (290). A genetically diverse sample of potato cultivars native to the Andes of South America has been obtained from a collection of nearly 1,000 genotypes and evaluated as a source of antioxidants and minerals (copper, iron, manganese, and zinc) (289, 292). International potato center (CIP) and HarvestPlus have developed high iron and zinc advanced breeding material after crossing diploid Andean landrace potatoes with high zinc and iron with disease resistant tetraploid clones. The main target countries for biofortified potato are Rwanda and Ethiopia. National Institute for Agrarian Innovation's (INIA) Potato Program has developed the INIA 321 Kawsay variety in Peru that has a high content of iron and zinc.

### Sweet Potato Breeding

Developing countries are growing 95% of the world's sweet potato crop, where malnutrition is the biggest problem. The sweet potato has been targeted for improvement in vitamin A. HarvestPlus and International Potato Centre (CIP) have developed and released several varieties of orange sweet potato with high vitamin A. Six varieties have been released in Uganda (Ejumula, Kakamega, Vita, Kabode, Naspot 12O, and Naspot 13O) and three in Zambia (Twatasha, Kokota, and Chiwoko). Zambia Agriculture Research Institute has successfully completed the development of 15 new varieties of vitamin A fortified sweet potatoes. The HarvestPlus orange sweet potato consumption had a significant effect on household food and nutritional security in Sub Saharan Africa, and for this contribution; they have been recently honored with World Food Prize-2016. Furthermore, researchers have identified several sweet potato genotypes that completely lack or have only traces of β-amylase in their storage roots. Such verities could facilitate the breeding of sweet potato for low β-amylase content which can be potentially used for processing and as a staple food (293).

### Cauliflower Breeding

*Brassica oleracea* including cauliflower gene pool has been screened for genetic variation of zinc concentration and sufficient natural variation has been identified (309). The provitamin A (beta-carotene) rich orange colored cauliflower variety (Pusa BetaKesari; 800–1,000 μg/100g) has been released by the Indian Agricultural Research Institute (IARI). Now numbers of colored cauliflower verities are known at world level, having orange and purple color rich in beta-carotene and anthocyanin, respectively. Colored cauliflower varieties, Purple Graffiti and Orange Cheddar, have been developed by Cornell University, USA.

### Cassava Breeding

Cassava is a staple vegetable root crop in developing countries, especially in Africa, Latin America, and the Caribbean. In the African continent, it has been targeted for alleviation in provitamin A (beta-carotene) by HarvestPlus in collaboration with International Institute of Tropical Agriculture. Under these collaborations, they have released six vitamin A fortified varieties in Nigeria (2011; TMS 01/1368—UMUCASS 36, TMS 01/1412— UMUCASS 37 and 2014; TMS 01/1371—UMUCASS 38 and NR 07/0220—UMUCASS 44, TMS 07/0593—UMUCASS 45, and TMS 07/539—UMUCASS 46) and one in DRC-Democratic Republic of Congo [Kindisa (TMS 2001/1661)]. Cassava also has a wide range of genotype differences for total carotene, proteins, and minerals (iron and zinc) which has led to the development of improved nutritive value cassava crop (294, 295).

### FRUITS

### Tomato Breeding

Tomato is a highly valuable crop and an important source of vitamin A and C. Genetically diverse wild population of tomato has been investigated intensively for specific traits and exploited in tomato breeding (310). Anthocyanin biofortified tomato "Sun Black" with deep purple fruit pigmentation due to high anthocyanin content in the peel has been developed by conventional breeding approach (296). Another variety "Black Galaxy" generated by similar approach has been reported from Israel.

### Banana Breeding

Breeding banana is difficult and expensive, as commercial varieties are sterile triploids (3×) and also a high degree of cross incompatibility can exist among the fertile groups. For combating this problem, large scale screening of several banana germplasm for the identification of high levels of provitamin A has been carried out in the Democratic Republic of Congo (DRC) and Burundi by Biodiversity International (BI) in collaboration with HarvestPlus. In this program, they released five varieties (Apantu, Bira, Pelipita, Lai, and To'o) rich in provitamin A in Eastern DRC and Burundi.

### Mango Breeding

Mango offers a natural source of beta-carotene, vitamin C, and valuable antioxidants but their nutrient levels vary with mango variety. It has been observed that most of the mango varieties provide more than recommended daily value of vitamin C and beta-carotene. Mango also contains a variety of phenolics like ellagic acid, gallotannin, and mangiferin (311). The Mexicangrown Ataulfo variety ranked highest in both vitamin C (ascorbic acid) and beta-carotene (USDA's Agricultural Research Service). In India, IARI introduced many varieties with enhanced nutritional and agronomical important characters.

### Grape Breeding

Grapes have high mineral content, including high vitamins C and K, and are a natural source of antioxidants and other polyphenols, and offer a variety of additional health benefits. Phenolic compounds and antioxidant properties of different grape cultivars grown in China have been assessed (312). The Indian Agricultural Institute has released an improved variety, i.e., Pusa Navrang which contains higher amount of total soluble solids (carbohydrates, organic acids, proteins, fats, and minerals) and antioxidants.

### LIMITATIONS OF BIOFORTIFICATION

### Limitations in Agronomic Biofortification

Application of fertilizers fortified with micronutrients is the simplest method among all biofortification methods. But the success of agronomical biofortification is highly variable due to the differences in mineral mobility, mineral accumulation among plant species, soil compositions in the specific geographical location of each crop (313). For example, a study involving diverse rice genotypes indicated that, in the phosphate deficient soils due to reduction in the root biomass, differences in the phosphate uptake among the genotypes were as high as 20-fold (314). Soil composition analysis has indicated that almost 1/2 of the agricultural soils of India, 1/3 of China, 14 Mha of Turkey, 8 Mha of Australia are zinc deficient (315). Agronomic biofortification is less cost-effective and labor intensive as it demands continuous inputs, through the application of micronutrient to the soil or plant regularly. Furthermore, it is not always possible to target the micronutrient into edible plant parts like seed or fruit and can sometimes result in the accumulation of desired nutrients in the leaves or other non-edible portions of plants; therefore, this technique is only successful in certain minerals and specific plant species. For instance, higher zinc efficiency in cereals grown in zinc deficient soils in Turkey was associated with higher uptake of zinc from the soil, but not with increased accumulation of zinc in the grain (208). Furthermore, mineral bioavailability hindered by antinutrient compound like phytic acid is another major challenge (316). In addition, the biggest of all constraints is that the fertilizers accumulation in soil and water poses adverse environmental effects (317).

### Limitations in Conventional Breeding Methods

The design of conventional plant breeding programs to improve micronutrient content has proved to be successful and is a sustainable and cost-effective solution in the long run; however, there are limitations with respect to the amount of genetic variability for the micronutrients in the plant gene pool and the time needed to generate cultivars with the desired trait(s). In some cases, this can be overcome by crossing to distant relatives and thus introgressing traits into commercial cultivars, but in many occasions, it would be impossible to breed for a specific trait using conventional means, and the timescale and effort involved may be quite unrealistic, e.g., improving Se concentration in wheat grains (318) and improvement of oleic, linoleic, and linolenic fatty acid content in soybean (319). In general, improvement in oil quality has been targeted with better results with transgenic-based approach (**Figure 3B**) due to limited variability, heritability, and linkage drag.

### Limitations in Transgenic Methods

Transgenic crops overcome the limitation of restricted genetic variation among plants as in the case of conventional breeding but the major limitation of this method is its low acceptance among masses. It is very important that the biofortified crops be readily adapted by farmers and community in significant enough numbers to improve the general nutritional health of a given community (320). Another limitation is that different countries have adopted different regulatory processes for the acceptance and commercialization of these transgenic crops. Regrettably, the current political and economic landscape is not receptive to this technology (321). Furthermore, these regulatory processes are very expensive and time consuming (322). Let us take the example of Bt Brinjal. It has been initially developed by Mahyco, an Indian seed company. Unfortunately, it was not released in Indian because some of the scientists, farmers, and anti-GMO activists, raised concerns and a moratorium on its release was imposed, until further tests were conducted. However, four varieties of Bt Brinjal were given approval for commercial release in Bangladesh in 2013–2014. Although the research efforts devoted to the transgenic-based approach are quite higher compared with breeding based, its success rate in terms of cultivar release in very low (**Figure 3A**) due to time required from target trait and gene identification, modification, expression, and assessment of agronomical traits to understanding the possible effect on other life forms. For example, after 8 years project, the scientific details of the Golden rice were first published in Science in 2000 (41), and since then different groups, including International Rice Research Institute scientists are working on it, but Golden Rice is still not ready for farmers due to issues with its yield. Its dissemination is also being held back due to inability to get approval from Governments.

### Other Limitations

The postharvest processing of each crop must be considered to optimize biofortification strategies. For example, the seeds of many cereals are often consumed after milling or polishing. Although the concentrations of some essential mineral elements, such as Se and S, are highest in the embryo, others, such as iron, zinc, and copper, are highest in the bran (269, 317). Milling or polishing cereal seeds can, therefore, remove large quantities of minerals from the diet; the extent of these losses is genotype dependent (269). In addition, the presence of certain antinutrients in crops reduces the bioavailability of certain nutrients in crops. For examples, antinutrients like phytate, tannins, oxalate, fiber, and hemaglutinins reduce the bioavailability of minerals in human gut (20, 101). Furthermore, in the context of global environmental change, approaches for improving food production, improvements in a crop's ability to maintain yields with lower water supply and quality will be critical. In addition, numerous genes are involved in controlling the amount of a mineral element that is absorbed by roots, translocated to shoot, remobilized from vegetative tissues, and deposited in edible portions of seeds and grains in forms that are utilizable in persons consuming the crop (323, 324). Considerations must also include the micronutrient concentrations in the edible portions of crops, and the amount of nutrients that can be absorbed by the consumer, after processing and cooking (325).

## CONCLUSION

It is well established that biofortification is a promising, cost-effective, agricultural strategy for improving the nutritional status of malnourished populations throughout the world. Biofortification strategies based on crop breeding, targeted genetic manipulation, and/or the application of mineral fertilizers hold great potential for addressing mineral malnutrition in humans. The generation of biofortified food crops with improved nutrient contents such as increases in iron, zinc, Se, and provitamin A content are providing sufficient levels of these and other such micronutrients that are frequently lacking in the diets of the developing and developed world. International initiatives, such as the HarvestPlus program and national initiatives, are acting as pillars to achieve these targets. These efforts have delivered crops with the potential to increase both the amounts and bioavailability of essential mineral elements in human diets, especially in staple cereal crops like

### REFERENCES


wheat, maize, cassava, beans, sweet potatoes, and millets. But biofortification of crops is a challenging endeavor. To achieve this, collaboration between plant breeders, nutrition scientists, genetic engineers, and molecular biologists is essential. Traditional breeding approaches are finding widespread and easy acceptance and have been used to enhance the nutritional qualities of foods. Although a greater emphasis is being laid on transgenic means success rates of breeding based approaches are much higher as transgenically fortified crop plants have to face hurdles due to acceptance constraints among consumers and different expensive and time consuming regulatory approval processes, adopted by different countries. Besides these challenges, biofortified crops hold a very bright future as these have the potential to remove micronutrient malnutrition among billions of poor people, especially in the developing countries.

### AUTHOR CONTRIBUTIONS

MG and NG built the layout of the article, collected literature, and wrote the article. SS and PK collected literature and helped in manuscript writing. AK and VC edited it. PA assisted in reference management.

### ACKNOWLEDGMENTS

This work was supported by the National Agri-Food Biotechnology Institute Core grant for improvement of nutrition and processing quality, for which the authors are deeply indebted.


produces amylose-only starch granules. *BMC Plant Biol* (2012) 12(1):223. doi:10.1186/1471-2229-12-223


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2018 Garg, Sharma, Sharma, Kapoor, Kumar, Chunduri and Arora. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Phosphorus Alters Starch Morphology and Gene Expression Related to Starch Biosynthesis and Degradation in Wheat Grain

Runqi Zhang† , Cheng Li † , Kaiyong Fu, Chao Li and Chunyan Li\*

Xinjiang Production and Construction Group, The Key Laboratory of Oasis Eco-Agriculture, College of Agriculture, Shihezi University, Shihezi, China

### Edited by:

Huixia Shou, Zhejiang University, China

### Reviewed by:

Miroslav Nikolic, University of Belgrade, Serbia Rumen Ivanov, Heinrich Heine Universität Düsseldorf, Germany

\*Correspondence:

Chunyan Li lichunyan82@aliyun.com † These authors have contributed equally to this work.

### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 01 July 2017 Accepted: 22 December 2017 Published: 12 January 2018

### Citation:

Zhang R, Li C, Fu K, Li C and Li C (2018) Phosphorus Alters Starch Morphology and Gene Expression Related to Starch Biosynthesis and Degradation in Wheat Grain. Front. Plant Sci. 8:2252. doi: 10.3389/fpls.2017.02252 Phosphorus is an essential plant macronutrient which profoundly affects the yield and quality of wheat starch. In this study, scanning electron microscopy showed that P fertilizer amount (0, 46, and 92 kg P ha−<sup>1</sup> ) had no significant effect on the shape of starch granules in wheat (cv. Xindong 20) grain. However, confocal laser scanning microscopy with 3-(4-carboxybenzoyl) quinoline-2-carboxaldehyde and methanolic merbromin stains indicated that P amount influenced the microstructure of the starch granules. Starch granules from the 46 kg P ha−<sup>1</sup> treatment released significantly more reducing sugars than those from the 0 and 92 kg P ha−<sup>1</sup> treatments during digestion with alpha-amylase and amyloglucosidase digestion. Phosphorus application (especially the 46 kg P ha−<sup>1</sup> treatments) significantly increased the relative expression of genes related to starch synthesis (especially during early to mid-grain filling) and starch degradation (especially during mid- and late grain filling). Phosphorus application also increased the transcript abundance of amylase genes at the periphery of the endosperm. We propose that P application, especially the 46 kg P ha−<sup>1</sup> treatment, enhanced channels in wheat starch granules. These channels facilitated the transport of substances required for starch biosynthesis, thus increasing starch accumulation in wheat endosperm. These results provide insight into the potential mechanisms through which P influences the microstructure and biosynthesis of wheat starch.

### Keywords: biosynthesis, degradation, phosphorus, starch, wheat

### INTRODUCTION

Phosphorus (P) is one of the main limiting factors for plant growth in natural ecosystems. The application of P fertilizer is often essential for crop production (Buchanan et al., 2004). Plants generally take up soil P in its inorganic forms. However, 50–80% of the total P in agricultural soils exists as organic phosphate, which is biologically unavailable (Wang et al., 2009). Wheat is grown under P-poor as well as P-rich conditions. In their efforts to maximize field, farmers often over apply P fertilizer. The fertilization rates in some areas are several times greater than the amount required by wheat. This can create serious environmental problems (Tang et al., 2008; Wang et al., 2017). It should also be noted that rock phosphate is a non-renewable resource that is being depleted (Wang et al., 2017). More understanding is needed about how P affects wheat yield and quality.

The main component of wheat endosperm is starch, which accounts for ∼70% of grain dry weight. Many enzymes are involved in starch biosynthesis including granule-bound starch synthase (GBSS), adenosine diphosphate glucose pyrophosphorylase (AGPase), starch branching enzyme (SBE), starch debranching enzyme (DBE), and soluble starch synthase (SSS). The AGPase is activated by 3-phosphoglycerate and is inhibited by inorganic P in leaves. The activity of AGPase can be changed by altering the ratio of 3-phosphoglycerate and inorganic P, which in turn regulates starch synthesis (MacDonald and Strobel, 1970).

Starch molecules are deposited as semi-crystalline structures in starch granules (Tang et al., 2006). Wheat starch can be classified as either large A-type granules (>10µm diam.) or small B-type granules (<10µm diam.). The two types of granules are differentiated by their morphological and chemical characteristics (Yu et al., 2015).

Pores and radial, tube-like channels have been observed at the surfaces of starch granules in wheat (Kim and Huber, 2008), corn, and sorghum (Huber and Bemiller, 2000). It has been hypothesized that the pores are not just surface features but might be openings to channels that provide access to the interior of starch granules (Huber and Bemiller, 2000). Starch granule architecture suggests that the pores and channels may be loosely assembled zones in "defective" blocklets (Tang et al., 2006). Pores and channels within starch granules have important influence on granule accessibility to reagents. This means that pores and channels can influence the reactivity of starch when it is chemically modified for specific purposes in industry (Huber and BeMiller, 1997; Han et al., 2005). Enzymatic digestion is also greatly influenced by pores and channels within native starch granules. Maize starch, which contains pores and channels, is more susceptible to enzymatic hydrolysis than potato starch (Han et al., 2005), which does not have pores (Fannon et al., 1992a). It is possible that starch granule channels can be manipulated to improve digestibility or alter the chemical characteristics of starch (Han et al., 2005). However, much more information is needed about these channels and their biological origin.

Environmental conditions such as high temperature (Li et al., 2017) and drought stress (Li et al., 2015) can alter the pits and channels in wheat starch granules. Commuri and Jones (1999) postulated that pitting is induced by imbalances in starch hydrolase and synthase, resulting in premature autolysis. Based on the protein constituents of channels in normal maize starch granules, it has been proposed that the channels are remnants of amyloplast microtubules and may facilitate starch polymer and granule biosynthesis (Fannon et al., 2004; Benmoussa et al., 2010).

Previous research in our laboratory indicated that the P fertilizer can also significant affect the characteristics of wheat starch, especially the presence of "pin holes" within starch granules (Li et al., 2013). The objective of this study was to increase understanding about the influence of P fertilizer on starch biosynthesis and granule structure in wheat.

### MATERIALS AND METHODS

### Plant Material and Cultivation

The study was conducted at the Shihezi University Experimental Farm, Shihezi, China (44◦ 17′ N, 86◦ 03′ E) from October 2014 to June 2015. The soil at the site is classified as gray desert soil by Chinese scientists and a Calcaric Fluvisol according to the FAO. The 0–20 cm depth had the following characteristics: 63 mg kg−<sup>1</sup> available (mineral) N (potassium chloride extraction), 15 mg kg−<sup>1</sup> available P (Olsen), and 208 mg kg−<sup>1</sup> available K (ammonium acetate extraction).

Seeds of the winter wheat cultivar "Xindong 20" were supplied by the Agriculture College of Shihezi University. On the day of sowing, 75 kg ha−<sup>1</sup> urea (46% N, Sinopec Group) was applied to the soil. The plots were drip irrigated at 10–12 d intervals three times before winter and six times after winter dormancy. Urea was applied via drip irrigation at elongation (45 kg ha−<sup>1</sup> ), heading (75 kg ha−<sup>1</sup> ), and flowering (120 kg ha−<sup>1</sup> ).

The experiment used a randomized block design with three replications. Three P treatments were applied to the plots 160 d after sowing, when about 5% plants of the plants had turned green after dormancy. The P treatments were as follows: 0 kg P ha−<sup>1</sup> (control, abbreviated P0); 46 kg P ha−<sup>1</sup> (normal P, abbreviated NP); and 92 kg P ha−<sup>1</sup> (high P, abbreviated HP). The P fertilizer (triple superphosphate, 45% P) was applied in a 10-cm-deep band between each ridge. The plots were 2.4 × 3 m. Each plot was separated by a 50-cm-wide bare strip.

### Sampling

Grain samples were collected from the middle region of the wheat spikes at 7, 14, 21, 28, and 35 days post anthesis (DPA). The samples were collected from each plot and then pooled to form one sample per treatment. Three subsamples were removed from each composite sample. These subsamples were frozen in liquid N for 5 min and then stored at −80◦C for RNA extraction. Another three subsamples were fixed in 4% paraformaldehyde (pH 7.2–7.6) and 0.1% DEPC for histochemical analysis and in situ hybridization. The remaining grains were dried at 70◦C to constant weight and weighed. Starch granules were isolated from grain (not oven-dried) collected at 35 DPA.

### Isolation of Starch Granules

Starch granules were isolated using a modified version of the method used by Peng et al. (1999). The embryos were excised from mature wheat grains with a scalpel. The deembryonated grains were soaked overnight in deionized water at 4◦C. Subsequently, the grains were ground in a mortar and pestle with deionized water. The slurry was centrifuged (4,000 g, 10 min). The sediment was treated twice with 80% (w/v) CsCl. The starch was then washed three times with washing buffer (62.5 mmol L−<sup>1</sup> Tris-HCl, pH = 6.8; 10 mmol L−<sup>1</sup> EDTA; 4% w/v

**Abbreviations:** AGPase, adenosine diphosphate glucose pyrophosphorylase; AGP, gene encoding AGPase; AMY, gene encoding α-amylase; BAM, gene encoding β-amylase; CLSM, confocal laser scanning microscopy; DPA, days post anthesis; GBSS, granule-bound starch synthase; GBSS, gene encoding GBSS; HP, high phosphorus; ISO, gene encoding isoamylase; NP, normal phosphorus; P0, phosphorus application control; SBE, gene encoding starch branching enzyme; SEM, scanning electron microscopy; SS, gene encoding soluble starch synthase.

SDS), three times with deionized water, and finally three times with acetone. The starch samples were then air dried.

### Starch Granule Morphology

The starch granules were sprinkled onto double-sided conductive adhesive tape attached to aluminum stubs and then coated with gold-palladium (60:40) particles (20 nm particle size) using a sputter coater (Denton Vacuum-Moorestown, NJ, USA). The morphology of the starch granules was examined using a field emission scanning electron microscope (JEOL JFC-1600, Japan) at an accelerating voltage of 5–10 kV.

### Enzyme Assays

Total amylase activity was measured as described by Liu (2011). Briefly, fresh, de-embryonated grains (1 g) were ground in 8 mL deionized water. The mixture was placed at room temperature for 15–20 min to extract total amylase. After centrifugation (3,000 rpm, 10 min), 1 mL of soluble starch (1%, w/v) was added to the supernatant (crude enzyme extract) and incubated at 40◦C for 5 min. Subsequently, 2 mL of 3, 5-dinitrosalicylic acid reagent (1% w/v 3,5-dinitrosalicylic acid, 30% w/v potassium sodium tartrate, and 0.4 mol L−<sup>1</sup> NaOH) was added to the mixture before heating in a boiling water bath. Absorbance was measured at 540 nm using a spectrophotometer (Shanghai Precision Scientific Instrument Co., Ltd.722G, China). A standard curve was prepared using malt sugar solutions with concentrations of 0, 0.1, 0.3, 0.5, 0.7, 0.9, and 1.0 mg mL−<sup>1</sup> .

To measure α-amylase activity, β-amylase in the crude enzyme extract was inactivated by heating at 70◦C for 15 min. All of the other steps were the same as the assay of total amylase activity described above. Beta-amylase activity was calculated as the difference between total amylase and α-amylase activity.

### Treatment of Starch Granules with Proteolytic Enzyme

The isolated starch granules (2 g) were suspended in 40 mL of proteolytic buffer [50 mmol L−<sup>1</sup> sodium acetate (pH 7.5), 1 mmol L −1 calcium chloride, and 0.02% (w/v) sodium azide]. Next, Protease Type XIV [196 units (g starch)−<sup>1</sup> , a mixture of various proteases, Sigma, Lot#051M1894V, USA, source: streptomyces griseus] was added to the suspension before incubating at 4◦C for 24 h on a shaking table. After centrifuging at 3,000 g for 20 min, Protease Type XIV-treated starch granules were washed with deionized water and ethanol and then air dried.

### Treatment of Starch Granules with Merbromin and 3-(4-carboxybenzoyl) quinoline-2-carboxaldehyde (CBQCA)

Protease (Type XIV)-treated starch granules were stained with a methanolic solution of merbromin and 3-(4-carboxybenzoyl) quinoline-2-carboxaldehyde (CBQCA) according to Kim and Huber (2008). Merbromin is a non-reactive fluorescent dye that is absorbed onto the surface of the starch granules. Merbromin can be used to highlight external surface of the granules, including channels and cavities connected with the granule's exterior (Huber and BeMiller, 1997). The protein-specific dye CBQCA covalently reacts with primary amines of amino acids, peptides, and proteins under weak basic conditions. The dye, which is fluorescent only after reaction, can be used to highlight the protein network of radially oriented, channel-like structures within starch granules (Han et al., 2005).

### Confocal Laser Scanning Microscopy (CLSM)

After staining with merbromin and CBQCA, the starch samples were transferred to glass slides and then photographed using a Zeiss LSM 510 CLSM system (Zeiss, Oberkoche, Germany). Excitation was achieved with an Argonlaser (488 nm) operating at 30% power. Emission was detected through an LP505 emission filter.

### Detection of Wheat Starch Content

Grain starch content was determined according to the method of Zhao (2005), with three replications. Oven-dried grain was ground using a pulverizer (Shanghai Jiading Grain and Oil Instrument Co., Ltd. JFSD-70, China). Ten milligrams of the powder was transferred into centrifuge tubes and then blended with 100 µL of ethanol and 900 µL of 1 mol L−<sup>1</sup> NaOH. The mixture was heated in a boiling water bath for 10 min. After centrifugation (800 rpm, 15 min), 500 µl of the supernatant was diluted 200 times with distilled water, 1 mL of 1 mol L−<sup>1</sup> acetic acid and 1 mL iodine reagents (0.2% I2, 2% KI, w/v) were added, color development conditions were room temperature for 10 min, absorbance was measured at 620 nm with a spectrophotometer (Shanghai Precision Scientific Instrument Co., Ltd. 722G, China). A standard curve was prepared using soluble starch solutions at 0, 20, 40, 60, and 80% (w/v).

### Enzymatic Hydrolysis of the Starch Granules

Alpha-amylase and amyloglucosidase digestion was conducted according to Tang et al. (2002). After isolation, the starch granules (25 mg) were suspended in 1 mL sodium acetate solution (0.1 mol L −1 , pH 4.8) containing 360U α-amylase (Sigma, A4551, USA) or 50U amyloglucosidase (Sigma, A7420, USA). The samples were incubated on a shaking incubator (200 rpm, 37◦C, 72 h). Next, 50 µL of 1 mol L−<sup>1</sup> HCl was added to the samples. The reaction was stopped by adjusting the pH to 7 using 1 mol L−<sup>1</sup> NaOH.

The extent of starch degradation was determined by measuring the concentrations of reducing sugar produced by starch hydrolysis. Reducing sugar concentrations were determined using a modification of the method described by Bernfield (1951). After centrifuging at 1,500 g for 10 min, 0.1 mL of DNS reagent (0.63% w/v 3,5-dinitrosalicylic acid, 0.524 mol L <sup>−</sup><sup>1</sup> NaOH, 18.5% w/v NaKC4H4O6·4H2O, 0.5% w/v phenol, 0.5% w/v NaHSO3) was added to 0.1 mL of supernatant. The mixture was heated in a boiling water bath for 5 min. The absorbance was measured at 540 nm with a spectrophotometer (Shanghai Precision Scientific Instrument Co., Ltd.722G, China). A standard curve was prepared at using glucose at concentrations of 0, 100, 200, 300, 400, 500, and 600 µg mL−<sup>1</sup> .

### Detection of Relative Expression of Genes Involved in Starch Biosynthesis and Degradation

### Designation of Primers

The primers of AGP1, AGP2, SS1, SS2, SS3, SS4, GBSS1, GBSS2, SBE1, SBE2A, SBE2B, ISO1, AMY1, AMY2, AMY3, AMY4, BAM1, BAM2, BAM3, BAM4, BAM5, BAM6, and BAM7 were designed using Primer Premier 5.0 software according to sequences published in the National Center for Biotechnology Information (NCBI). The primers were synthesized by Sangon Biotech (Shanghai) Co., Ltd. The reference control was wheat ACTIN gene. The specificity of the primers was tested and the PCR conditions were optimized using gradient PCR and agarose gel electrophoresis (Bio Rad, Power Pac 300, USA). The primer sequences are presented in **Supplementary Table 1**.

### RNA Extraction and cDNA Synthesis

The RNA was extracted from de-embryonated wheat grain using RNAiso Plus (Takara, Cat#9108, Japan) and Fruit-mate (Takara, Cat#9192, Japan) kits according to the manufacturer's instructions. Total RNA quality was tested using agarose gel electrophoresis. First strand cDNA was synthesized using a reverse transcription kit (Tiangen, Cat#KR104-02, China). The cDNA quality was tested by amplifying the wheat ACTIN gene.

### Quantitative Real-Time PCR

The rt-qPCR reaction solution was prepared with a SYBR Premix Ex Taq Kit (Takara, Cat#RR420A, Japan). The components are presented in **Supplementary Table 2**. The amplifications of the individual cDNA sequences were detected using real time qPCR (Roche LightCycler 480 II, USA) with three replications.

A mathematical model was used to determine the relative expression of target gene compared with the reference gene. The relative quantification was calculated with the following formulae:


The wheat ACTIN gene was used as the reference. The gene is the favored reference for studying wheat genes because it is highly conserved in cell integrity, motility, and structure.

### In Situ Localization of AMY4, BAM1, and BAM5 Transcripts

### Synthesis of the Probes

The AMY4, BAM1, and BAM5 gene fragments were amplified using PCR. The fragments were collected using an EasyPure Quick Gel Extraction Kit (Transgen, Code #EG101-01, China). The fragments were linked with pEASY-T3 cloning vectors, and then the vectors were transformed to pEASY-T1 competent cells using a pEASY-T3 cloning kit (Transgen, Cat#CT301- 1, China). After screening and culturing, the plasmids were isolated using a TIANprep Mini Plasmid Kit (Tiangen, Cat #DP103-02, China) and sequenced by Sangon Biotech (Shanghai) Co., Ltd. Based on the sequencing results, the plasmids were digested with a restriction enzyme, either Nco I (Takara, Code#1160A, Japan, 10U µl −1 , source: Escherichia coli carrying the plasmid encoding Nco I gene) or Pst I (Takara, Code#1073A, Japan, 15U µl −1 , source: Escherichia coli ED8654 carrying the plasmid encoding Pst I gene). The linearized plasmids were used to synthesize antisense and sense probes via in vitro transcription. This was performed using a Dig RNA Labeling Kit (Roche, REF11175025910, USA).

The antisense probe sequence of AMY4 was as follows: 5 ′ -UUGGUUUCCGAUGGUGUUGUCCAAGAACAG GCAGCUCGCAAUGGCGGGAUCAUUAAGAACGGGAGA GAAAUCCUAUUGCAGGCUUUUAAUUGGGAAUCCCA UAAACACAAUUGGUGGAGUAAUUUAGAGGGCAGAGU UGCCGACAUUGCUA-3′ .

The antisense probe sequence of BAM1 was as follows: 5 ′ -ACUCAGGAAUGCAAGGCCUCAUGGCAUCAACAAG AGCGGCCCUCCUGAGCACAAGCUGUUUGGAUU CACCUACCUCCGGCUGUCGAAUCAGUUGGUGGAG GGACAAAACUAUGUCAAUUUCAAGACCUUUGUUGACA GAAUGCAUGCCAACCUGCCUCAUGACCCAU-3′ .

