# HORMONAL CONTROL OF IMPORTANT AGRONOMIC TRAITS

EDITED BY : Chi-Kuang Wen, Yunde Zhao and Yong-Ling Ruan PUBLISHED IN : Frontiers in Plant Science and Frontiers in Physiology

#### Frontiers Copyright Statement

© Copyright 2007-2018 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 978-2-88945-686-4 DOI 10.3389/978-2-88945-686-4

#### 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

# HORMONAL CONTROL OF IMPORTANT AGRONOMIC TRAITS

Topic Editors:

Chi-Kuang Wen, Shanghai Institute of Plant Physiology and Ecology (CAS), China Yunde Zhao, University of California San Diego, United States Yong-Ling Ruan, The University of Newcastle, Australia

Rapeseed blossoms, a harmony of sustainable agriculture, agrodiversity, and environment (Qianxi County, Bijie City of Guizhou Province, China).

Image: Dr. Jianru Zuo (Institute of Genetics and Developmental Biology (CAS), China), reproduced with permission.

One of the goals of plant science is to improve agricultural sustainability, increasing yield, food diversity, and nutrition, while minimizing the negative impact on our environment. In response to internal and external cues, plant hormones control various aspects of plant growth and development. The wealth of our knowledge on plant hormones shall greatly advance sustainable agriculture.

Citation: Wen, C-K., Zhao, Y., Ruan, Y-L., eds. (2018). Hormonal Control of Important Agronomic Traits. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-686-4

# Table of Contents

*06 Editorial: Hormonal Control of Important Agronomic Traits* Chi-Kuang Wen, Yunde Zhao and Yong-Ling Ruan

#### CHAPTER 1

#### HORMONAL CONTROL OF PLANT DEVELOPMENT AND ARCHITECTURE

	- Yunhua Xiao, Dapu Liu, Guoxia Zhang, Hongning Tong and Chengcai Chu

Xiaozhao Xu, Xu Li, Xingwang Hu, Ting Wu, Yi Wang, Xuefeng Xu, Xinzhong Zhang and Zhenhai Han

# CHAPTER 2

#### INTERACTIONS OF PLANT HORMONES WITH EXTERNAL CUES


Swadhin Swain, Han-Wei Jiang and Hsu-Liang Hsieh


Adrien Corot, Hanaé Roman, Odile Douillet, Hervé Autret, Maria-Dolores Perez-Garcia, Sylvie Citerne, Jessica Bertheloot, Soulaiman Sakr, Nathalie Leduc and Sabine Demotes-Mainard

#### CHAPTER 3

#### PLANT HORMONES IN AGRICULTURE


Emuejevoke Vwioko, Onyekachukwu Adinkwu and Mohamed A. El-Esawi


Meiru Jia, Ping Du, Ning Ding, Qing Zhang, Sinian Xing, Lingzhi Wei, Yaoyao Zhao, Wenwen Mao, Jizheng Li, Bingbing Li and Wensuo Jia

*235 Gibberellins Promote Brassinosteroids Action and Both Increase Heterosis for Plant Height in Maize (*Zea mays *L.)*

Songlin Hu, Cuiling Wang, Darlene L. Sanchez, Alexander E. Lipka, Peng Liu, Yanhai Yin, Michael Blanco and Thomas Lübberstedt

*252 Genetic Regulation of GA Metabolism During Vernalization, Floral Bud Initiation and Development in Pak Choi (*Brassica rapa *ssp.* chinensis *Makino)*

Mengya Shang, Xueting Wang, Jing Zhang, Xianhui Qi, Amin Ping, Leiping Hou, Guoming Xing, Gaizhen Li and Meilan Li

*262 Identification, Classification, and Expression Analysis of* GRAS *Gene Family in* Malus domestica

Sheng Fan, Dong Zhang, Cai Gao, Ming Zhao, Haiqin Wu, Youmei Li, Yawen Shen and Mingyu Han

# CHAPTER 4

#### ACTIONS OF PLANT HORMONES AND GROWTH REGULATORS

*281 DWARF14, A Receptor Covalently Linked With the Active Form of Strigolactones, Undergoes Strigolactone-Dependent Degradation in Rice* Qingliang Hu, Yajun He, Lei Wang, Simiao Liu, Xiangbing Meng, Guifu Liu, Yanhui Jing, Mingjiang Chen, Xiaoguang Song, Liang Jiang, Hong Yu, Bing Wang and Jiayang Li

*292 Accumulation and Transport of 1-Aminocyclopropane-1-Carboxylic Acid (ACC) in Plants: Current Status, Considerations for Future Research and Agronomic Applications*

Lisa Vanderstraeten and Dominique Van Der Straeten

*310 N-Terminus-Mediated Degradation of ACS7 is Negatively Regulated by Senescence Signaling to Allow Optimal Ethylene Production During Leaf Development in* Arabidopsis

Gongling Sun, Yuanyuan Mei, Dewen Deng, Li Xiong, Lifang Sun, Xiyu Zhang, Zewen Wen, Sheng Liu, Xiang You, Nasrullah, Dan Wang and Ning Ning Wang

*323 Possible Interactions Between the Biosynthetic Pathways of Indole Glucosinolate and Auxin*

Siva K. Malka and Youfa Cheng

*337 The Mammalian Peptide Adrenomedullin Acts as a Growth Factor in Tobacco Plants*

Rafael Peláez, María Niculcea and Alfredo Martínez

*349 miRNA and Degradome Sequencing Reveal miRNA and Their Target Genes That May Mediate Shoot Growth in Spur Type Mutant "Yanfu 6"* Chunhui Song, Dong Zhang, Liwei Zheng, Jie Zhang, Baojuan Zhang, Wenwen Luo, Youmei Li, Guangfang Li, Juanjuan Ma and Mingyu Han

#### CHAPTER 5

#### INNOVATION IN RESEARCH TOOLS

*373 Single-Molecule Fluorescence Methods to Study Plant Hormone Signal Transduction Pathways*

Song Song, Jian Chang, Chongjun Ma and Yan-Wen Tan

*388 Dynamic Regulation of Auxin Response During Rice Development Revealed by Newly Established Hormone Biosensor Markers* Jing Yang, Zheng Yuan, Qingcai Meng, Guoqiang Huang, Christophe Périn,

Charlotte Bureau, Anne-Cécile Meunier, Mathieu Ingouff, Malcolm J. Bennett, Wanqi Liang and Dabing Zhang

*405 GSHR, a Web-Based Platform Provides Gene Set-Level Analyses of Hormone Responses in* Arabidopsis

Xiaojuan Ran, Jian Liu, Meifang Qi, Yuejun Wang, Jingfei Cheng and Yijing Zhang

# Editorial: Hormonal Control of Important Agronomic Traits

#### Chi-Kuang Wen<sup>1</sup> \*, Yunde Zhao<sup>2</sup> and Yong-Ling Ruan<sup>3</sup>

<sup>1</sup> National Key Laboratory of Plant Molecular Genetics, CAS Center for Excellence in Molecular Plant Sciences, Shanghai Institute of Plant Physiology and Ecology, Chinese Academy of Sciences, Shanghai, China, <sup>2</sup> Section of Cell and Developmental Biology, University of California, San Diego, La Jolla, CA, United States, <sup>3</sup> School of Environmental & Life Sciences, University of Newcastle, Callaghan, NSW, Australia

Keywords: plant hormones, light signaling, auxin, ethylene, ABA, brassinosteroids, strigolactones, agronomic traits

**Editorial on the Research Topic**

#### **Hormonal Control of Important Agronomic Traits**

From early 1950s to late 1960s, crop productivity was greatly increased by the adoption of the integrated agricultural practice involving application of chemical fertilizers, pesticides, irrigation, and high-yield variety crops with weakened gibberellin pathways. The high-intensity agricultural practice and the prevalent use of high-yield crop varieties, on the other hand, may bring profound harm to our environment and greatly impact agrodiversity (Pingali, 2012). With a rapid increase of the global population and depletion of natural resources, sustainable agricultural practice becomes critical for food security and environmental conservation. Plant hormones and their signaling networks are involved in almost all aspects of plant growth and development, and this Frontiers research topic covers a breadth of studies on hormonal controls of plant traits and biological activities, which may potentially be applied to improve agronomic traits, agricultural diversity, and sustainability.

The environmental signal light and plant hormones coordinately regulate many aspects of plant growth and development. One important light response is shade avoidance. Plants grown under canopy can sense the shade by detecting reduced red/far red light ratio and are able to adjust growth and development accordingly (Smith, 2000; Martínez-García et al., 2014). Chlorophyll biosynthesis upon exposure of etiolated plant tissues to light marks the initiation of photoautotrophic growth, and its degradation may facilitate nutrient reallocation upon senescence and fruit ripening. Three articles, by Yang and Li, Liu et al., and Zhu et al., respectively, reviewed regulations of shade response, chlorophyll biosynthesis, and degradation by light in concert with plant hormones. These articles addressed the integration of the external light signal with the internal hormone signaling network in the control of morphogenesis and other important biological processes such as stem elongation and flowering time.

Hormonal responses may interweave with nitrogen-use efficiency. In this context, the mutation of DELLA proteins involved in the green revolution rice varieties results in the inhibition of gibberellin responses and nitrogen assimilation efficiency (Li et al., 2018). Nitrate is a major nitrogen source for most land plants. Here, Guan reviewed the nitrate transport, signaling, and the interplay of nitrate with the biosynthesis and signaling of plant hormones. The intensive application

Edited and reviewed by: Catherine Bellini, Umeå University, Sweden

> \*Correspondence: Chi-Kuang Wen qgwen@sibs.ac.cn

#### Specialty section:

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

Received: 01 September 2018 Accepted: 26 September 2018 Published: 17 October 2018

#### Citation:

Wen C-K, Zhao Y and Ruan Y-L (2018) Editorial: Hormonal Control of Important Agronomic Traits. Front. Plant Sci. 9:1504. doi: 10.3389/fpls.2018.01504

**6**

of nitrogen fertilizers in modern agricultural practice may thus alter plant hormone homeostasis and responses, and the interplay may be a new research topic important for sustainable agriculture and environmental conservation.

Auxin plays pivotal roles in almost every major developmental process throughout the life cycle of a plant. Malka and Cheng reviewed the relationship between auxin biosynthesis and the production of the plant secondary metabolite glucosinolates (GLS). They proposed that auxin homeostasis might be modulated by GLS metabolism via the interaction of the biosynthetic pathways of indole glucosinolates and auxin. Plant cells can be programed to differentiate into root cells from explants and calli, namely direct and indirect de novo root regeneration, respectively. Yu et al. proposed that the steps for cell fate transition require auxin and are very similar between the two types of root regeneration.

ABA regulates several aspects of root development. Wang et al. reported that the promotion of rice root hair elongation by ABA is achieved through the positive control of auxin biosynthesis and transport. A related finding was that an elevated miR156 level is required for the auxin-induced adventitious root formation in Malus xiaojinensis.

Ethylene is a gaseous plant hormone, and ethylene response is regulated at both the biosynthesis and signaling levels. As the immediate ethylene biosynthesis precursor, ACC levels are tightly associated with ethylene production. Vanderstraeten and Van Der Straeten assessed the regulations of ACC pool and discussed a novel field of study on ACC as a potential signal molecule independent of ethylene action. The stability of Type I and Type II ACC Synthases (ACSs) is dependent on the phosphorylation status of the C-terminal region (Xu and Zhang, 2015). The Type III ACS7 does not have the C-terminal extension and its degradation is mediated by the RING-type E3 ligase XBAT32 (Lyzenga et al., 2012). Sun et al. reported involvement of the N-terminal 14 amino acid residues in ACS7 degradation, which is negatively regulated by leaf senescence, revealing a regulation of ACS7 stability by the senescence signal. As an important gaseous plant hormone, aspects of ethylene-related agronomic traits of regular rice varieties were reviewed by Yin et al. The nuclear localization signal of the ER protein EIN2 is essential to ethylene signaling activation (Zhang and Wen, 2015). Interestingly, Kessenbrock et al. reported an interaction of the tomato ethylene receptors LeETR4 and Never-ripe (Nr) with the synthetic peptide derived from EIN2 NLS sequence. The peptide may have agricultural value serving as a ripening inhibitor. The

#### REFERENCES


finding that the NLS-peptide affects ethylene responses may also shed light on how ethylene signaling is regulated.

Crop architecture is a key yield-determining trait (Wang et al., 2018), and the green revolution was directly resulted from using semi-dwarf crop varieties, which decreased lodging-related yield loss (Li et al., 2018). The plant hormones brassinosteroids (BRs) and strigolactones play prominent roles in shaping plant architecture. The study by Hu et al. revealed that elevated levels of gibberellins and BRs contribute to maize heterosis for plant height. The rice plant architecture can be regulated by the transcription factor OPF1, which interacts with the GRAS protein DLT in response to BRs (Xiao et al.). Strigolactones promote branching (tilling) of rice plants, and Hu et al. unveiled the underlying mechanism of a novel feedback regulation. The perception of strigolactones by the enzyme-receptor protein DWARF14 leads to its ubiquitination at Lys280 residue and subsequent degradation. Nie et al. reported the improvement of several major agronomic traits in tomato by overexpressing the BR receptor SlBRI1.

Research tools are continuously innovated to drive new discoveries. Yang et al. developed a series of auxin biosensors that facilitate studies on dynamic regulation of auxin responses throughout development of the rice plant. The GSHR web server (http://bioinfo.sibs.ac.cn/GSHR/) by Ran et al. facilitates the mining of biologically significant data for understanding plant hormone responses via cross-platform comparisons. Song et al. reported single-molecule fluorescence methods that can be widely applied to study the interaction kinetics of signaling components of a signal transduction pathway, and a competitive association of the BR signaling components BKI1 and BES1 with 14-3-3κ was exemplified.

The modern plant biology covers a wide variety of studies, such as genomics, transcriptomics, proteomics, genetics, and molecular biology. Its progress is often facilitated by the development of innovated tools. Plant growth and development are fine-tuned and dynamically regulated at many levels of hormonal control and interactions. Integration of the various fields of modern plant biology studies and understanding the complexity, plasticity, and the underlying mechanisms will aid diversifying agricultural practice and breeding to improve sustainable agriculture, agrodiversity, and environment.

#### AUTHOR CONTRIBUTIONS

C-KW wrote the manuscript. YZ and Y-LR made revisions.


**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 Wen, Zhao and Ruan. 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.

# Abscisic Acid Regulates Auxin Homeostasis in Rice Root Tips to Promote Root Hair Elongation

Tao Wang1,2, Chengxiang Li<sup>1</sup> , Zhihua Wu<sup>2</sup> , Yancui Jia<sup>2</sup> , Hong Wang<sup>2</sup> , Shiyong Sun<sup>2</sup> , Chuanzao Mao<sup>3</sup> and Xuelu Wang<sup>2</sup> \*

<sup>1</sup> National Key Laboratory of Genetic Engineering, School of Life Sciences, Fudan University, Shanghai, China, <sup>2</sup> National Key Laboratory of Crop Genetic Improvement, Center of Integrative Biology, College of Life Science and Technology, Huazhong Agricultural University, Wuhan, China, <sup>3</sup> State Key Laboratory of Plant Physiology and Biochemistry, College of Life Science, Zhejiang University, Hangzhou, China

Abscisic acid (ABA) plays an essential role in root hair elongation in plants, but the regulatory mechanism remains to be elucidated. In this study, we found that exogenous ABA can promote rice root hair elongation. Transgenic rice overexpressing SAPK10 (Stress/ABA-activated protein kinase 10) had longer root hairs; rice plants overexpressing OsABIL2 (OsABI-Like 2) had attenuated ABA signaling and shorter root hairs, suggesting that the effect of ABA on root hair elongation depends on the conserved PYR/PP2C/SnRK2 ABA signaling module. Treatment of the DR5-GUS and OsPIN-GUS lines with ABA and an auxin efflux inhibitor showed that ABA-induced root hair elongation depends on polar auxin transport. To examine the transcriptional response to ABA, we divided rice root tips into three regions: short root hair, long root hair and root tip zones; and conducted RNA-seq analysis with or without ABA treatment. Examination of genes involved in auxin transport, biosynthesis and metabolism indicated that ABA promotes auxin biosynthesis and polar auxin transport in the root tip, which may lead to auxin accumulation in the long root hair zone. Our findings shed light on how ABA regulates root hair elongation through crosstalk with auxin biosynthesis and transport to orchestrate plant development.

Keywords: ABA, auxin, crosstalk, root hair elongation, transport, biosynthesis, Oryza sativa

#### INTRODUCTION

Roots have important functions in uptake of nutrients and water, and anchoring plants in the soil. Root hairs, extensions from single root epidermal cells, constitute up to an estimated 70% of the root surface area in crops (Richardson et al., 2009; Pereg and McMillan, 2015). Root hairs help plants maintain sufficient levels of water and nutrients; for example, in different plant species under phosphate (P)-limiting conditions, up to 90% of the mineral nutrients appear to be taken up by root hairs (Föhse et al., 1991). In addition, root hairs play important roles in the uptake and transport of NO<sup>3</sup> <sup>−</sup> and NH<sup>4</sup> <sup>+</sup> (Gilroy and Jones, 2000). Compared to bald roots, a root 1 mm in diameter with root hairs 0.5 or 1 mm in average length, growing in sand, will improve the soil water uptake rate by 30 to 55% in barley (Segal et al., 2008).

#### Edited by:

Yunde Zhao, University of California, San Diego, United States

#### Reviewed by:

Chengbin Xiang, University of Science and Technology of China, China Stephan Pollmann, Centre for Plant Biotechnology and Genomics, Spain

> \*Correspondence: Xuelu Wang xlwang@mail.hzau.edu.cn

#### Specialty section:

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

Received: 10 March 2017 Accepted: 12 June 2017 Published: 28 June 2017

#### Citation:

Wang T, Li C, Wu Z, Jia Y, Wang H, Sun S, Mao C and Wang X (2017) Abscisic Acid Regulates Auxin Homeostasis in Rice Root Tips to Promote Root Hair Elongation. Front. Plant Sci. 8:1121. doi: 10.3389/fpls.2017.01121

**Abbreviations:** ABA, abscisic acid; IAA, indole-3-acetic acid; NAA, naphthaleneacetic acid; NPA, 1-N-naphthylphthalamic acid; RAM, root apical meristem; RHE; root hair regulatory element; SEM, scanning electron microscopy.

Phytohormones and abiotic stresses affect root hair formation and elongation in Arabidopsis and crops. ABA, a major abiotic stress-responsive hormone, plays an important role in root hair elongation. The application of exogenous ABA leads to root swelling and root hair formation in the tips of young and seminal rice roots, and this process requires de novo synthesis of proteins (Chen et al., 2006). Rice and Arabidopsis plants under moderate water stress (treated with polyethylene glycol) accumulate ABA and grow root hairs with high intensity (Xu et al., 2013).

Work in Arabidopsis has identified a core ABA signaling pathway (Ma et al., 2009; Park et al., 2009; Umezawa et al., 2009; Cutler et al., 2010). The ABA receptors PYRABACTIN RESISTANCE1 (PYR1)/PYRABACTIN RESISTANCE1-LIKE (PYL)/REGULATORY COMPONENTS OF ABA RECEPTOR (RCAR) bind to ABA and then interact with the subclass A type 2C protein phosphatases (PP2Cs), which can suppress SNF1-RELATED PROTEIN KINASE 2 (SnRK2s). As a result, SnRK2s phosphorylate and activate bZIP transcription factors to regulate ABA-responsive gene expression.

Bioinformatics work in rice has identified orthologs of these Arabidopsis ABA signaling components (Kim et al., 2012; He et al., 2014). Rice contains 10 members of the OsPYL/RCAR family, 9 members of the subclass A PP2Cs, 10 SAPKs (Stress/ABA-activated protein kinases), and 10 members of the group A bZIP transcription factors (Kim et al., 2012). OsABIL2, a rice ortholog of AtABI1 and AtABI2, plays a negative role in rice ABA signaling, as OsABIL2 can interact with and dephosphorylate SAPK10 to form an OsPYL1– OsABIL2–SAPK8/10 ABA signaling module (Li et al., 2015). Furthermore, the root hair length of OsABIL2 overexpression lines is significantly reduced (Li et al., 2015). However, little is known about the underlying cellular and molecular mechanisms of ABA in regulating root hair development.

Auxin also regulates root hair elongation. Exogenous auxin enhances root hair length, and inhibition of auxin signaling represses root hair elongation (Pitts et al., 1998; Rahman et al., 2002). In Arabidopsis, active polar auxin transport moves auxin to the root tip to regulate root hair elongation, and blocking auxin transport results in short root hair (Pitts et al., 1998; Rahman et al., 2002). Auxin efflux mediated by PIN2 facilitates auxin supply through basipetal auxin transport from the root apex to the root hair differentiation zone (Cho et al., 2007; Cheng et al., 2013). Modeling of auxin flow suggests that auxin influx carrier AUX1-dependent transport through non-hair cells can maintain auxin supply for developing hair cells and sustain root hair outgrowth (Jones et al., 2009). These studies showed that changes in the endogenous or exogenous auxin content through auxin transport in root tip and root hair cells affected root hair elongation.

Formation of the auxin gradient in root tips depends on auxin transport and requires local auxin biosynthesis. Auxin can be synthesized locally in the root tip (Ljung et al., 2005; Petersson et al., 2009), and a simple two-step pathway that converts tryptophan to IAA acts as the main auxin biosynthesis pathway in Arabidopsis (Mashiguchi et al., 2011; Zhao, 2012). Trp is first converted to indole-3-pyruvate (IPA) by the TAA family of amino transferases, and IPA is converted into IAA by the YUCCA (YUC) family of flavin monooxygenases. Accordingly, overexpression of the auxin biosynthesis gene YUCCA1 in Arabidopsis enhanced root hair growth compared to wild type (Zhao et al., 2001) and inhibition of auxin biosynthesis using L-amino-oxyphenypropionic acid (AOPP) blocked the formation of root hairs, which can be rescued by the application of exogenous IAA (Soeno et al., 2010).

Many studies have suggested that ABA interacts with auxin to regulate root growth and development (Kobayashi et al., 2005; Vanstraelen and Benková, 2012). For example, the mutants of AUXIN RESPONSE FACTOR 2 (ARF2), which affects auxinmediated responses, show enhanced ABA sensitivity during seed germination and primary root growth, and ABA treatment alters auxin distribution in Arabidopsis primary root tips (Wang et al., 2011). ABI4 mediates ABA's inhibition of lateral root formation via reduction of polar auxin transport, resulting in decreased auxin levels in roots (Shkolnik-Inbar and Bar-Zvi, 2010). Auxin may act as an organizer of hormonal signals for root hair growth (Lee and Cho, 2013). However, it remains unclear whether the ABA-induced root hair elongation in rice occurs through polar auxin transport or local auxin biosynthesis in the root tip.

In this study, we used transgenic lines with enhanced ABA signaling (SAPK10 overexpression) or attenuated ABA signaling (OsABIL2 overexpression) to study how ABA signaling regulates root hair elongation. We found that ABA signaling promotes root hair length in root tips and that the ABA-promoted root hair elongation requires polar auxin transport. Our RNA-seq analysis found that ABA enhances both auxin transport and auxin biosynthesis in root tips and identified a set of genes co-regulated by ABA and auxin to promote root hair length.

#### MATERIALS AND METHODS

#### Plant Materials and Growth Conditions

The wild-type rice Dongjin (O. sativa L. cv. japonica) was used in this work. The OsABIL2 and OsSAPK10 overexpression transgenic lines were in the Dongjin background, and the detail of OsABIL2 overexpression transgenic lines were introduced in previous study (Li et al., 2015). The auxin-related transgenic rice lines DR5-GUS (ZH11 background), OsPIN1b-GUS, OsPIN1c-GUS, OsPIN2-GUS, OsPIN5a-GUS, and OsPIN10a-GUS were from a previous study (Wang et al., 2009). For polar auxin transport assays by crown root GUS staining, the F<sup>1</sup> hybrid of OsABIL2-OE X DR-GUS and DJ X DR5-GUS lines was used.

For physiological analysis, the rice seeds were imbibed for 2 days at 30◦C, and the germinated seeds were then transferred to bottom-cut 96-well PCR plates. Seedlings were grown on water with a 16-h (light, 28◦C)/8-h (dark, 25◦C) rhythm for the indicated days, which was defined as normal conditions in this study.

#### Generation of Transgenic Rice Plants

For overexpressing SAPK10, the genomic sequence of Os03g0610900 was cloned into the binary vector pCAMBIA1306

fused with a FLAG-tag at the C-terminus. The primers used for cloning OsSAPK10 are shown in Supplementary Table 1.

#### Phytohormone and Chemical Treatments of Plant Materials

For measurement of SAPK10-OE expression level, roots of the 7-day-old wild-type and transgenic seedlings grown under normal conditions were cut, and the samples were frozen in liquid nitrogen and stored at −80◦C for RNA extraction.

For root hair induction assays, the 5-day-old normal grown seedlings were transferred to solutions with or without the following additions: ABA (100 mM dissolved in ethanol and in order to add the same amount of ethanol between treatment group and mock, we gradient dilution ABA to 20 mM, 10 mM, 5 mM, 2 mM, and 1 mM solution), NAA (100 mM dissolved in ethanol, and in order to add the same amount of ethanol, we gradient dilution NAA to 5 and 1 mM solution), the ABA biosynthesis inhibitor fluridon (FLU) (100 mM dissolved in ethanol), the auxin efflux inhibitor NPA (100 mM dissolved in DMSO), and the control added the same amount of ethanol and/or DMSO. After 24 h, the crown roots were fixed in FAA solution (ethanol: acetic acid: 37% formaldehyde: H2O = 50:5:10:35, v/v), and then photographed with a stereoscopic microscope (Carl Zeiss Discovery V20).

For β-glucuronidase (GUS) staining, crown roots were immersed in the GUS staining solution (1 mg/ml X-glucuronide in 100 mM sodium phosphate, pH 7.2, 0.5 mM ferricyanide, 0.5 mM ferrocyanide, and 0.1% Triton X-100), briefly subjected to a vacuum, and then incubated at 37◦C in the dark. The stained plant roots were photographed using Carl Zeiss Discovery V20 stereomicroscope.

#### Scanning Electron Microscopy (SEM)

Crown roots of the 6-day-old seedlings were cut and fixed in FAA solution overnight. The fixed samples were dehydrated in an ethanol series of 50, 60, 70, 80, 90, and 100% ethanol for 1 h each, and then treated with ethanol: tert-butyl alcohol (v/v), 3:1, 1:1, and 1:3 for 1 h each. Finally, the samples were kept in pure tert-butyl alcohol. After vacuum freeze-drying, the samples were sputter-coated with Au-Pd and further visualized with a scanning electron microscope (Hitachi TM3000). For observation of root hair initiation, we screened the root tip from the root apex to the root hair zone.

#### Root-Hair Length Measurement

Root hairs were observed, and images were captured with a Discovery V20 Stereomicroscope (Carl Zeiss). The zone of root hair growth is 4–5 mm from the root apex. At least 15 crown roots were measured, and 20 root hairs of each root were measured. Root-hair length measurements were performed with ImageJ<sup>1</sup> .

#### Quantification of IAA

To analyze IAA concentrations, firstly, according to the root hair phenotype, we divided the root tips of rice crown roots into three regions: short root hair zone (SRH), long root hair zone (LRH), and root tip zone (Tip) (Supplementary Figure 3). Each region was excised from at least 96 five-day-old seedlings treated with 0.5 µM ABA for 24 h, or not treated as a control. Samples were separated under Carl Zeiss Discovery V20 stereomicroscope and immediately frozen in liquid nitrogen, afterward stored at −80◦C. Samples were powdered in liquid nitrogen and at least 30 mg powder homogenized in 750 µl of cold extraction buffer 1 [methanol: H2O: acetonitrile = 80:19:1 (v/v), 10 ng/ml d5-IAA], then shade and shake at 4◦C 16 h (300 rpm). Samples were centrifuged at 4◦C at 13,000 rpm for 10 min, the supernatant is transferred to another Eppendorf tube. Add 450 µl of cold extraction buffer 2 [methanol: H2O: acetonitrile = 80:19:1 (v/v)] into precipitate, shade and shake at 4◦C 4 h (300 rpm), centrifuged at 4◦C at 13,000 rpm for 10 min, combine two supernatants, transfer supernatant through 0.22 µm filter to a new Eppendorf tube. Then each sample was dried using nitrogen flow and re-dissolved in 200 µl of 30% cold methanol 3–6 h. Quantification was performed in an ABI 4000Q-TRAR LC-MS system (Applied Biosystems, United States) with stable, isotope-labeled auxin as the standard (OlChemIm, Czech Specials) according to a method described previously (Liu et al., 2012).

# Quantitative Real-Time PCR Assays

The qRT-PCR assays were carried out as described previously with small modifications (Zhang et al., 2009). The primers for qRT-PCR were designed with NCBI Primer-Blast<sup>2</sup> to avoid the homologous regions, shown in Supplementary Table 1. Total RNA was extracted with the Plant RNAprep Kit (Tiangen). About 2 µg DNase-treated RNA was used for reverse transcription (M-MLV reverse transcriptase, TaKaRa). The RT-PCR amplifications were carried out with a Bio-Rad CFX system, and PCR products were monitored with SYBR green dye. The expression level was normalized by the expression of OsACTIN1 (LOC\_Os03g50885, internal control), and RT-qPCR results were analyzed by the 2−11CT method using Bio-Rad CFX Manager 3.1 software.

# Sample Collection and RNA Isolation for RNA-Seq

According to the root hair phenotype, we divided the root tips of rice crown roots into three regions: short root hair zone (SRH), long root hair zone (LRH), and root tip zone (Tip). Each region was excised from at least 96 five-day-old seedlings treated with 0.5 µM ABA for 24 h, or not treated as a control. Samples were separated and immediately frozen in liquid nitrogen, and afterward stored at −80◦C until RNA isolation. Total RNA was extracted from each tissue using TRIzol (Invitrogen). RNA purity was checked with a NanoPhotometer spectrophotometer (IMPLEN, Munich, Germany) and RNA integrity was assessed

<sup>1</sup>https://imagej.nih.gov/ij/index.html

<sup>2</sup>http://www.ncbi.nlm.nih.gov/tools/primer-blast/

with an Agilent 2100 Bioanalyzer (Agilent, Palo Alto, CA, United States).

## The cDNA Library Preparation and Transcriptome Sequencing

Two biological replicates were used for RNA-seq experiments for each tissue type. A total amount of 3 µg RNA per sample was used as input material for the RNA sample preparations. The samples were sent to Beijing Novogene Bioinformatics Technology Co., Ltd. (Beijing, China). The sequencing libraries were constructed using a NEBNext Ultra RNA Library Prep Kit for Illumina (NEB, Ipswich, MA, United States) following the manufacturer's recommendations, and index codes were added to attribute sequences to each sample. Briefly, mRNA was purified from total RNA using poly(T) magnetic beads. Quality of these libraries was assessed with an Agilent 2100 Bioanalyzer system. The clustering of the index-coded samples was performed with a cBot Cluster Generation System using a TruSeq PE Cluster Kit v3-cBot-HS (Illumina, San Diego, CA, United States) according to the manufacturer's instructions. After cluster generation, the libraries were sequenced on an Illumina Hiseq 4000 platform and 150-bp paired-end reads were generated.

In this study, the genes with a significant (P ≤ 0.05) fold change > 2 or < −2 between the control and ABA treatment were defined as differentially expressed genes responding to ABA stimulation.

#### Statistical Analysis

Statistical analysis was performed using IBM SPSS 20.0 software. One-way analysis of variance was used, and comparison between the two groups was performed using the least significant difference (LSD) test. Differences were considered significant if P < 0.05.

#### Functional Enrichment Analysis

Gene ontology analysis was carried out using the Singular Enrichment Analysis (SEA) tool offered by agriGO (Du et al., 2010) at default settings of Fisher t-test (P < 0.05), False Discovery Rate (FDR) correction by Hochberg method and five minimum number of mapping entries against the rice-specific precomputed background reference.

### RESULTS

## ABA Regulates Root Hair Growth in Rice

To understand how plants adapt to environmental stresses, we tested whether the stress-responsive hormone ABA affects root hair growth in the Dongjin (DJ) cultivar of rice. Crown roots of the 5-day-old seedling were treated with exogenous ABA at concentrations of 0.1, 0.5, 1, and 2 µM for 24 h. As shown in **Figure 1A**, in the seedlings treated with 0.1 µM ABA, the root hair length was significantly less than in the control seedlings not

treated with ABA. However, treatments with 0.5, 1, and 2 µM ABA significantly enhanced root hair elongation, indicating higher concentrations of ABA can promote root hair elongation, and we used the ABA concentrations of 0.5 and 2 µM for further studies.

We also used SEM to observe root hair morphology. We found that most root hairs initiated in the region about 3 mm from the root apex under normal growth conditions, but most root hairs initiated in the region about 0.2 mm from the apex in roots treated with 2 µM ABA for 24 h. We also further confirmed that the root hair length increased in response to ABA treatment (**Figure 1B** and Supplementary Figure 1). These results indicated that exogenous ABA not only enhances root hair length, but also promotes earlier root hair initiation. In addition, fluridone as an ABA biosynthetic inhibitor (10 µM) was used to investigate the effect of ABA on root tip responses under moderate water stress in rice and Arabidopsis (Yoshioka et al., 1998; Xu et al., 2013). The treatment of fluridon significantly repress the root hair length (**Figures 1C,D**). We measured the root hair length which located 4–5 mm distant from root apex in **Figure 1D**, this because considering the several ABA concentration treatment of root hair phenotype, we found that the best region to measure root hair length located at 4–5 mm distant from root tip.

#### ABA Signaling Promotes Root Hair Elongation

To investigate whether ABA promotes root hair elongation through the major ABA signaling components, we constructed rice lines overexpressing SAPK10, which plays a positive role in the ABA signaling. Three of the SAPK10 overexpression lines showed ∼60-fold, ∼100-fold, and ∼130-fold increases in SAPK10 expression in the roots of the 7-day-old seedlings (**Figure 2A**). Then we measured the length of root hairs in these lines and found that the SAPK10 overexpression significantly enhanced root hair elongation (**Figure 2B** and Supplementary Figure 2). However, root hairs were rarely observed in the transgenic lines expressing OsABIL2, which is a negative regulator of rice ABA signaling (**Figure 2C**). The next to detect the root hair of

SAPK10-OE and OsABIL2-OE response to ABA treatment, we would like to clarify the SAPK10-OE-1 was used in this study. Because this SAPK10-OE-1 line has moderate overexpression level of SAPK10 gene, which lead to moderate response to ABA treatment suitable for comparison with the wild type. While the other two lines SAPK10-OE-2, and -3 have much higher level expression of SAPK10 than the SAPK10-OE-1 line (**Figure 2A**), which lead to root hair elongation in the region very close to the root tip in the other two lines, rather than in the region 4–5 mm from the root tip after ABA treatment (Supplementary Figure 2). Considering these factors, we choose the SAPK10-OE-1 line for all other experiments for better comparison to the wild type after ABA treatments. With the 0, 0.1, 0.5, 1, and 2 µM ABA treatments (**Figures 2C,D**), the root hair length of the SAPK10- OE-1 line increased much more than that of the wild-type DJ. In contrast, ABA treatment did not increase root hair elongation of the OsABIL2-OE plants, which is a negative regulator of rice ABA signaling. This indicated that the effect of ABA on root hair elongation largely depends on the major PYR/PP2C/SnRK2 signaling pathway.