The antisense probe sequence of BAM5 was as follows: 5 ′ -UGAACCGGAACCUGUUCGACGGCGACAACU GGCGACGGUUCGUCGCGUUCGUGAAGACCAUG GCCGACGGCGGCGCGAGGACGGCGCUGCCCAGGU GCGACACUGGGCACUCGGAUCUGUACGUGGGGUUC GUUGA-3′ .

### Paraffin Sectioning

The method for making a paraffin section was modified from Ausubel et al. (1995). Wheat grains were crosscut and fixed in 4% paraformaldehyde (pH 7.2–7.6) and 0.1% DEPC under slight vacuum for 4 h at room temperature. The grains were dehydrated in a graded alcohol series (30–100%), and then cleared three times in solutions of alcohol and chloroform with ratios of 3:1, 1:1, and 1:3, each time for 3 h. Finally, the grains were cleared in absolute chloroform for 3 d. The cleared grains were infiltrated with a graded mixture of chloroform and paraffin wax at different temperatures (3:1, 40◦C; 1:1, 40◦C; 1:3, 45◦C), each time for 4 h. Then, the grains were infiltrated in paraffin wax at 55◦C either for 3 d (grain collected of 21 DPA and before) or for 5 d (grain collected at 28 and 35 DPA). The infiltrated grains were embedded in paraffin wax and then sectioned into 10–25µm thicknesses on a microtome (Kedee, 1508A, China). The exposed surfaces of grain collected at 28 and 35 DPA were soaked in DEPC-water for several hours before being sectioning. The sections were then affixed to adhesive microscope slides (Citoglas, REF188105W, China).

### In Situ Hybridization

The paraffin sections were dewaxed and rehydrated. Then, in situ hybridization was performed with an Enhanced Sensitivity ISH Detection Kit I, POD (Boster, MK1030, China). The paraffin sections were observed and photographed using a stereo microscope (Zeiss Discovery V20, Germany).

### Staining of Grain Median Transverse Sections with I2-KI

Paraffin sections containing wheat grain were stained with I2-KI (0.1%) for 5 min after dewaxing and rehydration. The sections were washed with deionized water, and then observed and photographed using a stereo microscope (Zeiss Discovery V20, Germany).

### Statistical Analysis and Image Processing

The data was analyzed by one-way ANOVA using Microsoft Excel and SPSS 13.0 software. Significance comparisons were made by Duncan's multiple range test at P < 0.05. Image processing was performed using Adobe Photoshop CS6.

### RESULTS

### Starch Content and Starch Granule Morphology

Grain weight increased across time and was significantly enhanced by P application (**Figure 1**). Similarly, the total starch content in the grain was very low during the early grain-filling stage and then increased with grain development. Beginning at 14 DPA, total starch contents were significantly greater in HP and NP than in P0 (note the difference between NP and P0 was not significant on 21 DPA). The grain also matured earlier in HP and NP than in P0 (**Supplementary Figure 1**).

The morphological characteristics of the starch granules was observed using SEM (**Figures 2A–C**). The A-type granules were disk-shaped with diameters >10µm. The B-type granules were spherical with diameters <10µm. The "pinholes" along the equatorial grooves of the granules and on their flat surfaces were more obvious in NP than in HP and P0.

To study the effects of P fertilizer on the micro-structure of starch granules, protease (Type XIV) -treated granules were stained with methanolic merbromin and CBQCA and then visualized by CLSM. The results showed that P fertilizer caused substantial changes in the histochemical patterns of the starch granules. Fluorescence was clearly visible and strong in large areas of many starch granules in NP (**Figures 2D–F**). In contrast, the P0 samples exhibited only faint fluorescence at the equatorial regions. The fluorescence of starch granules in HP was intermediate between NP and P0. The CBQCA staining (**Figures 2G–I**) patterns were similar those of membromin. These results suggested that P application influenced, presumably enhanced the pits and channels within starch granules.

### Reducing Sugars from Starch Granules after Exogenous Enzymatic Hydrolysis

Drought-induced microstructural changes to starch granules may facilitate the transfer of chemicals (water, enzymes, and acid) into the matrix of the starch granule and accelerate hydrolysis (Li et al., 2015). To ascertain whether the effects of P application were similar to those of drought, we measured the amounts of reducing sugars released from granules after hydrolysis for 72 h with amyloglucosidase and α-amylase. Reducing sugar concentrations after digestion were greater in HP and NP than in P0 (**Figure 3**, note the difference was not significant between HP and P0 in the amyloglucosidase treated samples). Together with the SEM and CLSM images, this result suggested that P fertilizer enhanced the pits and channels in starch granules and increased the starch surface area available for hydrolysis reactions.

### Patterns of α-Amylase and β-Amylase Activities during Grain Filling

The α- and β-amylase activities in the grain varied depending on sampling time and P treatment (**Figure 4**). The β-amylase activity was much higher than the α-amylase activity. The α-amylase activity under three treatments was gradually decreased from 7

FIGURE 2 | Starch granules isolated from mature wheat seeds (35 days post anthesis) and observed using SEM (×500 magnification) and CLSM (×400 magnification). Starch granules were isolated from the wheat grain in the 0 kg P ha−<sup>1</sup> (P0) treatment (A,D,G), 46 kg P ha−<sup>1</sup> (NP) treatment (B,E,H), and 92 kg P ha−<sup>1</sup> (HP) treatment (C,F,I). (A–C), starch granules observed using SEM; (D–F) starch granules stained with merbromin and then observed using CLSM; (G–I) starch granules stained with CBQCA and then observed using CLSM. The "pinholes" along the equatorial grooves and on flat surfaces of the granules were visualized with box. The magnified insets (×2,000) (A–C) were the starch in red box. Arrows indicate short channels and/or cavities (connected to the exterior by channels) (E,F) and radially oriented, channel-like, protein networks (G–I) within the granules.

to 21 DPA and then remained steady. The α-amylase activity was significantly greater in NP than in HP at 7, 21, 28, and 35 DPA. The P0 treatment had the lowest α-amylase activity among the treatments (except for 28 DPA). The β-amylase activity in all three P treatments gradually increased from 7 to 28 DPA and then declined. The β-amylase activity was significantly greater in HP and NP than in P0 at 21 and 28 DPA; however there was no difference between HP and P0 at 35 DPA.

### Patterns of Expression of Genes Involved in Starch Synthesis and Degradation during Grain Filling

The relative expressions of genes involved in starch synthesis and degradation are shown in **Figures 5**, **6**. The expression patterns of both AGP1 and AGP2 were similar across time in all three P treatments. The NP treatment had the highest AGP1 and AGP2 expression at 7 and 14 DPA.

The expression pattern of GBSS1 was different that of GBSS2. The GBSS1 transcripts were most abundant at 21 and 28 DPA, whereas the GBSS2 transcripts were most abundant at 7 and 14 DPA. The NP treatment had the highest GBSS1 expression (at 21 DPA) and the highest GBSS2 expression (at 7 DPA).

In NP, the SS1, SS2, and SS3 transcripts were greatest at 7 and 14 DPA and then decreased across time. In contrast, the SS4 transcripts remained steady between 7 and 21 DPA and then increased significantly at 28 DPA. In HP, the SS1, SS2, and SS3 transcripts were gradually increased from 7 to 14 DPA, whereas SS4 showed little expression on any sample date.

Among the genes encoding starch branching enzyme (SBE1, SBE2A, and SBE2B), SBE1 transcripts were most abundant. In NP, SBE1 transcript abundance was greatest at 21 DPA, whereas the

transcript abundances of SBE2A and SBE2B were both greatest at 7 DPA and then decreased. In HP, the transcripts of SBE1, SBE2A, and SBE2B were gradually increased from 7 to 14 DPA then decreased during the remaining time.

The NP and HP treatments generally upregulated the ISO1 transcripts, especially at 14 DPA. The NP treatment upregulated the AMY1 transcripts compared with P0, with peak expression at 21 DPA. In contrast, HP downregulated AMY1. The transcript levels of AMY2 were significantly greater in NP than in P0 and HP. There was no significant different in AMY2 expression between P0 and HP between 7 and 35 DPA. The relative expression of AMY3 in both P0 and NP was high and significantly greater than that in HP between 7 and 35 DPA. The relative expression of AMY4 increased suddenly at 28 DPA in all three P treatments. The AMY4 expression was greater in NP than in P0 and HP.

The seven BAM genes were differentially expressed among the P treatments. The BAM transcript levels were greater in NP than in P0 and HP. In NP and HP, BAM1 was mostly highly expressed at 28 DPA, whereas BAM2 was most highly expressed at 7 DPA. Of the expression patterns of BAM3 and BAM5 were almost identical both in NP and in HP. However, BAM3 and BAM5 genes were weakly expressed in P0. The transcript pattern of BAM4 significantly varied among the P treatments. The BAM4 expression at 7 and 14 DPA was less than that on the other sample dates. The transcript patterns of BAM6 and BAM7 were similar. The NP treatment upregulated both genes, with relative gene expression reaching a maximum at 21 DPA. The HP treatment downregulated BAM6 but upregulated BAM7 compared with P0.

These results indicated that 12 genes involved in starch synthesis and 11 genes involved in starch degradation were expressed in the developing wheat grains. Furthermore, the P treatments significantly influenced the expression patterns of these genes. Compared with P0 and HP, NP upregulated genes encoding starch synthesis enzymes (especially during early to mid-grain filling) and starch degradation enzymes (especially during mid- and late-grain filling).

### Spatial Profiling of Transcripts of AMY4, BAM1, and BAM5 during Grain Filling

As mentioned previously, β-amylase activity was much greater than α-amylase activity, and the transcription levels of BAM1 and BAM5 were the highest among the seven BAM genes. In addition, the relative expression of AMY4 increased sharply to high levels during late grain filling. For these reasons, AMY4, BAM1, and BAM5 mRNA were localized using in situ hybridization (anti sense: **Figure 7**, **Supplementary Figures 2**, **3**; sense control: **Supplementary Figures 4**–**6**). Starch accumulation in wheat caryopses (**Figures 7D,H,L**; **Supplementary Figures 3D,H,L**) indicated that the cavity in the ventral groove of caryopses was an intrinsic characteristic of wheat grains. AMY4, BAM1, and BAM5 transcripts were detectable in both the pericarp and early endosperm at 7 DPA in all three P treatments (**Supplementary Figure 2**). In P0, AMY4, BAM1, and BAM5 transcripts were detected in the entire endosperm from 7 to 35 DPA (**Figures 7A–C**, **Supplementary Figures 3A–C**). In NP and HP, AMY4, BAM1, and BAM5 transcripts had accumulated at the endosperm border at 28 DPA (**Figures 7E–G,I–K**) and at 35 DPA (**Supplementary Figures 3E–G,I–K**). However, the relative expression of AMY4, BAM1, and BAM5 at the edge of the endosperm was greater in NP (**Figures 7E–G**, **Supplementary Figures 3E–G**) than in HP (**Figures 7I–K**, **Supplementary Figures 3I–K**). This result showed that P fertilizer increased the transcript abundance of amylase genes at the edge of the endosperm. This phenomenon was more pronounced in NP than in HP.

### DISCUSSION

Wheat endosperm is the main tissue for biosynthesis and accumulation of starch. The A-type and B-type starch granules in mature wheat endosperm display bimodal distribution (Evers, 1971). Based on anatomical studies, Parker (1985) reported that A-type starch granules in wheat were initiated between ∼4 and 14 DPA. The B-type starch granules appeared from about 14 DPA until grain maturity. Previous research in our laboratory showed that P fertilizer significantly influenced the ratio of A/B starch granules and the average diameter of the granules (Li et al., 2013). However, the SEM images in this study showed that P fertilizer did not cause significant changes in the shape of the starch granules. Overall, these results indicate that P application may affect the timing and development of A- and B-type starch granules in wheat.

Analysis of the expression patterns of genes involved in starch synthesis is important to understand the mechanism of

starch biosynthesis. A previous study showed that AGPase is less sensitive to 3-PGA and inorganic P in cereal endosperm than in other tissues (Gómez-Casati and Iglesias, 2002). This suggests AGPase activity may be controlled at the transcriptional level in endosperm. McCue et al. (2002) found that GBSS I may control starch synthesis at the transcriptional and post-transcriptional level. Wang et al. (2014) studied the relationships among starch accumulation, the activities of key enzymes, and gene expression in wheat endosperm. The results indicated the amylose, amylopectin, and total starch accumulation rate were significantly and positively correlated with the activities of SBE, SSS, and GBSS. The SBE, SSS, and DBE may control starch synthesis at the transcriptional level, whereas GBSS1 may control starch synthesis at the post transcriptional level.

Phosphorus increases photosynthetic rates and promotes post anthesis dry matter accumulation (Zhu et al., 2012). Both of these factors play a vital role in starch biosynthesis and accumulation. In the present study, P application increased the expression of genes involved in starch synthesis, and these increases in gene expression coincided with greater starch content. This indicated that P application promoted starch biosynthesis not only by increasing photosynthetic rates to produce more substrate, but perhaps also by controlling starch synthesis at the transcriptional level. However, excess P application may also enhance respiration, which could increase sugar and energy loss (Shen, 2001). It is well-known that P application can cause earlier seed maturity, and we observed that the grain was darker (or more yellow) in HP than in NP and P0 at 35 DPA (**Supplementary Figure 1**). This indicates that the grain filling stage in HP was shorter than that in the other two treatments. This perhaps is one reason why HP had less starch accumulation than NP.

Whan et al. (2014) studied α-amylase levels in wheat grain and suggested the endosperm-specific over-expression of AMY3 resulted in an increase in total α-amylase activity in harvested wheat grain. However, increased α-amylase activity did not significantly influence starch content or composition. As mentioned previously, seed maturity can be promoted by P application (**Supplementary Figure 1**). In certain varieties of wheat, triticale, and barley, amylase activity always increases with grain maturity (Lindblom et al., 1989; Mares and Oettler, 1991; Radchuk et al., 2009). Thus, it is possible that increased amylase activity in this study resulted from early maturity induced by P application. On the other hand, the increased expression of amylase genes in NP and enhanced amylase activity suggested that P may also control starch degradation at the transcriptional level. Whan et al. (2014) observed that enhanced amylase activity did not reduce starch content. Therefore, there is no contradiction between high amylase activity and high starch content. Increases in amylase activity may cause variation in the channel structures of starch granules.

The channels and pores within starch granules are intrinsic characteristics of wheat. Using CLSM, Kim and Huber (2008) observed that the channel types are different in A- and Btype granules of wheat starch wheat. External factors [e.g., high temperature (Li et al., 2017) and drought stress (Li et al., 2015)] can enhance the number and size of the channels. In this study, "pinholes" and pits on flat surfaces along the equatorial groove of starch granules were more obvious in NP than in HP and P0. The CLSM images and the enzymatic digestion results also confirm such micro-structural changes. We speculate that the

increases in channel size and number enlarge the granule surface area available for hydrolysis reactions, resulting ultimately in the release of more glucose units. Earlier investigations in our laboratory revealed that pit micro-structures affected such starch characteristics as pasting properties, swelling power, solubility capacity, and enzymatic hydrolysis (Li et al., 2013, 2015). On the other hand, P-induced changes in the physicochemical properties of starch are complex, and other factors also have significant influence (e.g., the ratio of A- and B-type granules, as well as their volume, internal structure, and surface area; Hayfa et al., 2009).

Fannon et al. (2004) postulated that in the endosperm of sorghum and maize, amyloplasts contain microtubules which radiate outward from the initiation point of starch granule growth (i.e., granule hilum) to the plastid periphery. The authors

speculated that the granule develops around the radially-oriented microtubules, which become channels terminating at the outer surface of starch granules. Therefore, the channels within starch granules are the remnants of amyloplast microtubules. Benmoussa et al. (2010) observed that the protein constituents of channels in maize starch included actin-like and tubulin-like structural proteins, a membrane protein (adenylate translocator, Bt1), and the enzymes involved in starch biosynthesis. Further, Benmoussa et al. (2010) hypothesized that microtubules may possess at least two purposes in amyloplasts and starch granules: (1) they may facilitate starch polymer and granule biosynthesis; and (2) they may function to provide variation in the process of granule degradation during seed germination (Fannon et al., 1992a,b).

was 20µm.

We are not aware of any previous studies which have examined the biological importance of pits and channels in wheat starch biosynthesis. Based on previous findings in maize, we hypothesize that P fertilizer promotes the development of amyloplasts and influences the structure of amyloplast microtubules (Benmoussa et al., 2010). The microtubules may provide greater surface area for transporting starch-synthesizing enzymes and substrates needed for starch synthesis into the amyloplast. The hypothesis is supported by the observation that NP significantly increased wheat grain starch content as well as the expression of genes involved in starch synthesis. Furthermore, the observation that starch content was greater in NP than in HP suggests that NP optimized the exchange of substances for starch biosynthesis.

During late grain filling, increasing α-amylase and β-amylase activities may act on the channel ends to form pits on the granule surfaces. A previous study in our laboratory showed that endogenous hydrolysis (seed germination) was increased by P-induced increases in the number of pits and channels in wheat starch granules (Zu et al., 2017). Our observations appear to confirm the proposal by Benmoussa et al. (2010) that channels may influence granule degradation during seed germination. Further study is necessary to test this hypothesis. Such studies may provide significant information about the relationship between P application and starch biosynthesis.

Wheat endosperm cells are differentiated from the meristematic region at the periphery of the endosperm. Therefore, the youngest cells are found at the outer edge of the endosperm and the oldest at the center (Bradbury et al., 1956). During wheat grain development, starchy endosperm initiates a cell death program (Young and Gallie, 2000). In this study, P application increased (i) the expression of amylase genes; (ii) amylase activity; and (iii) AMY4, BAM1, and BAM5 transcript abundance at the periphery of the endosperm at 28 DPA. This indicates that during late grain filling in NP, the meristematic region at the periphery of the endosperm still maintained metabolic activities to support the relatively abundant transcripts of genes involved in starch synthesis and degradation. This explains the increased starch content in NP.

### CONCLUSIONS

The results of this study show that P fertilizer significantly altered microstructures in the starch granules. This is important because the channels may provide greater surface area for the transport of starch-synthesizing enzymes and substrates needed for starch synthesis. The study also indicated that P fertilizer significantly affected starch accumulation by influencing the expression of genes related to starch biosynthesis and degradation. Further study is necessary to understand the mechanism by which P influences starch morphology and biosynthesis. Such information may provide information helpful for increasing wheat yield and starch quality. The latter could have important implications for the food industry.

### AUTHOR CONTRIBUTIONS

RZ: Substantial contributions to the acquisition, analysis, and interpretation of data for the work; Drafting the work and revising it critically for important intellectual content; Final approval of the version to be published; Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. KF, CAL: Substantial contributions to the acquisition, analysis, and interpretation of data for the work; Revising the work critically for important intellectual content; Final approval of the version to be published; Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. CEL, CYL: Substantial contributions to the conception and design of the work; Drafting the work and revising it critically for important intellectual content; Final approval of the version to be published; Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

### ACKNOWLEDGMENTS

This study was financially supported by the National Natural Science Foundation of China (31360334, 31160256, 31360292, and 31560389), New Cultivar Breeding and Germplasm Enhancement of Wheat (2016AC027), Young Innovator

### REFERENCES


Cultivating Project of Shihezi University (CXRC201703), and Specific Project for Breeding of Shihezi University (YZZX201702).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2017. 02252/full#supplementary-material

Supplementary Table 1 | Characteristics of primers used to measure differential gene expression using Quantitative PCR.

Supplementary Table 2 | Components and volume of real time quantitative PCR reaction solution.

Supplementary Figure 1 | Fresh grains at 28 DPA (A–C) and 35 DPA (D–F) under P0 (A,D), NP (B,E), and HP (C,F) conditions. P0: 0 kg P ha−<sup>1</sup> ; NP: 46 kg P ha−<sup>1</sup> ; HP: 92 kg P ha−<sup>1</sup> .

Supplementary Figure 2 | In situ localization of AMY4, BAM1, and BAM5 transcripts in wheat caryopses at 7 DPA (×30). Hybridization sites of AMY4 (A,D,G), BAM1 (B,E,H), and BAM5 (C,F,I) transcripts were visualized as reddish-brown signals in median transverse sections of wheat grains under the P0 (A–C), NP (D–F), and HP (G–I) conditions, respectively. es, endosperm; np, nucellar projection; p, pericarp. P0: 0 kg P ha−<sup>1</sup> ; NP: 46 kg P ha−<sup>1</sup> ; HP: 92 kg P ha−<sup>1</sup> . The thickness of these sections was 20µm.

Supplementary Figure 3 | In situ localization of AMY4, BAM1, and BAM5 transcripts and starch accumulation in wheat caryopses at 35 DPA (×25, the magnified insets ×150). Hybridization sites of AMY4 (A,E,I), BAM1 (B,F,J), and BAM5 (C,G,K) transcripts were visualized as reddish-brown signals in median transverse sections of wheat grains under the P0 (A–C), NP (E–G), and HP (I–K) conditions, respectively. Starch granules were stained with I2-KI in median transverse sections of wheat grains under the P0 (D), NP (H), and HP (L) conditions, respectively. al, Aleurone; es, endosperm; np, nucellar projection. P0: 0 kg P ha−<sup>1</sup> ; NP: 46 kg P ha−<sup>1</sup> ; HP: 92 kg P ha−<sup>1</sup> . The thickness of these sections was 20µm.

Supplementary Figure 4 | Sense control of AMY4, BAM1, and BAM5 transcripts in wheat caryopses at 7 DPA (×25). Sense control of AMY4 (A,D,G), BAM1 (B,E,H), and BAM5 (C,F,I) transcripts in median transverse sections of wheat grains under the P0 (A–C), NP (D–F), and HP (G–I) conditions, respectively. es, endosperm; np, nucellar projection; p, pericarp. P0: 0 kg P ha−<sup>1</sup> ; NP: 46 kg P ha−<sup>1</sup> ; HP: 92 kg P ha−<sup>1</sup> . The thickness of these sections was 20µm.

Supplementary Figure 5 | Sense control of AMY4, BAM1, and BAM5 transcripts in wheat caryopses at 28 DPA (×25, the magnified insets ×150). Sense control of AMY4 (A,D,G), BAM1 (B,E,H), and BAM5 (C,F,I) transcripts in median transverse sections of wheat grains under the P0 (A–C), NP (D–F), and HP (G–I) conditions, respectively. al, Aleurone; es, endosperm; np, nucellar projection. P0: 0 kg P ha−<sup>1</sup> ; NP: 46 kg P ha−<sup>1</sup> ; HP: 92 kg P ha−<sup>1</sup> . The thickness of these sections was 20µm.

Supplementary Figure 6 | Sense control of AMY4, BAM1, and BAM5 transcripts in wheat caryopses at 35 DPA (×25, the magnified insets ×150). Sense control of AMY4 (A,D,G), BAM1 (B,E,H), and BAM5 (C,F,I) transcripts in median transverse sections of wheat grains under the P0 (A–C), NP (D–F), and HP (G–I) conditions, respectively. al, Aleurone; es, endosperm; np, nucellar projection. P0: 0 kg P ha−<sup>1</sup> ; NP: 46 kg P ha−<sup>1</sup> ; HP: 92 kg P ha−<sup>1</sup> . The thickness of these sections was 20µm.

endosperm amyloplasts. J. Cereal Sci. 52, 22–29. doi: 10.1016/j.jcs.2010. 02.013


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Zhang, Li, Fu, Li and Li. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Common Bean Fe Biofortification Using Model Species' Lessons

Raul A. Sperotto<sup>1</sup> \* and Felipe K. Ricachenevsky 2, 3 \*

*<sup>1</sup> Biological Sciences and Health Center, Graduate Program in Biotechnology, University of Taquari Valley - UNIVATES, Lajeado, Brazil, <sup>2</sup> Graduate Program in Agrobiology, Biology Department, Federal University of Santa Maria, Santa Maria, Brazil, <sup>3</sup> Graduate Program in Cell and Molecular Biology, Federal University of Rio Grande do Sul, Porto Alegre, Brazil*

Keywords: anti-nutrient, bean, biofortification, iron, model species, Phaseolus vulgaris, transgenic strategies

Common bean (Phaseolus vulgaris L.) is the most widely grown grain legume for direct human consumption and is highly preferred in many parts of Africa and Latin America, as well as in southern Europe (Broughton et al., 2003; Blair and Izquierdo, 2012). It is an important source of nutrients for more than 300 million people, representing 65% of total protein consumed, 32% of energy, and a major source of micronutrients e.g., iron (Fe), zinc, thiamin, and folic acid (Welch et al., 2000; Broughton et al., 2003; Blair et al., 2010a; Petry et al., 2015). It is known as the "poor men's meat," due to its high protein, minerals, and vitamins content (Blair, 2013). Fe is an essential micronutrient for almost all living organisms (Bashir et al., 2013), and Fe deficiency is the most common micronutrient deficiency worldwide, disproportionately affecting the poorest and most vulnerable populations in resource-limited settings, leading to Fe deficiency anemia (IDA; Stevens et al., 2013; Finkelstein et al., 2017). IDA is difficult to address through Fe supplementation or processed foods; therefore, several attempts are being made to enhance Fe accumulation into staples such as rice, maize, wheat, and legumes (Blair and Izquierdo, 2012) using biofortification, which is the process of breeding or genetic engineering to improve nutrient content in a crop (Blair, 2013). Biofortification is considered a sustainable and cost effective strategy to address malnutrition in developing countries because it targets staple foods that are consumed daily (Dwivedi et al., 2012).

### Edited by:

*Sebastien Thomine, Centre National de la Recherche Scientifique (CNRS), France*

#### Reviewed by:

*Louis Grillet, Academia Sinica, Taiwan*

\*Correspondence:

*Raul A. Sperotto rasperotto@univates.br Felipe K. Ricachenevsky felipecruzalta@yahoo.com.br*

#### Specialty section:

*This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science*

Received: *23 October 2017* Accepted: *12 December 2017* Published: *22 December 2017*

#### Citation:

*Sperotto RA and Ricachenevsky FK (2017) Common Bean Fe Biofortification Using Model Species' Lessons. Front. Plant Sci. 8:2187. doi: 10.3389/fpls.2017.02187*

Nutritional quality in common beans has been found to be higher than in cereals, with large amounts of minerals and vitamins accumulated in the seeds (Broughton et al., 2003). Common bean is estimated to have 4–10 times the amount of Fe, and 2–3 times the amount of Zn compared to rice (Pfeiffer and McClafferty, 2007). Also, these minerals and vitamins are generally retained after harvest and processing, while in polished cereal grains the Fe-rich tissues (embryo and aleurone layer) are lost (Wang et al., 2003). Although the average Fe concentration in beans is high, many people still suffer from IDA due to an insufficient level of bioavailable Fe in a monotonous cereal/bean-based diet without meat (Bouis, 2007). For Fe biofortification purposes, the use of common bean is advantageous because the baseline grain Fe content is high at 55 ppm and variability for the trait is great (Petry et al., 2015), ranging up to 110 ppm, allowing initial biofortification attempts to start from already high Fe levels (Blair et al., 2012; Blair, 2013). Another advantage of using common beans in biofortification programs is that seeds are consumed whole after boiling. Therefore, all major components of the common bean seed could be targets of biofortification: seed coat, cotyledons, and embryo (Blair et al., 2013).

The target Fe level of HarvestPlus, an international research program supporting the research and development of biofortified crops, is 94 ppm in whole bean seeds (Blair and Izquierdo, 2012; Petry et al., 2015). According to Vasconcelos et al. (2017), in order to achieve 30% of the estimated average daily dietary requirement for Fe on a dry weight (DW) basis, the recommended Fe levels in whole beans should be 107 ppm. The target level was quickly reached, and in some countries plant breeders have already developed and released new P. vulgaris bean varieties with Fe concentrations of about 100 ppm (Petry et al., 2015). These varieties show good micronutrient retention after processing, and equal or increased agronomic yield (Bouis and Welch, 2010). However, successful

**328**

bean Fe biofortification might be constrained due to the reported low Fe bioavailability (Ariza-Nieto et al., 2007) associated with high concentrations of Fe absorption inhibitors, also called antinutrients, such as polyphenols and phytate (Beninger et al., 2005; Petry et al., 2014). Here we propose multiple, complementary approaches to increase Fe concentration and bioavailability in common bean, based on the current knowledge on model species. These approaches are summarized in **Figure 1**.

### DECREASING ANTI-NUTRIENT CONCENTRATION AND CO-LOCALIZATION WITH FE IN SEEDS

Short-term human isotope studies indicate that phytate is the major Fe absorption inhibitor in beans, with polyphenols playing a minor role (Petry et al., 2012, 2014). Phytate increases with the Fe concentration in beans, and both are mainly found in the cotyledons. It should be possible to simultaneously breed for high Fe and low phytate since most phytate-related QTLs are independent of Fe concentration QTLs (Blair et al., 2012, 2013). Two main strategies for phytate reduction in seeds have been attempted: disruption of its biosynthetic pathway with knockout/knockdown of inositol pentakisphosphate 2-kinase (IPK1) in Arabidopsis and rice showing decreased phytate in seeds and normal yield (Stevenson-Paulik et al., 2005; Ali et al., 2013), but with possible defects in Pi homeostasis (Kuo et al., 2014); and mutations on phytate vacuolar transporters expressed in seeds, which reduced phytate concentration in other species (Shi et al., 2007; Nagy et al., 2009; Xu et al., 2009). In common bean, disruption of the orthologous transporter PvMRP6 resulted in 90% less phytate in seeds and normal

complementary, non-excludent approaches for bean biofortification. (A) Schematic representation of common bean seed and its main parts: seed coat, embryo and cotyledons. In cotyledons, iron (Fe) is shown with phytate (PA), whereas in the seed coat (detail), Fe is shown with polyphenols (PP). Each one act as an Fe absorption inhibitor in the human gut, with PA being likely a stronger anti-nutrient than PP. (1) Strategy aiming at increasing Fe concentration in the cotyledons to overcome PA anti-nutrient properties. (2) Strategy aiming at decreasing PA concentration in the cotyledons, making Fe in this tissue more bioavailable. (3) Strategy aiming at increasing Fe concentration in the seed coat to overcome PP anti-nutrient properties. (4) Strategy aiming at decreasing PP concentration in the seed coat, making Fe in this tissue more bioavailable. (B) Proposed candidate genes for genetic engineering in common bean, based on previous studies in model species. These genes are orthologous to genes found in *Arabidopsis thaliana* based on BLAST searches, except for Phvul.001G012200, which is the best hit using a soybean (*Glycine max*) *IPK* gene as query (Yuan et al., 2012). For each candidate gene, the type of manipulation is indicated.

agronomic performance (Panzeri et al., 2011; Campion et al., 2013). However, seeds were hard to cook and induced digestive problems in human subjects (Petry et al., 2016). Thus, further research is necessary to improve Fe bioavailability by decreasing phytate while maintaining agronomic performance and consumer preferences.

Biofortification in beans can target all seed tissues: the thick seed coat, two large cotyledons and a well-developed embryo (Blair et al., 2013), which comprise 7–10, 85, and 2–3% of seed weight, respectively (Ariza-Nieto et al., 2007). Remarkably, segregating populations derived from crosses between wild and cultivated parents showed that QTLs for Fe accumulation in each tissue segregate separately, and the Fe range and maximum amount observed in seed coat is higher than in cotyledons (Blair et al., 2013). Seed coat can contribute with as much as 26% of the total seed Fe, and polyphenols, not phytate, are the main anti-nutrients in the tissue (Ariza-Nieto et al., 2007). Thus, exploring seed coat biofortification is promising, as little is known about which specific polyphenol molecules reduce Fe bioavailability and how reduction in their concentration might affect plant and seed physiology (Petry et al., 2015).

### FURTHER INCREASING FE ACCUMULATION IN BEANS

Genetic engineering beans to accumulate higher Fe concentrations in seeds can benefit from work on model plants. Vacuolar Iron Transporter (VIT) family members are likely candidates, since they are involved in seed Fe localization and/or concentration in Arabidopsis and rice (Kim et al., 2006; Zhang et al., 2012). AtVIT1 localizes Fe to the provasculature, and changes in provasculature density have been proposed as a means to increase Fe content in seeds (Roschzttardtz et al., 2017). In rice, OsVIT1 and OsVIT2 are involved in flag leaf Fe pool regulation, and might also have a role in seed Fe localization (Zhang et al., 2012). Recent work showed that endosperm-specific overexpression of TaVIT2 increased Fe concentration in wheat endosperm (Connorton et al., 2017), indicating that VIT genes can increase tissue Fe sink strength.

In rice, overexpression of NICOTIANAMINE SYNTHASE (NAS) genes was shown to substantially increase Fe concentration in the endosperm, presumably increasing Fe translocation through the phloem (Johnson et al., 2011). In addition, OsNAS1 over-expression in rice plants enhance human Fe bioavailability from the high nicotianamine (NA) grains (Zheng et al., 2010). NA role in Fe long-distance transport is likely to be conserved in land plants (Schuler and Bauer, 2011), and therefore a similar approach could be applied to common bean. Wirth et al. (2009) overexpressed bean Ferritin, Arabidopsis Nicotianamine synthase, and Aspergillus fumigatus Phytase genes in rice plants, and detected 6.3 fold increase in Fe concentration on the polished seeds. Aluru et al. (2011) used a lpa maize mutant to overexpress soybean Ferritin gene, and found 2.7-fold increase in seed Fe concentration. Similar approaches could be certainly used in common bean plants in order to concomitantly decrease phytate levels and increase Fe accumulation and availability.

Another approach would be to explore genes that regulate Fe concentration. Over-expression of AtbHLH039 results in constitutive Fe deficiency response and increased Fe levels in leaves and seeds (Naranjo-Arcos et al., 2017). Interestingly, the bean genome has only one gene similar to all four subgroup Ib from Arabidopsis, which are known to be involved in Fe deficiency response (Brumbarova et al., 2015). Work in Arabidopsis and rice has shown that the negative regulators of Fe deficiency response BRUTUS/HRZ-like proteins could lead to increased Fe concentration in seeds of knockout/knockdown plants (Kobayashi et al., 2013; Hindt et al., 2017). Three BRUTUS/HRZ-like genes are found in the bean genome. Although promising, manipulation of regulatory proteins should be performed with caution, since plants might display undesired phenotypic changes besides increased Fe in seeds.

Common bean genetic transformation protocols are lengthy and of low reproducibility, while in vitro plant regeneration is especially difficult (Veltcheva et al., 2005; Rech et al., 2008). The Agrobacterium rhizogenes system allow for bean root transformation and could be used for characterization and selection of candidate genes for stable transformation (Estrada-Navarrete et al., 2007). Another solution is CRISPR-Cas9 mediated genome editing, which does not necessarily require transformation (Malnoy et al., 2016; Wolt et al., 2016) and could circumvent the problem in the near future. However, using this method, it would be easier to knockout a specific gene than overexpress it.

### EXPLORING BEAN NATURAL VARIATION AND WILD RELATIVES

The wide genetic Fe variability of beans has enabled plant breeders to develop varieties with twice Fe compared to normal beans (Blair et al., 2010b). Common bean is native to Latin America, and is one of the five cultivated species of the Phaseolus genus. It has two main genetic pools: Andean (large seeds) and Mesoamerican (small seeds). Andean and intergene-pool hybrids have higher Fe concentrations compared to Measoamerican ones, although the range of variation is similar (Blair, 2013). Large germplasm collection screenings for high Fe genotypes conducted in local and wild varieties of P. vulgaris have reported up to 110 ppm Fe. However, early analyses on closely related species such as P. coccineus and P. dumosus have found up to 127 ppm Fe, indicating that wild relatives might be useful (Blair et al., 2013). Even considering that high Fe wild genetic material showed poor agronomical performance (and introgression might not be straightforward in interspecific crosses), further screening of wild genotypes is promising. Moreover, wild beans accumulate more Fe in seed coats and less in cotyledons compared to domesticated genotypes, indicating that they can contribute differently for tissue-specific biofortification (Blair et al., 2013).

QTL studies show that multiple genes regulate seed Fe levels (Blair and Izquierdo, 2012; Blair et al., 2013). Interestingly, Fe concentration inheritance seems to be associated with Zn concentration, as found in other crops, indicating that similar genes are involved in micronutrient loading in seeds and that breeding for both minerals simultaneously is feasible (Blair et al., 2013). Based on QTL localization, Fe and Zn concentration might be associated with the seed storage protein Phaseolin, whereas the Fe storage protein Ferritin was also associated with a QTL (Blair et al., 2009). Indeed, engineering for increased Ferritin expression in endosperm of Poaceae species have been a relatively successful strategy (Sperotto et al., 2012), and thus Ferritin-associated QTLs are interesting candidates. Fe-chelate reductase, which is important for Fe uptake in roots, has also been suggested as a possible candidate gene (Blair et al., 2013).

### WHERE TO FOCUS NEXT?

Biofortification for any crop will benefit from multiple approaches, which can improve one another to achieve target Fe seed levels. For common bean, bioavailability tests are key due to the high level of anti-nutrients. The Caco-2 cell in vitro model has been widely used, with better results than previous in vivo absorption models (Ariza-Nieto et al., 2007; Blair et al., 2013; Petry et al., 2016). Recently, a new model using poultry (Gallus gallus) combined with Caco-2 cells showed that they can

### REFERENCES


be used as a robust, cost-effective two-step system to evaluate Fe bioavailability, which should be mandatory to generate as well as to monitor biofortified crop seeds after their release (Tako et al., 2016).

Another focus should be to independently increase Fe in cotyledons and in seed coat, and understand the physiological roles of phytate/polyphenols and the effects of their reduction on seed viability and seedling establishment. Fe in cotyledons accumulates at the vascular bundles (Cvitanich et al., 2010), similar to Arabidopsis where it depends on Vacuolar Iron Transporter (VIT1; Kim et al., 2006). Phytate is also likely to accumulate in vacuoles, based on the activity of MRP transporters (Nagy et al., 2009; Panzeri et al., 2011). It remains to be determined if the same cells accumulate Fe and phytate, and if the vacuole is the main site where phytatebound Fe is localized. Thus, analyses of cellular and subcellular distribution of Fe and phytate (using phosphorous as a surrogate) will be key for advances in cotyledon biofortification (Punshon et al., 2013). Moreover, understanding how polyphenols affect Fe homeostasis and how their levels could be manipulated will indicate new approaches for seed coat biofortification.

### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

common bean (Phaseolus vulgaris L.) and association with newly-mapped candidate genes. Mol. Breed. 30, 1265–1277. doi: 10.1007/s11032-012-9713-z


acid mutant reveals an exon-excluding splice-site mutation. Theor. Appl. Genet. 125, 1413–1423. doi: 10.1007/s00122-012-1922-7


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Sperotto and Ricachenevsky. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Exogenous Glycine Nitrogen Enhances Accumulation of Glycosylated Flavonoids and Antioxidant Activity in Lettuce (Lactuca sativa L.)

Xiao Yang<sup>1</sup> , Xiaoxian Cui 2, 3, Li Zhao<sup>1</sup> , Doudou Guo<sup>1</sup> , Lei Feng<sup>4</sup> , Shiwei Wei <sup>5</sup> , Chao Zhao2, 6 \* and Danfeng Huang<sup>1</sup> \*

### Edited by:

Marta Wilton Vasconcelos, Universidade Católica Portuguesa, Portugal

#### Reviewed by:

Ulrike Mathesius, Australian National University, Australia Francisco A. Tomas-Barberan, Consejo Superior de Investigaciones Científicas (CSIC), Spain

#### \*Correspondence:

Chao Zhao czhao@fudan.edu.cn Danfeng Huang hdf@sjtu.edu.cn

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 26 July 2017 Accepted: 24 November 2017 Published: 15 December 2017

#### Citation:

Yang X, Cui X, Zhao L, Guo D, Feng L, Wei S, Zhao C and Huang D (2017) Exogenous Glycine Nitrogen Enhances Accumulation of Glycosylated Flavonoids and Antioxidant Activity in Lettuce (Lactuca sativa L.). Front. Plant Sci. 8:2098. doi: 10.3389/fpls.2017.02098 <sup>1</sup> Key Laboratory of Urban Agriculture (South), Ministry of Agriculture, School of Agriculture and Biology, Shanghai Jiao Tong University, Shanghai, China, <sup>2</sup> Key Laboratory of Medical Molecular Virology, School of Basic Medical Sciences, Fudan University, Shanghai, China, <sup>3</sup> Institutes of Biomedical Sciences, Shanghai Medical College, Fudan University, Shanghai, China, <sup>4</sup> Instrumental Analysis Center, Shanghai Jiao Tong University, Shanghai, China, <sup>5</sup> Shanghai Agrobiological Gene Center, Shanghai, China, <sup>6</sup> National Clinical Research Center for Aging and Medicine, Huashan Hospital, Fudan University, Shanghai, China

Glycine, the simplest amino acid in nature and one of the most abundant free amino acids in soil, is regarded as a model nutrient in organic nitrogen studies. To date, many studies have focused on the uptake, metabolism and distribution of organic nitrogen in plants, but few have investigated the nutritional performance of plants supplied with organic nitrogen. Lettuce (Lactuca sativa L.), one of the most widely consumed leafy vegetables worldwide, is a significant source of antioxidants and bioactive compounds such as polyphenols, ascorbic acid and tocopherols. In this study, two lettuce cultivars, Shenxuan 1 and Lollo Rossa, were hydroponically cultured in media containing 4.5, 9, or 18 mM glycine or 9 mM nitrate (control) for 4 weeks, and the levels of health-promoting compounds and antioxidant activity of the lettuce leaf extracts were evaluated. Glycine significantly reduced fresh weight compared to control lettuce, while 9 mM glycine significantly increased fresh weight compared to 4.5 or 18 mM glycine. Compared to controls, glycine (18 mM for Shenxuan 1; 9 mM for Lollo Rossa) significantly increased the levels of most antioxidants (including total polyphenols, α-tocopherol) and antioxidant activity, suggesting appropriate glycine supply promotes antioxidant accumulation and activity. Glycine induced most glycosylated quercetin derivatives and luteolin derivatives detected and decreased some phenolic acids compared to nitrate treatment. This study indicates exogenous glycine supplementation could be used strategically to promote the accumulation of health-promoting compounds and antioxidant activity of hydroponically grown lettuce, which could potentially improve human nutrition.

Keywords: luteolin, organic nitrogen, nitrate, quercetin, ascorbic acid, H2O2 scavenging capability

### INTRODUCTION

A balanced diet is essential to ensure physical development and health. Numerous epidemiological studies have suggested high daily consumption of fruits and vegetables lowers the risk of several chronic diseases, such as cancer, cardiovascular disease and diabetes; the protective effects of fruit and vegetable consumption are mainly attributed to the presence of bioactive phytochemicals such as polyphenols, vitamin C and vitamin E (Arts and Hollman, 2005; Hooper and Cassidy, 2006; Russo et al., 2012; Chen and Chen, 2013; Wang et al., 2017). The economically valuable vegetable crop lettuce (Lactuca sativa) is a minimally processed food product available throughout the entire year, and is a significant source of natural healthpromoting compounds. Multiple factors, such as environmental conditions, agronomical manipulation, harvest time, watering and fertilization can strongly influence the levels of healthpromoting compounds in horticultural plants (Liu et al., 2007; Li and Kubota, 2009; Becker et al., 2014; Tavarini et al., 2015). Specifically, nitrogen fertilization plays an essential role in balancing the yield and quality of edible plants, especially the levels of secondary metabolites.

As the classical terrestrial nitrogen cycling paradigm asserts that organic nitrogen must be converted into nitrate or ammonium prior to becoming biologically available, the value of organic nitrogen (especially simple forms, such as amino acids) as a fertilizer has been largely ignored (Ge et al., 2009; Näsholm et al., 2009). Recently, several lines of evidence have suggested organic nitrogen (Ge et al., 2009; Gonzalez-Perez et al., 2015), which represents 96–99% of total nitrogen in soil, can be directly absorbed by plants and significantly influence plant physiology and nutritional quality (Paungfoo-Lonhienne et al., 2008). The simple amino acid glycine is regarded as a proof of life, was the original nutrient form for organisms (Xu et al., 2017) and is one of the most abundant free amino acids in horticultural soil; the glycine concentration of soil ranges from 1.14 to 2.39 µg N/g, corresponding to more than 30% of total free amino acids (Wang et al., 2013; Gonzalez-Perez et al., 2015). Compared to other amino acids, there is lower microbial demand for glycine and it is taken up more rapidly by plants (Lipson et al., 1999). Glycine is regarded as a model amino acid in plant organic nitrogen research.

There is growing interest in how nitrogen, especially its inorganic forms, influence antioxidant accumulation and bioactivity. Most studies support the notion that nitrate supply has a negative effect on the biosynthesis of phenolics and vitamin C, as well as antioxidant activity (Lee and Kader, 2000; Awad and de Jager, 2002; Staugaitis et al., 2008; Ibrahim et al., 2012; Yañez-Mansilla et al., 2014). However, recent evidence suggests glycine enhances tolerance to salinity (Badran et al., 2015), drought stress (Yang N. et al., 2016) and cold temperatures (Cao et al., 2017) via elevating the reactive oxygen species (ROS) scavenging system, nitrogen uptake and photosynthesis. In addition, glycine promotes the accumulation of carbohydrates (sucrose, glucose, fructose), which can provide a source of energy and carbon rings for polyphenol biosynthesis (Liu et al., 2016). L-phenylalanine, a flavonoid pathway precursor and phenylalanine ammonia lyase (PAL) substrate, was induced in pak choi by exogenous glycine supply (Wang X. L. et al., 2014). We previously assessed the main changes between lettuce cultured in glycine and nitrate without soil using a non-target metabolomics approach. Glycine nitrogen promoted the accumulation of glycosylated quercetin derivatives and luteolin derivatives (quercetin 3-O-glucoside, quercetin 3-O-malonylglucoside, luteolin 7-O-glucoside, and luteolin 7-O-glucronide), ascorbic acid and amino acids, but reduced the levels of some phenolic acid derivatives and some organic acids involved in the tricarboxylic acid cycle (Yang et al., 2018). Luteolin 7-O and quercetin 3-O glycosides are potent free radical scavengers/antioxidants and prevent ROS generation effectively (Agati et al., 2012; Brunetti et al., 2013). Therefore, we hypothesized glycine supply could promote the synthesis of health-promoting compounds in lettuce.

Thus, in the present study, the influence of different concentrations of organic nitrogen (as glycine) on the nutritional quality (i.e., total polyphenol, flavone, vitamin C and vitamin E contents, and antioxidative activity) of two lettuce cultivars was determined using a metabolomics approach and in vitro bioactivity assays. This work further explores the biological effects of organic nitrogen supply and indicates exogenous supply of glycine could potentially be used to enhance the nutritional quality of lettuce.

### MATERIALS AND METHODS

### Plants and Cultivation

Seeds of the lettuce cv. Lollo Rossa and cv. Shenxuan 1 were purchased from Atlas Seeds BJ Co., Ltd, (Beijing, China) and the Horticultural Research Institute, Shanghai Academy of Agriculture Sciences (China), respectively. Seeds were sown in white mesh pot net baskets (**Figure 1**) in nursing substrate (100% perlite) and germinated in a greenhouse at Shanghai Sunqiao Modern Agricultural Development Zone, China (latitude 31◦ 17′ N, longitude 121◦ 62′ E; altitude, 4 m above mean sea level). Then, 21-day-old seedlings were transferred to a water-cycled hydroponic experiment device (**Figure 1**). After recovering the seedlings in water for 3 days, the seedlings were cultivated for 30 days in nitrogenous nutrient solution (1.25 mM Mg, 3.5 mM K, 1.25 mM S, 2.05 mM Ca, 1 mM P, and 6.6 mM Cl, pH 5.8) containing different forms and concentrations of nitrogen: 9 mM nitrate (as NaNO3, control, 9Nit), 4.5 mM glycine (4.5Gly), 9 mM glycine (9Gly) or 18 mM glycine (18Gly). Ampicillin (10 mg/L) was added to the nutrient solution to prevent bacterial infections (Okamoto and Okada, 2004).

All treatments (90 seedlings per treatment, three replicates) were arranged randomly. During the experiment, environmental conditions were maintained at 22 ± 3 ◦C during the day and 15 ± 2 ◦C at night with 250–280 µmol·m−<sup>2</sup> ·s <sup>−</sup><sup>1</sup> during the 14 h photoperiod (natural and artificial lighting). The culture solutions were contained in a circulating water system and renewed every 2 days. At the end of cultivation, all leaves were collected. Each sample was divided in two: one half was used for physiological assessments; the other half was flash frozen in liquid nitrogen and stored at −80◦C until further analysis.

### Chemicals and Reagents

Ultra-pure water was prepared using a Milli-Q system (Millipore Laboratory, Bedford, MA, USA). Methanol and acetonitrile (LC-MS grade) were purchased from Fisher Scientific (Pittsburgh, PA, USA); luteolin 7-glucoside, quercetin glucoside and chicoric acid standards (HPLC grade, ≥ 98%), from the Chinese National Institute for Food and Drug Control (Beijing, China); methoxyamine hydrochloride, L-2-chlorophenylalanine, bis (trimethylsilyl) trifluoroacetamide (BSTFA), and 2′ ,7′ dichlorofluorescin diacetate (DCFH-DA), from Sigma-Aldrich (Merck KGaA, Darmstadt, Germany); Dulbecco's modified Eagle medium (DMEM) and fetal bovine serum, from Invitrogen (Thermo Fisher Scientific, Waltham, MA, USA); L-glutamine and penicillin, from Sangon Biotech (Shanghai, China); the Cell Counting Kit-8, from Dojindo (Kumamoto, Japan); and 2, 2′ -azobis (2-amidinopropane) dihydrochloride (ABAP), form Wako (Osaka, Japan). All other chemicals were of analytical grade and obtained from China National Medicines Co., Ltd. (Shanghai, China).

### Analysis of the Fresh Weight and Total Phenolic and Anthocyanin Contents of Lettuce Leaves

Fresh weight was measured using electronic scales (AUY220; Shimadzu, Kyoto, Japan). For total phenolic content analysis, 1 g of raw leaf was ground in 10 mL of 0.1 mM HCl-methanol (v/v 1:1), extracted ultrasonically for 30 min and centrifuged at 9,000 g for 30 min. The supernatant was diluted to 25 mL with methanol and filtered through a 0.45µm membrane (Złotek et al., 2014). Total polyphenol content was quantified using a UV-Vis spectrophotometer (U2900; Hitachi, Tokyo, Japan) at 725 nm as described by Złotek et al. (2014) and expressed as mg gallic acid equivalent (GAE) per g fresh weight (mg GAE·g <sup>−</sup><sup>1</sup> FW). The formula for calculating GEA was Y (GAE, mg) = 0.1266 × (OD725nm) − 0.0008.

Total anthocyanin content was measured via a nondestructive method, as described previously (Yang X. et al., 2016; Ferrandino et al., 2017), using a Dualex 4 Scientific<sup>+</sup> (Dx4, FORCE-A, Orsay, France) and expressed as: anthocyanin content (ng per cm<sup>2</sup> ) = log (red-excited infrared fluorescence/greenexcited infrared fluorescence) × 10<sup>3</sup> .

### UPLC-VION-IMS-QTOF-MS/MS Analysis

For UPLC-MS, lettuce leaf samples (200 mg) were weighed, ground into a powder in liquid nitrogen, extracted in 1 mL

TABLE 1 | Plant fresh and dry weights and total anthocyanin contents of lettuce cultivated in hydroponic solution containing nitrate or glycine.


For fresh weight analysis, values are means ± SD (n = 4); for total anthocyanidin content, values are mean ± SD (n = 100); Value with different letters are significantly different; p < 0.05, LSD analysis.

n.d. represents not detected.

TABLE 2 | Metabolites putatively identified by UHPLC-IMS-QTOF-MS in the leaf extracts of nitrate- and glycine-treated lettuce.


(Continued)

### TABLE 2 | Continued


RT, Retention time.

<sup>a</sup>5-Caffeoylquinic acid and 3-caffeoylquinic acid were the major forms.

<sup>b</sup>3,5-Di-O-caffeoylquinic acid was reported as the major form.

methanol/water (80:20, v/v), sonicated at 25◦C for 30 min, incubated at 4◦C for 12 h, centrifuged at 12,000 g for 10 min, and 0.5 mL supernatant was used for UPLC-MS analysis as previously described (Abu-Reidah et al., 2013a; Yang et al., 2018).

The composition and relative contents of polyphenols in lettuce leaves were analyzed using an Acquity class UPLC and Vion IMS QTOF MS (Waters, Corp. Milford, MA, USA) using an Acquity UPLC HSS T3 column (100 mm × 2.1 mm, i,d.: 1.7µm). The mobile phases were water containing 0.1% formic acid (A) and acetonitrile containing 0.1% formic acid (B). The injection volume was 3 µL, flow rate was 0.4 mL/min, with gradient elution (0–4 min, 20% B; 4–6 min, linear gradient from 20 to 25% B; 6– 8.5 min, 50% B; 8.5–12.5 min 50–85% B, 12.5–14 min, 85–100% B; 14–17 min, 100%), then initial conditions were restored for 5 min to equilibrate the column. The scan range was 50–1,000 m/z, and spectra were acquired in negative-ion mode. MS and MS/MS spectra were identified based on accurate mass, MS<sup>2</sup> fragments and isotopic distribution using online databases (e.g., ResPect, http://spectra.psc.riken.jp/) and bibliographies related to lettuce metabolites. MS and MS/MS tolerance were set at 3 mDa and 10 mDa, respectively. MS and MS/MS data were processed using Progenesis QI software (Waters Corp.).

To evaluate analytical reliability and reproducibly, quality control (QC) samples (a mixture of all samples) were analyzed at the start, middle and end of each batch, as previously described (Want et al., 2013). Principal component analysis (PCA) showed the PCA scores of the seven QC samples clustered together (Supplemental Figure 1), confirming the reliability and repeatability of the metabolomic analysis.

### GC-MS Analysis

For GC-MS analysis, 200 mg lettuce leaf tissue was ground in liquid N, extracted in 1 mL ice-cold methanol:chloroform (3:1 v/v), 20 µL of 0.3 mg·mL−<sup>1</sup> L-2-chlorophenylalanine (internal standard) was added, the samples were centrifuged at 15,000 g for 10 min, 0.3 mL of the supernatant was vacuum freeze-dried at 25◦C, 80 µL methoxyamine hydrochloride (15 mg mL−<sup>1</sup> in pyridine) was added, incubated at 37◦C for 1.5 h, then 80 µL bis (trimethylsilyl) trifluoroacetamide (BSTFA, containing 1% trimethylchlorosilane) was added and incubated at 80◦C for 1 h (Du et al., 2011).

Relative GC/MS quantification of ascorbic acid, α-tocopherol and γ-tocopherol were performed using an Agilent 7890 Gas Chromatograph coupled to a LECO Mass Spectrometer (PerkinElmer Inc., Waltham, MA, USA) using a DB-5MS capillary column (30 m × 0.25 mm × 0.25µm; Agilent J&W Scientific, Folsom, CA, USA). Inlet temperature, transfer line temperature and ion source temperature were 280◦C, 280◦C, and 230◦C, respectively. The gas (helium) flow rate was 1 mL·min−<sup>1</sup> and injection volume was 1 µL. After 6.5 min solvent delay, the initial GC oven temperature was 60◦C; 1 min after injection, the GC oven temperature was raised to 300◦C at 5◦C/min, then held at 300◦C for 11 min. Measurements were made via electron impact ionization (70 eV) in full scan mode (m/z 33–600). Ascorbic acid, α-tocopherol and γ-tocopherol were identified using LECO Chroma TOF (PerkinElmer Inc.) by comparison with reference spectra in the NIST 14 Mass Spectral Library (Scientific Instrument Services, Inc. NJ, USA). Relative response ratio was obtained by dividing the peak area by the peak area for L-2 chlorophenylalanine.

### Antioxidant Bioactivity Analysis Reducing Potential Assay

The antioxidant activity of the lettuce leaf extracts was determined using the ferric-reducing antioxidant power assay (FRAP) (Chan et al., 2007). Samples (2.5 mL, extracted as described for polyphenol analysis) were mixed with 2.5 mL phosphate buffer (0.2 mM, pH 6.6) and 2.5 mL potassium ferricyanide (1% w/v), incubated at 50◦C for 20 min, and then immediately transferred onto ice. Trichloroacetic acid solution (2.5 mL of 10% w/v) was added to stop the reaction, the mixture was centrifuged at 3,000 g for 10 min, then 2.5 mL of the supernatant was diluted with 2.5 mL water, 0.5 mL ferric chloride solution (0.1% w/v) was added, incubated for 30 min and absorbance was determined at 700 nm. The extraction solution used for polyphenol analysis was used as control sample. FRAP values were expressed as mg GAE·g <sup>−</sup><sup>1</sup> FW.

### Preparation of Extracts for Cellular Antioxidant Activity Assays

Lettuce leaf extracts were prepared as described for UPLC-MS analysis. Prior to the antioxidant bioactivity assays, 0.5 mL of supernatant from each sample was vacuum freeze-dried at 25◦C and resuspended in 200 µL water containing 0.1% DMSO.

### Cytotoxicity Assay

Hepatitis B virus-producing HepG2 cells were cultured in DMEM supplemented with 2 mM L-glutamine, 50 U mL−<sup>1</sup> penicillin and 10% fetal bovine serum at 37◦C in a 5% CO<sup>2</sup> atmosphere (Hong et al., 2013). The cytotoxicity of the lettuce leaf extracts toward HepG2 cells was assessed using the CCK-8 assay, as described by Shi et al. (2017). Cell viability (%) was calculated as (OD450 (sample) – OD450 (blank))/(OD (mock) – OD450 (blank)).

### Cellular Antioxidant Activity (CAA) Assay

Cellular antioxidant activity (CAA) was quantified as previously described (Wolfe and Liu, 2007). Briefly, HepG2 cells were seeded at a density of 1 × 10<sup>5</sup> cells per well into 96-well microplates in 100 µL media. After 24 h, the media was removed, cells were washed with PBS, then incubated with 100 µL of media containing 25µM DCFH-DA and 0.5 µL lettuce leaf extract for 1 h at 37◦C. Then the solution was removed, 100 µL of 600µM ABAP was added, and fluorescence values were read at 485 nm excitation and 538 nm emission using a VictorTM X3 Multilabel Plate Reader (Perkin Elmer) every 5 min for 1 h. CAA was expressed as CAA (unit) =100 – (R SA /R CA) × 100, where SA is the integrated area of sample fluorescence vs. time curve and CA is the integrated area of the control curve.

### Cellular H2O<sup>2</sup> Scavenging Assay

HepG2 cells were seeded at 5,000 cells/well in 96-well plates in 100 µL media, cultured for 24 h, incubated with 400µM H2O<sup>2</sup> containing 1 µL of lettuce leaf extract or an equivalent volume of media (mock) or H2O<sup>2</sup> (400µM H2O<sup>2</sup> solution plus 1 µL medium) as control treatments for 24 h, and cell viability was assessed using the CCK-8 assay as described by Shi et al. (2017).

### Statistical Analysis

Values are the mean ± SD of three biological replicates per treatment and three technical replicates per sample. ANOVA based on LSD analysis and Students t-tests were performed using IBM SPSS Statistics 22 (IBM, Armonk, NY, USA); p < 0.05 was considered significant. Pathway analysis was performed using ProcessOn (https://www.processon.com/) and R software (https://www.r-project.org/). PCA analysis was performed using SIMCA-P13.0 (Sartorius Stedim Biotech, Gottingen, Germany), Pearson correlation analysis was conducted using R software. Figures were created using R software or OriginPro 2016 (OriginLab, Northampton, MA, USA).

### RESULTS

### Effect of Glycine on Growth of Lettuce

The fresh weights of the aboveground parts and whole lettuce plants after 30 days cultivation in hydroponic solution containing 9 mM nitrate (control) or 4.5, 9, or 18 mM glycine are shown in **Table 1**. Glycine significantly reduced the fresh and aboveground weights compared to control lettuce. Among the glycine-treated plants, 9 mM glycine led to a significantly higher fresh weight (p < 0.05) than 4.5 or 18 mM glycine.

### Effect of Glycine on Accumulation of Antioxidative Compounds

In this study, the total anthocyanidin content was only assessed in the Lollo Rossa cultivar; the Shenxuan 1 cultivar is a green leafy lettuce, which does not contain detectable levels of anthocyanidins. As shown in **Table 1**, glycine supply increased (p < 0.05) the total anthocyanidin content in Lollo Rossa leaves compared to control plants. The anthocyanidin level peaked at 34.72 ng per cm<sup>2</sup> in Lollo Rossa leaves exposed to 18 mM glycine, which was significantly higher than the plants treated by 4.5 or 9 mM glycine. The highest exogenous concentration of glycine (18 mM) also significantly (p < 0.05) increased the total polyphenol content of the lettuce leaves compared to lettuce cultivated in 9 mM nitrate or 4.5 or 9 mM glycine (**Figure 2A**), with maximal levels of 1.48 and 2.53 mg g−<sup>1</sup> observed in Shenxuan 1 and Lollo Rossa, respectively.

In addition, relative ascorbic acid content increased significantly as glycine supply increased (from 4.5 to 9 mM glycine), but decreased at 18 mM glycine (**Figure 2B**) compared to control lettuce. The α-tocopherol content peaked in lettuce exposed to 18 mM glycine, corresponding to respective 3.4- and 1.7-fold increases in Shenxuan 1 and Lollo Rossa compared to the controls supplied with 9 mM nitrate (**Figure 2C**). Moreover, in both varieties, the levels of γ-tocopherol were significantly higher in lettuce supplied with 4.5 and 18 mM glycine (p < 0.05) than control plants (**Figure 2D**).

UPLC-MS analysis can separate co-effluents and enables robust and reproducible identification of the isomeric structures of polar metabolites (e.g., phenolic compounds) (Paglia et al., 2014). By comparison with online and in-house databases as well as published data, a total of 35 polyphenols were tentatively identified in the lettuce leaf extracts (level 2, putatively-annotated compounds), including 17 phenolic acid derivatives and 18 glycosylated flavonoids (**Table 2**). Metabolic pathway analysis was conducted to investigate the relationships between glycine supply and the accumulation of phenolic acids and flavonoids (**Figure 3**).