# Polar Auxin Transport Is Required for the ABA-Mediated Root Hair Elongation

Auxin has a positive effect on root hair elongation without affecting the determination of root hair cell fate in Arabidopsis (Masucci and Schiefelbein, 1994, 1996; Pitts et al., 1998; Cho and Cosgrove, 2002). Therefore, we asked whether the ABA-regulated root hair elongation in rice also requires auxin transport. Then we used the auxin efflux inhibitor NPA to examine this. The 5-dayold wild-type seedlings were treated with ABA, NPA, or ABA plus NPA for 24 h. As shown in **Figure 3A**, ABA plus NPA treatment inhibited root hair elongation as compared with ABA alone. Similarly, in the SAPK10-OE line, ABA plus NPA also severely inhibited the root hair elongation as compared with the wild type shown in **Figure 3B**. We then examined whether additional auxin can rescue the reduced root hair elongation in the OsABIL2-OE, using the 5-day-old seedlings of OsABIL2-OE treated with 0.1 µM or 0.5 µM NAA for 24 h. As shown in **Figure 3C**, compared with Mock, 0.5 µM of NAA can significantly rescue the root hair length of OsABIL2-OE. Taken together, these results indicated that auxin acts downstream of ABA to mediate the ABA-induced root hair elongation.

To address whether ABA influences local auxin concentrations, we used a DR5-GUS rice line, which has been widely used as a marker line for monitoring endogenous auxin levels in plants (Yamamoto et al., 2007). The 5-day-old DR5-GUS seedlings were treated with 0, 0.1, 0.5, 1, or 2 µM ABA for 24 h, then we stained their roots for GUS activity. We found that DR5 was widely expressed in the crown roots, especially in their steles, and with increased concentrations of exogenous ABA, the GUS staining gradually accumulated close to the root tip and in the outer layers of the roots (**Figure 3D**). With a higher magnification, we observed that the region with the strongest GUS signal produced the longest root hairs (**Figure 3E**), suggesting that the ABA-promoted local auxin accumulation may lead to the specified root hair elongation.

We further tested whether the ABA-induced auxin response is dependent on ABA signaling by crossing the DR5-GUS into the OsABIL2-OE background; the wild type crossed with DR5-GUS (ZH11 background) was used as a control. The GUS staining showed that ABA treatment did not cause auxin redistribution in the OsABIL2-OE crown roots (**Figure 3F**), indicating that ABA signaling promotes auxin accumulation in the root tip to induce root hair elongation. The DR5-GUS staining patterns look different between **Figures 3D,F**. There may be two reasons. First, in **Figure 3F**, F<sup>1</sup> seeds were used for the observation, which contains half dosage of the reporter gene GUS. Therefore, the GUS staining apparently is weaker in **Figure 3F** than in **Figure 3D**. In general, the relative staining density and the trends of DR5-GUS expression pattern between **Figures 3D,F** are similar.

### ABA Modulates the Expression Level and Pattern of Genes Involved in Auxin Transport

Auxin redistribution is mainly controlled by auxin transporters, which include the influx transporters of the AUXIN1/LIKE AUX1 (AUX1/LAX) family (Kerr and Bennett, 2007; Swarup et al., 2008; Zhao et al., 2015), and efflux transporters of the PIN-FORMED (PIN) families (Petrasek et al., 2006; Wang et al., 2009). To test how ABA signaling regulates auxin transport, we investigated the ABA-regulated expression of the major genes involved in auxin transport. We found that the mRNA levels of OsPINs and OsAUX1 were induced by treatment with 2 µM ABA at 6 and 24 h (**Figure 4A**).

To examine the spatial expression of these genes, we divided the root tip into three zones according to the root hair distribution and DR5-GUS activity: the short root hair zone (SRH), long root hair zone (LRH), and root tip zone (Tip) (**Figure 4B** and Supplementary Figure 3). First, we tested expression of the ABA marker genes OsLEA3, OsBZ8, OsZEP1, and OsRD22 after ABA treatments (**Figure 4C**), and the result showed that all of the ABA marker genes (except OsRD22 in Tip) were induced in the three different zones. We then detected the expression of OsPINs in the three zones and found that most of the OsPINs were induced by the ABA treatment (**Figure 4D**). Second, we used the GUS reporter lines driven by various OsPIN promoters (OsPIN1b-GUS, OsPIN1c-GUS, OsPIN2-GUS, OsPIN5a-GUS, and OsPIN10a-GUS) (Wang et al., 2009) to visualize the OsPIN expression pattern. We found that the expression of OsPIN1b-GUS, OsPIN1c-GUS, and OsPIN5a-GUS was not dramatically altered by ABA treatment, but the expression of OsPIN2-GUS and OsPIN10a-GUS was significantly enhanced by ABA treatment, especially in the outer layers of the root (**Figure 4E**). In this assay, we indeed found that there were some different between real-time RT-PCR and GUS staining. In general, real-time RT-PCR or promoter-GUS has their own advantages and disadvantages. First, real-time RT-PCR cannot distinguish the specific expression tissue, which may mask some tissue specific signals. Second, the promoter-GUS approach has a defect on the specificity of promoter, because some times, ones do not known what the exact promoter size in plants.

Therefore, the combination of these two approaches will show the expression pattern more accurately. This assay we wanted to show the genes expression fold change used the real-time RT-PCR, and GUS staining to show tissue specificity. In Arabidopsis, PIN2 is expressed in the epidermal cells (Wisniewska et al., 2006). We also checked the rice microarray expression database

hair zone (LRH), and root tip zone (Tip). Each region was cut and collected from at least 96 five-day-old seedlings with or without 0.5 µM ABA treatment for 24 h. (C) Expression of ABA-responsive genes OsLEA3, OsBZ8, OsZEP1, and OsRD22 in the Tip, LRH and SRH zones after ABA treatment. Data are means ± SD. (D) Quantitative RT-PCR analysis of OsPINs in the SRH, LRH and Tip treated with or without 0.5 µM ABA for 24 h. The expression level of each gene in the untreated Tips was defined as "1." Data are means ± SD. (E) The GUS activity of the OsPINs-GUS in the 5-day-old rice seedlings in response to 0.5 µM ABA for 12 h.

(RiceXPro<sup>3</sup> , OsPIN2 Locus ID: Os06g0660200, OsPIN10a Locus ID: Os01g0643300), and found that OsPIN2 and OsPIN10a were especially expressed in the epidermal cells (Supplementary Figures 4A,B). Therefore, we concluded that ABA-promoted root hair elongation likely requires functional basipetal auxin transport.

#### RNA-Seq Analyses of the ABA-Treated Root Tips

To obtain a global view of the differential expression of genes related to ABA-induced root hair elongation, the 5-day-old seedlings were treated without or with 0.5 µM ABA for 24 h, then the three different zones (SRH, LRH, and Tip) of root tips were collected to analyze transcript profiles (Supplementary Figure 3). We conducted two biological replicates and analyzed their repeatability by calculating the Pearson correlation coefficient (Supplementary Figure 5A), which indicates these results are consistent and repeatable.

To confirm the result that ABA promotes auxin transport and to validate the RNA-seq accuracy, we used RNA-seq to analyze the expression of OsPINs, OsAUX, and PINOID (PID). Arabidopsis PID encodes a serine/threonine protein kinase that

<sup>3</sup>http://ricexpro.dna.affrc.go.jp

regulates auxin redistribution through control of the subcellular localization of PINs (Robert and Offringa, 2008). As shown in Supplementary Figure 5B, ABA treatment promotes auxin transport by enhancing expression of OsPINs, OsAUX, and OsPID, especially in the Tip and LRH zones, consistent with the qRT-PCR (**Figure 4D**) and OsPINs-GUS staining results (**Figure 4E**). The results also showed a high correlation (R <sup>2</sup> = 0.84, Supplementary Table 2) between the RNA-seq and qRT-PCR data.

#### ABA Promotes Auxin Biosynthesis

In addition to increasing auxin transport to the LRH zone, enhanced local auxin biosynthesis or reduced local auxin degradation could also increase local auxin concentrations, as indicated by the enhanced DR5-GUS expression in the LRH zone. Therefore, we first used our RNA-seq data to measure the expression levels of genes involved in auxin biosynthesis, including OsYUCCAs (Yamamoto et al., 2005), OASA (Kanno et al., 2004), OASB, OsNIT, OsAO, and OsAMI. These data showed that almost all of these auxin biosynthesis genes were upregulated after ABA treatment in the root tips (**Figure 5A** and Supplementary Table 4) and that OsAMI1 (∼3.6-fold), OsYUCCA4 (∼1.8-fold), and OsYUCCA1 (∼1.4-fold) were upregulated in the LRH zone.

We then checked the expression of the genes encoding IAA-amido synthetases of the GH3 family, key enzymes involved in the conversion of active IAA to an inactive form via conjugation of IAA with amino acids, such as Asp, Ala, and Phe (Fu et al., 2011). We also examined the genes encoding members of the UDP glycosyltransferase (UGT) family and dioxygenase for auxin oxidation (OsDAO) (Bowles et al., 2006; Zhao et al., 2013; Kasahara, 2016). In rice, 10 of the 14 genes involved in IAA inactivation were upregulated in the tip zone, and 9 of them were also upregulated in the LRH but not expressed in the SRH (**Figure 5B** and Supplementary Table 4), suggesting that the enhanced auxin transport and biosynthesis in the Tip and LRH zones of rice roots are the major reasons for local auxin accumulation, and the enhanced auxin inactivation is most likely caused by negative feedback regulation.

We also measured the endogenous IAA levels in the Tip, LRH, and SRH zones of the rice root. As shown in **Figure 5C**, auxin accumulated in the LRH zone, but not in the Tip, and the IAA concentration in the LRH zone after ABA treatment was significantly higher than in the untreated roots. We should note that the DR5-GUS staining shows strong signal in the root tip, indicating high concentration auxin in the root tip. However, we measured the IAA concentration in Tip, LRH, and SRH, and found the different IAA concentration between "Tip" and the typical root tip as shown in Supplementary Figure 3. We would like clarify here that the defined "Tip" here includes not only the classical root tip region but also the meristematic zone,

transition zone, and root hair zone under the LRH, which covers a larger region. In addition, we used the roots from the 5-dayold seedling, and the majority of auxin may be from the shoot by transport.

## A Number of Genes Regulated by Auxin Are Involved in Root Hair Elongation in Response to ABA

To extend our analysis, we globally analyzed the genes differentially regulated in response to exogenous ABA in the three regions of root tips. We identified 1444, 838, and 855 genes that were differentially expressed before and after ABA treatment in Tip, LRH, and SRH, respectively. Of these, 1070, 353, and 477 genes were altered specifically in Tip, LRH, and SRH, respectively (**Figure 6A**). Further clustering analysis of these differentially expressed genes in each zone indicated that most of the ABA upregulated genes are in the Tip, followed by the LRH, and the fewest in the SRH (**Figure 6B**). These results suggest that the rice root tip cells rapidly sense and respond to ABA.

To understand the overall biological processes that occurred in the LRH zone in response ABA treatment, we performed gene ontology (GO) enrichment analysis of the 353 differentially expressed genes in the LRH; this identified 24 significantly (FDR < 0.05) enriched GO terms (**Figure 6C**) in two big categories. The first category includes GO terms involved in stress responses, such as "response to oxidative stress," "response to chemical stimulus," "response to stimulus," and "response to stress." The second category includes GO terms involved in "metabolic process," such as "glycine metabolic process," "small molecule metabolic process," "primary metabolic process," "cellular amino acid and derivative metabolic process," "lipid metabolic process," "cellular carbohydrate metabolic process," and "polysaccharide metabolic process." Then, we found 19 stress-responsive genes specifically responding to ABA in LRH (Supplementary Table 3). Nine of the 19 stress-responsive genes encode peroxidases, suggesting that reactive oxygen species have important roles in root hair growth in response to ABA treatment.

Because of the importance of auxin in root hair elongation, to understand whether these differentially expressed genes are related to ABA-regulated root hair elongation, we checked whether auxin affects the 353 genes that are differentially expressed in the LRH by comparing the genes to the auxinregulated genes in the RiceXpro database (Sato et al., 2013). As shown in **Figure 6D**, among the 192 ABA upregulated genes, 57 (∼29.7%) were also upregulated by auxin (**Table 1**), and 56 of the 161 ABA downregulated genes were also repressed by auxin (**Table 2**). This suggests that auxin plays an important role in ABA-promoted rice root hair elongation, and the 113 genes co-regulated by both ABA and auxin in the LRH zone may function directly in regulating root hair elongation. We found that OsPIN9 and OsPP2C 59 are upregulated, further supporting the importance of polar auxin transport in root hair elongation.

To identify root hair-specific genes, we also screened the promoter regions of the 113 genes co-regulated by ABA and auxin (2000 bp upstream of the start codon) for the RHE sequence "WHHDTGNNN(N)KCACGWH" (where W = A/T, H = A/T/C, D = G/T/A, K = G/T, and N = A/T/C/G), as previously described (Won et al., 2009). We found 69 RHEs in 51 genes, with 13 genes carrying two or more RHEs (**Table 3** and Supplementary Table 5). Because we lack root hairspecific gene expression data, we predicted whether these genes are specifically expressed in epidermal cells by searching the rice microarray expression database (RiceXPro) (Supplementary Table 5). These results indicated that 35 out of the 51 genes containing the RHE are highly expressed in epidermal cells in rice root tips.

# DISCUSSION

This study provides several lines of evidence to demonstrate that in rice, ABA promotes root hair elongation by regulating auxin transport and biosynthesis in specific zones of the roots. First, exogenous ABA treatment enhances root hair elongation in rice root tip and inhibit ABA biosynthesis also repress the root hair elongation. It has been reported that low concentration (0.1 µM) of exogenous ABA can promote root elongation, but high concentrations (>0.5 µM) of ABA inhibits root elongation (Ghassemian et al., 2000; Rowe et al., 2016). We also found that 0.1 µM ABA inhibit root hair elongation, while 0.5 µM and 2 µM ABA enhanced rice root hair elongation. One explanation is that feedback regulation is a common regulatory mechanism for phytohormonal signaling, so different concentrations of ABA may reflect the different levels and results of feedback. Second, examination of transgenic lines overexpressing OsABIL2 or SAPK10, key components in rice ABA signaling showed that ABA-promoted root hair elongation is dependent on the major PYR/PP2C/SnRK signaling pathway. The ABA-insensitive OsABIL2-OE line showed decreased root hair elongation, and the ABA-hypersensitive line SAPK10-OE showed enhanced root hair elongation. Third, our analysis demonstrated that auxin acts downstream of ABA signaling to promote root hair elongation, which depends on the ABA-regulated polar auxin transport and local auxin biosynthesis. Our results consistent with previous studies that ABA and auxin functionally interact in roots (Rock and Sun, 2005; Xu et al., 2013). For example, ABA accumulation promotes auxin transport in root tips and enhance proton secretion for maintaining root growth under moderate water stress (Xu et al., 2013). Fourth, RNA-seq analysis and RHE screening identified 35 genes that respond strongly to ABA and auxin and may also be specifically expressed in root hair cells in the rice LRH zone. These genes may have important functions in regulating root hair elongation in rice. The last few years have been seen significant progress being made in uncovering the mechanisms that are involved in root hair development in rice, but major knowledge gaps still persist, especially compared with Arabidopsis, in which 138 genes related to root hair development have already been identified. While 8 genes that are involved in root hair development<sup>4</sup> (Marzec et al., 2015). Our results provide

<sup>4</sup>www.iroothair.org

TABLE 1 | ABA and Auxin upregulated genes specially changed in LRH under ABA treatment.


(Continued)

#### TABLE 1 | Continued

fpls-08-01121 June 24, 2017 Time: 15:6 # 13


Shown are the log<sup>2</sup> fold change values (log<sup>2</sup> FC) for genes commonly induced (>1.5) by ABA with adjusted p-values < 0.05.

TABLE 2 | ABA and Auxin downregulated genes specially changed in LRH under ABA treatment.


(Continued)

#### TABLE 2 | Continued

fpls-08-01121 June 24, 2017 Time: 15:6 # 14


Shown are the log<sup>2</sup> fold change values (log<sup>2</sup> FC) for genes commonly repressed (< −1.5) by ABA with adjusted p-values < 0.05.

TABLE 3 | Distribution of RHE motif in the upregulated genes promoter between mock and ABA treatment in LRH.


35 RHE-contained genes as candidate genes to regulate rice root hair development.

Polar auxin transport apparently plays a critical role in the ABA signaling-regulated root hair elongation in rice. Treatment with the auxin transport inhibitor NPA and staining of DR5- GUS lines demonstrated that ABA-promoted root hair elongation depends on polar auxin transport. ABA signaling is required for the ABA-regulated redistribution of auxin. In addition, auxin homeostasis in the root hair cell is critical for root hair elongation. In Arabidopsis, root hair-specific expression of auxin efflux carriers such as PINs (PIN1-4, PIN7, and PIN8) strongly suppresses root hair length, suggesting that auxin efflux carriers inhibit root hair elongation by depleting auxin in the root hair cell (Lee and Cho, 2006, 2013; Cho et al., 2007). Redistribution of auxin from its concentration maximum to epidermal cells requires the activity of PIN2, AUX1 and other carriers (Swarup et al., 2005; Ikeda et al., 2009). We showed that ABA not only significantly induces OsPINs and OsAUX1 mRNA levels, as confirmed by qRT-PCR and RNA-seq analysis, but also strongly induces the ectopic expression of OsPIN2 and OsPIN10a in the LRH zone. Because OsPIN2 and OsPIN10a are specifically expressed in the epidermal cells, indicating that ABA promotion of root hair elongation requires functional basipetal auxin transport. Consistent with our results, other workers have reported that under osmotic stress in an ABA-regulated manner enhanced PIN2 expression level in Arabidopsis root (Rowe et al., 2016).