As shown in **Figures 4**, **5**, the relative contents of apigenin 7-O, luteolin 7-O and quercetin 3-O and caffeoylquinic acids derivatives in Lollo Rossa cultivar are higher than their in Shenxuan 1 lettuce. In the Shenxuan 1 cultivar, glycine supply significantly decreased the contents of dihydroxybenzoic acid derivatives (dihydroxybenzoic acid hexoside isomer 1 and 2), p-coumaroylquinic acid, dihydrocaffeic acid sulfate, tri-4-hydroxyphenylacetyl glucoside, ferulic acid methyl ester, caffeoyl hexose, hydroxybenzoyl dihydroxybenzoyl hexose and syringic acid hexose compared to control lettuce cultivated in nitrate; all of these metabolites were present at the highest levels in lettuce cultivated in 4.5 or 18 mM glycine and lowest levels in lettuce cultivated in 9 mM glycine (**Figure 4**). In addition, apigenin 7-O-glucuronide and some luteolin glycoside derivatives (luteolin 7-O-glucoside, luteolin 7-glucuronide) and quercetin glycoside derivatives (quercetin 3-glucuronide, quercetin 3-O-(6′′-O-malonyl)-glucoside 7-O-glucuronide, quercetin 3-O-(6′′ -O-malonyl)-glucoside 7-O-glucoside, quercetin glucose acetate isomer 1 and 2) were not detected in control Shenxuan 1 lettuce, but were induced by 9 and 18 mM glycine (**Figure 5**).

In the Lollo Rossa cultivar, glycosylated luteolin derivatives, quercetin derivatives, apigenin derivatives, dihydrocaffeic acid derivatives (dihydrocaffeic acid hexose isomer 2, dihydrocaffeic acid sulfate), tri-4-hydroxyphenylacetyl glucoside, syringic acid hexose, ferulic acid methyl ester, syringaresinol glucoside and esculetin hexoside were significantly induced by glycine, whereas the dihydroxybenzoic acid derivatives (dihydroxybenzoic acid hexose isomer 1 and 2) and hydroxybenzoyl dihydroxybenzoyl hexose were significantly reduced by 9 and 18 mM glycine (**Figures 4**, **5**).

### Effect of Glycine on Antioxidant Activity

To assess the effect of glycine supply on antioxidant activity, in vitro assays were performed to directly evaluate simple ferric reducing ability and cellular antioxidant activity (**Figures 6A,B**). Both nitrate- and glycine-treated lettuce leaf extracts had low cytotoxicity toward HepG2 cells in the CCK-8 assay (**Figure 6C**). However, supply of 9 or 18 mM glycine significantly increased the antioxidant activity of the Lollo Rossa cultivar extracts,

FIGURE 3 | Effect of glycine and nitrate supply on the composition and concentrations of polyphenols in lettuce leaf extracts. Relative abundance of metabolites is indicated from red (high) to green (low). The dotted lines in the metabolic pathway represent possible relationships that have not yet been proven experimentally and solid lines indicate pathways in the KEGG or PlantCyc databases.

which peaked in lettuce supplied with 18 mM glycine with a 3.3 fold (CAA) and 1.93-fold (FRAP) increase compared to control plants. In the Shenxuan 1 cultivar, 18 mM glycine significantly increased FRAP by 1.4-fold and CAA by 1.8-fold compared to nitrate-treated control plants (**Figures 6A,B**).

Moreover, the CCK-8 assay was conducted to evaluate the viability of HepG2 cells and verify the effect of extraction on H2O<sup>2</sup> scavenging capability (**Figure 7**). Cell viability was significantly higher in all treatments compared to H2O<sup>2</sup> treatment, suggesting the extraction process is superoxidescavenging. In addition, no significant differences were observed between the mock treatments and all Lollo Rossa samples supplied with glycine and Shenxuan 1 samples supplied with 18 mM glycine. Glycine-treated lettuce exhibited higher scavenging capability than nitrate-treated control lettuce. In the Shenxuan 1 cultivar, the cell viability of lettuce exposed to 18 mM glycine was significantly higher than control lettuce, while the extracts of lettuce treated with 4.5 and 9 mM glycine increased cell viability compared to the 9 mM nitrate-treated extracts.

### DISCUSSION

### Glycine Nitrogen Supply Reduces the Fresh Weight of Lettuce Plants

The ability of higher plants to use organic nitrogen (amino acids, peptides and proteins) as a nitrogen source has been demonstrated in laboratory studies and field experiments (Näsholm et al., 2009). Compared to nitrogen deficiency, exogenous glycine supply increases production of biomass in Arabidopsis plants (Forsum et al., 2008). However, glycine could not support plant growth to the same extent as the same concentration of nitrate in pak choi (Wang, X. et al., 2014). In this study, glycine nitrogen supply significantly decreased the fresh and dry weight of lettuce compared to plants supplied with 9 mM nitrate. In horticulture, 9 mM is suggested as a standard reference concentration in commercial hydroponic lettuce production (Brechner and Both, 2017, Grower's Handbook: Lettuce). It is not surprising that glycine supply decreases plant growth compared to plants provided with the appropriate nitrate concentration in agricultural practice. These results are in accordance with our previous studies of pak choi (Wang X. et al., 2014), which indicated glycine may limit plant root growth (Dominguez-May et al., 2013) and induce differential proteomic responses associated with plant defense or stress and energy and nitrogen metabolism (Wang X. et al., 2014).

### Appropriate Concentrations of Glycine Promote Accumulation of Antioxidants

Primary antioxidants, such as flavonoids, ascorbic, acid and tocopherols, are abundant in plants, exert various physiological functions (Dixon et al., 2002; Singh and Singh, 2008) and play significant roles in the human diet (Chen and Chen, 2013; Tomas-Barberan et al., 2016). In this study, 18 mM glycine supply significantly (p < 0.05) increased total polyphenols and anthocyanidin content compared to 9 mM nitrate or lower concentrations of glycine (4.5 or 9 mM). The influence of

nitrogen on the accumulation and bioactivity of antioxidants in plants are controversial. Generally, high levels of inorganic nitrogen (nitrate or ammonium) exert negative, dose-dependent effects on plant flavonoid biosynthesis and anthocyanin accumulation and activity (Patil and Alva, 1999; Awad and de Jager, 2002; Yañez-Mansilla et al., 2014; Becker et al., 2015). In addition to environmental conditions (e.g., temperature, light), the dose-dependent variations observed in plants exposed to different nitrogen sources may also be due to the complexity of plant responses to nutrient availability. According to the growthdifferentiation balance hypothesis, plants with sufficient nitrogen supply (e.g., 9 mM nitrate in this study) tend to allocate nitrogen to vegetative growth rather than biosynthesis of phenolics; low supply (e.g., 4.5 mM glycine) limits both growth and secondary metabolite accumulation, In contrast, plants with intermediate resources (e.g., 9 or 18 mM glycine, as N use efficiency is lower for glycine compared with nitrate) accumulate high levels of phenolic acids with an intermediate increase in biomass (Herms and Mattson, 1992; Glynn et al., 2007).

Ascorbic acid is a major antioxidant in lettuce (Nicolle et al., 2004). In this study, ascorbic acid was significantly induced by 9 mM glycine compared to control nitrate and 4.5 or 18 mM glycine. The effects of different concentrations of nitrogen on ascorbic acid synthesis remain controversial. Some studies have indicated increased nitrogen supply increases the vitamin C content in some plants, though most studies reported ascorbic acid decreased or did not significantly change (see review by Mozafar, 1993; Flores et al., 2004). This controversy may be related to inter-plant variations in the optimum nitrogen concentration required for maximal vitamin C accumulation.

Tocopherol is a lipid-soluble natural antioxidant; the α- and γforms are the major isomers in lettuce (Nicolle et al., 2004; Cruz et al., 2014). Exposure to 18 mM glycine significantly increased the α-tocopherol content compared to lettuce cultivated in 9 mM nitrate and 4.5 or 9 mM glycine. In both varieties of lettuce, the concentration of γ-tocopherol was significantly higher for plants supplied with 4.5 and 18 mM glycine than control plants. Similarly, previous research reported inorganic nitrogen fertilization increased the concentration of tocopherols in rapeseed, increasing the levels of urea more than the levels of ammonium (Hussain et al., 2014).

### Glycine Supply Promotes Accumulation of Apigenin-3-O, Quercetin-3-O and Luteolin-7-O Glycoside Derivatives

The phenylpropanoid and flavonoid pathways synthesize a wide range of secondary metabolites including phenolic acid derivatives, lignins and flavonoids, which play important roles

in both plant growth and human nutrition (Tzin and Galili, 2010). Glycine supply decreased the contents of several phenolic acids (e.g., hydroxycinnamic and hydroxybenzoic derivatives in the Shenxuan 1 cultivar; hydroxybenzoic derivatives in Lollo Rossa), but led to accumulation of luteolin, apigenin and quercetin glycoside derivatives. In general, phenolic acid derivatives and flavonoid biosynthesis share the same precursor, p-coumaroyl CoA. The induction of flavonoid biosynthesis and reductions in the content of some phenolic acids and derivatives observed in the presence of glycine indicate altered precursor availability induced metabolic flux from phenolic acid biosynthesis to flavonoid pathways by altering the expression of chalcone synthesis and auxin polar transport (Besseau et al., 2007; Taulavuori et al., 2016). In addition, glycine significantly promoted the accumulation of sugars, which may positively stimulate the biosynthesis of flavonol glycosides by increasing the supply of carbon rings and glycosides (Liu et al., 2016).

Luteolin and quercetin derivatives have a greater capacity to scavenge ROS than most other flavonoids (Brunetti et al., 2013), thus an increase in the luteolin to apigenin glycosides ratio and kaempferol to quercetin glycosides ratio are a component of plant responses to light quality and intensity; luteolin (or quercetin) glycoside derivatives increased significantly, while apigenin (or kaempferol) glycosides derivatives increased only slightly in response to light (Markham et al., 1998; Tegelberg and Julkunen-Tiitto, 2001; Oh et al., 2009). In this study, apigenin glycosides were not detected (Shenxuan 1) or only present at trace levels (Lollo Rossa) in the control lettuce supplied with nitrate, whereas a high concentration of glycine (18 mM) induced accumulation of apigenin glycosides. In addition, the downstream metabolites luteolin glycoside derivatives and

another dihydroxy B-ring-substituted flavonoid (quercetin 3-O glycoside derivatives) were also significantly induced by glycine compared to control lettuce. For example, 15-fold (Lollo Rossa) and 2-fold (Shenxuan 1) increases in luteolin 7-glucuronide were observed in lettuce supplied with 18 mM glycine compared to the respective control lettuce supplied with 9 mM nitrate. Moreover, 10- and 3-fold increases in quercetin glucoside were observed in Lollo Rossa and Shenxuan 1 supplied with 18 mM glycine compared to control lettuce. Thus, we hypothesize that apigenin-3-O, quercetin-3-O, and luteolin-7-O glycoside derivatives may represent signals of the response to glycine supply and indicate a metabolic switch from accumulation of small quantities of glycosylated flavonoids to synthesis of both monohydroxy and dihydroxy B-ring-substituted flavonoid derivatives.

### Appropriate Concentrations of Glycine Promote Antioxidant Bioactivity

Genotype and growing conditions influence antioxidant compositions and bioactivity in lettuce. Red leafed lettuce cultivars have higher average total polyphenol contents and antioxidant capacities than green leafed cultivars (Liu et al., 2007). In this study, extracts from the Lollo Rossa cultivar exhibited significantly stronger ferric-reducing antioxidant power, cellular antioxidant activity and H2O<sup>2</sup> scavenging ability than the Shenxuan 1 cultivar. The Lollo Rossa cultivar is likely to contain significantly higher levels of polyphenols (particularly glycosylated quercetin, apigenin, and luteolin), vitamin C and anthocyanins, which correlate positively with antioxidative activity.

We performed Pearson Correlation analysis to investigate the possibility of an inter-relationship between the metabolites detected and antioxidant activity, as indicated by FRAP, CAA, and H2O<sup>2</sup> scavenging capability (**Figure 8**). Antioxidant bioactivity was significantly (p < 0.05) and positively (r > 0.75) correlated with total polyphenol content and the levels of apigenin 7-O-glucuronide, luteolin 7-glucoside, quercetin 3-O-(6′′ -O-malonyl)-glucoside 7-O-glucoside, quercetin 3-O- (6′′ -O-malonyl)-glucoside 7-O-glucuronide, quercetin glucose acetate isomer 2, quercetin glucoside and quercetin hexoside glucuronide. These results are in agreement with a previous study of Stevia rebaudiana leaves treated with nitrogen, which found antioxidant bioactivity positively correlated with total phenolic acids and the levels of glycosylated quercetin, apigenin, and luteolin (Tavarini et al., 2015).

Glycine-treated lettuce extracts exhibited higher scavenging capability than nitrate-treated control lettuce extracts. The antioxidant bioactivity of Shenxuan 1 lettuce exposed to 18 mM glycine was significantly higher than that of control lettuce, while the extracts of lettuce treated with 9 and 18 mM glycine had higher antioxidative activities than the 9 mM nitratetreated extracts. These results can mainly be attributed to the significantly higher total levels of polyphenols, particularly luteolin, quercetin and apigenin glycosides, in the lettuce treated with 18 mM glycine. A luteolin or quercetin-rich diet is related to reduced risks of specific types of cancer (Ekström et al., 2010; Lam et al., 2012; Lin et al., 2014) and cardiovascular disease (Duthie et al., 2000; Lee et al., 2011), and plays a protective effect in diabetes (Babu et al., 2013). Thus, exogenous glycine supply may promote the accumulation of healthpromoting compounds and increase the antioxidative activity of lettuce, which could potentially be beneficial for human nutrition.


### CONCLUSION

The appropriate concentration of glycine (18 mM for Shenxuan 1; 9 mM for Lollo Rossa) significantly enhanced the levels of antioxidants, including total polyphenols and α-tocopherol, and antioxidant activity (as indicated by FRAP, CAA, and H2O<sup>2</sup> scavenging capability) compared to lettuce supplied with nitrate. Most glycosylated flavonoids detected, including apigenin, quercetin and luteolin, were also induced by 9 and 18 mM glycine, whereas glycine decreased the levels of some phenolic acids. This study indicates exogenous glycine supply could be used strategically to promote the accumulation of health-promoting compounds and increase the antioxidative activity of hydroponically grown lettuce; this strategy may have potential relevance to human nutrition.

## AUTHOR CONTRIBUTIONS

XY, XC, and LZ performed all the experimental measurements, analyzed the data, and drafted the manuscript. DG, LF, and SW helped with the figures and samples. CZ and DH designed experiment and supervised all the results, and contributed to writing the manuscript.

### FUNDING

This work was supported by the National Natural Science Foundation of China (No. 61233006 and No. 81370046); the Seed Industry Development Project of Shanghai, China (Grant No. 2016, 1-8); and the Agriculture Research System of Shanghai, China (Grant No. 201702).

### ACKNOWLEDGMENTS

We acknowledge lab members Miss. Yifei Zhao, Miss. Yanwen Gu, and Mr. Jiaxin Zheng for assisting with the

### REFERENCES


experiments, and colleague Dr. Bin Liu, Mr. Muhammad Khalid, Mr. Kai Dou, Dr. Xiaosong Liu (Chinese Academy of Sciences), and Mr. Hongkai Zhu (University of Copenhagen) for providing some advice in preparation of the manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2017. 02098/full#supplementary-material


to enhanced ultraviolet-B radiation. Physiol. Plant. 113, 541–547. doi: 10.1034/j.1399-3054.2001.1130413.x


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Yang, Cui, Zhao, Guo, Feng, Wei, Zhao and Huang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Genome-Wide Identification and Expression Analysis of NRAMP Family Genes in Soybean (Glycine Max L.)

Lu Qin1†, Peipei Han1†, Liyu Chen<sup>2</sup> , Thomas C. Walk <sup>3</sup> , Yinshui Li <sup>1</sup> , Xiaojia Hu<sup>1</sup> , Lihua Xie<sup>1</sup> , Hong Liao<sup>2</sup> and Xing Liao<sup>1</sup> \*

<sup>1</sup> Key Laboratory of Biology and Genetics Improvement of Oil Crops of the Ministry of Agriculture, Oil Crops Research Institute of Chinese Academy of Agricultural Sciences, Wuhan, China, <sup>2</sup> Root Biology Center, Fujian Agriculture and Forestry University, Fuzhou, China, <sup>3</sup> Golden Fidelity LLC, St. Louis, MO, United States

### Edited by:

Raul Antonio Sperotto, Centro Universitário Univates, Brazil

#### Reviewed by:

Jitender Giri, National Institute of Plant Genome Research, India Rupesh Kailasrao Deshmukh, Laval University, Canada Victoria Fernandez, Universidad Politécnica de Madrid (UPM), Spain

> \*Correspondence: Xing Liao liaox@oilcrops.cn

† These authors have contributed equally to this work.

### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 09 April 2017 Accepted: 03 August 2017 Published: 18 August 2017

#### Citation:

Qin L, Han P, Chen L, Walk TC, Li Y, Hu X, Xie L, Liao H and Liao X (2017) Genome-Wide Identification and Expression Analysis of NRAMP Family Genes in Soybean (Glycine Max L.). Front. Plant Sci. 8:1436. doi: 10.3389/fpls.2017.01436 The NRAMP (natural resistance-associated macrophage protein) family of genes has been widely characterized in organisms ranging from bacteria to yeast, plants, mice, and humans. This gene family plays vital roles in divalent metal ion transport across cellular membranes. As yet, comprehensive analysis of NRAMP family genes has not been reported for soybean. In this study, bioinformatics analysis was conducted to identify 13 soybean NRAMP genes, along with their gene structures, phylogenetic relationships, and transmembrane domains. Expression analysis suggests that GmNRAMP genes function in numerous tissues and development stages. Moreover, soybean NRAMP genes were differentially regulated by deficiencies of N, P, K, Fe, and S, along with toxicities of Fe, Cu, Cd, and Mn. These results indicate that GmNRAMP genes function in many nutrient stress pathways, and might be involved in crosstalk among nutrient stress pathways. Subcellular localization analysis in Arabidopsis protoplasts confirmed the tonoplast or plasma membrane localization of selected soybean NRMAP proteins. Protein-protein interaction analysis found that the networks of three GmNRAMP proteins which putatively interact with nodulin-like proteins, almost distinct from the network that is common to the other 10 soybean NRAMP proteins. Subsequent qRT-PCR results confirmed that these three GmNRMAP genes exhibited enhanced expression in soybean nodules, suggesting potential functions in the transport of Fe or other metal ions in soybean nodules. Overall, the systematic analysis of the GmNRAMP gene family reported herein provides valuable information for further studies on the biological roles of GmNRAMPs in divalent metal ion transport in various soybean tissues under numerous nutrient stresses and soybean-rhizobia symbiosis.

Keywords: soybean, NRAMP gene family, nutrient deficiency, divalent metal toxicity, nodules

### INTRODUCTION

Iron (Fe) is an essential element for plant development and growth, with functions in several basic cellular processes, including photosynthesis, respiration, and chlorophyll biosynthesis (Kobayashi and Nishizawa, 2012). Furthermore, Fe is also a vital component in heme, the Fe-sulfur (S) cluster, and other Fe-binding sites (Kobayashi and Nishizawa, 2012). Given these requirements and Fe deficiency is common in soils, plants have evolved highly efficient systems to acquire

**350**

Fe from the soil (Kim and Guerinot, 2007). Furthermore, the uptake, utilization, and storage of Fe are also tightly controlled by the coordination of multiple mechanisms regulated to at the transcriptional and post-translational levels (Kobayashi and Nishizawa, 2012). Mechanisms contributing to Fe acquisition in a number of plant species can be divided into two categories (Hell and Stephan, 2003; Morrissey and Guerinot, 2009; Conte and Walker, 2011). Strategy I, which is found in non-graminaceous plants, utilizes IRT1 as the primary transporter responsible for uptake of Fe from soil into roots (Eide et al., 1996; Hell and Stephan, 2003; Walker and Connolly, 2008). Meanwhile, YSL is the main transporter responsible for uptake of Fe from siderophore-Fe complexes into Strategy II graminaceous plants (Curie et al., 2001; Inoue et al., 2009; Thomine and Vert, 2013). Beyond these Strategy I and II transporters, the NRAMP family represents another transporter family associated with Fe uptake and transport (Thomine and Vert, 2013).

The NRAMP family, with its highly conserved domain, is widespread in genomes ranging from bacteria to humans (Nevo and Nelson, 2006). It is known to mediate transport of divalent metal ions, such as Fe and manganese (Mn) across cellular membranes. The first known NRAMP protein (NRAMP1) was discovered in mice phagosomal membranes, and was found to function in natural defense against infections by intracellular parasites (Vidal et al., 1993). In contrast to NRAMP1, mice NRAMP2 (also called DMT1), yet it still acts as a divalent metal ion transporter in the absorption of Fe, Mn, zinc (Zn), copper (Cu), cadmium (Cd), and lead (Pb) (Garrick et al., 2006). Mutations in NRAMP2 have been associated with defects in Fe absorption and result in Microcytic anemia in mice and the anemic Belgrade rat (Fleming et al., 1998). NRAMP homologs with similar function also were found in human (Cellier et al., 1994; Beaumont et al., 2006; Illing et al., 2012).

Several NRAMP gene family members have also been functionally characterized in plants. In Arabidopsis, there are six NRAMP proteins (Mäser et al., 2001). AtNRAMP1 regulates Fe homoeostasis (Curie et al., 2000), and function as a high-affinity transporter for Mn uptake (Cailliatte et al., 2010). AtNRAMP3 and AtNRAMP4 are both localized on the vacuolar membrane and participate in vacuolar Fe mobilization during seed germination (Lanquar et al., 2005). AtNRAMP6 is targeted to a vesicular-shaped endomembrane compartment and functions as an intracellular metal transporter, with possible involvement in Cd tolerance (Cailliatte et al., 2009). In rice, three NRAMP proteins participate in Fe, Cd, and Mn uptake (Takahashi et al., 2011; Sasaki et al., 2012; Yang et al., 2014), while another, OsNrat1, encodes a transporter mediating aluminum (Al) uptake from root tip cell walls into the cell, which contributes to rice Al tolerance (Li et al., 2014). In recent years, several NRAMP genes have been identified in legumes. For example, a peanut NRAMP gene, AhNRAMP1, is significantly induced by Fe deficiency in roots and leaves, and heterologous expression of AhNRAMP1 in tobacco leads to Fe accumulation in young leaves and tolerance to Fe deprivation (Xiong et al., 2012). Moreover, in the model legume Medicago truncatula, MtNRAMP1, is mainly localized to the plasma membrane, with expression levels highest in roots and nodules, suggesting it was the major transporter responsible for apoplastic Fe uptake in rhizobia-infected cells (Tejada-Jiménez et al., 2015).

Provided the commercial significance of Soybean (Glycine max L.) worldwide and the detrimental effects of Fe deficiency on yield and quality, it is key to improve our understanding of Fe transport as tool for improving soybean Fe utilization. However, little data is available concerning the NRAMP gene family in soybean until now. In the present study, bioinformatics analysis was conducted to identify 13 soybean NRAMP genes. Subsequently, tissue-specific expression of GmNRAMP genes and their responses to various nutrient stresses were all analyzed. The genome-wide analysis of soybean NRAMP genes herein provides a basis to further investigate detailed functions of NRAMP genes in soybean.

### MATERIALS AND METHODS

### Identification and Bioinformatics Analyses of NRAMP Genes in Soybean

To identify NRAMP homologs in soybean, nucleic acid, and amino sequences of all reported NRAMPs in Arabidopsis, Rice and Medicago, were used as query sequences in BLASTN (Target type: Genome) and BLASTP (Target type: Proteome) searches of the G. max cultivar Williams 82 in the Phytozome genome database (https://phytozome.jgi.doe.gov/pz/portal.html#) using default settings for E-value and the number of hit sequences. Then, all returned genes and proteins were further examined for inclusion of the conserved Nramp domain (PF01566) by querying in the Uniprot (http://www.uniprot.org/) and Pfam (http://pfam.xfam.org/search) databases. The nucleic acid and amino sequences of identified soybean NRAMP genes were downloaded from the Phytozome website. Soybean NRAMP genes were named according to phylogenetic relationships among the proteins. The chromosomal localization map and duplication of soybean NRAMP gene was determined by using ORTHOMCL (http://orthomcl.org/orthomcl/) and SVG softwares (http://search.cpan.org/~ronan/SVG-2.28/SVG/ Manual.pm). Protein molecular weights and theoretical pI values were computed using ProtParam (http://web.expasy. org/protparam/). Sequence identity among soybean NRAMP proteins was determined using BLASTP with each sequence queried against the other soybean NRAMP sequences in standalone BLAST downloaded from NCBI (https://blast.ncbi. nlm.nih.gov/Blast.cgi). Transmembrane helices in proteins were predicted using the TMHMM Server v. 2.0 (http://www. cbs.dtu.dk/services/TMHMM/). Predictions of subcellular localization for soybean NRAMP proteins were generated with ProtComp 9.0 (http://linux1.softberry.com/berry.phtml?group= programs&subgroup=proloc&topic=protcomppl). Multiple sequence alignment was performed with Clustal W and drawn in Genedoc, with the logo of consensus transport residues then generated by WebLogo 3 (http://weblogo.threeplusone. com/). Phylogenetic trees based on full length protein sequence alignments of NRAMPs from soybean and several other species were constructed by the neighbor-joining method with 1,000 bootstrap replicates in MEGA 6.0 (http://www.megasoftware. net/download\_form). Downloaded CDS and genomic sequences of soybean NRMAP genes were used to construct gene structures on the Gene Structure Display Server 2.0 (http://gsds.cbi.pku. edu.cn/index.php). PlantCARE (http://bioinformatics.psb.ugent. be/webtools/plantcare/html/) was used to cis-element analysis in the 1,500 bp region upstream of the start codon for each NRAMP gene.

### Plant Materials and Treatments

Soybean cv. Williams 82 was employed in this study. For tissue specific expression analysis of GmNRAMPs, soybean plants were cultured in hydroponics and harvested at a number of developmental stages for qRT-PCR assays. Specifically, soybean seeds were surface-sterilized in 10% H2O2, then, after germination for 1 week, seedlings were transplanted into fullstrength nutrient solution as previously described (Li et al., 2012) which contained 250 mM KH2PO4, 3,000 mM KNO3, 2,000 mM Ca(NO3)2, 250 mM MgSO4, 25 mM MgCl2, 12.5 mM H3BO3, 1 mM MnSO4, 1 mM ZnSO4, 0.25 mM CuSO4, 0.25 mM (NH4)6Mo7O24, and 25 mM Fe-Na-EDTA. The pH value of the nutrient solution was adjusted to 5.8, and nutrient solution was changed weekly. Seedlings were grown in a growth chamber with a 16 h light and 8 h dark cycle at 28◦C for 40 days before separately harvesting young leaves, older leaves, roots, stems, and flowers. At 55 days, young pods and seeds were also separately harvested. All tissue samples were stored at −80◦C for further RNA extraction and qRT-PCR analysis.

To investigate possible responses of GmNRAMPs to nutrient deficiency, 10-day-old seedlings were exposed to low-nitrogen (LN), -phosphorus (LP), -potassium (LK), -iron (LFe), or -sulfur (LS) conditions for 14 days, in which time nutrient deficiency symptoms became evident. For the LN treatment, KNO<sup>3</sup> and Ca(NO3)<sup>2</sup> were replaced by K2SO<sup>4</sup> and CaCl2, respectively. For the LP treatment, KH2PO<sup>4</sup> was replaced by K2SO4. For LK, KNO3, and KH2PO<sup>4</sup> were replaced by Ca(NO3)<sup>2</sup> and NaH2PO4, respectively. For LFe, Fe-Na-EDTA was not added to the nutrient solution. In the LS treatment, MgSO<sup>4</sup> was replaced by MgCl2. Seedlings grown in full-strength nutrient solution were sampled as the control. Each treatment had four biological replicates. Leaves and roots were separately sampled for further RNA extraction and qRT-PCR analysis.

To elucidate the probable functions of GmNRAMPs in response to divalent metal toxicity stresses, 10-day-old seedlings were treated with excess Fe (1,000 µM EDTA-Fe), Cu (200 µM CuSO4·5H2O), Cd (100 µM CdCl2), and Mn (200 µM MnSO4·H2O) treatments for 24 h. Each treatment had four biological replicates. Leaves and roots were separately sampled for further analysis.

To further study the responses of GmNRAMPs to rhizobia inoculation, 7-day-old seedlings were inoculated with the effective rhizobial strain Bradyrhizobium sp. BXYD3 (Cheng et al., 2009), and then transplanted into low nitrogen (500 µM N added) nutrient solution. Each treatment had four biological replicates. Young leaves, stems, roots, and nodules were separately collected at 30 days after inoculation and then stored at –80◦C for RNA extraction and qRT-PCR analysis.

### RNA Extraction and qRT-PCR Analysis

Total RNA was extracted from different soybean samples using RNAisoTM Plus reagent (Takara Bio, Otsu, Shiga, Japan) according to the manufacturer's instructions. RNA samples were treated with RNase-free DNaseI (Invitrogen, Grand Island, NY, USA) to remove contaminating genomic DNA. The quality of total RNA was checked via spectrophotometry (TGem Plus, Tiangen, China). Then, first strand cDNA was synthesized using the PrimeScriptTM RT Master Mix (Takara, Tokyo, Japan) according to the manufacturer's protocol. For qRT-PCR analysis, the soybean housekeeping gene TefS1 (encoding the elongation factor EF-1a: X56856) was used as a reference gene, and specific primers for GmNRAMPs and TefS1 were designed with Primer-NCBI (https://www.ncbi.nlm.nih.gov/tools/primerblast/index.cgi?LINK\_LOC=BlastHome) as listed in Table S1. In addition, the specific primers for nutrient deficiency responsive genes were also designed and listed in Table S2. qRT-PCR reactions were carried out in a CFX connect Real-Time PCR Detection System (Bio-Rad, Hercules, USA) with SYBR <sup>R</sup> Premix Ex TaqTM II (TaKaRa, Tokyo, Japan). The PCR reaction volume was 20 µL in total, which included 2 µL diluted cDNA, 10 µL SYBR Premix Ex TaqTM reagent, 0.6 µL primers and 6.8 µL RNAfree water. PCR Reactions were performed under the following conditions: 95◦C for 1 min, followed by 40 cycles of 95◦C for 15 s, 60◦C for 15 s, and 72◦C for 30 s. The expression of NRAMP genes was calculated by the 2−1Cq and 2−11Cq methods (Livak and Schmittgen, 2001).