Besides polar auxin transport, local auxin biosynthesis was also shown to modulate gradient-directed planar polarity in root hair development in Arabidopsis (Ikeda et al., 2009). We found that the ABA-enhanced accumulation of auxin in the LRH is also a consequence of the upregulation of local auxin biosynthesis in the Tip and LRH zones. First, ABA upregulates expression of almost all auxin biosynthetic genes in the Tip zone and upregulates OsAMI1, OsYUCCA4, and OsYUCCA1 in the LRH zone. Second, examination of negative feedback mechanisms showed that 10 of the 14 genes involved in the inactivation of IAA were upregulated in the Tip zone and 9 were also upregulated in the LRH zone, indicating that auxin biosynthesis was enhanced in the Tip and LRH zones. Third, direct measurement of the endogenous IAA levels in the Tip, LRH, and SRH zones of the rice roots directly demonstrated the ABA-enhanced auxin accumulation in rice roots. However, the IAA concentration in the Tip region was identical before and after ABA treatment. A reasonable explanation is that ABApromoted basipetal auxin transport may lead to quick auxin flow into the LRH region. Alternatively, the drastic increase of the expression of genes encoding IAA inactivation enzymes

in the Tip region promoted a rapid conversion of IAA to inactive forms.

Our current data and previous discoveries suggested a model for the root hair elongation regulated by ABA in rice (**Figure 7**). ABA signaling promotes auxin biosynthesis in the Tip and LRH regions, and further enhanced the redistribution auxin through auxin basipetal transport to accumulate auxin in the LRH region. The high concentration of IAA activates downstream gene expression to promote root hair elongation. In the future, we plan to investigate the specific components in the ABA signaling pathway that directly regulate expression of genes involved in auxin transport and biosynthesis. In addition, the zone-specific genes identified in this study provide a great opportunity and resource to further understand how ABA regulates root hair elongation. In addition, the root hair region seems moving down during ABA treatment. We predicted this maybe caused by local auxin accumulation. According to our results, ABA promotes auxin acropetal and basipetal transport in rice tip and results in auxin local accumulation. As the applied ABA concentration increases, auxin may accumulate in the region much close to rice root tip, which lead to the move down of long root hair region. In addition, the inhibited root elongation by high concentration ABA may also result in the long root hair region closer to the root tip.

#### AUTHOR CONTRIBUTIONS

TW and XW conceived the research and planned the experiments; TW performed all of the experiments; CL provided

#### REFERENCES


SAPK10 and OsABIL2 transgenic plants; ZW performed the RNA-seq data analysis; YJ and HW helped to prepare material for RNA-seq; CM provided OsPIN-GUS transgenic plants; SS put forward improvement advise. TW and XW wrote the manuscript.

#### FUNDING

This work was supported by grants 91535104 and 31430046 (to XW), and 2016YFD0100403, 31271684 and 31540080 (to SS) of the National Natural Science Foundation of China, grant 2012CB114304 of the Ministry of Science and Technology of China (to XW and SS), and grants 2662015PY020 and 2014RC002 of Huazhong Agricultural University (to XW).

#### ACKNOWLEDGMENT

We thank Dr. Jianjun Jang of Fudan University, and Dr. Shiyong Sun and Changxi Yin of Huazhong Agricultural University for proofreading the text.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2017.01121/ 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 © 2017 Wang, Li, Wu, Jia, Wang, Sun, Mao and Wang. 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.

# Auxin Control of Root Organogenesis from Callus in Tissue Culture

Jie Yu1 †, Wu Liu1 †, Jie Liu1, 2, Peng Qin<sup>3</sup> \* and Lin Xu1, 2 \*

*<sup>1</sup> National Key Laboratory of Plant Molecular Genetics, CAS Center for Excellence in Molecular Plant Sciences, Institute of Plant Physiology and Ecology, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai, China, <sup>2</sup> University of Chinese Academy of Sciences, Beijing, China, <sup>3</sup> Department of Instrument Science and Engineering, Shanghai Jiao Tong University, Shanghai, China*

Keywords: plant regeneration, tissue culture, callus, adventitious root, de novo root regeneration

## INTRODUCTION TO DIRECT AND INDIRECT DE NOVO ROOT REGENERATION

During post-embryonic development, roots can be initiated by a programmed developmental order or by environmental and wound stimulation (Bellini et al., 2014; Xu and Huang, 2014; Birnbaum, 2016; Ikeuchi et al., 2016; Kareem et al., 2016; Lup et al., 2016; Rellan-Alvarez et al., 2016; Steffens and Rasmussen, 2016). De novo root regeneration (DNRR) is a type of plant regeneration to produce adventitious roots upon wounding or stress (Liu et al., 2014; Xu and Huang, 2014). For example, using leaf explants of Arabidopsis (Arabidopsis thaliana), adventitious roots could usually be regenerated by two ways: adventitious roots could be formed directly from detached leaf explants when cultured on B5 medium without added hormones (Chen et al., 2014; Liu et al., 2014), hereafter called direct DNRR (**Figure 1A**); or adventitious roots could be formed from callus in tissue culture, hereafter called indirect DNRR. In indirect DNRR, leaf explants are first cultured on callus-inducing medium (CIM) with high auxin levels to induce callus formation, and then the callus is transferred to root-inducing medium (RIM) with low auxin levels or even on B5 medium without auxin supplement to allow root formation (**Figure 1B**).

Many plants such as Arabidopsis and rice (Oryza sativa) can form adventitious roots from callus (**Figures 1C,D**). While the cell fate transition during direct DNRR from Arabidopsis leaf explants has been carefully studied in Arabidopsis (Liu et al., 2014; Chen et al., 2016b,c; Hu and Xu, 2016; Sheng et al., 2017), in indirect DNRR the cell fate transition is still not clear. How roots are formed from callus remains unanswered. In this paper, we present our analyses of the cell lineage of indirect DNRR and discuss the similarities and differences between direct and indirect DNRR.

# CELL FATE TRANSITION DURING DIRECT DNRR

Here, we summarize the four steps of cell fate transition involved in direct DNRR from leaf explants together with adventitious rooting in other systems (**Figure 1G**). In the first step "priming," the endogenous auxin is transported into regeneration-competent cells (i.e., procambium and vascular parenchyma cells) in the vasculature near the wound and activates WUSCHEL-RELATED HOMEOBOX11 (WOX11) expression for the fate transition from regenerationcompetent cells to root founder cells (Liu et al., 2014; Chen et al., 2016a,b). In the second step "initiation," WOX11 and auxin coordinately activate WOX5 and LATERAL ORGAN BOUNDARIES DOMAIN16 (LBD16) expression for the fate transition from root founder cells to root primordium (Liu et al., 2014; Hu and Xu, 2016; Sheng et al., 2017). WOX11 expression then decreases in this step (Liu et al., 2014; Hu and Xu, 2016). Auxin keeps a high level in the root primordium. In the third step "patterning," cell division continues in the root primordium, which begins to differentiate into a root apical meristem (RAM). The auxin level is tuned down and auxin distribution is restricted to the tip of the meristem to

#### Edited by:

*Yunde Zhao, University of California, San Diego, United States*

#### Reviewed by:

*Shuang Wu, Fujian Agriculture and Forestry University, China Akira Iwase, RIKEN, Japan*

#### \*Correspondence:

*Peng Qin pqin@sjtu.edu.cn Lin Xu xulin01@sibs.ac.cn These authors have contributed equally to this work.*

*†*

#### Specialty section:

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

Received: *04 January 2017* Accepted: *25 July 2017* Published: *08 August 2017*

#### Citation:

*Yu J, Liu W, Liu J, Qin P and Xu L (2017) Auxin Control of Root Organogenesis from Callus in Tissue Culture. Front. Plant Sci. 8:1385. doi: 10.3389/fpls.2017.01385*

confine the region of the stem cell niche (De Klerk et al., 1999; Della Rovere et al., 2013; Druege et al., 2016). WOX5 is gradually restricted into the stem cell niche and LBD16 expression decreases (Hu and Xu, 2016). In the fourth step "emergence," the mature root tip and stem cell niche are formed and the root tip grows out of the leaf explant (Chen et al., 2016c; Hu and Xu, 2016).

#### CELL FATE TRANSITION DURING INDIRECT DNRR

In tissue culture, adventitious roots could be obtained via indirect DNRR (**Figure 1H**). On CIM, callus is induced from leaf explants by a high level of auxin. Recent theory suggests that callus formation is via the rooting pathway (Che et al., 2007; Atta et al., 2009; Sugimoto et al., 2010; Fan et al., 2012; He et al., 2012; Liu et al., 2014) and also involves two cell fate transition steps in Arabidopsis (Liu et al., 2014). In the first step (the priming step) of cell fate transition from regeneration-competent cells to founder cells, WOX11 is specifically induced in founder cells (Liu et al., 2014). In the second step (the initiation step) of cell fate transition from founder cells to callus, WOX11 expression decreases while WOX5 expression increases in the newly formed callus (Liu et al., 2014; **Figure 1E**). LBD16 expression is also observed in the newly formed callus (Fan et al., 2012). Therefore, the newly formed callus seems to be a group of root primordiumlike cells that is under the control of the high auxin level from CIM (see "newly formed callus" in **Figure 1H**).

Ideally, under continuous stimulation with a high auxin level, the status of callus is maintained at the root primordiumlike status. However, in tissue culture, auxin might not be evenly distributed in the callus mass and there is always partial differentiation of callus as some callus cells try to enter the patterning step. Many root meristem genes were observed in diverse domains of the fast dividing and partially differentiated callus mass (Sugimoto et al., 2010; Kareem et al., 2015). WOX5 and LBD16 may not be ubiquitously expressed in the partially differentiated callus. Therefore, the partially differentiated callus could be at any stage from root primordium to root meristem and could be composed of many different types of meristem cells with diverse gene expression patterns (see "partially differentiated callus" in **Figure 1H**). We believe that this is a balanced result from the tug of war between the exogenous auxin stimulation and the endogenous developmental program. On one side, the high level of exogenous auxin attempts to maintain the callus at the root primordium-like status because the root primordium has a high auxin level (Sabatini et al., 1999; Benkova et al., 2003; Okumura et al., 2013; Liu et al., 2014) and consists of precursor cells of the stem cell niche (Hu and Xu, 2016). On the other side, the endogenous developmental program tries to force this group of root primordium-like callus cells into the patterning step to differentiate into the RAM. As a result of the balance of these two forces, the callus mass maintains some of the root primordium features while there is also partial differentiation with some RAM traits.

When callus is moved to RIM or B5 medium, the removal of auxin in the medium results in the loss of the ability to keep the callus at the root primordium-like status. The endogenous developmental program then drives the callus to finish the patterning step during which WOX5 is gradually restricted to the stem cell niche of newly formed RAMs (**Figure 1F**). As there are plenty of root primordium-like cells present during callus formation, we could observe the formation of numerous adventitious root tips from the callus on RIM or B5 medium.

#### COMPARISON OF THE TWO ROOTING TYPES

Here, we summarize the similarities and differences between the two types of adventitious rooting from leaf explants (see models in **Figures 1G,H**). Based on the above hypothesis, we propose that direct DNRR and indirect DNRR have very similar cell fate transition steps. They all experience priming, initiation, patterning, and emergence steps to finally form adventitious roots. Additionally, the molecular markers are also similar, involving WOX11 for founder cells and WOX5/LBD16 for root primordium and newly formed callus. WOX11 is generally involved in adventitious root formation and callus formation in both Arabidopsis and rice (Zhao et al., 2009; Liu et al., 2014; Hu et al., in press).

There are two major differences between direct DNRR and indirect DNRR. First, the auxin source for the two types of rooting is different. In direct DNRR, endogenous auxin is mainly produced in mesophyll cells, leaf margin cells and some other cells in the leaf explant and then transported into regenerationcompetent cells (Liu et al., 2014; Chen et al., 2016b), while in indirect DNRR, exogenous auxin is mainly provided from the medium. Therefore, mesophyll and many other cells in the leaf explant are required for direct DNRR (Chen et al., 2016a,b) but may not be required for indirect DNRR.

Second, the auxin behavior is different in the two types of rooting. In direct DNRR, the cell fate transition is strictly controlled by endogenous auxin and developmental programs. The auxin concentration is focused just in a few cells (the root primordium) in direct DNRR, and therefore the regenerated root number is limited (usually 1–3 roots). The endogenous auxin keeps a high level in the root primordium and its level decreases during patterning. The rooting process usually does not stop between initiation and patterning. However, during indirect DNRR, the priming and initiation steps of cell fate transition are quite dramatic and fast, and numerous regeneration-competent cells are induced to form founder cells that then divide to become newly formed callus following stimulation with high levels of exogenous auxin. Auxin is enriched in many callus cells, and therefore numerous adventitious roots could be observed when auxin is removed in indirect DNRR. In the newly formed callus, the high level of exogenous auxin in CIM prevents callus from patterning, resulting in disruption of the recognition of the tissue anatomy and keeping callus cells without pre-specific fate. Removal of auxin from the medium allows callus into the patterning step.

Overall, we believe that direct DNRR and indirect DNRR share similar cell fate transition steps but have different auxin

FIGURE 1 | Direct and indirect DNRR. (A) A system to study direct DNRR. Leaf explants were cultured on B5 medium without added hormones (Chen et al., 2014). (B) A system to study indirect DNRR. Leaf explants were cultured on CIM to induce callus and were then transferred to RIM or B5 medium to produce roots. (C,D) Indirect DNRR from rice (C) and Arabidopsis (D). Leaf explants were first cultured on CIM for 11 d (C) or 6 d (D) and then transferred to B5 medium for another 6 d (C) or 5 d (D). (E,F) *WOX5pro:GUS* (He et al., 2012) in callus on CIM (E) and in roots after transferred to B5 medium (F) during indirect DNRR. Leaf explants were first cultured on CIM for 4 d before being transferred to B5 medium for another 2 d. Notably, the GUS signal was strong in newly formed callus cells on CIM (E) and was gradually restricted to the stem cell niche in root tips after transferred to B5 medium (F). (G,H) Proposed cell lineage in direct DNRR (G) and indirect DNRR (H). Scale bars, 1 mm in (C,D) and 100µm in (E,F).

sources and behavior. Further studies on genetic and epigenetic regulations of direct and indirect DNRR will improve our understanding of the two ways of adventitious rooting at the molecular level. We cannot exclude the possibility that adventitious roots can also initiate from differentiated cells via cell fate reprogramming. Cell lineage analysis using WOX11, WOX5, LBD16 and other markers and phenotype analysis of mutants are important in studying different types of rooting in the future.

#### AUTHOR CONTRIBUTIONS

JY, WL, JL, PQ, and LX designed the research. JY, WL, and JL conducted the research. LX wrote the manuscript.

#### REFERENCES


#### ACKNOWLEDGMENTS

This work was supported by grants from the National Natural Science Foundation of China (31630007/31422005), National Basic Research Program of China (973 Program, 2014CB943500), the Key Research Program of CAS (QYZDB-SSW-SMC010), the Strategic Priority Research Program "Molecular Mechanism of Plant Growth and Development" of CAS (XDPB0403), and Youth Innovation Promotion Association CAS (2014241).


**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 Yu, Liu, Liu, Qin and Xu. 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.

# Arabidopsis MYB24 Regulates Jasmonate-Mediated Stamen Development

Huang Huang1,2† , Hua Gao<sup>1</sup>† , Bei Liu<sup>3</sup>† , Tiancong Qi<sup>1</sup> , Jianhua Tong<sup>4</sup> , Langtao Xiao<sup>4</sup> , Daoxin Xie<sup>1</sup> \* and Susheng Song<sup>3</sup> \*

<sup>1</sup> School of Life Sciences, Tsinghua University, Beijing, China, <sup>2</sup> College of Biological Science and Engineering, Beijing University of Agriculture, Beijing, China, <sup>3</sup> Beijing Key Laboratory of Plant Gene Resources and Biotechnology for Carbon Reduction and Environmental Improvement, College of Life Sciences, Capital Normal University, Beijing, China, <sup>4</sup> College of Bioscience and Biotechnology, Southern Regional Collaborative Innovation Center for Grain and Oil Crops in China, Hunan Agricultural University, Changsha, China

#### Edited by:

Chi-Kuang Wen, Shanghai Institutes for Biological Sciences (CAS), China

#### Reviewed by:

Yang Do Choi, Seoul National University, South Korea Hsu-Liang Hsieh, National Taiwan University, Taiwan Claus Wasternack, Leibniz-Institut für Pflanzenbiochemie (IPB), Germany

#### \*Correspondence:

Daoxin Xie daoxinlab@tsinghua.edu.cn Susheng Song songsslab@163.com

†These authors have contributed equally to this work.

#### Specialty section:

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

Received: 30 June 2017 Accepted: 21 August 2017 Published: 05 September 2017

#### Citation:

Huang H, Gao H, Liu B, Qi T, Tong J, Xiao L, Xie D and Song S (2017) Arabidopsis MYB24 Regulates Jasmonate-Mediated Stamen Development. Front. Plant Sci. 8:1525. doi: 10.3389/fpls.2017.01525 The phytohormone jasmonates (JAs) regulate various defense responses and diverse developmental processes including stamen development and fertility. Previous studies showed that JA induces CORONATINE INSENSITIVE 1-mediated degradation of JA ZIM-domain (JAZ) proteins, and activates the MYB transcription factors (such as MYB21 and MYB24) to regulate stamen development. In this study, we further uncover the mechanism underlying how MYB24 interacts with JAZs to control JA-regulated stamen development. We show that N-terminus of MYB21/24 interacts with 10 out of 12 JAZ proteins while both N-terminus and C-terminus of MYB24 are involved in dimerization of MYB21 and MYB24. Interestingly, male sterility of the JA-deficient mutant opr3 can be rescued by suitable level of the MYB24 overexpression but not by excessive high level of MYB24. Surprisingly, overexpression of MYB24NT, but not MYB24CT, could cause male sterility. These results provide new insights on MYB factors in JA-regulated stamen development.