### Subcellular Localization and Predicted Protein Interaction Networks

To determine the predicted subcellular localization of soybean NRAMP proteins, six GmNRAMP proteins were selected to generate subcellular localization constructs, then transient expression in Arabidopsis protoplasts, which were widely used in subcellular localization analysis of genes not only for soybean, but also for other plant species (Zhang et al., 2016; Chong et al., 2017; Chu et al., 2017; Péron et al., 2017; Xu et al., 2017). Specifically, the coding region of each GmNRAMP gene was amplified with gene-specific primers shown in Table S3. These CDS sequences were cloned into the pMDC43 vector to express GmNRAMP-GFP fusion proteins driven by the CaMV 35S promoter. The constructs of 35S::GmNRAMP-GFP and 35S:GFP (control) were separately transformed into Arabidopsis protoplasts. Arabidopsis protoplasts were isolated from the leaves of 4-week-old Arabidopsis plants and subsequently transformed according to previously published protocols (Yoo et al., 2007). After transfection using polyethylene glycol and incubation in a plate under weak light for 12–16 h, protoplasts were observed with an Olympus FV10-ASW laser scanning confocal microscope (Olympus, Japan). Corresponding markers used in coexpression experiment were selected according to predicted subcellular localizations of GmNRAMP proteins, which were also verified in rice protoplasts isolated from the stems of 10-day-old rice plants under dark culture conditions. The coexpression of two marker genes in rice protoplasts were performed same as the protocol for Arabidopsis protoplasts.

To further investigate possible protein interactions involving GmNRAMPs, putative interaction networks were identified in the interaction section of the UniProt protein knowledgebase (http://www.uniprot.org/), with interactions originating from STRING 10.0 protein-protein interaction databases (http://www. string-db.org/). The default settings of association networks were applied in these analyses.

### Statistical Analysis

All data were analyzed using Microsoft Excel 2010 (Microsoft Company, USA) for calculating mean and standard error. Comparisons of gene expression among genes and tissues using analysis of variance (ANOVA) and Duncan's Multiple Range Test (DMRT) for mean separation, as well as, in response to nutrient deficiencies using t-tests were performed in RStudio (RStudio, USA) using standard R packages (https://www.R-project.org). Resulting p-values from t-tests were corrected for false discovery in multiple hypotheses testing by manually calculating adjusted p (Q) values in Excel using the Benjamini-Hochberg method (Benjamini and Hochberg, 1995). Comparisons of gene expression in response to metal toxicity and rhizobia inoculation were also performed using analysis of variance (ANOVA) in Excel.

### RESULTS

### Genome-Wide Identification and Bioinformatic Analysis of Soybean NRAMP Family Genes

Thirteen putative GmNRAMP genes were identified in BLAST searches of the G. max cv. Williams 82 reference genome in the Phytozome database using arabidopsis, rice, and medicago NRAMP as query sequences. All identified GmNRAMPs were named based on phylogenetic relationships among soybean NRAMP family members (**Figure 1A** and **Table 1**), with the tree being comprised of two main branches (**Figure 1A**). The NRAMP family domain (PF01566) and 10–12 putative transmembrane domains (TMDs) were also found in each putative GmNRAMP protein (**Table 1** and Figure S1). The CDS regions of putative GmNRAMP genes range in length from 1,521 to 1,767 bp and encode proteins with lengths of 506–588 amino acid residues, molecular weights of 55.44– 64.39 KDa, and pI values of 4.77–9.04 (**Table 1**). Gene structures were similar within each of the two main subfamilies as illustrated in **Figure 1A**. Subfamily I is comprised of 8 members, each with 4 exons and 3 introns, while the five Subfamily II GmNRAMPs each have 13–14 exons and 12– 13 introns. Among the Subfamily II proteins, GmNRAMP7, with a relatively short length, appears to be divergent from the other four members. It is also worth mentioning that GmNRAMP6a and GmNRAMP6b each contain 13 introns, with one intron located in the 5′ UTR (**Figure 1A**). Chromosome mapping showed that the 13 GmNRAMPs are distributed on 11 chromosomes. Chromosomes 6 and 7 each contained two NRAMPs, while chromosomes 1, 4, 5, 8, 11, 13, 15, 16, and 17 each contained one NRAMP (**Figure 1B**). From the phylogenetic tree of soybean NRAMP proteins, we noticed that all but GmNRAMP7 appeared in pairs, implying possible gene duplication occurred during evolution of NRAMP gene family, therefore, synteny analysis also performed to determine the potential gene duplication with soybean NRAMP family. As shown in **Figure 1B**, six pairs of soybean NRAMP genes were found to be located in duplicated blocks, suggesting the duplication event also happened in the soybean NRAMP gene family.

In BLASTP analysis, all alignments included at least 73% of each GmNRAMP sequence, while 102 of the 169 alignments incorporated over 90% of the protein sequence (data not shown). Sequence identity in aligned regions ranged from 38 to 98% outside of self hits (**Table 2**), with the highest percentage of identity occurring between GmNRAMP3a and GmNRAMP3b which also fell very close to each other in the phylogenetic tree (**Figure 1A**). The lowest identity, on the other hand, occurred between GmNRAMP4b and GmNRAMP7 (**Table 2**). The reported consensus transport residues, GQSSTITGTYAGQY(/F)V(/I)MQGFLD(/E/N) were present in all identified soybean NRAMP sequences (**Figure 2**).

Further phylogenetic investigation was conducted with the inclusion of NRAMPs from other plant species, bacteria, fungi and humans. As expected (**Figure 3**), NRAMPs from Deinococcus radiodurans (DraNRAMP), Staphylococcus capitis (ScaDMT), and Saccharomyces cerevisiae (ScSMF1, ScSMF2, and ScSMF3) were closely related and separated from higher plant and human NRAMPs (HsDMT1 and HsNRAMP1). The tested higher plant NRAMPs from soybean, arabidopsis, rice, medicago, barley, peanut, apple, and mustard, with one exception, fall into two large groups, subfamily I and subfamily II, with both monocots and dicots represented in each of these two groups. Arabidopsis carries four subfamily I members and two subfamily II members, while the corresponding subfamily I and II numbers were two and five for rice, and four and three from medicago. The 13 soybean NRAMP members sorted into eight subfamily I members and five subfamily II members, which matched the phylogenic and gene structure results described above. In addition, the tested NRAMP proteins from mustard clustered into subfamily I, while the NRAMP proteins from peanut, barley, and apple clustered into subfamily II (**Figure 3**). Interestingly, the subcellular localization predictions for GmNRAMPs indicate that subfamily I members localize to vacuoles, while subfamily II members localize to the plasma membrane (**Table 1**). To better understand the potential regulation of GmNRAMP genes, cis-element analysis was performed and shown in Table S4. A number of cisacting regulatory elements involved in light responsiveness, were frequently identified in soybean NRAMP genes promoter regions (Table S4). In addition, cis-regulatory elements in GmNRAMPs were also associated with various stress factors. Notably, the promoter regions of 9 NRAMP genes contained cis-elements related to defense and stress responsiveness (Table S4). Most NRAMP gene promoters contained at least one HSE element (heat stress responsiveness), which was followed in prevalence by MES (drought-inducible) and LTR (low temperature responsiveness) elements. Furthermore, several

identified elements in GmNRAMP promoter regions have been reported to be involved in hormone responsiveness. Specifically, all GmNRAMPs except GmNRAMP6b were associated with the cis-acting element ABRE, which is involved in abscisic acid (ABA) responsiveness. Plus, the GmNRAMP2a and GmNRAMP4b promoter regions contained a series of elements responsive to nearly all types of hormone, including ABA, Methyl Jasmonate (MeJA), ethylene (ETH), gibberellin (GA), and auxin (IAA). These results indicate that GmNRAMP genes may be involved in complex regulatory networks and could be regulated by various environmental, developmental, and physiological factors.


TABLE 1 | Summary of NRAMP family genes in soybean.

### Tissue-Specific Expression of GmNRAMPs

In order to investigate tissue-specific expression profiles of GmNRAMPs, qRT-PCR analysis was performed with seven soybean tissues, namely young leaves, older leaves, stems, roots, flowers, pods, and seeds. Using an ANOVA significance threshold of p = 0.05, expression varied among tissues for each of the 13 tested GmNRAMPs, as well as, among GmNRAMP genes within each tissue. Quantified expression levels for each GmNRAMP gene in each of the seven tissues are displayed in **Figure 4**, with those in the most highly expressed category according to Duncan's Multiple Range Test marked by an "a" for the tissues in which each gene was most highly expressed, and an "∗" for the most highly expressed GmNRAMPs in each tissue. Where GmNRAMP transcription was above the detection limit, expression levels varied by over 50-fold among tissues and GmNRAMP genes. Expression was low for GmNRAMPs 4b and 5b in all tissues, but was highest for both of these genes in flowers, as well as in pods for GmNRAMP4b. All GmNRAMPs were detected in each tissue, except seeds. The most highly expressed GmNRAMP in each tissue was GmNRAMPs 1a, 1b, 2b, 3b, or 7, with GmNRAMPs 1a, 1b, 2b, and 3b being notable for relatively high expression in multiple tissues. For leaves, relatively high expression was also observed for GmNRAMPs 2a, 3a, and 6a in young leaves, and for GmNRAMP 5a in old leaves. Higher expression was also observed for GmNRAMPs 6a and 6b in stems, and for GmNRAMP 5a in roots. In flowers and pods, appreciable expression was observed for all GmNRAMPs, except for 5a and 5b in pods. One notable result was that the most highly expressed GmNRAMPs typically belonged to subfamily II in all tissues, except in roots, where the subfamily I GmNRAMP 7 was the most highly expressed GmNRAMP. Taken together, structures, phylogenies (**Figure 1**) and expression patterns (**Figure 4**) demonstrate that structurally similar GmNRAMPs also exhibit similar expression patterns.

### Expression of GmNRAMPs in Response to Macronutrient Deficiency

To evaluate potential responses of GmNRAMPs to macronutrient deficiencies, expression was assessed by qRT-PCR in soybean plants exposed to deficiencies of nitrogen (N), phosphorus (P), or potassium (K). GmNiR, which was repressed by N deficiency (Qin et al., 2012), together with GmPLDZ (a low P responsive gene) and GmHAK (a low K responsive gene), which had been demonstrated to respectively enhanced by P or K deficiency (Qin et al., 2012), were used to verify the nutrient deficiency treatments in this study (Figures S2A–C). Significant effects of macronutrient deficiencies were determined using FDR corrected t-tests for each GmNRAMP within each tissue for each nutrient treatment. As shown in **Figure 5**, the ratio of expression in deficient conditions relative to sufficient conditions significantly varied from constant expression for each GmNRAMP in one or more conditions and tissues. More specifically, N deficiency led to decreased expression of GmNRAMPs 1a, 2a, 2b, and 6a in leaves, and GmNRAMPs 5a, 6b, and 7 in roots, while GmNRAMP5a was up-regulated in leaves, and GmNRAMPs 3b and 6a were up-regulated in roots. Under P deficiency conditions, GmNRAMP expression was dramatically altered. Nine GmNRAMPs, 1a, 1b, 3b, 4a, 5a, 5b, 6a, 6b, and 7, were up-regulated by P deficiency in leaves. In roots, GmNRAMPs 3b, 5a, and 6a were up-regulated, while GmNRAMP7 was down-regulated. In comparison to N and P deficiency responses, K deficiency resulted primarily in decreased expression of GmNRAMP genes. GmNRAMPs 6a, 6b, and 7 were down-regulated in both leaves and roots, while GmNRAMP5b was down-regulated in roots, and, in leaves, GmNRAMPs 1a and 3b were down-regulated and GmNRAMPs 1b and 5a were upregulated. Among treatments and tissues, a few GmNRAMPs exhibited more notable responses. GmNRAMPs 5a, 6a, and 7 responded to all treatments, except for GmNRAMP7 in low N leaves and GmNRAMP5a in low K roots. Furthermore,


GmNRAMP7 responses were in the negative direction, except in low P leaves, while GmNRAMP5a responses were in the positive direction, except in low N roots. Two GmNRAMPs, 3a and 4b, displayed considerable variation in relative expression among tissues and nutrient treatments (**Figure 5**), yet this variation was not significant due to the overall low level of expression for each of these genes (**Figure 4**).

### Expression of GmNRAMPs in Response to Iron or Sulfur Deficiency

To further investigate potential roles for soybean NRAMP genes in nutrient homeostasis, GmNRAMPs were tested for alterations in expression in response to Fe or S deficiency. Two known marker genes, GmIRT (for low Fe) and GmSULTR1;2b (for low S) (Qin et al., 2012; Ding et al., 2016), were also applied to confirm the Fe or S deficiency in this study (Figures S2D,E). As shown in **Figures 6**, **9** GmNRAMP genes were significantly down-regulated by Fe deficiency. The abundance of GmNRAMP6a was downregulated in both Fe-deficient leaves and roots. GmNRAMPs 1a, 3a, 3b, 4a, 4b, and 6b were also down-regulated only in leaves, and GmNRAMPs 5a and 5b were also down-regulated only in roots. Four NRAMP genes responded to Fe starvation with enhanced expression levels. Among them, GmNRAMP2a and GmNRAMP2b exhibited increased expression both in leaves and roots, while GmNRAMP7 expression was enhanced in roots, and GmNRAMP1b was up-regulated in leaves. Furthermore, except for GmNRAMP2a and GmNRAMP5b, all soybean NRAMP genes also responded to S deficiency (**Figure 6**). Interestingly, responses to S deficiency were opposite of those observed for Fe deficiency, with the exception of three GmNRAMPs. Specifically, GmNRAMP5a was down-regulated in both Fe- and S-deficient roots, GmNRAMP1b was up-regulated in both Fe- and Sdeficient leaves, and GmNRAMP2a, responded to Fe, but not to S (**Figure 6**).

### Expression of GmNRAMPs in Response to Divalent Metal Toxicity Stresses

In order to evaluate the probable functions of GmNRAMPs in responses to divalent metal toxicity stresses, expression of these genes was also assayed by qRT-PCR in soybean seedlings exposed to excess supply of Fe, Cu, Cd, or Mn. Due to low abundances of GmNRAMPs 4a, 4b, and 5b in leaves and roots under these treatments, only 10 NRAMP genes were analyzed in this experiment. In general, the expression of soybean NRAMP gene family members was most sensitive to Cd toxicity, followed by Cu toxicity. The expression of six GmNRAMP genes was greatly enhanced by excess Cd, with four of them were being up-regulated in both leaves and roots, and two of them were being enhanced in roots (**Figure 7** and **Table 3**). While three NRAMP genes responded to Cd toxicity with reductions in expression levels (**Figure 7** and **Table 3**). Under Cu toxicity stress, GmNRAMP5a was upregulated in both leaves and roots, and GmNRAMP1a was upregulated in roots, whereas the expressions of GmNRAMP6a and GmNRAMP2a were down-regulated in leaves and roots, respectively. In addition, two NRAMP genes, GmNRAMP2b

FIGURE 2 | Multiple alignment of soybean NRAMP family proteins. Multiple alignment was performed with Clustal W and the residues were colored using Genedoc software. Red lines underneath alignments indicate reported consensus transport residues. "\*" above the sequence mean every ten amino acid residues. The logo of these residues was then generated in WebLogo 3 online.

and GmNRAMP3a exhibited opposite trends in soybean leaves and roots under excess Cu stress. In another, the expression of soybean NRAMP genes seemed less influenced by excess Fe supply in comparison to the other toxicity treatments or to Fe deficiency, with only three genes, GmNRAMP2a, 2b, and 7 being suppressed by Fe toxicity (**Figure 7** and

**Table 3**). Finally, in the Mn toxicity treatment, expression was altered for only four NRAMP genes, with three being up-regulated, while one was down-regulated (Figure S3 and **Table 3**).

### Subcellular Localization of GmNRAMP Proteins

To explore the subcellular localization of the GmNRAMP proteins, putative subcellular localizations were first identified by ProtComp analysis. As shown in **Table 1**, eight NRAMP proteins were predicted to localize to the vacuole, whereas the other five GmNRAMP proteins were predicted to target the plasma membrane. Subsequent to this computational analysis, subcellular localization for six selected GmNRAMPs proteins was empirically observed though transient expression of GFP: GmNRAMP fusions in Arabidopsis protoplasts containing the membrane marker OsMCA1 (Kurusu et al., 2012) or the tonoplast marker AtTPK1 (Voelker et al., 2006), with verification of these two specific localization makers conducted in co-transformed rice protoplasts (Figure S4). Microscopic observation revealed that the 35SGFP::GmNRAMP1a, 35SGFP::GmNRAMP2a, 35SGFP::GmNRAMP2b, and 35SGFP::GmNRAMP3a fusions were exclusively localized to the tonoplast, as evidenced by co-localization with the known tonoplast marker AtTPK1 (**Figure 8A**). The 35SGFP::GmNRAMP5a and 35SGFP::GmNRAMP7 fusions

localized on the plasma membrane along with the membrane marker OsMCA1 (**Figure 8B**), and 35SGFP empty vector controls yielded whole cell fluorescence (**Figure 8C**). These results indicate that GmNRAMP proteins localize to different subcellular compartments. Furthermore, localization might be associated with specific biological functions in plant cells.

### Bioinformatic Analysis of Protein-Protein Interactions Involving GmNRAMPs

To explore potential interactions among GmNRAMP members, protein-protein interaction analysis was conducted computationally in the Uniprot and STRING database. Predicted interacting proteins were nearly identical for 10 of the 13 GmNRAMP query proteins (network I in Figure S5A), including for GmNRAMP1a, GmNRAMP2a, and GmNRAMP4b to GmNRAMP7. All of the proteins predicted to interact with these 10 NRAMPs contain multicopper oxidase domains, and may be involved in the metabolism of ascorbate and aldarate (Table S5). In contrast to network I, GmNRAMP2b, GmNRAMP3a, and GmNRAMP3b were predicted to interact in another set of similar interaction networks as shown in Figure S5B. The putative networks for GmNRAMP3a, GmNRAMP3b, and GmNRAMP2b are very similar to each other, with seven identical interacting proteins labeled with a star in Figure S5B. These seven proteins include a Zn transporter, a vacuolar Fe transporter and a ferric reductase (Table S5). The interaction networks involving GmNRAMP3a and GmNRAMP3b share two proteins in common with network I as labeled with check marks, both of which contain multicopper oxidase domains. Meanwhile, GmNRMAP2b is predicted to interact with RBCS-1 (Ribulose bisphosphate carboxylase) and HDL56 (Transcription factor HEX, containing HOX, and HALZ domains). Interestingly, nodulin-21 and a nodulin-like protein were also found in the interaction networks of GmNRAMP2b, GmNRAMP3a, and GmNRAMP3b, suggesting potential functions for these GmNRAMPs in Fe or other metal ion transport in soybean nodules.

### Expression of GmNRAMPs in Soybean Nodules with Rhizobia Inoculation

In order to determine whether GmNRAMP genes function in soybean nodules, the expression of GmNRAMPs were tested in different tissues of soybean 30 days after rhizobia inoculation. Due to low expression of GmNRAMP4a and GmNRAMP4b under low N conditions, only 11 GmNRAMP genes were evaluated in this experiment. As shown in **Figure 9**, expression of GmNRAMP1a, GmNRAMP1b, GmNRAMP6a, and GmNRAMP6b were lower in soybean nodules than in other tissues upon inoculation with rhizobia. Compared with non-inoculated soybean, expressions of GmNRAMPs 2a, 2b, 5a, 5b, and 6b were significantly reduced in soybean roots after inoculation with rhizobia (**Figure 9**), during which time expression in nodules ramped up considerably. Moreover, under rhizobial-inoculation conditions, expression of GmNRAMP2b,

GmNRAMP3a, GmNRAMP3b, and GmNRAMP7also scaled up in soybean nodules, as indicated by the respective 21.66-, 13.96-, 11.96-, and 6.8-fold differences compared to expression in soybean roots (**Figure 9**). Higher expression of GmNRAMP2b, GmNRAMP3a, and GmNRAMP3b in nodules compared to other GmNRAMPs suggests that these genes might be important for the transport of Fe or other metal ions in soybean nodules.

## DISCUSSION

NRAMP proteins are exist in a wide range of bacteria, animals and plants, to date, the NRAMP gene family has been reported in a number of plant species, but information on this family is limited for soybean, the most important cultivated legume. In this study, the soybean genome was comprehensively searched for NRAMP genes, which resulted in the identification of 13 putative GmNRAMP genes. Phylogenetic analysis clustered these 13 NRAMP proteins into two distinct subfamilies (**Figure 1A**). Interestingly, 12 of the 13 GmNRAMP proteins further clustered into six branches of paired proteins (**Figure 1A** and **Table 2**). These homologous pairs of GmNRAMPs might be the products of duplication events in soybean evolutionary history (**Figure 1B**). Altogether, the combination of homologous pairs and two subfamilies based on structural similarities guided the phylogenetically based naming of GmNAMPs employed herein. In Arabidopsis, the analysis of six NRAMP proteins has also revealed two subfamilies (Thomine et al., 2000; Mäser et al., 2001). Similarly to the current findings for soybean NRAMP genes, AtNRAMP2 through 5 in subfamily I have 2–3 introns, while AtNRAMP1 and AtNRAMP6 in the other subfamily have 10 and 12 introns (data not shown in paper). On the other hand, unlike AtNRAMP1 and AtNRAMP6, two soybean NRAMP genes in Subfamily II have an intron located on the 5′ UTR (**Figure 1A**). Previous studies have associated the presence of an intron within the 5′ UTR with enhanced RNA and protein accumulation (Rethmeier et al., 1997; Chung et al., 2006). In the current study, the two GmNRAMPs, GmNRAMP6a and GmNRAMP6b, which containing an intron within the 5′ UTR, were not among the most highly expressed genes in any of the tested tissues or conditions. Whether these 5′ UTR introns play roles in the regulation of GmNRAMP6a or GmNRAMP6b remains to be determined. Next, phylogenetic analysis of the 13 GmNRAMP proteins identified here, along with NRAMP protein sequences from other plant species (**Figure 3**), revealed that NRAMP proteins representing both subfamilies are common in both dicots and monocots.

This suggests that the NRAMP family fulfills similar and basic functions in widely divergent plant species. Interestingly, soybean NRAMP proteins in the same subfamily were predicted and confirmed to have similar subcellular localizations (**Figure 8** and **Table 1**). Such compartmentalization is likely related to specific functions, namely, uptake of metals on the plasma membrane, or release from vacuolar stores on the tonoplast as previously reported in Arabidopsis and rice. Specifically, AtNRAMP3 and AtNRAMP4 have been reported to be localized in Arabidopsis vacuoles, which is in consistent with roles in the release of metals from vacuolar stores (Lanquar et al., 2005).

Meanwhile, OsNRAMP1, OsNRAMP3, and OsNRAMP5 have been identified as plasma membrane-localized proteins in rice participate in metal uptake (Takahashi et al., 2011; Sasaki et al., 2012; Yamaji et al., 2013). Phylogenetic analysis with these known NRAMP proteins implies potentially similar functions for soybean NRAMP proteins.

NRAMP genes functions in metal ions uptake, especially Fe, are widely found in mice, humans and plants, particularly under Fe-deficiency conditions. Although the IRT/FRO system seems to be a major component of Fe-uptake system in the non-graminaceous plants, several previous researches revealed

FIGURE 7 | Expression of GmNRAMPs in response to different metal toxicity stresses. Ten-day-old seedlings were treated with excess Fe (1,000 µM EDTA-Fe), Cu (200 µM CuSO4·5H2O), and Cd (100 µM CdCl2) treatments for 24 h. Each bar is the mean of four biological replicates with standard error. "\*" indicates effects of toxicity treatments relative to controls in one-way analysis of variance, \*P < 0.05, ns: not significant.


TABLE 3 | Summary of the response of soybean NRAMP genes to different divalent metal toxicities.

Significant differences between treatment and control are marked as arrows. "↑" stands for up-regulated expression in response to the treatment, while "↓"stands for down-regulated expression in response to the treatment.

that other genes could be involved in this process, such as NRAMP genes. In Arabidopsis, AtNRAMP1 can complement the Fe uptake mutant of yeast, and appears to be expressed preferentially in Fe-deficient roots, indicating its role in Fe uptake and transport (Curie et al., 2000). In soybean, 13 GmNRAMP genes were significantly affected by Fe deficiency in leaves or roots of soybean (**Figure 6**), indicating that GmNRAMP genes also are involved in Fe nutrition. For root-expressed GmNRAMP genes, GmNRAMP7 was remarkably up-regulated by Fe deficiency in roots (**Figure 6**), combining with its plasma membrane localization (**Figure 8**), implying that GmNRAMP7 might participate in the acquisition of Fe on the plasma membrane in root cell, perhaps also coupling with IRT and FRO as integral part of plant root cell Fe uptake machinery under Fe starvation stress. However, this hypothesis would be further investigated by the tissue and cell specific localization of GmNRAMP7 by GUS staining and GFP fluorescence in future trails. Besides, previous study showed that overexpression of AtNRAMP1 leads to an increased resistance to toxic Fe level (Curie et al., 2000), indicating it may participates in Fe remobilization under Fe deficiency, despite this function was not consistent with its plasma membrane localization. However, it is uncertain that this kinds of gene always going to be present in cell membranes? It's interesting to speculate that if AtNRAMP1 or GmNRAMP7 was not always present in cell membrane, whether it might performed potential functions in Fe or other metals transport in root cell, for example, remobilization of Fe stored in organelles under Fe-deficient conditions as MxNRAMP1 which mainly exists in the plasma membrane and vesicles (Pan et al., 2015), or regulate homeostasis of free ions which might induced by Fe starvation, mediate sequestration of free ions into a cellular compartment, such as plastid or vacuole, all of these speculations also need further study.

Beyond functioning in Fe transport, NRAMP genes have been also demonstrated to perform wide ranging transport activities for divalent transition metals, including Mn2+, Fe2+, Co2+, Ni2+, Cu2+, and Zn2<sup>+</sup> (Illing et al., 2012). Recently, several studies have reported the substantial role of NRAMP family members in Mn uptake. As shown in **Figure 3**, four NRAMP proteins, GmNRAMP5a, GmNRAMP5b, GmNRAMP6a, and GmNRAMP6b cluster together into a small phylogenetic branch with AtNRAMP1, a plasma membrane localized high-affinity Mn transporter (Cailliatte et al., 2010). This small branch is also closely related to OsNRAMP3, which is a known Mn transporter that is involved in shoot Mn distribution, and is constitutively expressed in nodes, stems and panicles, where it mediates adaptation of rice to a wide change of external Mn conditions (Yamaji et al., 2013; Yang et al., 2014). Interestingly, one soybean NRAMP protein, GmNRAMP5a, clustered in the phylogenetic tree with AtNRAMP1 and OsNRAMP3 (**Figure 3**), and was also affected by Mn as indicated by the significant increase in its expression in response to Mn toxicity (Figure S3 and **Table 3**), suggesting it is probably important for Mn homeostasis in soybean. In addition, NRAMP genes also seem to affect the intracellular remobilization of divalent toxic heavy metals, such as Cd (Cailliatte et al., 2009), which might contribute to increase plant tolerance to heavy metal toxicity. In this study, the responses of GmNRAMP genes to different divalent metal toxicities were also investigated (**Figure 7** and **Table 3**). As expected, the remarkable changes in expression were observed for most GmNRAMPs under excess Cd treatment, suggesting that these genes might contribute to Cd tolerance in soybean plants under Cd toxicity stress. Interestingly, among 13 soybean NRAMP genes, the expressions of GmNRAMP1a, GmNRAMP3a, GmNRAMP5a in roots were both dramatically enhanced by Cu and Cd toxicities (**Figure 7** and **Table 3**), furthermore, the expression of GmNRAMP5a in roots were significantly increased by Cu, Cd, and Mn toxicities (**Figure 7**, Figure S3 and **Table 3**). In consequence, we propose that soybean NRAMP genes widely participated in Fe, Mn, Cu, and Cd transport, might be involved in the uptake and homeostasis regulation of these metal ions.

Moreover, considering the fact that macronutrient deficiencies often exist in field conditions, the responses of soybean NRAMP genes to N, P, and K deficiencies were also evaluated in this study. Most GmNRAMP responses to N deficiency involved down regulation, while, in contrast, a majority of GmNRAMP

genes were remarkably enhanced by P deficiency (**Figure 5**). This enhancement of NRAMP gene expression under low P conditions is similar to patterns observed for soybean Fe-S assembly genes (Qin et al., 2015). A considerable amount of research has outlined interactions between P and Fe response pathways in plants. Phosphorus deficiency can increase carbon flux through glycolysis for the synthesis of organic acids, change lipid metabolism, and affect the abundance of genes involved in Fe and Zn metabolism (Wasaki et al., 2003; Zheng et al., 2009). Notably, increased Fe concentrations are often observed in P-deficient plants (Misson et al., 2005; Hirsch et al., 2006). The correlated Fe and P responses described in the current and previous studies indicate the existence of linkages between P and Fe metabolism through one or more pathways, including those involved with transport, homeostasis and accumulation. In previous work, P deficiency affected Fe storage, as indicated by the accumulation of Fe associated with ferritin in chloroplasts (Hirsch et al., 2006). More recently, cross-talk between P and Fe homeostasis pathways has been demonstrated in Arabidopsis, where the ferritin gene, AtFer1, is regulated by the phosphate starvation response transcription factor AtPHR1 (Bournier et al., 2013). In contrast, P deficiency does not appear to affect Arabidopsis genes encoding strategy I Fe-uptake system proteins, such as IRT1 (Hirsch et al., 2006). Whether the expression of other Fe uptake transporters is triggered in P-deficient Arabidopsis remains unknown. In this study, significant changes in the expression of GmNRAMP genes in P-deficient soybean plants reinforces the idea that plant P and Fe response pathways are linked and interact, and further imply that low P impacts Fe uptake and homeostasis through the activities of GmNRAMP transport systems. However, comprehensive elucidation of linkages between P and Fe metabolic pathways and the underlying mechanisms requires further investigation. Interactions between P and Fe pathways in plants has also been observed in jasmonate ZIM domain (JAZ) genes, which act as transcriptional repressors of jasmonateresponsive genes, as indicated by highly altered expression of such genes in rice and chickpea experiencing mineral nutrient deficiency (Singh et al., 2015). What's more, observations of the involvement of JAZ genes in nutrient deficiency responses contribute to further understanding of the roles jasmonates play in regulating nutrient deficiency response adaptations. It is noteworthy that several cis-acting elements involved in hormone-responsiveness were found in the putative promoter regions of soybean NRAMP genes (Table S4), but the regulatory mechanisms remain unclear. Whether NRAMP involvement in the uptake or homeostasis of metal ions is related to hormoneregulated morphological and physiological responses to nutrient and toxin status requires further study. One final interaction relevant here is the coordination of Fe with S in Fe-S proteins, which are the biggest Fe sink in plants. Unsurprisingly, therefore, interplay between Fe and S has been noted in recent years (Zuchi et al., 2015). Interestingly, most soybean NRAMP genes displayed contrasting responses to Fe and S deficiencies (**Figure 6**). Since Fe-S clusters are significant Fe sinks within cells, the deprivation of Fe or S will depress the synthesis of Fe-S clusters. Changes in GmNRAMP expression in response to Fe- or S-deficiency suggest links between these two stress responses. More definitive demonstrations of whether NRAMP genes are involved in coordinated regulation of Fe and S homeostasis, as well as, the synthesis of Fe-S clusters requires further investigation.

It is worth mentioned that several soybean NRAMP proteins were not only found to interact with metal ion transporters, such as the Zn/Fe transporter, but several were also observed in interactions with nodulin and nodulin-like protein (Figure S5 and Table S5). The high abundance of GmNRAMP2b, GmNRAMP3a, and GmNRAMP3b in soybean nodules confirmed a predicted interaction network (Figure S5B) and suggested possible functions for these GmNRAMP genes in soybean-rhizobia symbiosis. Besides these three NRAMP genes, GmNRAMP2a and GmNRMAP7 were also detected in nodules at relatively high expression levels (**Figure 9**). However, these two proteins were not predicted to interact with nodulationrelated proteins. The expression patterns of GmNRAMP2a in various soybean tissues subsequent to rhizobia inoculation did not correspond to those of its homolog GmNRAMP2b, which might be due to the evolutionary divergence. As shown in **Figure 3**, the nodule-expressed subfamily II gene GmNRAMP7 has diverged from the other four nodule-expressed NRAMP genes. GmNRAMP7 is the only soybean NRAMP protein in a small branch with several monocot NRAMPs (including OsNRAMP1, OsNRAMP5, and HvNRAMP5), which are mainly expressed in roots (Sasaki et al., 2012; Wu et al., 2016). This is consistent with the tissue-specific expression of GmNRAMP7 (**Figure 4**). It also clusters with AhNRAMP1, a Fe transporter proven to be involved in Fe acquisition (Xiong et al., 2012). Most notably, GmNRAMP7 also clusters closely with MtNRAMP1, which has been reported to be most highly expressed in roots and nodules, as well as, the main transporter responsible for Fe uptake in nodule cells (Tejada-Jiménez et al., 2015). Taken together, the results herein imply that GmNRAMP7 might function in Fe uptake during symbiotic nitrogen fixation. Finally, GmNRAMP7 is predicted to be localized to the plasma membrane, while the four other nodule-expressed NRAMP genes were predicted to localize on the tonoplast. Any specific roles for these genes in the regulation of metal ions homeostasis requires further study.

### CONCLUSION

In the present study, 13 NRAMP genes were identified from the soybean genome and, subsequently, named according to the phylogenetic relationships inferred among them. Expression profiles of GmNRAMP genes varied among soybean tissues and in response to a series of nutrient stresses, which suggests that GmNRAMPs perform a range of functions in specific tissues throughout growth and development. Furthermore, this gene family also likely participates in crosstalk among different nutrient stress pathways. Then the subcellular localization analysis in Arabidopsis protoplasts confirmed the tonoplast or plasma membrane localization of soybean NRMAP proteins. Taken together, the results reported here comprise a systematic genome-wide analysis of the soybean NRAMP gene family. These results supply basic and important information for understanding the putative functions of NRAMP genes in soybean. Moreover, protein-protein interaction network and qRT-PCR analysis in rhizobia-infected soybean further revealed that 3 NRAMP proteins putatively interact with nodulin-like proteins and are markedly up-regulated in soybean nodules. These results suggest potential functions for a subset of GmNRAMP proteins in the uptake of Fe or other metals, as well as, regulation of homeostasis in soybean nodules. Overall, this study provides valuable information for further functional studies on the biological roles of NRAMP genes in soybean. Plus, this report provides a basis for further understanding crosstalk between different nutrient response pathways in plants.

### AUTHOR CONTRIBUTIONS

LQ and XL conceived the study, analyzed the data, and drafted the manuscript. LQ, XH, and LX cultivated the soybean materials and collected the soybean samples. PH, LC, and YL extracted

### REFERENCES


RNA and performed the qRT-PCR experiments. LC designed the vectors for subcellular localization. TW conducted the statistical analyses of raw data. HL and TW revised the manuscript. All authors read and approved the final manuscript.

### ACKNOWLEDGMENTS

This research was financially supported by National Natural Science Foundation of China (Grant No. 31301833), Agricultural Science and Technology Innovation Program (CAAS-ASTIP-2013-OCRI) and Excellent Young Scientist Fund (1610172015004) of Chinese Academy of Agricultural Sciences.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2017. 01436/full#supplementary-material


in Arabidopsis with homology to Nramp genes. Proc. Natl. Acad. Sci. U.S.A. 97, 4991–4996. doi: 10.1073/pnas.97.9.4991


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Qin, Han, Chen, Walk, Li, Hu, Xie, Liao and Liao. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Harnessing Finger Millet to Combat Calcium Deficiency in Humans: Challenges and Prospects

Swati Puranik <sup>1</sup> , Jason Kam<sup>1</sup> , Pranav P. Sahu<sup>1</sup> , Rama Yadav <sup>1</sup> , Rakesh K. Srivastava<sup>2</sup> , Henry Ojulong<sup>3</sup> and Rattan Yadav <sup>1</sup> \*

1 Institute of Biological, Environmental and Rural Sciences, Aberystwyth University, Aberystwyth, United Kingdom, 2 International Crops Research Institute for the Semi-Arid Tropics, Patancheru, India, <sup>3</sup> International Crops Research Institute for the Semi-Arid Tropics, Nairobi, Kenya

Humans require more than 20 mineral elements for healthy body function. Calcium (Ca), one of the essential macromineral, is required in relatively large quantities in the diet for maintaining a sound overall health. Young children, pregnant and nursing women in marginalized and poorest regions of the world, are at highest risk of Ca malnutrition. Elderly population is another group of people most commonly affected by Ca deficiency mainly in the form of osteoporosis and osteopenia. Improved dietary intake of Ca may be the most cost-effective way to meet such deficiencies. Finger millet [Eleusine coracana (L.) Gaertn.], a crop with inherently higher Ca content in its grain, is an excellent candidate for understanding genetic mechanisms associated with Ca accumulation in grain crops. Such knowledge will also contribute toward increasing Ca contents in other staple crops consumed on daily basis using plant-breeding (also known as biofortification) methods. However, developing Ca-biofortified finger millet to reach nutritional acceptability faces various challenges. These include identifying and translating the high grain Ca content to an adequately bioavailable form so as to have a positive impact on Ca malnutrition. In this review, we assess some recent advancements and challenges for enrichment of its Ca value and present possible inter-disciplinary prospects for advancing the actual impact of Ca-biofortified finger millet.

Keywords: finger millet, calcium, osteoporosis, bioavailability, food processing, biofortification, genetic improvement, plant breeding

### IMPORTANCE OF CALCIUM IN HUMAN DIET

Calcium (Ca) is the fifth most abundant element present in the human body, accounting for up to 1.9% of the body weight in adults (Nordin, 1976). Its main functions are to provide rigidity and structure, mediating vascular and muscular contractions or dilations and nerve signal transmission (Institute of Medicine (US) Standing Committee on the Scientific Evaluation of Dietary Reference Intakes, 1997; Nordin, 1997). Ca may also serve in the protective role against various types of cancer viz. colorectal (Institute of Medicine (US) Committee to Review Dietary Reference Intakes for Vitamin D and Calcium, 2011), ovarian (Goodman et al., 2002), breast (Lin et al., 2007), and prostate (Gao et al., 2005). Although not supported by clinical trials, observational studies have associated higher Ca intakes to lower body weight and reduced adiposity, which may be due to lower intracellular Ca in fat cells leading to a higher fat breakdown (Parikh and Yanovski, 2003). Thus, it may reduce the risk of cardiovascular diseases by lowering intestinal lipid absorption, promoting lipid excretion and decreasing cholesterol concentrations in the blood

#### Edited by:

Raul Antonio Sperotto, Centro Universitário UNIVATES, Brazil

#### Reviewed by:

Sukhwinder Singh, International Maize and Wheat Improvement Center, Mexico Prashant Vikram, International Maize and Wheat Improvement Center, Mexico Charu Lata, National Botanical Research Institute (CSIR), India

### \*Correspondence:

Rattan Yadav rsy@aber.ac.uk

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

> Received: 17 May 2017 Accepted: 12 July 2017 Published: 26 July 2017

#### Citation:

Puranik S, Kam J, Sahu PP, Yadav R, Srivastava RK, Ojulong H and Yadav R (2017) Harnessing Finger Millet to Combat Calcium Deficiency in Humans: Challenges and Prospects. Front. Plant Sci. 8:1311. doi: 10.3389/fpls.2017.01311

**369**

(Institute of Medicine (US) Committee to Review Dietary Reference Intakes for Vitamin D and Calcium, 2011). Ca is also known to be important for those who have diabetes (Levy et al., 1994; Vestergaard, 2006; Pittas et al., 2007), particularly type 1 diabetics, who in general have lower bone mineral density than healthy subjects (Ma et al., 2012; Oei et al., 2013).

Given its importance, authorities like the Food and Agriculture Organization (FAO) of the United Nations have set up a recommended daily intake (RDI) of Ca based on age, life stage and gender (Food Agricultural Organization of the United Nations, 2002). During the phases of active growth, Ca equilibrium in the body maintains a stable bone mass. Therefore, FAO recommends that children of 1–3 years consume 500 mg Ca/day, 4–6 years consume 600 mg Ca/day and 7–9 years consume 700 mg Ca/day, which should be increased to 1,300 mg/day during 10–18 years (Food Agricultural Organization of the United Nations, 2002). About 1,000 mg Ca/day is recommended between the ages of 19–65 years in males. The organization also advocates that women should take 1,000 mg Ca each day from 19 years onwards until menopause raising it to 1,200 mg during the last trimester of pregnancy, and to 1,300 mg from 65 years and above (Food Agricultural Organization of the United Nations, 2002).

However, after 50 years in men and in menopausal women, the onset of bone decalcification and demineralization leads to reductions in bone mass causing the disease osteoporosis (Michaelsson et al., 2005). According to International Osteoporosis Foundation, it is a significant problem both in the developed as well as in developing nations (http:// www.iofbonehealth.org/facts-statistics). The World Health Organization (WHO) has declared osteoporosis as the next main public healthcare concern globally, after cardiovascular diseases (CVDs), inflicting almost 75 million people in Europe, the United States of America and Japan alone (Consensus Development Statement, 1997; Haldipur, 2003). With a growing elderly population, osteoporosis is threatening to become a major global economic burden for the already stretched healthcare system across the globe. By 2050, the worldwide cost of treating osteoporosis is forecasted to USD 131.5 billion (Lindsay et al., 2001). Thus, in order to prevent Ca deficiency at this age, it is necessary to sustain the recommended daily intake (RDI) of 1,300 mg/day (especially after 65 years; Food Agricultural Organization of the United Nations, 2002) to ensure maximal bone mass during the developmental stages and to have reduced bone mass loss during old age (Michaelsson, 2009).

Despite the importance of adequate Ca intake for human health and wellbeing, the WHO estimates that low dietary intake of Ca is common across the world (WHO, 2006). On a global scale, it presents a large risk which gets graver in underdeveloped regions of the world. However, the absence of reliable and practical indicators in these areas provides insufficient data, which is a challenge for resolving the actual global status for the prevalence of Ca deficiency. It has recently been determined (mainly based on food supply) that 3.5 billion people were at the risk of Ca deficiency in 2011, with approximately 90% of the affected individuals in Africa and Asia (Kumssa et al., 2015). Most of these regions have an agriculture-based economy and large segments of these populations are typically dependent on what they grow and produce for their Ca need. In such situations, staple crops that can offer adequate Ca requirements, especially for people of low income groups in these countries, are highly recommended. One such Ca-rich, traditional and locally well-adapted crop is finger millet. As opposed to nutritionally deficient cereals, such as rice, its regular consumption has a vast potential to curb the incidences of Ca deficiency. Finger millet also possesses many other health-benefitting traits. For example, it has been highlighted that as a model nutraceutical crop, finger millet can provide excellent solutions to food and health security issues (Kumar et al., 2016a). Being a stress resilient crop requiring minimal inputs for growth it is especially suited for sustainable agriculture (Gupta et al., 2017). Despite this, it has received very little scientific attention, relative to other crops, such as rice, wheat and maize. Some very recent reviews have provided a comprehensive account of molecular mechanisms that may be involved in calcium nutrition in finger millet and how various approaches, including molecular breeding, functional genomics and transgenic technology can elevate Ca accumulation in its grains (Ganapathy, 2017; Sharma et al., 2017). However, beyond this, in order to advance biofortification of finger millet, challenges associated with its impact on human health also needs to be critically evaluated. With this as focus, we have not only attempted to highlight its potential but also the prospects of advancing the bioaccessibility of grain Ca content even further.

### STRATEGIES TO PREVENT CALCIUM DEFICIENCY

Calcium (Ca) deficiency is almost physically undetectable and difficult to diagnose during preliminary stages (Wang et al., 2013), however, if detected, it is usually easy to treat by increasing dietary Ca intake or absorption. Currently, there are many strategies that can be undertaken to manage the problem of Ca deficiency. These mainly include diversification of diet, food fortification, external supplementation and crop biofortification.

Dietary changes by including foods that are naturally rich in Ca, like dairy products, seems to be the most efficient way to prevent Ca deficiency. However, it is not easy to persuade people to go for diversified diets and certain foods cannot be included in the diets. For example, 65% of the world's population is lactoseintolerant and therefore they cannot rely on dairy products for their Ca requirement. Incidentally, most of these lactoseintolerant people live in Asian and African regions (Curry, 2013), which are predominantly dependent on agriculturebased economies. Additionally, due to resource constrain, they may not be able to afford livestock or may solely raise cattle for supplementing income and deny themselves of milk and milk products. Therefore, many communities in these regions, including vegetarians and vegans, need alternative sources to meet their Ca needs.

Another strategy can be an industry-based fortification of foods consumed by target groups. For example, policy decisions in the western world have supported fortification of breakfast cereals, fruit juices, flours and sports drinks with Ca. However, fortification may alter flavor, bioavailability, shelf life or safety of the product, thus making it unpalatable to the consumers. Further, consumers in under-developed nations, with limited education and economic resources, often do not always have access to such fortified foods.

Supplementation is an alternative in which Ca tablets are used to prevent Ca deficiency. Although more accessible to people, this strategy suffers from major drawbacks, mostly associated with their side-effects (Institute of Medicine (US) Committee to Review Dietary Reference Intakes for Vitamin D and Calcium, 2011). The most common content of supplementation pills are the inorganic forms: calcium carbonate and calcium citrate. Large amounts of Ca supplements cause excessive Ca accumulation in vascular and soft tissues like arteries or organs, such as kidneys which can lead to heart attack or kidney stones, respectively (Bolland et al., 2010; Institute of Medicine (US) Committee to Review Dietary Reference Intakes for Vitamin D and Calcium, 2011). Some studies have also linked supplement intake with higher risk of breast cancer, prostate cancer and cardiovascular diseases (Institute of Medicine (US) Committee to Review Dietary Reference Intakes for Vitamin D and Calcium, 2011). Further, taking Ca supplements with meals may also reduce absorption of other minerals like zinc and iron (Cook et al., 1991). Therefore, even though this strategy can potentially reach many people, the associated medical problems make it a non-viable long-term option.

To combat Ca deficiency for a wider and more sustainable impact, alternative solutions that are cost-effective and can easily be adopted especially in the developing countries settings need to be pursued. Genetic biofortification is one such very powerful plant breeding or transgenics-based strategy which can combat the global challenge of micronutrient deficiency (www.copenhagenconsensus.com). As it involves the already established agricultural systems to grow, breed and distribute nutrient-dense staple crops, it can be the most economically and socially feasible approach to integrate nutrition into the diets of the impoverished. Biofortified staple crops have been quantified to have high potential benefits and cost effectiveness as their cost per disability-adjusted life years (DALYs) saved is less than the national per capita income, when compared to other interventions, such as fortification and supplementation (Stein et al., 2005; Asare-Marfo et al., 2014).

Over the last decade several projects, such as those initiated by Harvest Plus, have released nutritionally enhanced rice and wheat varieties (Bouis et al., 2011). Recently, efforts have also commenced for biofortification breeding of small cereals, such as millets. Miles ahead of widely consumed rice and wheat, this group of hardy cereals contribute up to 75% of the total calorie intake for poorest of the poor (O'Kennedy et al., 2006). Therefore, millets hold stronger promise to provide economic, food and nutritional security to people surviving on marginal and resource-poor soils (Muthamilarasan et al., 2016). In this respect, pearl millet has become a target for biofortification of iron and zinc (Manwaring et al., 2016) and has entered the spotlight in combating diabetes (Kam et al., 2016). Similarly, finger millet is also gaining popularity as an ideal target crop for Ca biofortification.

### WHY USE FINGER MILLET AS A MODEL FOR CALCIUM BIOFORTIFICATION?

Finger millet possesses all the quantitative and qualitative traits to serve as a model for Ca biofortification. It stands out as the richest source of Ca among all the cereals (**Table 1**). It has three times more Ca than milk and 10-fold higher Ca than brown rice, wheat or maize (Kumar et al., 2016a). Besides Ca, finger millet is also very rich source of iron, amino acids like methionine, slowly digestible starch and phytochemicals like polyphenols. It is a gluten-free, low fat cereal which is non-allergic and easily digestible. For these characteristics, it is often termed as a "super cereal" (Kumar et al., 2016a). Apart from its nutritional attributes, finger millet has excellent environmental sustainability credentials. It can easily withstand harsh climatic conditions, low soil fertility, requires very little inputs with a short growing season (Kumar et al., 2016a). It can reach the yield potential of up to 10 tons/ha under optimum irrigated conditions (Padulosi et al., 2015). It has excellent storage quality traits and can be valuable in areas where farmers suffer losses due to dearth of post-harvest management.

Therefore, integration of a naturally Ca-rich crop like finger millet in global biofortification programs can be a good starting point to alleviate Ca malnutrition (Sharma et al., 2017). Given that women share a significantly higher proportion of osteoporotic morbidity ("Facts and Statistics," International Osteoporosis Foundation, http://www.iofbonehealth.org/factsstatistics), regular consumption of finger millet during and after pregnancy as well as lactation can provide significant benefits to maternal and child bone health. Another advantage of Ca-enriched finger millet over expensive commercially fortified foods is its affordability to these malnourished areas. For lowincome households which mostly subsist on starchy and bulky foods like rice and cassava for their calorie requirements (https:// www.devex.com/news/better-crops-for-better-nutrition-86583), finger millet ensures a pragmatic solution that no family member (especially women and children) suffers from Ca deficiency.

### PROSPECTING BIOAVAILABILITY OF CALCIUM FROM BIOFORTIFIED FINGER MILLET TO ACHIEVE ADEQUATE INTAKES

Determining the efficacy and biological impact of Ca-biofortified finger millet on better nutrition and improved health is very challenging. It depends mainly on two processes: bioaccessibility and bioavailability of Ca in the seeds. Bioaccessibility is a measure of the nutrient fraction available for absorption after its release from food matrix in the gastrointestinal tract. On the other hand, bioavailability is a utilization-based definition, where the ingested, digested and absorbed nutrient reaches the systemic circulation and exerts a positive effect on health (Carbonell-Capella et al., 2014). Solubility, dialysability and gastrointestinal model are generally used as in vitro methods for measuring Ca bioaccessibility. On the other hand, evaluation of Ca bioavailability is ideally evaluated through in vivo human studies. However, considering the complexity of large-scale human



trials and ethico-legal procedures, Caco-2 cell culture, which behave like human intestinal cells, can provide an alternative analysis.

It is well known that only <30% of the consumed Ca is effectively absorbed (Heaney, 2006). Interestingly, using the in vitro bioaccessibility methods, uncooked finger millet has been found to have 36.6% soluble and 28% dialyzable and bioavailable Ca (Amalraj and Pius, 2015). This is higher than rice (30.4% soluble Ca; 24.7% dialyzable Ca), sorghum (31.9% soluble Ca; 26.0% dialyzable Ca) and maize (25.4% dialyzable Ca). Therefore, finger millet in itself is an effective source of bioavailable Ca than many other staple cereals and its improvement through biofortification is an effective strategy that can relegate Ca deficiency.

However, the fate of actual Ca bioavailability from finger millet relies on and is challenged by several other factors. These include intrinsic grain property (solubility, interaction with other constituents of the food matrix) and extrinsic factors (condition of the host, food processing and storage). Therefore, a better understanding of how these factors influence and impact Ca efficacy becomes essential before the biofortified finger millet can carve a path in the farmer and consumer markets.

### Grain's Intrinsic Factors That Impact Ca Bioavailability in Finger Millet

Many plant-based Ca sources have limited accessibility of Ca for absorption due to the formation of insoluble complexes. Phytate and oxalate are two such bioavailability limiters that can impede Ca absorption as they exhibit a strong negative correlation with Ca bioaccessibility (Kamchan et al., 2004; Gibson et al., 2010; Krishnan et al., 2012). Several studies in legumes and cruciferous vegetables have reported that high in vitro Ca solubility and dialysability corresponds to low levels of phytate, oxalate and dietary fiber (Lucarini et al., 1999; Kamchan et al., 2004). In cereals, phytate and oxalate were shown to account for 7 and 15–20% inhibition of Ca bioavailability, respectively (Amalraj and Anitha Pius, 2015). In wheat and barley, phytate, but not fiber, has been proclaimed as having the major inhibitory effect on Ca absorption (Kennefick and Cashman, 2000). Phenolic compounds like tannins reduce the bioavailability of minerals by forming insoluble complexes with divalent metal ions (Rao and Prabavathi, 1982). An in vivo digestibility trial on birds fed on low (1%), medium (2%), and high (3%) tannin sorghum diets showed that as compared to control, the Ca absorption reduced by 1.22, 1.67 and 2.22 fold, respectively (Mahmood et al., 2014).

Likewise, finger millet also contains these antinutrients that negatively affect grain palatability and can be a constraint to its Ca bioaccessibility. There is a wide range of phytate and oxalate content in finger millet based on the genotypes. The phytate content in finger millet ranges from 679 mg/100 g to 1,419.4 mg/100 g (Antony and Chandra, 1999; Makokha et al., 2002). The grains have been found to contain higher phytate content (783.5 mg/100 g) than rice (289.9 mg/100 g), pearl millet (518.5 mg/100 g) and sorghum (571.1 mg/100 g) but lower than wheat (792.1 mg/100 g) and maize (851.5 mg/100 g) (Amalraj and Pius, 2015). Similarly, finger millet grains have been reported to contain oxalic acid to the extent of 45.7 mg/100 g (Rachic and Peters, 1977). Out of the total oxalate fraction present in the food matrix, soluble oxalate has the ability to bind Ca and reduce its absorption. In a recent study, it was found that finger millet has higher total oxalate content (11.3 mg/100 g) than other cereals (except pearl millet; 20.0 mg/100 g) but had the lowest percentage of soluble oxalate (45.9%) among other cereals (Amalraj and Pius, 2015). Even though the phytate and total oxalate content of finger millet are higher than many other cereals, it still contains more bioavailable Ca percentage (28%) than rice (24.7%), maize (25.4%), and sorghum (26%) (Amalraj and Pius, 2015).

Finger millet grains also have a wide range of total phenolics and tannins content (Devi et al., 2014). Tannin content estimation has revealed that the African varieties of finger millet have about three times more tannin percent than the Indian varieties (Ramachandra et al., 1977). Finger millet has been shown to contain up to 264.1 mg/100 g tannin (Amalraj and Pius, 2015). This is much higher than maize (25.5 mg/100 g) and rice (14.3 mg/100 g) but lower than pearl millet (275.8 mg/100 g) and wheat (287.3 mg/100 g) (Amalraj and Pius, 2015). Despite the knowledge of the extent of varietal variations in tannin content of finger millet, their direct effect on inhibition of Ca absorption and bioavailability has yet not been investigated.

Calcium absorption may also be affected by the non-digestible oligosaccharides and dietary fibers in the food matrix. These compounds reduce the activity of digestive enzymes and slow down the digestion process. Dietary fibers were found to be significantly associated to Ca bioavailability in commercially available rice flakes (Suma et al., 2007). The indigestible starch component (resistant starch) has been found to increase Ca absorption in rats, probably by enhancing its solubility (Schulz et al., 1993). Finger millet is also a great source of resistant starch (Devi et al., 2014). It also has the highest total, soluble and insoluble dietary fiber when compared with wheat, rice, maize, sorghum and pearl millet (Amalraj and Pius, 2015). The role of such factors present in finger millet and other grain crops is generally considered positive in Ca absorption but the magnitude of their stimulatory effect requires further validation by in vitro or in vivo methods.

For enhanced Ca bioavailability from finger millet, grain Ca content needs to be improved with a concomitant but conscious effort for the reduction of antinutrient compounds. This is because these compounds play a vital role in plant development and survival. For example, finger millet tannins are effective in reducing pre- and post-harvest losses as they provide protection against molds, insects and other abiotic stress (Gull et al., 2014). Similarly, phytic acid acts as the main phosphorus store for the seeds (Singh and Raghuvanshi, 2012). These compounds have also called attention due to their nutraceutical value and protective effects against many chronic diseases (Kumar et al., 2016a). Thus, their importance can never be completely disregarded. Engineering their content to become a non-limiting factor in Ca absorption from finger millet must be done in a way that does not negatively affect crop performance. A justified way to accomplish this is by employing efficient and suitable grain processing techniques.

### Influence of Grain Processing on Ca Bioavailability

Processing techniques of the grain can affect the total mineral content and factors associated with their bioaccessibility. Finger millet is a very versatile cereal and can be processed and utilized in numerous ways while retaining the available Ca. Popping by high temperature and short time (HTST) treatment was found to have no effects on the total Ca content in finger millet but lowered the Ca bioaccessibility and polyphenol content by as much as 19 and 22%, respectively (Krishnan et al., 2012). Microwave cooking by boiling also does not greatly improve the percentage of soluble or bioavailable Ca (Amalraj and Pius, 2015). This implies that HTST-based processing and microwave cooking methods do not favor improved bioaccessible Ca from finger millet.

On the other hand, sprouting of finger millet improves the extractability of Ca and lowers antinutrients like phytate and tannins to an undetectable level after 4 days (Mbithi-Mwikya et al., 2000). Flour made from whole grain finger millet was found to have higher Ca content (325 mg/100 g) than those made from decorticated (222 mg/100 g) one (Hemanalini et al., 1980). Decortication is a process of removal of the seed coat matter which is responsible for lowering Ca bioaccessibility. It is interesting to note that processing by decortication significantly improved Ca bioavailability in rats and this was attributed to its low fiber and phytic acid content (Hemanalini et al., 1980). In an evaluation of various processing methods of finger millet on Ca bioaccessibility, seed decortication and malting were found to be the most efficient techniques (Krishnan et al., 2012). Decortication improves Ca bioaccessibility by 37.5%, in spite of lowering the total Ca content by 40%. This increase was attributed to a direct decrease in inhibitory contents present in the seed coat like phytic acid (31% reduction) and polyphenols (70% reduction). Malting, which involves germination and thermal treatment, also influences the bioaccessible Ca content of finger millet in a positive way (Platel et al., 2010; Krishnan et al., 2012). Again, this is because germination process greatly reduces the concentration of phytic acid and polyphenols. Fermented flour or sprouting followed by fermentation also showed marked enhancement in Ca availability (20%) with a concomitant decline in phytates, phenols, tannins, and trypsin inhibitor activity (Sripriya et al., 1997; Antony and Chandra, 1998; Makokha et al., 2002).

The above examples establish beyond doubt that bioaccessibility of Ca from finger millet can be even further improved by simple processing methods which can be scaled up to industrial levels. These methods have potential to add values to both traditional as well as to contemporary value-added food products improving their edible and sensory properties (Hotz and Gibson, 2007; Shobana et al., 2013; Verma and Patel, 2013).

### Host (Extrinsic) Factors That Influence Ca Bioavailability

Apart from the Ca bioavailability parameters, the capability to determine the effect on Ca status on target populations is another specific challenge. Host factors, such as age, gender, dietary patterns may show differential effects of finger millet-based diets on the net Ca contribution. These factors must be considered in controlled feeding community-based studies to determining the biological impact of biofortified crops.

In the past, various attempts have been put together to assess the Ca bioavailability in vivo. Early nutrition studies have shown that rats fed with a diet composed of 70% finger millet retain 68% Ca (Giri, 1940). A further reduction of the finger millet content to 20–40% in diets contributed to increased Ca retention to 84–88% levels (Giri, 1940). This shows that even a low dietary component of finger millet is sufficient to maintain Ca availability because of its high Ca content. In fact, in a more recent study, Ca from finger millet had shown to have a better uptake as compared to commercial Ca supplementation tablet (Bhide et al., 2013). In this in vivo study, the serum Ca level of rats fed with finger millet extract and a finger millet based ready-to-drink formulation was >35% higher as compared to conventional Ca tablet supplemented group.

However, for human metabolism studies, host factors like age and gender are important parameters to estimate daily requirement, intake and retention of dietary Ca. Many nutrition reports have estimated the contribution of finger millet for Ca homeostasis in humans. A study by Subrahmanyan et al. (1955) had found that a finger millet variety, H22, with Ca content 440 mg/100 g can on an average provide 3.4 g Ca/ day to healthy adult males aged 22–32 years. This amounted to Ca retention of 98 mg (approximately 3%) from a total daily intake of 3.4 g/day. This was higher than a brown bread or Ca carbonate fortified brown bread diet providing only 0.5–1.2 g Ca/day, respectively (McCance and Widdowson, 1942). It is recommended that diets should provide at least 200 mg/100 g of Ca to counteract the anticalcifying effect of phytic acid (McCance and Widdowson, 1942). Interestingly, 86% of phytate ingested from the fingermillet-based diet was hydrolysed during digestion and absorption process (Subrahmanyan et al., 1955). As most of the phytate is broken down during digestion, therefore, regular inclusion of finger millet in diet can efficiently maintain a positive Ca balance. In another study, young girls (aged 9–10) were fed on four different diets containing only rice, 75% rice + 25% finger millet, 50% rice + 50% finger millet and only finger millet as the primary cereal (Joseph et al., 1959). About 19% Ca was retained from just rice-based diets which increased to 22.5– 25.3% when a portion of rice was replaced with 25–50% finger millet. Therefore, finger millet can naturally contribute to boost the Ca status across ages evidently surpassing the other cereals. One serving of finger millet-based Ca-rich products, which were processed to increase Ca bioavailability, was shown to provide >0.2 g Ca contributing to 25% of Indian RDA of Ca for children and adolescents (Sanwalka et al., 2011). It is interesting to note that the Ca retention capacity of children seems to be much higher than adult subjects. This may be because growing age lowers digestibility and retention capacity of Ca and hydrolysis of antinutrients like phytate (Yoshida et al., 1983). It needs to be stressed here that most of these studies have been conducted more than 50 years ago. However, lifestyle and dietary patterns have drastically changed in recent years. Therefore, evaluating positive Ca impacts of finger millet diet across various groups needs to be measured keeping in view the current scenarios as well.

In a case study in rural India, the dietary patterns of women self-help groups was assessed for their nutrient adequacy (Vijayalakshmi et al., 2010). The families usually consumed one portion of finger millet preparation two times a day along with one portion of pulses and vegetables. It was found that irrespective of the socio-demographic profile of the subjects (like age, education, family income, family size), Ca levels met the recommended RDA and their adequacy was attributed to regular finger millet inclusive diet. As no major replacement of diet is necessary, finger millet and its derived food products have the advantage to be more acceptable to the people. This makes it a more viable option to be effective providing adequate Ca intakes and prevent Ca deficiency. These reports provide an idea that designing food products from biofortified varieties of finger millet can easily supplement and add-on to the daily Ca intake across ages and genders with various dietary practices. Such information can allow the acceleration of finger millet biofortification programs.