Keywords: fertility, interaction, JAZs, MYB24, OPR3, stamen

# INTRODUCTION

Jasmonates (JAs), a class of lipid-derived phytohormones (Browse, 2009; Wasternack and Song, 2017), are crucial players in various aspects of plant developmental processes (Huang et al., 2017), including root growth (Fernandez-Calvo et al., 2011; Lischweski et al., 2015), stamen development (Mandaokar et al., 2006; Mandaokar and Browse, 2009; Song et al., 2011; Reeves et al., 2012; Song et al., 2013b), flowering (Zhai et al., 2015), trichome initiation (Qi et al., 2011), and leaf senescence (Qi et al., 2015b); they also mediate plant abiotic stress tolerance and defenses against herbivores and necrotrophic pathogens (Howe and Jander, 2008; Campos et al., 2014; Kazan, 2015; Goossens et al., 2016).

In response to developmental signals or environmental cue-triggered JA biosynthesis, the JA receptor CORONATINE INSENSITIVE 1 (COI1) (Xie et al., 1998; Yan et al., 2009) perceives bioactive molecules of JA (Fonseca et al., 2009; Yan et al., 2016) to recruit JA ZIM-domain (JAZ) proteins for ubiquitination and subsequent degradation via the 26S-proteasome (Chini et al., 2007; Thines et al., 2007; Yan et al., 2007), thereby de-repressing JAZ-inhibited transcription factors,

**31**

such as MYC2/3/4/5 (Cheng et al., 2011; Fernandez-Calvo et al., 2011; Niu et al., 2011; Song et al., 2014a; Figueroa and Browse, 2015; Qi et al., 2015a; Gimenez-Ibanez et al., 2017; Major et al., 2017), MYB21/24 (Song et al., 2011), IIId bHLH factors (Nakata et al., 2013; Nakata and Ohme-Takagi, 2013; Song et al., 2013a), and TTG1/bHLH/MYB complexes (Qi et al., 2011) to modulate distinct JA responses.

Jasmonate-deficient mutants (e.g., aos and opr3), the JA receptor mutant coi1-1, and JAZ dominant-negative transgenic plants (JAZ113A) are all male sterile with defects in filament elongation, anther dehiscence, and pollen maturation (Xie et al., 1998; Stintzi and Browse, 2000; Park et al., 2002; Thines et al., 2007). The R2R3–MYB transcription factors MYB21 and MYB24 associate with IIIe bHLH factors (MYC2, MYC3, MYC4, and MYC5) to form MYB–MYC complexes, and interact with JAZs to mediate late stamen development (Qi et al., 2015a).

We previously showed that JAZ1/8/11 interact with MYB21/MYB24 in yeast and plants (Song et al., 2011). In this study, we further showed that MYB21 and MYB24 interact with most JAZs via their N-terminal R2R3 domains, and both the N-terminus and C-terminus of MYB24 mediate the dimeric interactions of MYB21 and MYB24. Proper overexpression of MYB24 partially restores male fertility of opr3. Overexpression of N-terminus of MYB24, but not C-terminus, causes male sterility in wild-type. Furthermore, young flower buds from myb21 myb24 myb57 accumulate more jasmonic acid than that of wild-type.

# MATERIALS AND METHODS

#### Plant Materials and Growth Conditions

The Arabidopsis mutants opr3 (Stintzi and Browse, 2000) and myb21 myb24 myb57 (Cheng et al., 2009) were described previously. Arabidopsis thaliana seeds were disinfected, germinated on Murashige and Skoog (MS) medium, stored at 4◦C for 3 days, and transferred to a growth room for another 7 days before being transferred to soil under a 16 h (22–24◦C)/8 h (17–19◦C) light/dark photoperiod. Nicotiana benthamiana seeds were sown in soil and grown under a 16 h (26◦C)/8 h (22◦C) light/dark photoperiod.

#### Yeast Two-Hybrid Assay

The full lengths of coding regions of MYB21, MYB24, 12 JAZs, and truncated domains of MYB24 were individually fused with the activation domain (AD) in pB42AD, or the DNA binding domain (BD) in pLexA. The primer pairs used for vector construction are listed in Supplementary Table 1. Yeast transformation and protein–protein interaction assays were performed according to the Matchmaker LexA Two-Hybrid System (Clontech) as described previously (Song et al., 2011).

# Firefly Luciferase (LUC) Complementation Imaging (LCI) Assay

JAZ5, MYB21, MYB24, MYB24NT, and MYB24CT were individually inserted into pCAMBIA-nLUC or pCAMBIA-cLUC for fusion with the N-terminal half of LUC (nLUC) or C-terminal half of LUC (cLUC). The primers used to construct the LCI vector are listed in Supplementary Table 1. Agrobacterium tumefaciens cells containing the indicated plasmids were co-infiltrated into N. benthamiana leaves and LUC activity was detected as described previously (Song et al., 2011).

# Generation of Transgenic Plants

To obtain transgenic plants overexpressing MYB24, the coding sequence of MYB24 was amplified and cloned using XbaI and SacI into pCAMBIA1301 under the control of the CaMV 35S promoter. The construct was transformed into OPR3/opr3 heterozygous plants using the Agrobacterium-mediated floral dip method. The primers used for vector construction are presented in Supplementary Table 1. pCAMBIA-MYB24NT-nLUC and pCAMBIA-MYB24CT-nLUC were transformed into Arabidopsis Col-0 wild-type to generate MYB24NT- and MYB24CToverexpressing plants.

## Quantitative Real-time PCR

Young flower buds of plants were collected for total RNA extraction and subsequent reverse transcription. Quantitative real-time PCR was performed with RealMasterMix (SYBR Green I; Takara; Bio Inc., Otsu, Japan) using the ABI 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA, United States), and ACTIN8 was used as an internal control. The primers used are listed in Supplementary Table 1.

### Flower Phenotype Analysis and Pollen Germination Assay

For flower phenotype analysis, flowers of each genotype at stage 13 were photographed under a microscope. Pollen germination assay was conducted as described previously (Song et al., 2011). Pollen grains were germinated on pollen germination medium [1 mM MgSO4, 5 mM CaCl2, 5 mM KCl, 10% (w/v) sucrose, 0.01% boric acid, and 1.5% agar at pH 7.5], incubated for 10 h at 22◦C in the dark, and observed under a microscope.

#### Jasmonic Acid Measurement

Five hundred milligrams of young flower buds before floral stage 13 from 5-week-old wild-type and the myb21 myb24 myb57 mutant was harvested, and jasmonic acid was extracted and quantified as described previously using a liquid chromatography–tandem mass spectrometry system (Cheng et al., 2009).

#### Accession Numbers

The Arabidopsis Genome Initiative numbers for genes mentioned in this article are as follows: JAZ1 (At1g19180), JAZ2 (At1g74950), JAZ3 (At3g17860), JAZ4 (At1g48500), JAZ5 (At1g17380), JAZ6 (At1g72450), JAZ7 (At2g34600), JAZ8 (At1g30135), JAZ9 (At1g70700), JAZ10 (At5g13220), JAZ11 (At3g43440), JAZ12 (At5g20900), MYB21 (At3g27810), MYB24

(At5g40350), MYB57 (At3g01530), OPR3 (AT2G06050), and ACTIN8 (At1g49240).

#### RESULTS

#### MYB21 and MYB24 Interact with Multiple JAZs via Their N-terminus

MYB21 and MYB24 were divided into MYB21NT and MYB24NT harboring the R2R3 DNA BD, and MYB21CT and MYB24CT containing the C-terminal NYWG/<sup>S</sup> <sup>M</sup>/VDD<sup>I</sup> /LW<sup>S</sup> /P transcriptional activation motif (**Figures 1A,B**). Binding domain-fused MYB21CT and MYB24CT exhibited autoactivation, while MYB21NT and MYB24NT did not. Binding domain-fused MYB21NT and MYB24NT were used to detect interactions with AD-fused JAZ proteins. The results (**Figure 1C**) showed that MYB21NT and MYB24NT interact with JAZ1, JAZ2, JAZ3, JAZ4, JAZ5, JAZ6, JAZ8, JAZ10, JAZ11, and JAZ12, but not with JAZ7 or JAZ9, demonstrating that MYB21 and MYB24 interact with multiple JAZs through their N-terminus.

We next performed a firefly LCI assay to test the interactions of JAZ5 with MYB21/24 in plant. JAZ5 was fused with nLUC, while MYB21 and MYB24 were fused with cLUC. The results showed that co-infiltration of JAZ5-nLUC/cLUC-MYB21 or JAZ5-nLUC/cLUC-MYB24 in N. benthamiana leaves resulted in strong LUC signals while the negative control did not (**Figure 1D**), suggesting that MYB21 and MYB24 interact with JAZ5 in plant.

### MYB24 Overexpression Restores Stamen Development in opr3

We next examined whether MYB24 overexpression could escape from inhibition by multiple JAZs to rescue stamen development and fertility in the JA-deficient mutant opr3. As shown in **Figure 2A**, opr3 exhibited unelongated filaments, indehiscent anthers, and inviable pollen grains at floral stage 13. MYB24 expression decreased in young opr3 flower buds (**Figure 2D**).

respectively. LUC signals were detected 60 h after co-infiltration of the indicated constructs into N. benthamiana leaves.

pollen germination (A) at floral stage 13, main inflorescence (B), and seed setting (C) in Col-0, opr3, and opr3 MYB24OE37. MYB24 overexpression partially restored the fertility of opr3 (A–C). (D) Quantitative real-time PCR analysis of MYB24 expression in young flower buds of Col-0, opr3, and opr3 MYB24OE37. ACTIN8 was used as the internal control. Data are means (±SE) of three biological replicates. Asterisks represent Student's t-test significance between pairs indicated with brackets (∗∗p < 0.01).

MYB24 overexpression in opr3 (opr3 MYB24OE37, with fivefold to sixfold of wild-type level) restored filament elongation, anther dehiscence, and pollen viability (**Figure 2A**). Further, the opr3 MYB24OE37 plants were able to set seeds (**Figures 2B,C**). These results suggest that overexpression of MYB24 could partially restore stamen development and fertility in opr3.

(A) Main inflorescences of Col-0, opr3, MYB24OE2, and opr3 MYB24OE2. (B) Quantitative real-time PCR analysis of MYB24 expression in young flower buds. ACTIN8 was used as the internal control. Data are means (±SE) of three biological replicates. Asterisks represent Student's t-test significance between pairs indicated with brackets (∗∗p < 0.01).

## Excess Expression of MYB24 Cannot Rescue the Fertility of the opr3 Mutant

Previous studies showed that strong MYB24 overexpression inhibits stamen development (Yang et al., 2007; Song et al., 2011). We therefore tested whether excess expression of MYB24 could restore fertility in opr3. As shown in **Figure 3**, transgenic plants that expressed excessive amounts of MYB24 (∼75-fold of the wild-type level) were male sterile. Further, transgenic opr3 plants that expressed excessive amounts of MYB24 (∼70-fold of the wild-type level) were still male sterile, suggesting that excess expression of MYB24 could not restore stamen development and fertility in opr3 plants.

#### The N-terminus and C-terminus of MYB24 Are Involved in Dimeric Interactions

We next investigated the dimeric interactions of MYB21 and MYB24 in detail. As shown in **Figure 4A**, Y2H analysis showed that BD-fused MYB24NT interacted strongly with AD-fused MYB24NT, and weakly with MYB24CT. Further, LCI assay exhibited that the co-expression of MYB24NT-nLUC/cLUC-MYB21, MYB24CT-nLUC/cLUC-MYB21, MYB24NT-nLUC/ cLUC-MYB24, and MYB24CT-nLUC/cLUC-MYB24 resulted in strong LUC activity, while the negative controls did not (**Figures 4B,C**). These results demonstrate that both N-terminus

and C-terminus of MYB24 are involved in dimeric interactions of MYB21 and MYB24.

# MYB24NT Overexpression Causes Male Sterility

We next examined whether the overexpression of MYB24NT and MYB24CT could dominantly repress stamen development and male fertility. MYB24NT overexpression (∼20–40-fold of the wild-type level) inhibited stamen development, including filament elongation, anther dehiscence, and male fertility (**Figures 5A,B** and Supplementary Figure 1A), while all MYB24CT transgenic lines showed no obvious influence on male fertility (**Figure 5C** and Supplementary Figure 1B). These data indicate that overexpression of N-terminus of MYB24 could dominantly repress stamen development and fertility.

# JA Concentration Increased in Young Flower Buds of myb21 myb24 myb57

MYB21, MYB24, and MYB57 are all responsive to JA (**Figure 1**; Mandaokar et al., 2006; Cheng et al., 2009; Mandaokar and Browse, 2009). We thus tested whether MYB21, MYB24, and

MYB57 in turn regulate JA biosynthesis in young flower buds (before floral stage 13). The jasmonic acid content of young myb21 myb24 myb57 mutant flower buds was approximately twofold of the wild-type level (**Figure 6**), suggesting that MYB21, MYB24, and MYB57 negatively regulate JA biosynthesis as part of a negative feedback loop.

#### DISCUSSION

Jasmonate ZIM-domain proteins serve as repressors that target specific transcription factors and control their downstream pathways to modulate distinct JA responses for coordinated regulation of development, growth, and defense (Song et al., 2014b; Chini et al., 2016; Gimenez-Ibanez et al., 2017; Major et al., 2017). Our previous study showed that JAZ1, JAZ8, and JAZ11 interact with MYB21 and MYB24 (Song et al., 2011). In this study, we further showed that MYB21 and MYB24 act through N-terminus to interact with 10 out of 12 JAZs (**Figure 1**), suggesting that most JAZs may act through interfering the DNA binding function of MYB21/24 to attenuate their function in regulating stamen development. Excess expression of MYB24 is unable to restore the stamen development of opr3, whereas suitable overexpression of MYB24 could recover stamen development and male fertility (**Figures 2**, **3**). Exploring the downstream pathways of MYB24 would help to understand that the suitable MYB24 expression level is essential for proper stamen development.

Both the N-terminal DNA BD and C-terminal transcriptional activation motif mediate dimerization of MYB21 and MYB24 (**Figure 4**). Determination of crystal structure of MYB21/24 will help to further elucidate the interaction. Interestingly, overexpression of N-terminus of MYB24, but not the C-terminus of MYB24, attenuates stamen development and male fertility (**Figure 5**). MYB24NT-OE2 with the highest expression level of MYB24NT confers the most severe male sterility (**Figures 5A,B** and Supplementary Figure 1A), suggesting that the expression level of MYB24NT is correlated with male sterility. It remains to study whether MYB24NT affects the dimeric interaction of MYBs to affect stamen development.

We also found that young flower buds of myb21 myb24 myb57 accumulated more jasmonic acid (**Figure 6**), suggesting that MYBs negatively regulate JA biosynthesis to attenuate JA-induced expression of MYBs and to elaborately regulate stamen development, and that the restored fertility in opr3 (Stintzi and Browse, 2000; Chehab et al., 2011) by suitable MYB24 overexpression (**Figure 2**) is not due to recovery of JA biosynthesis. It would be useful for understanding the MYB21/24 module in stamen development if the links between MYB21/24 and JA biosynthetic genes are elucidated.

### AUTHOR CONTRIBUTIONS

DX and SS designed the study; HH, HG, BL, TQ, JT, LX, and SS performed the experiments; HH, HG, BL, TQ, JT, LX, and SS analyzed the data; and HH, HG, and SS wrote the manuscript.

#### FUNDING

This work was financially supported by the National Natural Science Foundation of China (31670315 and 31570372), the Ministry of Science and Technology of China (2016YFA0500501), and Beijing Nova Program (Z171100001117037).

#### ACKNOWLEDGMENTS

We thank Dr. John Browse for providing the opr3 mutant and Dr. Jianmin Zhou for providing the vectors used in the LCI assay.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2017.01525/ full#supplementary-material

#### REFERENCES

fpls-08-01525 September 1, 2017 Time: 16:33 # 7


Zhai, Q., Zhang, X., Wu, F., Feng, H., Deng, L., Xu, L., et al. (2015). Transcriptional mechanism of jasmonate receptor COI1-mediated delay of flowering time in Arabidopsis. Plant Cell 27, 2814–2828. doi: 10.1105/tpc.15.00619

**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 Huang, Gao, Liu, Qi, Tong, Xiao, Xie and Song. This is an openaccess 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.

# Brassinosteroids Regulate OFP1, a DLT Interacting Protein, to Modulate Plant Architecture and Grain Morphology in Rice

#### Yunhua Xiao1,2† , Dapu Liu1,2† , Guoxia Zhang1,2, Hongning Tong<sup>3</sup> \* and Chengcai Chu1,2 \*

<sup>1</sup> State Key Laboratory of Plant Genomics and Center for Plant Gene Research, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing, China, <sup>2</sup> University of Chinese Academy of Sciences, Beijing, China, <sup>3</sup> National Key Facility for Crop Gene Resources and Genetic Improvement, Institute of Crop Sciences, Chinese Academy of Agricultural Sciences, Beijing, China

#### Edited by:

Chi-Kuang Wen, Shanghai Institutes for Biological Sciences (CAS), China

#### Reviewed by:

Lei Wang, Institute of Botany (CAS), China Mingyi Jiang, Nanjing Agricultural University, China

#### \*Correspondence:

Hongning Tong tonghongning@caas.cn Chengcai Chu ccchu@genetics.ac.cn

†These authors have contributed equally to this work.

#### Specialty section:

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

Received: 30 July 2017 Accepted: 15 September 2017 Published: 27 September 2017

#### Citation:

Xiao Y, Liu D, Zhang G, Tong H and Chu C (2017) Brassinosteroids Regulate OFP1, a DLT Interacting Protein, to Modulate Plant Architecture and Grain Morphology in Rice. Front. Plant Sci. 8:1698. doi: 10.3389/fpls.2017.01698 Brassinosteroids (BRs) regulate important agronomic traits in rice, including plant height, leaf angle, and grain size. However, the underlying mechanisms remain not fully understood. We previously showed that GSK2, the central negative regulator of BR signaling, targets DLT, the GRAS family protein, to regulate BR responses. Here, we identified Ovate Family Protein 1 (OFP1) as a DLT interacting protein. OFP1 was ubiquitously expressed and the protein was localized in both cytoplasm and nucleus. Overexpression of OFP1 led to enlarged leaf angles, reduced plant height, and altered grain shape, largely resembled DLT overexpression plants. Genetic analysis showed that the regulation of plant architecture by OFP1 depends on DLT function. In addition, we found OFP1 was greatly induced by BR treatment, and OsBZR1, the critical transcription factor of BR signaling, was physically associated with the OFP1 promoter. Moreover, we showed that gibberellin synthesis was greatly repressed in OFP1 overexpression plants, suggesting OFP1 participates in the inhibition of plant growth by high BR or elevated BR signaling. Furthermore, we revealed that OFP1 directly interacts with GSK2 kinase, and inhibition of the kinase activity significantly promotes OFP1 protein accumulation in plant. Taken together, we identified OFP1 as an additional regulator of BR responses and revealed how BRs promote OFP1 at both transcription and protein levels to modulate plant architecture and grain morphology in rice.

#### Keywords: DLT, grain shape, GSK2, leaf angle, OFP, plant height, rice

# INTRODUCTION

Brassinosteroids (BRs) are a class of phytohormones playing important roles in regulating various aspects of plant growth and development. In the past decades, rapid progresses have been made regarding the hormone signaling in the model plant Arabidopsis. A nearly complete primary signaling pathway has been established from the receptor to the transcriptional factors involving multiple components including BRI1 receptor kinase, BAK1 co-receptor kinase, BKI1 receptor inhibitor kinase, BSK1 and CDG1 kinases, BSU1 phosphatase, BIN2 kinase, PP2A phosphatase, BZR1/BES1 transcriptional factors, and so on (Gruszka, 2013; Vriet et al., 2013). Among them, BIN2 is the central negative regulator which directly inhibits BZR1/BES1 activity (He et al., 2002;

Yin et al., 2002). Recently, KIB1 was identified as a negative regulator of BR signaling responsible for BIN2 degradation, and multiple components including SINAT and DSK were identified as positive regulator of BR signaling participated in BZR1/BES1 degradation (Nolan et al., 2017; Yang and Wang, 2017; Yang et al., 2017; Zhu et al., 2017). In addition, a number of additional BIN2 targets as well as BES1 interacting proteins were reported, which are involved in diverse biological processes, further complemented the signaling network and greatly advanced our understanding of BR functional mechanisms (Hao et al., 2013; Youn and Kim, 2015).

In crop plant rice, BRs play critical roles in regulating plant height, leaf angle, and grain size. Because these characteristics are important agronomic traits that mostly considered in breeding activity, BRs are thought to have great potentials in agricultural improvement (Divi and Krishna, 2009). Several studies have tried to utilize BR-related plants or genes for crop engineering and obtained promising results. For example, a weak BRdeficient plant osdwarf4, or an OsBRI1-cossuppression plant could have greatly enhanced population yield when planting at high density due to their favorable erect leaves (Morinaka et al., 2006; Sakamoto et al., 2006). However, since rice has a distinct plant architecture from Arabidopsis, it remains largely unclear how BRs specifically regulate these traits in detail in rice. Consistently, although some counterparts of the BR primary signaling components including OsBRI1, GSK2 (OsBIN2), and OsBZR1 were identified and their function were verified or partially verified, several rice specific BR components were also reported, including DLT, LIC, SMOS1/RLA1, and OFP8 (Wang et al., 2008; Tong et al., 2009; Tong and Chu, 2012; Zhang et al., 2012; Yang et al., 2016; Hirano et al., 2017; Qiao et al., 2017).

OFPs, for short of Ovate Family Proteins, are plant specific transcription factors containing a common domain named OVATE or DUF623 (domain of unknown function) (Wang et al., 2007, 2011). In rice, it should be noted that there were two genome-wide characterization studies of OFPs, which adopted different numbering systems of the family members (Liu et al., 2014; Yu et al., 2015). Founding member of the family, OVATE, was cloned in tomato, which plays crucial roles in regulating fruit shape (Liu et al., 2002). A naturally existed nonsense mutation in the gene in tomato varieties leads to the shift of roundshape fruit to pear-like shape. Surprisingly, consequent studies in Arabidopsis revealed that loss-of-function plants of many single or multiple OFP members are phenotype-silent (Wang et al., 2007, 2011; Schmitz et al., 2015). In contrast, overexpression of some OFP members produced evident morphology changes, suggesting the high redundancy among the family members. Overexpression of AtOFP1 in Arabidopsis produced dwarf statue due to the repression of gibberellin (GA) synthesis (Wang et al., 2007). AtOFP1 directly targets GA20ox-1, a GA biosynthetic gene, to inhibit its expression. Similarly, overexpression of OsOFP2 in rice also inhibited GA synthesis and decreased plant height likely by regulating GA20ox-7 (Schmitz et al., 2015). In addition, OFPs tend to generally interact with BELL- and KNOX-like proteins to regulate plant growth and development (Hackbusch et al., 2005; Schmitz et al., 2015). Recently, OFP8, corresponding to OFP30 in Liu et al. (2014), was shown to be involved in BR responses (Yang et al., 2016). Enhanced expression of the gene caused by T-DNA insertion containing 35S enhancer produced obviously enlarged leaf angles and increased BR responses. The protein localization was modulated by GSK2 phosphorylation, similar as the regulation of BZR1/BES1 by BIN2 (Yang et al., 2016).

More and more studies suggested that functions of BRs could be variable depending on species, tissues, hormone levels, developmental stages, and cell types (Singh and Savaldi-Goldstein, 2015; Tong and Chu, 2016; Unterholzner et al., 2016). For example, BRs were reported to repress stomata development in cotyledons but promote the process in hypocotyls in Arabidopsis (Gudesblat et al., 2012; Kim et al., 2012). As growth-promoting hormones, BRs promote GA accumulation by inducing the expression of GA biosynthetic genes to stimulate cell elongation under physiological conditions (Tong et al., 2014; Unterholzner et al., 2015), however, sufficient high BR level or signaling will inhibit cell elongation and plant growth partially by repressing GA synthesis in rice (Tong et al., 2014). The mechanisms underlying this kind of functional specificities of BRs remain largely unclear.

In this study, we identified OFP1 as a DLT interacting protein by yeast two-hybrid screening, and found that OFP1 can also interact with GSK2 kinase. We showed that OFP1 was regulated by BRs at both transcription and protein levels, and the regulation was closely linked to the core BR signaling components involving GSK2, OsBZR1, and DLT. Our results revealed the positive roles of OFP1 in regulating BR responses to modulate plant architecture and grain morphology.

#### MATERIALS AND METHODS

#### Plant Materials, Growth Conditions, and Chemical Treatment

Japonica cultivars Zhonghua 11 (ZH11) or Dongjin (DJ) was used as the wild type (WT) for the transgenic analyses. Plants were grown on soil in field or in greenhouse or on 0.5x Murashige and Skoog (1/2MS) medium in a conditioned growth chamber at 30◦C for 10 h (day) and 24◦C for 14 h (night). For BR induction experiments, young seedlings were transferred to 1/2MS medium supplemented with various concentrations of brassinolide (BL, one of the active BRs) for 48 h. For lamina inclination assay, ethanol (1 µL) containing various amount of BL was spotted onto the top of lamina after 2-day germination and 3-day growth at 30◦C (Tong and Chu, 2017). Images were taken after 3-day incubation, and the angles of lamina joint bending were measured using IMAGEJ software<sup>1</sup> . For bikinin treatment, 1-week-old seedlings were grown in 1/2MS supplemented with or without 30 µM chemical for 3 days before sampling.

# Vector Constructions and Transgenic Analysis

The DNA sequence containing the promoter and coding sequence of OFP1 was cloned into pCAMBIA2300 vector for

<sup>1</sup>http://rsb.info.nih.gov/ij

OFP1 overexpression (OFP1-OE or 1o). A pCAMBIA2300- 35S-GFP vector was used for OFP1-GFP construction. A pCAMBIA1300-35S-Flag vector was used for OFP1-Flag overexpression (OFP1-Flag-OE or 1Fo). OFP1 promoter (2-kb) was introduced into pCAMBIA2391Z vector for OFP1p:GUS construction. CRISPR/Cas9 knock-out vectors were constructed following previous reports (Ma and Liu, 2016). Primer sequences used for vector constructions and additional plasmid information are listed in the Supplementary Table 1. Sequences were introduced into vectors by either in-fusion cloning strategy (Clontech) or traditional cut-ligation cloning method. These constructs were used to transform rice plant or to infiltrate tobacco leaf epidermis cells by Agrobacterium-mediated method (Sparkes et al., 2006).

#### Protein: Protein Interaction Analyses

Yeast two-hybrid tests were performed following standard procedures descried by manufacturer (Clontech). BiFC (bimolecular fluorescence complementation) analyses were performed following the method described previously (Waadt et al., 2008). Rice protoplast preparation and plasmid transfection were performed according to the previous method (Bart et al., 2006). Fluorescence was observed on a confocal fluorescence microscope (Leica TCS SP6). Split-luciferase complementation assays were performed in tobacco leaves as described (Chen et al., 2008). Chemiluminiscence was photographed using a imaging system equipped with a cold CCD (NightOWL II LB983 with indigo software). See Supplementary Table 1 for the information of the primer sequences and plasmids used for the vector constructions prepared for the above analyses.

### Gene Expression and Promoter Activity Analysis

RNA was isolated using Trizol reagent (Invitrogen) according to manufacturer's instruction. The first-strand cDNA was synthesized using a Revertase Transcription kit (Toyoba). Quantitative RT-PCR (qRT-PCR) was performed on a realtime PCR detection system following manufacturer's instructions (Bio-Rad CFX96). Rice Ubiquitin2 gene (UBQ) was used as internal control for all analyses. Primer sequences are listed in the Supplementary Table 2. GUS staining was performed according to previous description (Tong et al., 2009).

#### Immunoblotting and ChIP-qPCR Analysis

Commercialized anti-Flag (Sigma) and anti-ACTIN (Abmart) antibodies were used for immunoblotting analyses. Commercialized anti-OsBZR1 antibody (BPI) was used for ChIP (chromatin immunoprecipitation) analysis. ChIP-qPCR (ChIP followed by quantitative PCR analysis) was performed according to previous description (Tong et al., 2014). Primer sequences are listed in the Supplementary Table 2.

#### GA Measurement

About 4 g of shoots was harvested from the rice seedlings for GA measurements. Quantification of endogenous GAs was performed as described (Chen et al., 2012).

# RESULTS

## OFP1 Interacts with DLT in Yeast

We previously showed that GSK2 kinase, the central negative regulator of BR signaling, targets DLT, the GRAS family transcription factor, to regulate downstream BR responses (Tong et al., 2012). To further explore the functional mechanism of the proteins, we performed extensive yeast two-hybrid screenings of the interacting proteins using DLT as bait. By this approach, we identified OFP1 (Ovate Family Protein 1) as one of DLTinteracting proteins from the yeast library prepared using mixed rice tissues. Full length OFP1 prey vector (OFP1-AD) was further constructed and the interaction was confirmed in yeast (**Figure 1A**). In addition, we found that when OFP1 was used as both bait and prey, the interaction also occurred (**Figure 1A**), suggesting that OFP1 may form homo-dimer or oligomer to function.

#### Subcellular Localization of OFP1 and Verification of the Interaction in Plant Cells

OFP1 contains both the putative NLS (nuclear localization signal) and NES (nuclear export signal) (**Figure 1B**). To analyze the protein subcellular localization, we fused the protein with GFP and introduced the corresponding vector into rice protoplasts. Consistently, microscopic observation revealed that OFP1-GFP fusion proteins were localized in both cytoplasm and nucleus, and can be co-localized with DLT-RFP distributed in nucleus (**Figure 1C**), supporting the possibility that they may interact in plant cells.

We then further verified their interactions by BiFC analysis in rice protoplasts. Different combinations of the proteins fused with either C-terminal or N-terminal of the split Venus YFP (VC or VN) were introduced into the rice protoplast cells, and the reconstituted fluorescence signals were detected using a laser confocal microscope. While as negative controls no obvious signal was detected between any of the fusion proteins and the corresponding empty split YFP partial protein, we observed the strong interaction between VC-OFP1 and VN-DLT in nucleus (**Figure 1D**). Interestingly, the interacting signals of both protein pairs were mostly spotted in punctate-structures, reminiscent of nuclear bodies which could be formed due to the assembling and spatial-specific distribution of protein complex (**Figure 1D**). This distinct pattern also indicated that the fluorescence was not false-positive signal and the interaction was indeed occurred. In addition, we also detected the clear interaction between OFP1 itself (**Figure 1E**). Notably, in most observed cells, the interacting signals were diffusely distributed in part of cells, which appeared to be mainly in cytoplasm, with unclear reason (**Figure 1E**).

#### Molecular Characterization of OFP1

According to rice genome annotation project<sup>2</sup> , OFP1 (LOC\_Os01g12690) is intronless and contains 1185-bp coding

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

sequence, 44-bp 5<sup>0</sup> -UTR and 244-bp 3<sup>0</sup> -UTR (**Figure 2A**). We analyzed the gene expression pattern by qRT-PCR, and found OFP1 was ubiquitously expressed in various rice tissues, preferentially higher in young panicles (**Figure 2B**). We also evaluated the gene promoter activity using GUS reporter and successfully detected the signals in various tissues of OFP1p:GUS transgenic plants, but mainly at young stages of the tissues (**Figure 2C**).

### Knock-out ofp1 Mutants and ofp1 ofp2 Double Mutants Showed No Obvious Morphological Phenotype

We produced the loss-of-function mutants of OFP1 by CRISPR/Cas9 gene-editing technology. A segment located at the 5<sup>0</sup> -end of the gene coding sequence was targeted for the editing, and a number of independent homozygous mutant lines were obtained. Two distinct alleles, ofp1-1 and ofp1-2, were selected for the further analysis. These alleles contained 35- and 65-bp deletions around the initial codon respectively, leading to the frameshift afterward of the sequence, suggesting that both are knockout mutants (**Figure 2A**). However, we failed to observe the morphological difference between ofp1 mutants and WT at both seedling stage and mature stage (**Figure 2D**). Both the mutants also exhibited indistinguishable BR sensitivity compared with the WT in lamina bending assay (**Figure 2E**). We further generated ofp1 ofp2 double mutant by similar strategy, yet the homozygous mutant plants still showed no obviously phenotypes under normal growth conditions as well as unaltered BR sensitivity in lamina bending assay (**Figures 2F,G**).

# OFP1 Transgenic Plants Exhibit Significantly Altered Plant Architecture and Grain Morphology

There are 33 OFPs in rice, and different OFP members could function with high redundancy, as strongly suggested in Arabidopsis (Wang et al., 2011; Liu et al., 2014). In this case, transgenic analysis was generally used to analyze the protein function. To this end, the whole OFP1 gene containing 2-kb upstream promoter and coding region was introduced into either Dongjing (DJ) or Zhonghua 11 (ZH11), two different Japonica WTs. Strikingly, although with different extents due to the various gene expression levels, most of the transgenic plants (OFP1-OE, designated as 1o for short in Figures) showed obvious phenotypes in both WT backgrounds (**Figure 3**). Unless specified, we majorly used transgenic lines in DJ background for our analysis, since we obtained the stable homozygous lines at the first place.

OFP1-OE had narrow leaves, enlarged leaf angles, decreased plant height, reduced tiller number, and long and narrow seeds (**Figures 3A–F,H**). The grain morphology was remarkably changed due to the greatly increase of the length/width ratio (**Figures 3C,E,F**). The severe lines are sterile, and have roll leaves and significantly decreased tiller numbers (**Figure 3H**). Most of these phenotypes, including the typical BR-enhanced phenotypes, largely resembled DLT-overexpression plants (Do), featured with the significantly narrowed leaves and seeds which

and ofp1 ofp2 knockout mutants. (G) Lamina bending analysis of the plants in (F) in response to various amount of BL. n = 10, bar = SD.

could be, at least partially, BR-independent (Tong et al., 2012). Severities of the phenotypes were obviously consistent with the gene expression levels in the transgenic plants (**Figures 3A–I**), suggesting the involvement of OFP1 in controlling these phenotypes. However, we surprisingly found that the gene was remarkably over-expressed in the positive transgenic plants (**Figures 3G,I**). While the reason is unclear, one possibility is that there exists additional suppressor elements outside the selected promoter region.

## OFP1 Enhanced BR Responses and Required DLT to Regulate Plant Architecture

We further tested the BR sensitivity of the plants by lamina inclination assay. OFP1-OE showed increased leaf angles than WT either with or without BR (**Figures 4A,B**). To exclude the possibility that the enhanced leaf angles were caused by increased BR levels, we analyzed the expression of BR synthetic genes including D2, D11, and DWARF and found that all of them have markedly decreased expression in OFP1-OE (**Figure 4C**), similar to that in Do plants (Tong et al., 2012). This result also implied that the BR synthesis was inhibited by enhanced BR responses in a feedback manner in OFP1-OE. Combined with the phenotypes, these results demonstrated that OFP1-OE plants have activated BR responses.