### MAJOR CHALLENGES TO DEVELOP FINGER MILLET AS A MODEL FOR CALCIUM BIOFORTIFICATION

In order to develop Ca biofortified finger millet, nutritionists must have available resources (superior Ca-rich varieties), a welldeveloped methodology to evaluate the bioaccessibility (in vitro and in vivo), awareness of the limiting factors (enhancers and inhibitors), and prior assessment of efficacy and effectiveness of Ca biofortified food. However, currently the development in this area is extremely limited and developing more nutritious varieties is challenging. Given that millet biofortification has recently been strategized, development of research programs for Ca-biofortified finger millet will have to address some arguments as discussed below.

### Challenges to Efficiently Utilize the Available Germplasm Resources and Genetic Diversity

In recent years, various efforts have been made by geneticists and breeders to identify naturally occurring genetic diversity in finger millet. However, the major challenge at present is how these resources could be exploited to develop Ca-biofortified finger millet. Currently, finger millet genebanks across the globe conserve more than 37,000 accessions with India, Kenya, Ethiopia, Uganda, and Zambia housing the major collections (Vetriventhan et al., 2015). As of now, the entire genetic diversity present among the finger millet germplasm is available as small sets (core) and sub-sets (mini core) collections (Vetriventhan et al., 2015). Using these collections, 15 accessions were identified as most promising (3.86–4.89 g/kg) for further improving grain Ca content in cultivated finger millet (Upadhyaya et al., 2011). Recently, another core set of 77 germplasm of Indian and African origin has been formed using the base germplasm of finger millet 1,000 accessions (Chandrashekhar et al., 2012). In addition, finger millet composite collections (1,000 accessions) and a derived reference set (300 accessions) representing regionand race-based available diversity of the entire collection, is also available (Upadhyaya et al., 2005; Upadhyaya, 2008). Although, these large collections of finger millet germplasm serve as an ideal resource to be utilized in improving its Ca concentration, a majority of it remains largely underutilized for breeding high Ca finger millet varieties. Some of the main reasons for this lag are due to factors, such as weak and insufficient strategies for harnessing the useful genetic diversity available in these collections, barriers related to introduction and crossing of exotic germplasm, few pre-breeding programs to facilitate introgression of desirable nutrition quality into breeding lines and recirculation of same working collections by breeders (Dwivedi et al., 2009; Upadhyaya et al., 2014). Although such precious germplasm collections exist, sometimes there are also practical barriers associated with their availability for use. For example, restricted global exchange of accessions due to legal aspects of seed transfer agreements poses a limitation for verification of adaptability to multi-location or multi-environment trials (Nass et al., 2012). Even a dearth of trained finger millet breeders to meet the demand can be an issue for making use of this treasure.

In addition, a huge range of diversity for grain Ca also exists within the gene pool of finger millet which remains to be exploited in targeted breeding. Two main gene pools exist for this species, namely Eleusine coracana sub-species africana (wild progenitor) and E. coracana sub-species coracana (domesticated cultivars/varieties and landraces (Agarwal and Maheshwari, 2016). Considerable variation in grain Ca contents has been observed in both the genepools. A detailed systematic analysis of grain Ca content in E. coracana sub species africana, from Ethiopia, revealed significantly higher Ca (515 mg/100 g) than the domesticated E. coracana sub species coracana from Kenya (401 mg/100 g) and India (375 mg/100 g) (Barbeau and Hilu, 1993). Many pieces of evidence for genotype effect in finger millet Ca content also exist. Vadivoo et al. (1998) found a large heritable genetic variation in relation to Ca content in 36 genotypes of finger millet, with white seeded varieties containing moderate levels of Ca. Based on their results, Malawi 1915 (486.7 mg/100 g Ca) and CO 11 (487 mg/100 g Ca) genotypes were proposed to be employed as crossing parents in breeding for the improvement of Ca content. Similarly, the dark red to very dusty red colored finger millet genotypes, CO 10, KM 1 and MI 302 sourced from the Dry zone Agricultural Research Station, Sri Lanka were shown to contain 240–250 mg% Ca (Ravindran, 1991). In another report, the white seeded finger millet varieties showed higher average Ca (329 mg%) than the brown seeded ones (296 mg%) (Seetharam, 2001). In spite of the extensive screening of finger millet germplasm for grain Ca content, the identified potential candidate genotypes have remained unused for developing higher Ca containing varieties breeding.

At the same time, just selecting suitable donor lines for selective breeding based on variation in grain Ca content is not sufficient and may not even be successful as such variations may often be regulated at various other levels. Therefore, determination of genetic stability and adaptability of this trait across multiple environments is one of the prerequisite to develop effective strategies for breeding elite lines. However, a severe gap exists in our knowledge about accuracy by which genetic variation for Ca content can be reproduced by finger millet genotypes grown across various agro-ecological conditions.

### Challenges Associated with Breeding-Based Genetic Improvement of Finger Millet

As finger millet is a naturally self-pollinating crop, artificial hybridisation by crossing of suitable parental lines is often a difficult task. Mass and pure-line selection practices have come in handy for inter-varietal improvement for grain yield, early maturity and disease resistance (Harinarayana, 1986; **Table 2**). For example, using pure line selection from the germplasm accession, finger millet culture WWN-25 has been released as a high yielding variety, GNN-7, for cultivation in Gujrat state of India (Patil et al., 2016). This is a promising development as this variety contains higher Ca (468.0 mg/100 g) than the national check variety VR-708 (398.0 mg/100 g) without compromising on the yield. However, optimum deployment of other breeding methods, such as recombination breeding, for generating stable hybrids, breeding progeny and inbred lines has been delayed due to challenging biparental cross, difficult emasculation and artificial hybridization in finger millet. To overcome these challenges, induced mutations, such as genetic male sterile systems (viz., INFM 95001 reported by ICRISAT; http://oar.icrisat.org/618/1/PMD\_71.pdf) have proved to be another efficient breeding tool for yield and disease resistance in finger millet. These systems and their subsequent breeding can be used effectively to increase the genetic variance by creating new recombinants and segregating populations by exploiting the genetic background. Therefore, developing genetic resources for finger millet, such as mapping populations, breeding lines and male-sterile mutant lines (Gupta et al., 1997; Krishnappa et al., 2009; Parashuram et al., 2011), deserves attention. Such material will be immensely valuable for tagging nutritional quality traits, especially grain Ca content, and thus facilitate genetic biofortification of finger millet.

### Underutilization of Available Genetic and Genomic Resources for Molecular Breeding Applications

Efforts have been made to generate molecular markers for characterizing important traits like grain Ca and protein content and resistance to blast infection in finger millet. Genomic tools like SSR markers have helped to assess the range of genetic diversity for grain Ca content in various finger millet genotypes (Panwar et al., 2010; Nirgude et al., 2014; Yadav et al., 2014; Kumar et al., 2015a). Nine SSR markers derived from candidate Ca transporter and sensor genes have been found to be significantly associated with the Ca trait. They could serve as an important functional resource for the genetic improvement of finger millet's nutritional value through markerassisted breeding (MAB; Kumar et al., 2015a; Sharma et al., 2017). However, the inbreeding nature, limited recombination rates and a historical genetic bottle-neck during isolated domestication of this crop significantly impacts the extent of available genetic diversity in finger millet. Such loss of genetic diversity is a challenge for geneticists and breeders working with a limited number of finger millet accessions. Further, until recently, there has been no progress in application of the finger millet genetic map in trait mapping despite the assembly of the only molecular marker-based linkage map a decade ago (Dida et al., 2007; Srinivasachary et al., 2007). It still remains under-utilized for tagging and identification of genes/quantitative trait locus (QTL) governing grain Ca content probably due to an insufficient number of informative markers.

The unavailability of sufficient markers and genome sequence information in finger millet has resulted in limited breeding efforts for nutritional improvement. Nonetheless, advances

#### TABLE 2 | Modern finger millet varieties released in the last decade.


#### TABLE 2 | Continued


(Continued)

### TABLE 2 | Continued


in large-scale genomics technology have now streamlined production of genome-wide markers which can be used for large-scale identification of candidate genomic loci. This advancement has also been capitalized to generate single nucleotide polymorphism (SNP) markers in finger millet using genotyping-by-sequencing (GBS; Kumar et al., 2016b) and Roche 454 and Illumina sequencing (Gimode et al., 2016). Inspite of the low level of polymorphisms in cultivated finger millet genotypes (Salimath et al., 1995), these SNPs may provide some explanation for variation in grain Ca content among finger millet genotypes. However, before their utilization, it is crucial to differentiate true SNPs among different genotypes from the homeologous SNPs within an individual genotype due to allotetraploidy (AA and BB sub-genomes) of finger millet.

Advances in cereal genomics have elaborated that genomic level similarities are conserved in the relative physical positions across species both on a fine scale (co-linearity) as well as on a chromosomal scale (synteny). For example, comparative mapping of finger millet to rice has shown a fairly high level of collinearity among these crops (Srinivasachary et al., 2007). The availability sequence information for several members of the grass family is now assisting in the development of inter-species molecular markers and gene discovery in unexplored crop genomes, such as finger millet (Wang et al., 2005; Kalyana Babu et al., 2014a,b). Until finger millet whole genome sequence becomes available, such comparative genomics approaches can benefit finger millet improvement programs. However, identification of common quantitative trait loci (QTLs) that control grain Ca content among cereals remains to be explored. This will further facilitate identification of orthologous regions and transfer of genetic information across species for improving Ca content of other millets and nonmillets.

### Inadequate Understanding of Ca Homeostasis Mechanisms in Finger Millet

Emphasizing on the molecular status of Ca accumulation in finger millet grains, various genes have recently been identified (Sood et al., 2016). More recently, a molecular model for Ca transport from soil to seed has been proposed (Sharma et al., 2017). This hypothetical model is based upon genes that are differentially expressed in contrasting finger millet cultivars (Mirza et al., 2014) or grain filling and developing spike transcriptome studies in this crop (Singh et al., 2014, 2015; Kumar et al., 2015b). These genes correspond to Ca sensing and binding, Ca transport and seed storage, such as, type IIB ATPase, Ca2+/H<sup>+</sup> antiporter (CAX1), two pore channel (TPC1), calmodulin (CaM), CaM-dependent protein kinases (CaMK1 and CaMK2) and 14-3-3 (**Table 3**). High Ca accumulation in finger millet has been mainly attributed to the Ca sensor genes which have been proposed as candidates for targeted Ca enhancement in finger millet varieties (Singha et al., 2016). Alteration in the genes expression levels only indicates a pattern among the contrasting genotypes which does not always validate the translated protein products involved in Ca accumulation. A few recent studies have characterized accumulation of Cabinding protein (calreticulin) and CaM protein during grain filling stages of finger millet (Kumar et al., 2014; Singh et al., 2016). Unfortunately, large-scale protein profiling to identify the complete set of proteins involved in finger millet Ca homeostasis is still unavailable.

Even though the past and current developments have generated a huge wealth of transcriptomic datasets linked with the Ca uptake, translocation and accumulation in finger millet, without functional characterisation, these candidate genes are merely speculation. With a lack of evidence for efforts being made to functionally characterize these genes/proteins, it is possible that these potential candidates may not have expected TABLE 3 | List of important genes identified for calcium uptake, transport and storage in finger millet.


impact on enhancing grain Ca content. While, it is practically impossible to functionally characterize all genes linked with Ca acquisition, sensing and accumulation, transgenic technology along with progress in gene editing (such as CRISPR-Cas and TALEN) can aid in a better functional understanding of many candidate genes. Such advancements in gene editing technology along with well-established genetic transformation protocols for generation of transgenic finger millet (Kumar et al., 2016a) can significantly improve our understanding of this process.

Besides this, the role of other factors (for example, hormones, root morphological traits, endophytes, soil fertility, evapotranspiration pull, translocation distance of Ca) on Ca homeostasis in finger millet remains unknown. This is a crucial question as the soil-plant interactions substantially affect the available proportion of micronutrient to the roots. In this context, soil medium supplemented with growth promoting rhizobacterium (Azotobacter, Azospirillium, phosphorus solubilizing bacteria) and vesicular-arbuscular mycorrhizal (VAM) fungi are generally practiced to enhance the grain yield and growth in finger millet (Ramakrishnan and Bhuvaneswari, 2014; Thilakarathna and Raizada, 2015). VAM are also known to increase the content and uptake of minerals, such as phosphate, nitrogen, zinc and copper (Tewari et al., 1993). The nutrient supply to the plant can be augmented either by efficiently mobilizing nutrient forms available in the soil or by extending the nutrient absorption surface by designing better roots system. However, how these associations impact Ca absorption and uptake in finger millet has received little experimental consideration. Further, the relationship between growth environments and climates may also alter xylem water flow, thus indirectly determining Ca distribution. So far, the established mechanisms of Ca accumulation in finger millet have been developed on the basis of controlled growth conditions. However, under field conditions, the unique soilroot interaction can influence the Ca-sensing and transport differently. In addition, any confounding effects of agronomic traits, such as vegetative growth, yield, stress resilience and disease resistance on grain Ca accumulation are not clearly established. Therefore, to develop finger millet as a model for Ca biofortification, we need a comprehensive understanding of the mechanisms and other factors that may influence this trait.

### Potential of Next Generation Sequencing (NGS) For Improved Finger Millet Varieties

Enormous progress has been made in the genomics technology through application of high-throughput, economical and quicker next generation sequencing (NGS) platforms. Extending the benefits of NGS to finger millet, a recent effort of de novo sequencing has allowed whole genome sequence assembly covering approximately 82% of total estimated genome size (Hittalmani et al., 2017). Evidence of higher collinearity with foxtail millet and rice as compared to other Poaceae species, and the available genome sequencing information may help allele discovery and candidate gene identification for agronomically important traits (Hittalmani et al., 2017) leading to faster development of improved varieties. In addition, GBS, which is a NGS platform-based highly multiplexed genotyping system, has also been applied for SNP generation (Kumar et al., 2016b). Thus, now it is feasible to generate a higher density of markers by genotyping core collections of finger millet thereby increasing the level of genetic diversity explored. This is crucial for a predominantly self-fertilized crop like finger millet because it is expected to have low recombination rates and high linkage disequilibrium (LD) which would otherwise narrow the genetic diversity. Thus, the more genetically diverse populations in finger millet core collections, together with the huge amount of relevant marker information generated through NGS platforms can directly contribute to improved mapping resolution of traits, such as Ca content through genome-wide association studies (GWAS). By statistically reconnecting variation in grain Ca content back to its underlying genetic polymorphism, it is possible to identify functional common variants in LD and genomic regions where major-effect genes and QTLs (that serve as the targets of marker-assisted selection) are located. The GWAS studies can confirm previously identified genes involved in Ca homeostasis mechanisms as well as spot putative novel candidates. However, the efficiency of GWAS depends upon accurate grain Ca content phenotyping data over multilocation/multi-year trials.

An extension of MAS, genomic selection (GS) is an upcoming methodology in the area of genomics-assisted breeding (Meuwissen et al., 2001). In this approach, genomewide marker genotype data along with available phenotypic data for a tested (reference/training) population are used to predict the performance of an untested (breeding) population based on genomics estimated breeding values (GEBV). Thus, instead of identifying few large-effect loci associated with Ca content, the GEBV model can more accurately predict the expected phenotype of a broader breeding population. This significantly reduces the time and costs associated with phenotyping a trait like grain Ca content. Finger millet enjoys the availability of germplasm resources, such as the core collections, which can be utilized as test populations to build genomic prediction models. As GS eliminates the need for previous identification of major QTLs and their use in selection, it can substantially speed up the genetic gain in this "orphan" crop. Another advantage is that if GEBVs are efficiently evaluated, finger millet breeders can make appropriate selection choices even earlier in the program, thereby significantly saving time on the generation cycle (Heffner et al., 2010). However, the applicability of GS in finger millet and selection of superior genotypes are dependent upon precise measurement and heritability of Ca content, sufficient marker density, the extent of LD decay, effective design of training population and its genetic relationship with the breeding population (Varshney et al., 2014). Nevertheless, NGS can contribute in exploring the depth and breadth of genetic diversity across germplasm sets bringing forward a huge wealth of genetic information. This will eventually lead to a new horizon for finger millet Ca biofortification.

### FUTURE PROSPECTS AND CONCLUSIONS

By virtue of its health benefitting properties and environmental sustainability, a traditional but less popular crop like finger millet offers excellent opportunities for biofortification breeding. A foremost priority from geneticists and breeders viewpoint is capturing and utilizing genetic diversity for Ca content in the elite finger millet gene pools (for example, by bringing new sources of variation through rare and unique alleles). For trapping such useful variations, advances in the next generation sequencing technology must be utilized in generating sufficient number of markers for characterizing marker-trait associations and genomics-assisted breeding. With the implementation of such high-throughput approaches, it will be much easier to investigate the genetic architecture of this trait through comparative mapping in other millets and non-millet species. Mining of markers tightly linked to other traits governing grain Ca content and discovery of underlying genes can be an alternate strategy to develop high Ca finger millet varieties through traditional or modern breeding approaches and transformation-based methods. For example, rather than just arbitrarily increasing grain Ca content, future direction should target improvement in efficiency to mobilize, acquire, transport and store Ca in more bioavailable forms in the edible portions.

From the health perspective, at this point, we almost completely lack the understanding of interplay among grain Ca, other micronutrients and antinutrients metabolism in human body. Any potential risks of reducing the antinutrient content in the grains should be evaluated with a view of their longterm effects on human health. Therefore, the focus should be on demonstrating finger millet's bio-efficacy in order to monitor any potential negative trade-offs and unintended effects. In addition, experimental confirmation of in vivo Ca bioavailability needs to be cautiously attempted through designing pilot feeding studies for vulnerable groups (like children, nursing or post-menopausal women). Although the biological effects of improved finger millet may be relatively modest, it will have potential benefits

### REFERENCES


in improving healthcare cost savings by reducing the risk of osteoporotic fractures and DALYs lost.

While agronomic factors can influence acceptability of the improved finger millet varieties by farmers, other parameters are important from the view of consumer acceptance. The extent of Ca recommended dietary allowance proportion as well as the sensory satisfaction supported by finger millet is essential in terms of developing food products keeping in line with recent lifestyle changes. Even after successful Ca biofortification of finger millet, its introduction and success as a functional food still entails knowledge of adequate food processing strategies (to minimize the nutrient loss) and study of consumer preferences. Further, communication support and the creation of market demand for its value-added products will be necessary. Therefore, a multi-disciplinary research approach, incorporating nutrition, health, agriculture, along with policy and market research, is needed to ensure the impact of high Cabiofortified finger millet. Overall, it is worthwhile to conclude that finger millet biofortification will improve the quality of life for both the rural subsistence farming families as well as the consumers.

### AUTHOR CONTRIBUTIONS

SP and RY conceptualized the manuscript. SP wrote the manuscript. JK, PS, HO, and RY assisted and edited the manuscript. SP, JK, PS, and RS contributed in critically revising the draft and updating the manuscript for publication.

### ACKNOWLEDGMENTS

The authors wish to express their thanks to IBERS, UK for the support in writing this manuscript. IBERS receives strategic funding from BBSRC. SP acknowledges Marie Skłodowska-Curie Individual Fellowship from Horizon 2020 of European Commission (Project 657331; CaMILLET). We would also like to acknowledge the editor and reviewers whose constructive comments and suggestions helped to improve the manuscript.

on Biofortification: Getting Nutritious Foods to People. March 31–April 2, 2014 (Kigali).


of bioactive compounds from fruits and vegetables: a review. Compr. Rev. Food Sci. Food Saf. 13, 155–171. doi: 10.1111/1541-4337.12049


in differential accumulation of calcium in developing grains. Appl. Biochem. Biotechnol. 172, 2955. doi: 10.1007/s12010-013-0714-0


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Puranik, Kam, Sahu, Yadav, Srivastava, Ojulong and Yadav. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Genome-wide Identification, Characterization, and Expression Analysis of PHT1 Phosphate Transporters in Wheat

Wan Teng<sup>1</sup>† , Yan-Yan Zhao<sup>1</sup>† , Xue-Qiang Zhao<sup>1</sup> , Xue He<sup>1</sup> , Wen-Ying Ma<sup>1</sup> , Yan Deng<sup>2</sup> , Xin-Ping Chen<sup>3</sup> and Yi-Ping Tong<sup>1</sup> \*

<sup>1</sup> The State Key Laboratory for Plant Cell and Chromosome Engineering, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing, China, <sup>2</sup> Institute of Tropical Agriculture and Forestry, Hainan University, Haikou, China, <sup>3</sup> Research Center of Resource, Environment and Food Security, China Agricultural University, Beijing, China

#### Edited by:

Raul Antonio Sperotto, Centro Universitário UNIVATES, Brazil

### Reviewed by:

Stefano Cesco, Free University of Bozen-Bolzano, Italy Hatem Rouached, Institut National de la Recherche Agronomique (INRA), France Soren K. Rasmussen, University of Copenhagen, Denmark

\*Correspondence:

Yi-Ping Tong yptong@genetics.ac.cn †These authors have contributed equally to this work.

#### Specialty section:

This article was submitted to Plant Nutrition, a section of the journal Frontiers in Plant Science

Received: 11 February 2017 Accepted: 27 March 2017 Published: 11 April 2017

#### Citation:

Teng W, Zhao Y-Y, Zhao X-Q, He X, Ma W-Y, Deng Y, Chen X-P and Tong Y-P (2017) Genome-wide Identification, Characterization, and Expression Analysis of PHT1 Phosphate Transporters in Wheat. Front. Plant Sci. 8:543. doi: 10.3389/fpls.2017.00543 The PHT1 family of phosphate (Pi) transporters mediates phosphorus (P) uptake and re-mobilization in plants. A genome-wide sequence analysis of PHT1 genes in wheat (Triticum aestivum) was conducted, and their expression locations and responses to P availability were further investigated. We cloned 21 TaPHT1 genes from the homologous alleles at TaPHT1.1 to 1.10 through screening a BAC library and amplifying genomic sequences. The TaPHT1 transporters were clustered into five branches in the phylogenetic tree of PHT1 proteins, and the TaPHT1 genes from a given branch shared high similarities in sequences, expression locations, and responses to P availability. The seven tested PHT1 genes all showed Pi-transport activity in yeast (Saccharomyces cerevisiae) cells grown under both low Pi and high Pi conditions. The expression of TaPHT1.1/1.9, 1.2, and 1.10 were root specific. The expression of these TaPHT1 genes at flowering positively correlated with P uptake after stem elongation across three P application rates and two wheat varieties in a field experiment. Therefore, modification of PHT1 expression may improve P use efficiency in a broad regime of P availability.

Keywords: wheat (Triticum aestivum), PHT1 genes, genome-wide analysis, phosphate transporter, phosphorus uptake, phosphate-starvation response

## INTRODUCTION

Phosphorus (P) is one of the essential macronutrients for plant growth and development, and it takes part in cellular macromolecules, energy transfer reactions, and cellular metabolism. Efficient acquisition of phosphate (Pi) from soil combined with efficiency translocation of Pi within plants is essential for plants to maintain adequate levels of cellular Pi necessary for normal function (Raghothama and Karthikeyan, 2005). Although total P in soils is abundant, the soluble phosphate (Pi) is often low (Bieleski, 1973; Rausch and Bucher, 2002), and therefore plants often encounter a scarcity of Pi in soils of both agricultural and natural systems (Raghothama, 1999, 2000). As there is a large difference between Pi levels in plant cells (mM) and soil solution (µM), plants need to acquire Pi against a steep concentration gradient across the plasma membrane (Smith et al., 2003; Raghothama and Karthikeyan, 2005). The transmembrane transport of Pi from soils into plant cells requires a high-affinity, energy-driven transport mechanism (Smith et al., 2003). The PHT1 family of plant Pi transporters is assumed to play the predominant roles in this transmembrane transport

process. These proteins are characterized by 12 membranespanning domains which are similar to PHO84, a high-affinity Pi transporter from yeast (Saccharomyces cerevisiae) (Muchhal et al., 1996; Rausch and Bucher, 2002).

There are four PHOSPHATE TRANSPORTER (PHT) families in plants: PHT1, PHT2, PHT3, and PHT4 which are located on plasma membrane, plastid inner membrane, mitochondrial inner membrane, and Golgi-compartment, respectively (Lopez-Arredondo et al., 2014). Under the stress of P-starvation, the expression of PHT1 genes are strongly induced to increase the ability of the roots in acquiring P from soils and remobilize P within plants (Smith et al., 2003; Raghothama and Karthikeyan, 2005). A large number of PHT1 transporters have been identified in many plant species and show differences in expression locations and affinities for Pi (Nussaume et al., 2011). Nine PHT1 genes in Arabidopsis (Arabidopsis thaliana) have been identified. AtPHT1.1 and AtPHT1.4 are highly expressed at the root–soil interface, including the epidermis, root hair cells, and the root cap under low P conditions (Mudge et al., 2002), and they are the major genes responsible for Pi acquisition by roots in both high and low P supplies (Misson et al., 2004; Shin et al., 2004; Catarecha et al., 2007). AtPHT1.8 and AtPHT1.9 are likely to act sequentially in the interior of the plant during the root-to-shoot translocation of Pi and are involved in rootto-shoot translocation of Pi (Lapis-Gaza et al., 2014). There are 13 PHT1 members in the rice (Oryza sativa) genome, and some of them have been functionally characterized, including OsPHT1.1 (Sun et al., 2012), OsPHT1.2 and OsPHT1.6 (Ai et al., 2009), OsPHT1.4 (Ye et al., 2015), and OsPHT1.8 (Jia et al., 2011; Li et al., 2015). For example, OsPHT1.6 is expressed in both epidermal and cortical cells of the younger primary and lateral roots and encodes as a high-affinity transporter with a broad role in Pi uptake and translocation throughout the plant, whereas OsPHT1.2 is localized exclusively in the stele of primary and lateral roots and functions as a low-affinity transporter responsible for Pi translocation (Ai et al., 2009). Barley (Hordeum vulgare) is a close relative to wheat (Triticum aestivum). To date, 11 PHT1 genes have been reported in barley. The HvPHT1.1 and HvPHT1.2 promoters drive the expression of β-glucuronidase (GUS) and green fluorescent protein (GFP) reporter genes in epidermal and cortex cells as well as vascular tissues of roots (Schunmann et al., 2004). When expressed in Xenopus laevis oocytes, HvPHT1.1 is confirmed to be a high-affinity transporter with a very low K<sup>m</sup> value (1.9 µM) for Pi transport (Preuss et al., 2011). The expression locations and K<sup>m</sup> value for Pi transport indicate the possible role of HvPHT1.1 in P uptake. HvPHT1.6 is expressed in both roots and shoots (Huang et al., 2008). Also, it is highly expressed in old leaves compared to young leaves, especially in the leaf phloem tissue (Rae et al., 2003). HvPHT1.6 shows the linear transport activity for Pi-stimulated inward current over a concentration range of 5 to 30 mM in Xenopus laevis oocytes (Preuss et al., 2010). These results suggest that HvPHT1.6 function as a low-affinity Pi transporter responsible for Pi remobilization in the whole plant. Huang et al. (2011) investigated the expression of PHT1 genes and its relationship with P acquisition efficiency and P utilization efficiency (the amount of biomass produced per unit of acquired P) in four barley genotypes. They did not find a clear pattern in the expression of the four HvPHT1.1 paralogs (HvPHT1.1, 1.2, 1.9, 1.10) among the four barley genotypes, but observed that the expression of HvPHT1.3 and 1.6 positively correlated with P utilization efficiency. HvPHT1.8 and HvPHT1.11 (known as HvPT11) have been demonstrated to be specifically activated by arbuscular mycorrhizal (AM) fungi (Glassop et al., 2005; Sisaphaithong et al., 2012), indicating their possible roles in the mycorrhizal pathway of Pi uptake.

Wheat is one of the most important crops. However, limited attempts have been made to dissect the role of Pi transporters in wheat (Secco et al., 2017). Davies et al. (2002) isolated the first full-length sequence of a wheat PHT1 gene (TaPHT1.10-U, formerly known TaPT2) and partial clones of several other putative PHT1 genes. TaPHT1.10-U was induced by P-deficiency in roots, and had higher transcript abundance in P-efficient wheat varieties than in inefficient ones (Davies et al., 2002). In yeast, TaPHT1.10-U can complement highaffinity phosphate transporter gene PHO84 function (Zeng et al., 2002) and shows an apparent mean K<sup>m</sup> of 23.6 µM Pi (Guo et al., 2014). Overexpression of TaPHT1.10-U increases plant dry weight and Pi acquisition, whereas knock-down of TaPHT1.10-U has the opposite effect (Guo et al., 2014). These results suggest that TaPHT1.10-U functions as a high-affinity Pi transporter and mediates Pi uptake. A recent study observed that TaPHT1.12-7A (former name TaPHT1.4) was root-specific and P-deficiency inducible. Yeast complement analysis showed that TaPHT1.12-7A encodes a high-affinity Pi transporter with an apparent K<sup>m</sup> of 35.3 µM. Overexpressing TaPHT1.12-7A significantly improves growth traits and accumulates more Pi than the wild-type plant and those with downregulated TaPHT1.12-7A expression (Liu et al., 2013). A recent study also revealed the relationships between PHT1 expression and P use efficiency in wheat (Aziz et al., 2014). The highly P-efficient wheat cultivar Chinese 80-55 has a higher Pi acquisition in the presence of Pi and accumulates higher Pi concentrations in all organs upon Pi withdrawal compared with the less-efficient cultivar Machete. These differences correlate with differential organspecific expression of Pi transporters TaPHT1.10-4A (reported as TaPHT1.2, GenBank: AY293828), TaPHT1.6-5A (reported as TaPHT1.5, GenBank: AF110180) and TaPHT1.4-5B (reported as TaPHT1.8, GenBank: AK333026) (Aziz et al., 2014). Shukla et al. (2016) found that aleurone accumulates more Pi with higher expression of TaPHT1 genes compared to endosperm. TaPHT1.8- 6A (known as TRIae; Pht1; myc, Glassop et al., 2005, GenBank: AJ830009), TaPHT1.11-4A (known as TRIae; Pht1; 12, GenBank: AB753271), TaPHT1.11-4B (known as TRIae; Pht1;11, GenBank: AB753270), and TaPHT1.11-4D (known as TRIae; Pht1; 10, GenBank: AB753269) have been found to be induced by AM fungi (Glassop et al., 2005; Sisaphaithong et al., 2012). Although expression of some PHT1 genes has displayed correlation with P use-related traits in wheat and its close relative barley under controlled conditions, an on-farm field-scale investigation is required to explore the PHT1 genes contributing to P uptake and utilization, as the response of PHT1 genes to P supply level under controlled conditions greatly differed from that under field conditions. Our recent study showed that the expression

of TaPHT1.1, 1.2, 1.9, and 1.10 in roots at the flowering stage under low P conditions was lower than that under high P conditions in a field experiment (Teng et al., 2013). The inhibition of these four wheat genes by P-deficiency could be, at least partially, explained by the upregulated AM colonization under P-deficiency, considering that AM colonization has been found to inhibit the response of HvPHT1.1 and HvPHT1.2 to P deficiency in barley (Glassop et al., 2005).