To further analyze the OFP1 functions, we introduced OFP1 into d2, a BR synthesis mutant (Hong et al., 2003), by transgenic method, and found that d2 OFP1-OE showed obviously enlarged leaf angles and decreased grain width compared with d2 (**Figures 4D–F**). We also introduced OFP1 into dlt mutant, and found dlt OFP1-OE showed similar morphology as dlt,

differences between groups (Fisher's LSD, p < 0.05). (H) Phenotypes of three representing OFP1 overexpression plants in ZH11 background, designated as 1o-3, 1o-4, and 1o-5 for short. Enlarged picture showed the development of sterile seeds and roll leaves in the severe line. (I) Relative OFP1 expression in the plants tested by qRT-PCR. UBQ gene was used as internal reference. n = 3, bar = SD. Different letters above the columns indicate statistically significant differences between groups (Fisher's LSD, p < 0.05).

suggesting that OFP1 required DLT to regulate plant architecture (**Figure 4G**). It should be noted that OFP1-OE can obviously suppress the grain width of dlt mutant (**Figures 4G,H**), demonstrating that the protein is functional in the transgenic plant. Considering BR has relatively minor roles in regulating grain width (Tanabe et al., 2005), this result indicated that OFP1 and DLT have differential functional relationship in different tissues or biological processes. Taken together, these results strongly suggested that OFP1 positively regulates BR responses.

above the columns indicate statistically significant differences between groups (Fisher's LSD, p < 0.05). OFP1 overexpression in d2 background (D) and statistical data of the leaf angle (E) and grain width of the plants (F). n = 5 in (E), n = 20 in (F), bar = SD, ∗∗p < 0.01 in Student's t-test. OFP1 overexpression in dlt background (G) and statistical data of the grain width of the plants (H). n = 20, bar = SD. Different letters above the columns indicate statistically significant differences between groups (Fisher's LSD, p < 0.05).

#### BR Promotes OFP1 Expression and OsBZR1 Binds to OFP1 Promoter

We have performed RNA-seq analysis to compare the differentially expressed genes between high BL and low BL treated rice plants. In our available data, we found OFP1 was differentially over-expressed in 10−<sup>6</sup> M BL treated rice seedlings, but not in 10−<sup>9</sup> M BL treated materials. To confirm this result, we treated WT plants using various concentrations of BL, and then analyzed the expression of OFP1 by qRT-PCR. Consistently, the results showed that OFP1 was markedly induced by application of BR above 10−<sup>7</sup> M, but not induced under lower BR concentrations (**Figure 5A**). In addition, our available ChIP-seq data using OsBZR1 antibody enriched OsBZR1 protein on the promoter region of OFP1 with 3.28 fold increase in WT rice seedlings. Accordingly, we found OFP1 promoter contains multiple cis-elements that have been suggested to be bound by BZR1 (**Figure 5B**) (Sun et al., 2010; Yu et al., 2011). The association between OsBZR1 and the OFP1 promoter was further confirmed by ChIP-qPCR (**Figure 5C**). Compared with the samples with no antibody supplemented, OsBZR1 antibody can significantly pull down the DNA segments containing the putative BZR1 binding elements (**Figure 5C**).

# OFP1 Is Involved in BR Inhibition of GA Synthesis and Plant Growth

In Arabidopsis, it was shown that AtOFP1 directly suppresses GA synthetic gene to repress plant growth (Wang et al., 2007). Our previous study also suggested that high BR levels or enhanced BR signaling will repress GA level to inhibit plant growth in rice (Tong et al., 2014). The specific induction of OFP1 by high BR (**Figure 5A**) and the strongly repressed growth of OFP1-OE plants (**Figures 3A,B**) prompted us to test whether OFP1 was involved in this process. Indeed, we found GA3ox-2 and GA20ox-2, two GA synthetic genes, had decreased expression, whereas GA2ox-3, the GA inactivation gene, had increased expression in OFP1-OE (**Figure 6A**). Accordingly, GA1, the major bioactive GA form in young seedlings, was significantly decreased in the plants (**Figure 6B**), consistent with their decreased seedling height (**Figure 3B**). Actually, all the GA forms quantified were decreased in OFP1-OE plants (**Figure 6B**), reminiscent of the hormonal profiles obtained

from high-BL-treated plants as shown previously (Tong et al., 2014). These results indicated that OFP1 is involved in the BLinduced GA repression and growth inhibition, depending on BR concentrations or plant tissues.

### OFP1 Protein Stability Was Regulated by BR and Development Stages

To test whether OFP1 was regulated at protein level, we fused the coding sequence with Flag tag and introduced into the WT under 35S promoter (OFP1-Flag-OE, designated as 1Fo for short in Figures). We obtained a collection of the overexpression plants which showed similar phenotypes as above described OFP1-OE plants, including enlarged leaf angle, reduced plant height, and altered grain shape (**Figure 7A**). Immunoblotting analysis using Flag antibody can detect the protein expression in the corresponding plants, demonstrating the fusion proteins are functional (**Figure 7B**). We treated the plants with different concentrations of BL, and found the protein was gradually accumulated (**Figure 7C**). We also tested the OFP1 protein in different leaves representing the different developmental stages, and revealed that the protein was highly accumulated in young leaves, but gradually decreased in old leaves (**Figure 7D**). These results suggested that OFP1 was dynamically regulated by hormone levels and development stages.

## GSK2 Interacts with OFP1 and Regulates OFP1 Stability

GSK3-like kinases, as the central negative regulators of BR signaling, can interact with many transcriptional factors to modulate their stability and activity (Youn and Kim, 2015). The promotion of OFP1 stability by BR prompted us to test whether GSK2 can directly interact with OFP1 to regulate this process. Indeed, yeast two-hybrid analysis revealed the obviously interaction between GSK2 bait and OFP1 prey (**Figure 8A**). Surprisingly, we can rarely or only weakly observe the interaction between GSK2 and OFP1 by BiFC analysis in rice protoplasts. One possibility is that the interaction is unstable in rice, considering GSK2 might promote OFP1 degradation. Alternatively, we adopted split-luciferase complementation assay to analyze their interaction in tobacco leaves (**Figure 8B**). When the GSK2 fused with N-terminal of luciferase protein (GSK2- NLuc) was co-expressed with the OFP1 fused with C-terminal of luciferase protein (CLuc-OFP1), the interaction signals were evidently detected. No signal was detected when the GSK2-NLuc was co-expressed with the empty CLuc protein (**Figure 8B**). Thus, GSK2 indeed can interact with OFP1 in plant.

To explore whether GSK2 regulates OFP1 stability, we treated OFP1-Flag-OE plants using bikinin, a specific chemical inhibitor of GSK3-like kinase activity, and analyzed the OFP1 protein level. Compared with the plants without treatment, we found OFP1 was significantly accumulated (**Figure 8C**). Taken together, these results suggested that BR enhances OFP1 protein stability by inhibiting GSK2.

# DISCUSSION

Our study identified OFP1 as an additional positive player of BR responses by interacting with GSK2, OsBZR1, and DLT. We proposed that (1) at transcription level, BR induces OFP1 expression through OsBZR1, (2) at protein level, BR enhances OFP1 stability by inactivate GSK2, and (3) OFP1 activity partially relies on the physical interaction with DLT. In addition, our results suggested the novel roles of OsBZR1-induced OFP1 that required for responding to elevated BR.

Functional characterization of OFPs progressed slowly, largely due to the high redundancy among the family members (Wang et al., 2011). To our knowledge, there is no phenotype-evident knockout plants of any OFP members reported in both rice and Arabidopsis. In Arabidopsis, T-DNA insertion ofp mutants

including ofp1, ofp4, ofp8, ofp10, ofp15, and ofp16, all showed no visible phenotype under normal growth condition. In addition, the double mutants including ofp1 ofp4 and ofp15 ofp16 also showed no obvious phenotype (Wang et al., 2011). This seemed to be also the case in rice. The ofp1 knockout mutant, as well as ofp1 ofp2 double mutant, didn't produce the expected phenotypes (**Figure 2**). However, knock-down of OFP8/OFP30 by RNAi strategy indeed led to BR-related phenotypes (Yang et al., 2016). Although it's unclear whether other members were non-specifically targeted, this result at least suggested the native function of OFPs in regulating BR responses. On the other hand, OFP genes tend to be flexibly regulated by various hormones (Yu et al., 2015), and overexpression of OFPs usually resulted in obvious morphological changes, suggesting their roles in regulating hormone responses. The extensively interaction between OFPs and KNOX proteins further implied this speculation, considering the crucial roles of KNOXs in regulating hormone balance to maintain meristem

statistically significant differences between groups (Fisher's LSD, p < 0.05).

activity (Hackbusch et al., 2005; Tsuda et al., 2014). Given the distinct phenotypes of the OFP1 overexpression plants, and the remarkable regulation of OFP1 transcription and protein by BRs, our study should, at least partially, revealed the function of the native protein in plant.

OFP1 function is closely related to DLT, both positively regulate BR responses. The overexpression plants of the two genes are highly similar. It should be mentioned that although dlt mutant showed decreased tiller number, the severe DLT-overexpression lines could also have decreased tiller number, similar as OFP1-OE. Genetic analysis showed that OFP1 function in regulating plant architecture, but not the grain width, is dependent on DLT (**Figure 4**). It should be mentioned that the significant regulation of grain width by DLT could be partially independent of BR response, since the typical BR mutants have little alteration of grain width (Tanabe et al., 2005). Accordingly, OFP1 also has marked effect on grain width, further implying the functional relevance between OFP1 and DLT. Notably, the

OFP1-Flag fusion proteins in different leaf blades in 1Fo-2. Rice ACTIN protein was blotted as loading reference.

complementation analysis of the interaction between GSK2 and OFP1. (C) Effects of bikinin, the GSK3-like kinase inhibitor, on the protein stability of OFP1-Flag in 1Fo-2. Rice ACTIN protein was blotted as loading reference.

regulation of grain shape (length/width ratio) by OFP1 and DLT in rice was somewhat reminiscent of OVATE or OVATE-like gene in regulating the fruit shape in both tomato and pepper (Liu et al., 2002; Tsaballa et al., 2011), potentially suggesting a general role of OFPs in regulating seed development. A bold speculation is that OFPs could have indispensable regulatory roles in plant growth and development, thus plant has developed highly redundant system to maintain the normal function of the family.

Although BRs were known to be a class of growth-promoting hormones, it was found that most BR-enhanced plants didn't have increased plant height, and many of them actually exhibited decreased plant height in rice (Wang et al., 2008; Li et al., 2009; Tanaka et al., 2009; Tong et al., 2012). Consistently, exogenous high BR treatment also inhibits plant seedling growth. We previously discovered that this is because high BR or increased BR signaling turned to repress GA synthesis (Tong et al., 2014; Che et al., 2015). These observations suggested that BR balance is very important for plant growth. In this study, we revealed that OFP1 should play a role in this process. Overexpression of AtOFP1 in Arabidopsis, or OsOFP1 or OsOFP2 or OsOFP8 in rice, all leads to greatly reduced plant height (Wang et al., 2007; Schmitz et al., 2015; Yang et al., 2016). AtOFP1 directly targets GA20ox-1 to suppress GA synthesis, and OsOFP2 was also suggested to regulate GA20ox-7 to suppress GA synthesis. Measurement of GA contents in rice OFP1 overexpression plants demonstrated the decreased GA levels of basically all GA forms (**Figure 6B**). Notably, this profile is very similar to that in the high BR treated plants, which also contained decreased levels of basically all quantified GA forms (Tong et al., 2014). Actually, the OFP1 overexpression plants largely mimicked the high BR treated WT plants, including enlarged leaf angles and decreased plant height. Altogether, OFP1 should be involved in inhibition of GA levels and plant growth elicited by either high BR levels or elevated BR signaling.

#### REFERENCES


#### CONCLUSION

Elevated BR signaling will induce OFP1 expression by OsBZR1, and also promote protein stability by inhibiting GSK2, leading to activation of OFP1, which interacts with DLT factors and targets downstream genes, including GA metabolism genes, to regulate plant architecture and grain morphology in rice.

### AUTHOR CONTRIBUTIONS

YX and DL performed the study with the assistance of GZ and HT. HT, YX, and CC analyzed the data and wrote the manuscript. HT and CC conceived and supervised the study.

#### ACKNOWLEDGMENTS

We thank Prof. Jiayang Li and Prof. Jianmin Zhou (Institute of Genetics and Developmental Biology) for providing BiFC and split-LUC vectors, respectively, and Prof. Yaoguang Liu (South China Agricultural University) for providing CRISPR/Cas9 tools. This work was supported by National Natural Science Foundation (Nos. 91435106 and 91335203) and CAAS Elite Youth Program start-up funding (to HT).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2017.01698/ full#supplementary-material


roles in brassinosteroid signaling in rice. Plant J. 58, 803–816. doi: 10.1111/j. 1365-313X.2009.03825.x


GSK3-like kinases in Arabidopsis. Mol. Cell 66, 648–657. doi: 10.1016/j.molcel. 2017.05.012

**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 Xiao, Liu, Zhang, Tong and Chu. 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.

# High miR156 Expression Is Required for Auxin-Induced Adventitious Root Formation via *MxSPL26* Independent of *PINs* and *ARFs* in *Malus xiaojinensis*

Xiaozhao Xu, Xu Li, Xingwang Hu, Ting Wu, Yi Wang, Xuefeng Xu, Xinzhong Zhang\* and Zhenhai Han\*

*Institute for Horticultural Plants, College of Horticulture, China Agricultural University, Beijing, China*

#### *Edited by:*

*Yunde Zhao, University of California, San Diego, United States*

#### *Reviewed by:*

*Lee Jeong Hwan, Chonbuk National University, South Korea Guodong Wang, Shaanxi Normal University, China*

#### *\*Correspondence:*

*Xinzhong Zhang zhangxinzhong999@126.com Zhenhai Han rschan@cau.edu.cn*

#### *Specialty section:*

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

*Received: 25 February 2017 Accepted: 31 May 2017 Published: 19 June 2017*

#### *Citation:*

*Xu X, Li X, Hu X, Wu T, Wang Y, Xu X, Zhang X and Han Z (2017) High miR156 Expression Is Required for Auxin-Induced Adventitious Root Formation via MxSPL26 Independent of PINs and ARFs in Malus xiaojinensis. Front. Plant Sci. 8:1059. doi: 10.3389/fpls.2017.01059* Adventitious root formation is essential for the vegetative propagation of perennial woody plants. During the juvenile-to-adult phase change mediated by the microRNA156 (miR156), the adventitious rooting ability decreases dramatically in many species, including apple rootstocks. However, the mechanism underlying how miR156 affects adventitious root formation is unclear. In the present study, we showed that in the presence of the synthetic auxin indole-3-butyric acid (IBA), semi-lignified leafy cuttings from juvenile phase (Mx-J) and rejuvenated (Mx-R) *Malus xiaojinensis* trees exhibited significantly higher expression of miR156, *PIN-FORMED1* (*PIN1*), *PIN10*, and *rootless concerning crown and seminal roots*-like (*RTCS*-like) genes, thus resulting in higher adventitious rooting ability than those from adult phase (Mx-A) trees. However, the expression of *SQUAMOSA-PROMOTER BINDING PROTEIN-LIKE26* (*SPL26*) and some *auxin response factor* (*ARF*) gene family members were substantially higher in Mx-A than in Mx-R cuttings. The expression of *NbRTCS*-like but not *NbPINs* and *NbARFs* varied with miR156 expression in tobacco (*Nicotiana benthamiana*) plants transformed with *35S:MdMIR156a6* or *35S:MIM156* constructs. Overexpressing the miR156-resistant *MxrSPL* genes in tobacco confirmed the involvement of *MxSPL20*, *MxSPL21*&*22,* and *MxSPL26* in adventitious root formation. Together, high expression of miR156 was necessary for auxin-induced adventitious root formation via *MxSPL26*, but independent of *MxPINs* and *MxARFs* expression in *M. xiaojinensis* leafy cuttings.

Keywords: adventitious rooting, auxin, leafy cutting, *Malus xiaojinensis*, miR156

#### INTRODUCTION

Adventitious rooting is a cornerstone of proliferation for most fruit and forest species that are vegetatively propagated from elite genotypes. In the apple rootstocks, Sequoia sempervirens and Pinus radiata for example, propagation via both hard wood and leafy cuttings are constrained by adventitious rooting recalcitrance in the reproductively competent donor trees (Chen et al., 1981; Huang et al., 1992; Sanchez et al., 2007). To date, several techniques have been developed to improve rooting ability in some woody perennials, such as cycles of in vitro culture of apple rootstocks, or the repeated grafting of adult scions onto juvenile rootstocks in English ivy (Hedera helix L.) and chestnut (Castanea sativa; Giovannelli and Giannini, 2000; Xiao et al., 2014). The regulation of adventitious rooting is, however, relatively unexplored.

Adventitious root formation depends on multiple factors, such as genetic background, developmental stage, hormones, and other internal and external cues (Geiss et al., 2009; da Costa et al., 2013). In most tree species, the ability to form adventitious roots decreases during the transition from juvenile to adult development phases. In S. sempervirens, the rooting rate of cuttings from juvenile trees was up to 100% compared to 30% in cuttings from adult trees (Huang et al., 1992). Similarly, the rooting rate of cuttings from 3-year-old juvenile plants was significantly higher (88%) than from 36-year-old adult plants (17%) of Carolina Buckthorn (Rhamnus caroliniana Walt.; Graves, 2002). In our previous study, rooting rates of cuttings from juvenile, juvenile-like, and rejuvenated donor plants were significantly higher (77%) than those of cuttings from adult trees (11%) in Malus xiaojinensis (Xiao et al., 2014).

Numerous genes are differentially expressed between juvenile and adult cuttings prior to root induction. In loblolly pine (Pinus taeda L.), 5NG4, a nodulin-like gene, is highly and specifically induced by auxin in juvenile shoots prior to adventitious root formation, but is then substantially down-regulated in physiologically mature shoots that are adventitious rooting incompetent (Busov et al., 2004). A gene encoding a nitrate reductase involved in nitric oxide production in Eucalyptus grandis is up-regulated in juvenile cuttings as compared to that in mature cuttings, which might lead to increased ability to produce nitric oxide and form adventitious roots (Abu-Abied et al., 2012). Transcriptome data showed that the expression of E. grandis homologs of Peroxidase 72, PIN3, and Aux/IAA 19 (IAA19) were higher at a certain time point in auxin treated juvenile cuttings compared to mature ones (Abu-Abied et al., 2014); however, the insight effect of juvenility on adventitious root formation is not clear.