In this study, we aimed to identify the sequences of PHT1 genes in the whole genome of wheat, and to analyze the correlation between the PHT1 expression and P uptake under field conditions. We isolated 21 full length sequences of PHT1 genes in wheat, and further analyzed their functions, expression location and response to P supply level. We observed that the expression of TaPHT1.1, 1.2, 1.9, and 1.10 in roots at the flowering stage contributed to P uptake of different wheat varieties under field conditions.

### MATERIALS AND METHODS

### Wheat Varieties

The winter wheat (Triticum aestivum) variety Xiaoyan 54 was commercially released in 2000, and was used to isolate TaPHT1 sequences, and to analyze gene expression location and response to P availability. The winter wheat varieties Kenong 9204 (KN9204) and Shijiazhuang 8 (SJZ8) were commercially released in 2003, and were used in the field experiments to analyze the relationship between TaPHT1 expression and P uptake.

### Isolation of PHT1 Pi Transporters in Wheat

To isolate PHT1 sequences from the wheat variety Xiaoyan 54, we performed BAC library screening and genomic sequence amplification by using the primers in Supplementary Table S1. After several rounds screening the BAC library of Xiaoyan 54 (Dong et al., 2010), we obtained 28 BAC clones which contained PHT1 genes. These BAC clones were sequenced commercially by using a Roche/454 GS-FLX Titanium System (Roche Diagnostics, Germany) at SinoGenoMax Co., Ltd. (Chinese National Human Genome Center, Beijing, China). The resultant sequences were examined for the promoter and protein-coding sequences of PHT1 genes, and consequently the primers were designed to isolate the coding regions of the PHT1 genes in these BAC clones. The PCR products amplified from BAC clones and genomic DNA were sub-cloned into a pMD18-T Vector (Takara Bio, Dalian, China), and then sequenced commercially at SinoGenoMax Co., Ltd. The putative cis-elements in the promoters were predicted by RSAT::Plants software<sup>1</sup> . We used the neighbor-joining method to generate a phylogenetic tree of PHT1 proteins from wheat, Triticum urartu, Aegilops tauschii, barley, maize (Zea mays), rice, and Arabidopsis, and the phylogenetic tree was drawn using MEGA 5.0 (Tamura et al., 2011). Sequence alignment was performed by DNAMAN6.0 (Lynnon BioSoft, San Ramon, CA, USA).

#### <sup>1</sup>http://floresta.eead.csic.es/rsat/

### Functional Complementation Assay of Pi Transporters in Yeast

The yeast manipulations were performed as previously described (Ai et al., 2009). For the complementation assay, the coding sequences of the TaPHT1 genes were amplified by PCR and subcloned into the yeast expression vector p112A1NE to create TaPHT1-p112A1NE constructs. These constructs and the empty vector p112A1NE were transformed into the yeast Pi uptakedefective mutant MB192 (Bun-Ya et al., 1991). Because the PHT1 transporters are members of the H+/Pi symporter family, we firstly evaluated the optimal pH value for the growth of the transformed and control yeast strains. After measuring the optical density of the yeast cell lines at pH values ranging from 4 to 8 in yeast nitrogen base (YNB) liquid medium, we observed that the optimal pH value for most of the yeast mutant cells carrying TaPHT1s was 6, whereas the optimal pH value for the wild-type ranged from 4 to 6. Therefore, the pH value was set to 6 in the subsequent studies. To measure the kinetic growth profiles of the yeast strains, the yeast cells were grown in YNB liquid medium to the logarithmic phase (when the absorbance at 600 nm was 0.8), and were then harvested and washed in Pi-free YNB medium. Then, the yeast cells were grown at 30◦C for 24 h in the YNB liquid media containing 200 µM Pi (high Pi) and 20 µM Pi (low Pi). The absorbance at 600 nm (OD600) was recorded every 6 h. MB192 and p112A1NE were kindly provided by Prof. Shubin Sun from Nanjing Agricultural University, Nanjing, China.

### Plant Growth Conditions

A hydroponic culture and three field experiments were conducted. The winter wheat variety Xiaoyan 54 was used in the hydroponic culture. The nutrient solution and growth conditions of the hydroponic culture were described by Wang et al. (2013). The seedlings, after 6 days of germination, were grown in nutrient solutions that contained 200 µM Pi (high P) or 5 µM Pi (low P). The plants were grown at 20◦C for 3 weeks, and the roots and shoots were collected separately for gene expression analysis.

The field experiment in the experimental station of the Institute of Genetics and Developmental Biology in Beijing was carried out in the 2012–2013 growing season. The plant density and P fertilizer treatments was described by Wang et al. (2013). Briefly, the low P and high P treatments, i.e., 0.0 g m−<sup>2</sup> and 13.5 g m−<sup>2</sup> of P as calcium superphosphate, respectively, were applied before sowing. The seeds of Xiaoyan 54 were sown at the end of September in 2012. At the re-greening stage (March 18, 2013), the roots in 0–30 cm depth soil and shoots were collected separately. At the flowering stage (May 3, 2013), the stems, spikes, flag leaves, and aging leaves (top third leaf) were sampled. At the grain filling stage (14 days after flowering), the stems, grains, flag leaves, and aging leaves were collected. In each sampling time, 10 plants were randomly selected in each of the three replications. The plant samples were stored at −80◦C for gene expression analysis.

Two field experiments at the Quzhou Experiment Station (36.5◦ N 115.0◦ E, 40 m above sea level) of the China Agricultural

University have been described by Teng et al. (2013). These two experiments were conducted in the 2009–2010 growing season (referred as the 2010 field experiment) and the 2010–2011 growing season (referred as the 2011 field experiment). The data were collected from the winter wheat varieties KN9204 and SJZ8 at the P application rates 0, 100, and 400 kg ha−<sup>1</sup> of P as calcium superphosphate (referred as P0, P100, and P400, respectively). The data for P use-related traits and expression levels of TaPHT1 genes in KN9204 have been reported by Teng et al. (2013). The P application rates P0, P100, and P400 represented deficient, optimal, and excessive P supply, respectively (Teng et al., 2013).

### RNA Extraction and Quantitative Real-time PCR

Total RNA extraction and real-time quantitative reverse transcription PCR (qRT-PCR) were performed according to the methods of Teng et al. (2013). The primer sequences are listed in Supplementary Table S2. The gene expression levels were normalized to the internal control of TaActin.

### Measurement of Total P Concentration in Plant Samples

To determine plant total P, dried samples were milled and subsequently digested with concentrated H2SO<sup>4</sup> and H2O<sup>2</sup> using the molybdate-blue colorimetric method (Murphy and Riley, 1962).

### Statistical Analysis

The SPSS statistical software (SAS Institute, Cary, NC, USA) was used to perform analysis of variance using one-way analysis of variance (ANOVA). Comparisons of means were performed using Duncan's multiple range analysis test and paired samples t-test (α = 0.05).

### RESULTS

### Sequence Analysis of PHT1 Transporters in Wheat

We cloned 21 TaPHT1 genes from common wheat through screening a BAC library of Xiaoyan 54 and amplifying genomic sequences (Supplementary Table S3). None of these genes contained intron, 19 of them contained full length ORFs, and their deduced protein sequences varied from 521 to 539 amino acids (Supplementary Table S3). One nucleotide deletion occurred at 368 bp downstream of the start codon in TaPHT1.10- 4B and thus resulted in a frame shift mutation, and TaPHT1.9-4A had a premature stop codon mutation at 810 bp downstream of the start codon, but this premature stop mutation was not found in the Chinese spring. We mapped the cloned TaPHT1s on chromosomes by sequence analysis of BAC contigs and the reference sequence of Chinese spring<sup>2</sup> . The five clones, BAC48, BAC470, BAC674, BAC1217, and BAC1779, formed a BAC contig which contained TaPHT1.1-4B, 1.2-4B, 1.9-4B, and 1.10-4B (Supplementary Figure S1A). TaPHT1.9-4B and 1.2-4B matched with the sequences from 210,450 to 212,013 bp and from 299,422 to 300,999 bp in the scaffold TGACv1\_scaffold\_320302\_4BL, respectively (Supplementary Figure S1B), Therefore TaPHT1.1- 4B and 1.10-4B were assigned to chromosome 4B. Further sequence analysis showed that the 1011 bp fragment from 220136 to 221146 bp and the 1316 bp fragment from 258319 to 259634 bp of the scaffold TGACv1\_scaffold\_320302\_4BL matched with TaPHT1.10-4B and 1.1-4B, respectively, but both fragments had low sequence quality. The former fragment contained 585 unknown nucleotides, and the later fragment contained 1004 unknown nucleotides; this was possibly why these two fragments were not annotated. TaPHT1.10-4B also showed 99.7% of sequence identity with the sequence from 1 to 1465 bp in the scaffold TGACv1\_scaffold\_684896\_U (Overlapping gene TRIAE\_CS42\_U\_TGACv1\_684896\_AA2159320, Supplementary Table S3). TaPHT1.10-4B from Xiaoyan 54 seemed to be the allele of TRIAE\_CS42\_U\_TGACv1\_684896\_AA2159320 from Chinese spring, as both genes had the nucleotide deletion at 368 bp downstream of the start codon.

Genome-wide analysis of the genome sequence in Triticum\_aestivum\_CS42\_TGAC\_v1 assembly for Chinese spring<sup>2</sup> totally identified 32 Gene IDs for TaPHT1 (Supplementary Table S3). The Gene ID TRIAE\_CS42\_4BL\_ TGACv1\_320302\_AA1034400 matched with TaPHT1.2-4B and TaPHT1.9-4B; and no Gene ID was found to match with TaPHT1.1-4B and TaPHT1.10-U cloned in the current study or TaPHT1.11-4B (former name TRIae; Pht1;11) cloned by Sisaphaithong et al. (2012). Therefore, we identified a total of 36 TaPHT1 genes (Supplementary Table S3). The 31 genes of TaPHT1.1-TaPHT1.11 were named according to their similarity with barley PHT1 transporters and chromosome location, and the remaining five genes were sequentially named TaPHT1.12, TaPHT1.13, and TaPHT1.14, together with the chromosome location (Supplementary Table S3). The TaPHT1 genes were unevenly distributed on the chromosomes, as there were 17 and 8 PHT1 genes on the chromosomes of homologous group 4 and 5, respectively (Supplementary Table S3). This uneven distribution was mainly due to the PHT1 clusters on the chromosomes of these homologous groups. For example, we found five TaPHT1 genes (TaPHT1.5-4B, TaPHT1.1-4B, 1.2-4B, 1.9-4B and 1.10-4B) within a 150-kb region on the long arm of chromosome 4B (Supplementary Figure S1). The scaffold TGACv1\_scaffold\_407907\_5BL on the long arm of chromosome 5B conferred TaPHT1.3-5B and TaPHT1.4-5B within an approximate 18-kb region (Supplementary Table S3). We also cloned the promoter sequences of 10 TaPHT1 genes, and all these promoters were found to contain several putative Pi-starvation response regulator PHR1 binding cis-element P1BS and WRKY transcription factor binding element W-Box (Supplementary Figure S2).

We calculated the relatedness of TaPHT1s using the ClustalX 2.1 software, with the results suggesting that the protein sequence identities ranged from 46 to 99%. The highest identities were found between the protein sequences of TaPHT1.1/1.2/1.9/1.10, and for that of TaPHT1.3/1.4. There were more than 98% of protein sequence identities between each other of the

<sup>2</sup>http://plants.ensembl.org/Triticum\_aestivum/Info/Index

homologous alleles at a given TaPHT1 locus from genomes A, B, and D (e.g., TaPHT1.8-6A, -6B and -6D). A neighborjoining tree was constructed using a multiple sequence alignment according to TaPHT1 proteins and the PHT1 sequences from Triticum urartu, Aegilops tauschii, barley, maize, rice, and Arabidopsis (Supplementary Table S3). TaPHT1.10-4B was not included in the phylogenetic analysis, as it contained a frame shift mutation. The 35 TaPHT1s were clustered into five of the six branches (**Figure 1**). Branch I only contained PHT1s from Arabidopsis. TaPHT1.13-2A, TRIurPHT1.13, and OsPHT1.4/1.5 formed Branch II. The six TaPHT1.6/1.7 genes fell into Branch III which contained HvPHT1.6/1.7 and OsPHT1.6/1.7. The nine TaPHT1.3/1.4/1.5 genes belonged to Branch IV, and showed a close relationship with HvPHT1.3/1.4/1.5. The nine TaPHT1.1/1/2/1/9/1.10 genes belonged to Branch V, and they closely related to HvPHT1.1/1.2/1.9/1.10 and OsPHT1.1/1.2/1.3. The 10 TaPHT1.8/1.11/1.12/1.14 genes were grouped into Branch VI, which contained the AM fungiinducible PHT1s from cereals as well as AtPHT1.6/1.8/1.9 from Arabidopsis.

### Analysis of Pi Transport Activities of TaPHT1s in a Yeast Strain Defective in Pi Uptake

We analyzed the Pi transport activities of TaPHT1.1-4D, 1.10-4A, 1.4-5D, 1.5-4A, 1.6-5D, 1.7-4D, and 1.8-6B genes using the yeast mutant MB192 strain (pho84 mutant; Bun-Ya et al., 1991), which is defective in Pi uptake. TaPHT1.1-4D and 1.10-4A were selected to represent the closely related TaPHT1.1/1.2/1.9/1.10 which encoded two types of protein length, 521 amino acids and 525 amino acids. TaPHT1.4-5D was chosen to represent the closely related TaPHT1.3/1.4. The coding regions of the seven selected TaPHT1 genes were separately inserted into the yeast expression vector p112A1NE under the control of the yeast alcohol dehydrogenase promoter. The constructs were separately transformed into a yeast Pi transporter mutant MB192. An empty vector was also transformed to be used as a control (Yp112). We first analyzed the complementation of MB192 by TaPHT1 genes by using dilution based plate assays. All the yeast transformants harboring the candidate TaPHT1 genes grew better than the Yp112 (empty vector control), but poorer than the wild-type in the plates which contained 20, 60, 100, and 140 µM Pi when the yeast cells were diluted to 1/100 OD value (Supplementary Figure S3C). This result indicated that the seven tested TaPHT1 genes could partially restore the growth of MB192 mutant cells. Staining test for acid phosphatase activity also showed that TaPHT1.6-5D and TaPHT1.10-4A partially restore the growth of MB192 mutant (Supplementary Figures S3A,B). We then assessed the kinetic growth of the yeast cells in YNB liquid medium that contained 200 µM Pi (high Pi) and 20 µM Pi (low Pi). The wild-type yeast strain grew much quicker than the Yp112, MB192, and the yeast cells transformed with TaPHT1 genes (Yp112-TaPHT1s), whereas Yp112 and MB192 exhibited a growth defect on both high Pi and low Pi media (**Figure 2**). All the yeast mutant cells carrying Yp112-TaPHT1s transformants grew faster than Yp112 and MB192 under high Pi and low Pi conditions (**Figure 2**), suggesting that these selected TaPHT1s had Pi transport activity.

### Responses of TaPHT1 Expression to P Availability

Quantitative real-time RT-PCR was used to analyze the responses of TaPHT1 genes to P supply levels at the seedling stage in a hydroponic culture and at the re-greening stage in a field experiment. Primers were designed to amplify the homologous alleles at a particular locus; for example, the relative expression level of TaPHT1.2 might represent that of all three homologous alleles of TaPHT1.2 (TaPHT1.2-4A, -4B, and -4D). In both the hydroponic culture and the field experiment, the expression of TaIPS1.1, a molecular indicator of plant Pi status (Teng et al., 2013), was upregulated by the low P treatment (Supplementary Figures S4A–C), indicating that the plants in the low P treatment in both of the experiments were P-starved. TaPHT1.1/1.9, TaPHT1.2, and TaPHT1.10 were predominantly expressed in roots in both experiments (**Figures 3A**, **4A**), and their expression was dramatically induced by low P treatment in the hydroponic culture (**Figure 3A**), but not in the field experiment (**Figure 4A**). Of these four root-specific genes, TaPHT1.10 displayed the highest expression and TaPHT1.1/1.9 the lowest (**Figures 3A**, **4A**). TaPHT1.3/1.4 and TaPHT1.6 were expressed in both roots and shoots, and TaPHT1.6 exhibited stronger expression than TaPHT1.3/1.4 in both experiments (**Figures 3B**, **4B**). These three genes differed in the response

600, Optical density at 600 nm. Data are mean ± SE of three biological replications.

to P supply. Compared to high P treatment, low P treatment upregulated TaPHT1.3/1.4 in roots and TaPHT1.6 in shoots in the hydroponic culture (**Figure 3B**), and upregulated TaPHT1.3/1.4 in roots and shoots and TaPHT1.6 in shoots in the field experiment (**Figure 4B**). TaPHT1.5, 1.7 and 1.8 were presented at very low expression levels in both roots and shoots in both of the experiments (**Figures 3C**, **4C**). Upregulation by low P treatment was observed for TaPHT1.7 in shoots in the hydroponic culture (**Figure 3C**), TaPHT1.7 in roots and TaPHT1.8 in roots and shoots in the field experiment (**Figure 4C**).

Since TaPHT1.6 had the most abundant transcripts in shoots among the investigated TaPHT1 genes, we further analyzed the expression of TaPHT1.6 in different aerial parts at the flowering and grain filling stages (14 days after flowering) in the field experiment. The expression of TaPHT1.6 was much higher in leaves than in stems, spikes, and grains, and was higher in aging leaves than in flag leaves (Supplementary Figure S5). Significant upregulation by low P treatment was observed in aging leaves, stems, spikes, and grains (Supplementary Figure S5).

### Relationship of TaPHT1s Expression with P Uptake

We measured P uptake of two commercial wheat varieties at stem elongation, flowering, and maturity stages in two consecutive

field experiments (2010 experiment and 2011 experiment). Data were collected at the P application rates of 0 kg P ha−<sup>1</sup> (P0), 100 kg P ha−<sup>1</sup> (P100), and 400 kg P ha−<sup>1</sup> (P400). In most cases, the wheat variety KN9204 had higher total P concentration in shoots at stem elongation and flowering and in straws and grains at maturity than the wheat variety SJZ8, except for that of stem elongation in the 2011 experiment (**Table 1**). Comparison of aerial P accumulation between these two varieties showed that KN9204 absorbed more P than SJZ8 after stem elongation at all the P application rates in both of the field experiments (**Figure 5**).

As the differences in P uptake between KN9204 and SJZ8 were mainly observed at flowering and maturity, we analyzed the TaPHT1s expression at the flowering stage in the 2011 field experiment. The higher expression of TaIPS1.1 at P0 than at P100 and P400 indicated that the wheat plants grown under P0 conditions were P-starved (Supplementary Figure S4D). TaPHT1.1/1.9, 1.2, and 1.10 were expressed more abundantly in the roots of KN9204 than in those of SJZ8 at all the three P rates (**Figures 6A–C**), whereas SJZ8 had higher expression of TaPHT1.8 in roots at P100 and P400 (**Figure 6D**) and higher expression of TaPHT1.6 in roots at P0 and P100 (**Figure 6E**) and in shoots at P0 than KN9204 (**Figure 6F**). The paired t-test showed that the mean values across the three P application rates for P uptake after stem elongation in 2010 and 2011 field experiments and the expression of TaPHT1.1/1.9 and 1.10 at flowering in 2011 field experiment were significantly higher in KN9204 than in SJZ8 (Supplementary Table S4). We further analyzed the correlations between gene expression at flowering and P uptake after stem elongation (difference between stem elongation and maturity). P uptake after stem elongation showed a positive correlation with the expression of TaPHT1.1/1.9 (**Figure 7A**), TaPHT1.2 (**Figure 7B**), and TaPHT1.10 in roots (**Figure 7C**), but a negative correlation with the expression of TaPHT1.8 in roots (**Figure 6D**) and TaPHT1.6 in roots and shoots (**Figures 6E,F**).

### DISCUSSION

We identified a total of 36 TaPHT1 genes named from TaPHT1.1 to TaPHT1.14 in wheat. Of the 32 PHT1 genes with chromosome location information, 12, 11, and 9 were from the A, B, and D genomes (Supplementary Table S3), respectively. In order to evaluate the PHT1 number in wheat, we also identified 13 PHT1 genes in Triticum urartu and 14 PHT1 genes in Aegilops tauschii, and these PHT1 genes were named from PHT1.1 to PHT1.16 (Supplementary Table S3). We did not find the wheat Gene IDs which are orthologous to PHT1.15 and PHT1.16 of Triticum urartu and Aegilops tauschii (**Figure 1**), but several scaffolds from the short arms of group 2 chromosomes in wheat contained PHT1.15 and PHT1.6 like fragments which were not annotated yet. Although we identified a PHT1.14 gene (TaPHT1.14-U) in wheat, there were three closely related PHT1.14 genes (AEGtaPHT1.14-1, 1.14-2, and 1.14-3) in Aegilops tauschii (**Figure 1**). Further detailed analysis of the Chinese spring genome sequence found that the forward orientation of seven fragments showed high similarity with TaPHT1.14- U in the scaffold TGACv1\_scaffold\_642582\_U (Supplementary Figure S6). The second, fourth, fifth, and sixth fragments showed similarity only with the 3<sup>0</sup> -end of TaPHT1.14-U (Supplementary Figure S6). However, the first fragment showed 96–98% identity with 1–1604 bp of the 1656 bp coding region in TaPHT1.14-U (the third fragment 3, Supplementary Figure S6), whereas the seventh fragment located in the last 550 bp of the scaffold showed 99% identity with 1–550 bp of the coding region in TaPHT1.14-U (Supplementary Figure S6). As such, TGACv1\_scaffold\_642582\_U may contain three closely related PHT1.14 genes. Taking the information together, there may have as many as 16 (if with one PHT1.14 gene)-18 (if with three PHT1.14 genes) PHT1 genes in each of the three subgenomes in wheat.

The cloned 10 TaPHT1 promoters contained the putative Pi-starvation response regulator PHR1 binding cis-element P1BS


Different letters (a and b) indicate significant difference between SJZ8 and KN9204 at p < 0.05 level.

and the WRKY transcription factor binding element W-Box (Supplementary Figure S2), indicating that TaPHT1s may be regulated by PHR1 and WRKY transcription regulatory factors. In fact, our previous study has documented that TaPHR1 can form homodimers to activate TaPHT1.10-U expression in vitro (Wang et al., 2013). It has been reported that PHR1 regulates Pi starvation-inducible genes by binding as a dimer to the cis-element P1BS in the promoter region of its downstream gene (Rubio et al., 2001) and the majority of Pi starvation-inducible genes contain the P1BS element (Muller et al., 2007; Nilsson et al., 2010). As such, the P1BS elements in the promoters might contribute to the observed upregulation of TaPHT1 genes by low P treatment (**Figure 3**). Several WRKY transcription factors have been found to bind to the W-box to regulate the expression of Pi-starvation response genes in Arabidopsis (Devaiah et al., 2007; Chen et al., 2009; Wang et al., 2014). Whether WRKY transcription factors involved in regulating the response of TaPHT1 genes to P-deficiency is needed to be studied in the future.

The seven tested TaPHT1 genes showed Pi-transport activity in yeast cells grown under low Pi and high Pi conditions (**Figure 2**). The genes from the same branch of the phylogenetic tree shared similar tissue-specific expression and response to P-deficiency (**Figures 3**, **4**). The expression of TaPHT1.1/1.2/1.9/1.10 in Branch V was root-specific and upregulated by low P treatment in the hydroponic culture (**Figure 3A**), but their upregulations by low P treatment was abolished in the field experiments (**Figures 4A**, **6A–C**). These abolished upregulations by low P treatment were possible due to the increased AM colonization in roots under P deficiency (Teng et al., 2013), as AM colonization has been found to inhibit the response of HvPHT1.1 and HvPHT1.2 to P-deficiency in

roots of barley (Glassop et al., 2005). In Branch V, HvPHT1.1 has been found to encode high-affinity transporters of Pi (Preuss et al., 2011). Taking information together, TaPHT1 transporters in Branch V may function as high-affinity Pi transporters mediating Pi acquisition from soils. TaPHT1.6 in Branch III and TaPHT1.3/1.4 in Branch IV were expressed in both roots and shoots and were upregulated by low P treatment in the hydroponic culture and in the field experiment (**Figures 3B**, **4B**). In aerial parts, TaPHT1.6 was expressed in stems, leaves, spikes, and grains (Supplementary Figure S5). HvPHT1.6 in Branch III was expressed in both roots and shoots (Rae et al., 2003). OsPHT1.8 in Branch IV was expressed in various tissue organs from roots to seeds and plays an important role in Pi homeostasis and P redistribution from source to sink organs (Jia et al., 2011; Li et al., 2015). These results indicate that PHT1s in Branches III and IV may mediate Pi remobilization in whole plant. However, they may have diverse affinities for Pi, as OsPHT1.8 has been shown high-affinity for Pi (Jia et al., 2011), and HvPHT1.6 low-affinity for Pi (Rae et al., 2003). TaPHT1.5 in Branch IV and TaPHT1.7 in Branch III were expressed at very low levels in both of the hydroponic culture and field experiment (**Figures 3C**, **4C**). The reported AM fungi inducible PHT1s were grouped into Branch VI (**Figure 1**), including TaPHT1.8 and TaPHT1.11 from wheat, HvPHT1.8 and HvPHT1.11 from barley, ZmPHT1.6 from maize (Glassop et al., 2005; Sisaphaithong et al., 2012), and OsPHT1.11 from rice (Paszkowski et al., 2002). Here, we found that TaPHT1.8 was upregulated by low P treatment in the field experiments (**Figures 4C**, **6D**),

and P < 0.01, respectively.

but not by low P treatment in the hydroponic culture (**Figure 3C**). Since we observed that low P treatment increased AM colonization rate in roots of KN9204 compared to high P treatment in field experiments (Teng et al., 2013), the upregulation of TaPHT1.8 by low P treatment in the field experiments might reflect the fact that TaPHT1.8 was exclusively induced by AM fungi (Glassop et al., 2005).

Previous studies state that transgenic modifying expression of PHT1 genes altered P uptake and re-distribution in wheat (Liu et al., 2013) and rice (Ai et al., 2009; Jia et al., 2011; Yan et al., 2014). The transcript abundance of PHT1 genes has been shown to relate with P uptake in wheat (Aziz et al., 2014) and P utilization efficiency in barley (Huang et al., 2011) under controlled conditions. These results indicate that mRNA levels of PHT1 genes affect the capacities of P uptake and remobilization. Our current on-farm field-scale study showed that the expression of TaPHT1.1/1.2/1.9/1.10 correlated with the differences in P uptake between different wheat varieties. The positive correlations between P uptake after stem elongation and the expression levels of TaPHT1.1/1.2/1.9/1.10 at the flowering stage (**Figures 7A–C**) might result from two factors: P supply level and wheat variety. Firstly, both P uptake and expression of these TaPHT1 genes increased with P application rate (**Figures 5**, **6A–C**). Secondly, KN9204 had higher P uptake after stem elongation and the relative expression levels of TaPHT1.1/1.2/1.9/1.10 at the flowering stage than SJZ8 at a given P application rate (**Figures 5**, **6A–C**). In contrast to the positive correlations between the expression of TaPHT1.1/1.2/1.9/1.10 and P uptake after stem elongation, the expression of TaPHT1.8 and TaPHT1.6 negatively correlated with P uptake after stem elongation (**Figures 7D–F**). The negative correlation between TaPHT1.8 expression and P uptake after stem elongation resulted from the decreased TaPHT1.8 expression with P application rate and lower TaPHT1.8 expression in roots of KN9204 compared to that of SJZ8 at P100 and P400 (**Figure 7D**). However, this negative correlation did not support that AM colonization inhibited P uptake, as we did not analyze the expression of TaPHT1.11 yet. It has been reported that TaPHT1.11-A1, -B1, and -D1 were AM-inducible and were expressed at much higher level than TaPHT1.8 (Sisaphaithong et al., 2012). The negative correlation between TaPHT1.6 expression and P uptake after stem elongation mainly resulted from the decreased TaPHT1.6 expression with P application rate (**Figures 7E,F**), as KN9204 and SJZ8 had similar expression levels of TaPHT1.6 in roots at P400, and in shoots at P0, P100 and P400 (**Figures 6E,F**). Although TaPHT1.6 may mediate P redistribution, the similar expression levels of TaPHT1.6 in shoots at flowering did not explain the differences in grain P concentration between these two varieties (**Table 1**). This was possibly because that the transport of P to grains occurs during grain filling. As such, further research is needed to investigate the expression of TaPHT1 genes including TaPHT1.6

### REFERENCES

Ai, P. H., Sun, S. B., Zhao, J. N., Fan, X. R., Xin, W. J., Guo, Q., et al. (2009). Two rice phosphate transporters, OsPht1;2 and OsPht1;6, have different functions during grain filling, the research may identify the TaPHT1 genes which contribute to the differences in grain P concentration between different wheat varieties.

In summary, the hexaploid wheat has many more PHT1 genes than the diploid cereal crops such as barley and rice. Although we performed genome-wide analysis of PHT1 genes, we did not isolate all the PHT1 genes in wheat. The on-going wheat genome sequencing project will help us to understand the complexity of the Pi transport system in wheat. Although there were a large number of PHT1 genes in wheat, the TaPHT1 transporters from a given branch of the phylogenetic tree shared high similarities in sequences, expression locations, and responses to P-availability, this finding will help us to predict the roles of TaPHT1 genes in mediating Pi uptake and re-distribution. Our research also provided useful cues to understand the influences of PHT1 genes on the genotypic differences in P uptake. Further studies on mechanisms underlying the genotypic differences in PHT1 expression will facilitate the breeding of wheat varieties with improve P use efficiency.

### AUTHOR CONTRIBUTIONS

Y-PT, WT, and Y-YZ designed this study; Y-YZ screened BAC clones, WT and Y-YZ analyzed PHT1 sequences; Y-YZ and WT assayed expression and function of PHT1 genes; all authors carried out the field experiments; WT and Y-YZ wrote the manuscript under the supervision of Y-PT. All authors have read and approved this manuscript.

### FUNDING

This research was supported by the National Key Research and Development Program of China from Ministry of Science and Technology of China (2016YFD0100706) and the National Transgenic Key Project from the Ministry of Agriculture of China (2016ZX08002-005).

### ACKNOWLEDGMENT

The yeast mutant MB192 and vector p112A1NE were kindly provided by Prof. Shubin Sun from Nanjing Agricultural University, Nanjing, China.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2017.00543/ full#supplementary-material

and kinetic properties in uptake and translocation. Plant J. 57, 798–809. doi: 10.1111/j.1365-313X.2008.03726.x

Aziz, T., Finnegan, P. M., Lambers, H., and Jost, R. (2014). Organ-specific phosphorus-allocation patterns and transcript profiles linked to phosphorus

efficiency in two contrasting wheat genotypes. Plant Cell Environ. 37, 943–960. doi: 10.1111/pce.12210



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Teng, Zhao, Zhao, He, Ma, Deng, Chen and Tong. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.