The juvenile to adult phase change is initiated by a decrease in the expression of the miR156. In Arabidopsis, maize (Zea mays) and various woody plants including Acacia confusa, Acacia colei, E. globulus, H. helix, Quercus acutissima, and Populus × Canadensis, miR156 is highly abundant in seedlings and decreases in adult plants (Wu and Poethig, 2006; Chuck et al., 2007; Wang et al., 2011). In the Congrass1 maize mutant that overexpresses miR156, prop roots are produced at all nodes in the plant, while these roots only grow from shoot-born meristem at the juvenile nodes in wild-type plants (Chuck et al., 2007). Similarly, tomato and tobacco plants overexpressing miR156 exhibit dense aerial roots on their stems, while none appear on the stems of wild type plants (Zhang et al., 2011; Feng et al., 2016). In contrast, Arabidopsis thaliana plants transformed with 35S:MIM156 produce significantly fewer adventitious roots from the base of the hypocotyl than wild-type plants (Xu et al., 2016).

MiR156 acts by repressing the expression of a group of SQUAMOSA-PROMOTER BINDING PROTEIN-LIKE (SPL) genes. Elevated levels of miR156 promote adventitious root formation in maize, tomato, and tobacco, indicating that SPL proteins inhibit adventitious root formation (Chuck et al., 2007; Zhang et al., 2011; Feng et al., 2016). EgSPL2 and EgSPL5 are up-regulated in mature cuttings compared to juvenile cuttings of E. grandis, and the juvenile cuttings exhibited a higher adventitious rooting percentage (Abu-Abied et al., 2012). These results suggest that miR156 and miR156 putative targeted MxSPL genes may play important roles in adventitious root formation during the juvenile to adult phase change. Twenty-seven SPL gene family members were identified in the apple genome and 13 were predicted to be targets of miR156 using degradome sequencing (Li et al., 2013; Xing et al., 2014). However, which SPL gene members are involved in adventitious root formation is unknown.

Auxin is an effective inducer of adventitious root formation; the dosage and gradient of, and response to, auxin are all important for plant root growth. Synthetic auxins like indole-3-butyric acid (IBA) have been used for almost 80 years to induce adventitious rooting (Zimmerman and Wil-Coxon, 1935). Without IBA treatment, leafy cuttings from juvenile M. xiaojinensis do not form adventitious roots (Xiao et al., 2014). The gradient of auxin controls adventitious root formation; in detached Arabidopsis leaves, an auxin gradient is required in the procambium cells during adventitious root formation (Liu J. C. et al., 2014). Auxin efflux carriers, such as PIN-FORMED (PIN) proteins, establish auxin concentration gradients (Yang and Murphy, 2009; Adamowski and Friml, 2015). Transgenic data suggests that OsPIN1 plays an important role in auxindependent adventitious root emergence in rice (Xu et al., 2005). Furthermore, PtoPIN1c is induced and maintained at a 20-fold level during the adventitious root initiation phase in Populus (Liu B. B. et al., 2014). In Arabidopsis, the pin1- 1 mutant formed 40% fewer adventitious roots than wild type (Sukumar et al., 2013). All of these results demonstrate that regeneration of adventitious roots requires polar auxin transport.

Beyond auxin gradients and dosage, genes that participated in auxin signaling pathways are also involved in adventitious root formation. The transgenic Arabidopsis line overexpressing the auxin response factor 17 (ARF17) gene develops fewer adventitious roots, while plants overexpressing the ARF6 or ARF8 develop more adventitious roots than wild-type plants (Sorin et al., 2005; Gutierrez et al., 2009). It was previously reported that to promote adventitious root formation, ARF proteins directly regulate LATERAL ORGAN BOUNDARIES DOMAIN PROTEIN (LBD) genes by binding to their promoter sequences (Okushima et al., 2007; Majer et al., 2012). Crown rootless 1 (Crl1) encodes a member of the LBD genes in rice (Oryza sativa), which positively regulate adventitious root formation, and its expression is directly regulated when an ARF binds to the 5′ flanking sequences of Crl1 (Inukai et al., 2005). In maize, rootless concerning crown and seminal root (Rtcs) encodes a LBD protein that positively regulates adventitious root growth, and its expression is regulated by ZmARF34, which binds to an auxin response element in the promoter region of Rtcs (Majer et al., 2012). However, during the adventitious root formation in recalcitrant woody plants, whether and how miR156 interacts with auxin remains unknown.

Xu et al. miR156 Regulates Adventitious Root Formation

To evaluate how juvenility mediates adventitious rooting in woody plants and using apple rootstock M. xiaojinensis as an example, we analyzed the transcript level of miR156, miR156 putative target SPL gene expression, and MxPIN and MxARF gene family members during the adventitious rooting process. Then, the function of miR156 in adventitious root formation was validated by generating transgenic tobacco lines. The results elucidate the role of miR156 in adventitious root formation and will be useful in horticultural and forestry industries.

### MATERIALS AND METHODS

### Plant Material

M. xiaojinensis (Mx) was used in these experiments because of its high apomictic rate to ensure stable and robust juvenile materials (Li et al., 2004). The apomictic origin of donor trees used for leafy cutting collection has been confirmed by simple sequence repeat (SSR) and single nucleotide polymorphism (SNP) markers (Wang et al., 2012). Cuttings were excised from basal suckers (juvenile phase, Mx-J), shoots from the canopy of reproductively mature trees (adult phase, Mx-A), and rejuvenated plants via in vitro apical meristem culture (rejuvenated or juvenile like, Mx-R; Xiao et al., 2014).

Semi-lignified leafy cuttings (8–10 cm in length) were excised from actively growing shoots of donor plants, dipped 1 cm in depth into a solution containing 3000 mg/L indole butyric acid (IBA) (V900325, Sigma-Aldrich, St. Louis, MO, USA) for 1 min, and then the cuttings were plugged into 50 cell trays containing fine sand as a rooting medium. The cuttings were incubated in a solar greenhouse under 90–95% relative humidity (Xiao et al., 2014). Cuttings treated with an IBA-free medium solution were used as controls. The experimental errors were managed by using a complete randomized design for three biological replicates, each including at least 50 leafy cuttings. Rooting parameters were scored at 35 d after application (Xiao et al., 2014).

#### Plant Expression Vector Construction and Transgenic Tobacco Generation

Nine miR156 precursor genes were previously identified in apple genome (Ma et al., 2014). MdMIR156a6 (Genome location: MDC018927.245), which relative transcript level was higher than the other members, was chosen for overexpression study (**Supplementary Figure 1**). The apple MdMIR156a6 DNA fragment was amplified from Malus domestica genomic DNA. Artificial target mimics were generated by modifying the sequence of the AtIPS1 gene to knock-down miR156 expression (Franco-Zorrilla et al., 2007). The PCR products were sequenced by BGI (Shenzhen, China). All constructs were cloned behind the constitutive CaMV 35S promoter in the pBI121 vector, and then introduced into Nicotiana benthamiana by Agrobacterium tumefaciens-mediated transformation (Horsch et al., 1985).

MxSPLs coding sequences were amplified from M. xiaojinensis cDNA. The miR156-resistant SPLs (rSPLs) were made by two rounds of mutagenic PCR using KOD-Plus Nero DNA polymerase (KOD-401, TOYOBO LIFE SCIENCE, Japan), and sequencing confirmed the mutations by BGI (Shenzhen, China). The rSPLs variants were driven by the CaMV 35S promoter in the pBI121 vector. These constructs were introduced into 35S:MdMIR156a6 transgenic tobacco leaves by A. tumefaciens-mediated transformation. The infected leaves were selected on MS medium supplemented with 100 mg/L kanamycin and 300 mg/L cefotaxime sodium to generate rSPL and miR156 co-overexpressing transgenic lines. All primers are listed in **Supplementary Table 1**. Transgenic tobacco plants were proliferated by subculture on hormone-free MS medium. Rooting rate and the number of adventitious roots were recorded at 5, 7, 9, 11, and 13 d after subculture. Plants were incubated at 25◦C in long-day light conditions.

#### Histological Analysis

Mx-J, Mx-A, and Mx-R cutting samples were collected at 0, 14, and 28 d after IBA treatment; cuttings from 35S:MdMIR156a6 and 35S:MIM156 transgenic tobacco lines were excised at 0, 3, and 6 d after subculture on hormone-free MS medium. All samples were collected from three biological replicates and n = 10 in each replicate. A one-centimeter section from the bottom of each cutting was excised and fixed in a 5:50:5 (v/v/v) formaldehyde/ethanol/acetic acid (FAA) solution overnight at room temperature. Cuttings were dehydrated in an ethanol series (70, 85, 95, and 100%), infiltrated with xylene, and embedded in paraffin. Then, 15µm-thick transverse sections were cut with a rotatory microtome (KD-2258, KEDEE, China) and stained with toluidine blue (Rigal et al., 2012).

#### Gene Expression Analysis

For relative gene expression assays with M. xiaojinensis, stem bark from 0.5 to 1 cm basal sections of 20 cuttings were frozen in liquid nitrogen at 0, 6, 12, 24, 72, 120, and 168 h after IBA treatment. For gene expression profile analysis in tobacco, stems from 0.5 to 1 cm basal sections of 20 cuttings were sampled at 0, 3, 6, 12, 24, 48, and 72 h after subculture on hormonefree MS medium. Total RNA was isolated from ∼500 mg of frozen tissue using a modified cetyltrimethylammonium bromide (CTAB) method (Gasic et al., 2004). The RNA was digested by DNaseI (2313A, Takara, Dalian, China) and reverse-transcribed using oligo-dT18 primers and reverse transcriptase according to the manufacturer's instructions (2641A, Takara, Dalian, China). Semi-quantitative RT-PCR results were quantified by using ImageJ 1.47v (Wayne Rasband, National Institutes of Health, USA) according to the commands in "gels submenu" to analyze one-dimensional electrophoretic gels (https://imagej. nih.gov/ij/docs/menus/analyze.html#gels). Quantitative RT-PCR were performed using SYBR green reagents (RR820A, Takara, Dalian, China) in an Applied Biosystems 7500 real-time PCR system. M. xiaojinensis EF1α or N. benthamiana EF1α was used as the expression control for these experiments. The relative expression level was calculated according to the 2−11CT method (Livak and Schmittgen, 2001). Three independent biological replicates and technical replicates were performed.

MicroRNA was extracted by using the RNAiso for Small RNA kit (9753Q, Takara, Dalian, China) according to the manufacturer's instruction. MiR156 expression level was analyzed by qRT-PCR as described previously (Xiao et al., 2014). The apple PIN and ARF family genes were previously identified (Devoghalaere et al., 2012; Zhang H. et al., 2015). The apple and N. benthamiana RTCS-like gene were selected by using a BLASTP search of known Maize RTCS (GenBank accession number EF051732) gene against the Apple Genome Database (https://www.rosaceae.org/) and Tobacco Genome Database (https://www.solgenomics.net/), respectively. N. benthamiana PIN and ARF genes were selected by using a BLASTP search of known Arabidopsis auxin-related genes against the Tobacco Genome Database (https://www.solgenomics.net/). All primers used for qRT-PCR are listed in **Supplementary Tables 2**,**3**.

#### Statistical Analysis

Statistical analysis was performed using the Statistical Product and Service Solutions (SPSS) software (IBM Co., Armonk, USA). All experimental data were tested by Student's t-test or Duncan's multiple-range test.

# RESULTS

#### Adventitious Rooting Ability of Mx-A, Mx-J, and Mx-R Leafy Cuttings

Juvenile cuttings (Mx-J) exhibited a highly adventitious rooting ability in M. xiaojinensis (**Figure 1A**). Leafy Mx-J cuttings exhibited a 78.63% rooting percentage at day 35 after IBA treatment (**Figure 1B**). In contrast, the rooting percentage of adult cuttings (Mx-A) was significantly lower (4.38%) than that of Mx-J. The adventitious root number per cutting was also lower in Mx-A cuttings than in Mx-J cuttings (**Figure 1C**). Similarly, the adventitious root length in Mx-J was significantly longer than that in Mx-A cuttings. Mx-R cuttings exhibited significantly enhanced adventitious rooting percentage and adventitious root number than Mx-A or Mx-J cuttings (**Figure 1**).

To determine whether adult leafy cuttings have defects in the initiation of adventitious root formation primordia, cuttings were processed for histological analysis after IBA treatment. There were no obvious differences in anatomical structure between juvenile and adult softwood cuttings before IBA treatment (**Supplementary Figure 2**). At 14 d after IBA treatment, root primordia appeared in Mx-J and Mx-R but not Mx-A cuttings. At 28 d, root primordia penetrated the stem cortex and epidermis, and projected well beyond the stem surface in Mx-J and Mx-R (**Figure 1E**). In contrast, although meristematic cambium cells were observed between the phloem and xylem layers, except for callus, there was almost no differentiated root primordia detected in Mx-A cuttings even after 28 d of auxin treatment (**Figure 1E**).

### The Relationship between miR156 Expression and Adventitious Root Formation

Consistent with the rooting ability, the expression level of miR156 in Mx-J and Mx-R cuttings were significantly higher than that in Mx-A cuttings (**Figure 2**). To validate if miR156 regulates adventitious root formation, we generated transgenic tobacco plants expressing 35S:MdMIR156a6 or 35S:MIM156 to enhance or inhibit miR156 activity, respectively (**Figure 3A**). Mature miR156 expression levels were significantly up-regulated in 35S:MdMIR156a6 plants and significantly reduced in 35S:MIM156 plants (**Figure 3B**). Similarly, the expression levels of some miR156 target genes, including NbSPL2a, NbSPL2b, NbSPL5a, NbSPL5b, NbSPL15a, and NbSPL15b were down-regulated at least 2-fold in 35S:MdMIR156a6 stems, but at least 5-fold upregulated in 35S:MIM156 stems (**Figure 3C**). 35S:MdMIR156a6 plants developed almost 20 adventitious roots per cutting, which was 4-fold more than the wild type after 13 days of culture on MS medium. Almost no adventitious roots were observed in 35S:MIM156 cuttings (**Figure 3D**). In addition to the difference in adventitious root number, the rate of adventitious root development in 35S:MdMIR156a6 plants was significantly faster than that measured in the wild-type plants (**Figure 3E**).

To check whether miR156 affects the initiation of adventitious root primordia, serial cross sections of the stems of wild-type, 35S:MdMIR156a6 expressing transgenic lines, and 35S:MIM156 expressing transgenic tobacco plants were stained with toluidine blue. When compared with the wild-type, the initiation of adventitious root primordia was accelerated and well-developed in 35S:MdMIR156a6 stem cuttings only 3 days after subculture on MS (**Figure 4**). By contrast, no adventitious root primordia were observed in 35S:MIM156 stem cuttings throughout the experiment (**Figure 4**).

# Interaction between miR156 and Auxin during Adventitious Root Formation

To determine whether miR156 regulates adventitious root formation through modulating endogenous auxin levels or auxin polar transport, we examined adventitious rooting capacity in wild-type, 35S:MdMIR156a6, and 35S:MIM156 stem cuttings grown on MS medium supplemented with different concentrations of indole-3-acetic acid (IAA) or 1- N-naphthylphthalamic acid (NPA). The percent rooting was significantly increased in wild-type and 35S:MdMIR156a6 stem cuttings when cultured in the presence of 0.1 or 1 µM IAA (**Figures 5A,B**). The 10µM IAA treatment delayed adventitious root formation in both wild-type and 35S:MdMIR156a6 plants; however, 35S:MdMIR156a6 plants exhibited a significantly higher adventitious rooting percent and adventitious root number than wild-type. In addition, with IAA application, the adventitious root number was clearly increased in both wild-type and 35S:MdMIR156a6 plants (**Figures 5A,C**) but IAA treatment did not rescue the adventitious rooting capacity defect in 35S:MIM156 plants (**Figure 5**).

Treatments with 20µM NPA significantly inhibited the adventitious root formation in wild-type (**Supplementary Figure 3**); adventitious rooting was obviously delayed and adventitious root number was reduced in both wild-type and 35S:MdMIR156a6 plants treated with 20µM NPA (**Figure 6A**). Indeed, rooting percent and root number in 35S:MdMIR156a6 plants were still consistently higher than that in wild-type (**Figures 6B,C**).

Scale bars = 20 mm in (A). (B–D) Bars show *SD* from three biological replicates. *n* = 50 individuals in each replicate. Asterisks indicate significant difference from the Mx-A (Student's *t*-test, \*\**P* < 0.01, \**P* < 0.05). (E) Histological features of Mx-A, Mx-J, and Mx-R cuttings during adventitious root formation at 14 days (upper) and 28 days (bottom) after IBA treatment. Toluidine Blue-stained cross section of the stem of Mx-A, Mx-J, and Mx-R at 14 and 28 days after IBA treatment. Arrows indicate emerging adventitious root initials. Scale bars = 200 µm in (E).

(Mx-J), and rejuvenated (Mx-R) leafy cuttings of *Malus xiaojinensis*. MicroRNA was extracted from Mx-A, Mx-J, and Mx-R cuttings before IBA treatment. Relative expression was measured with quantitative real time PCR and normalized to that *5S rRNA*. Asterisks indicate statistical significance (\**P* < 0.05) in comparison with Mx-A. Bars show *SD* from three biological replicates.

# *PIN* Genes Are Not Altered by miR156 Expression

The expression levels of MxPIN3, MxPIN4, MxPIN6, MxPIN8, MxPIN9, MxPIN12, and MxPIN13 did not show obvious differences between Mx-A and Mx-R leafy cuttings during the adventitious root-induction process (**Figure 7A** and **Supplementary Figure 4**). The expression of MxPIN2, MxPIN5, MxPIN7, and MxPIN11 was not detected in any of the RNA samples tested. As shown in **Supplementary Figure 4**, the expression of MxPIN1 and MxPIN10 was higher in Mx-R than in Mx-A with or without IBA treatment. Both MxPIN1 and MxPIN10 expression was induced more than 2-fold after 6 h in the presence of IBA treatment in Mx-R, but not in Mx-A (**Figure 7B**). The relatively high expression level of MxPIN1 and MxPIN10 may contribute to adventitious rooting competitive in Mx-R cuttings. However, the relationship between miR156 and PIN gene expression was not robust in transgenic tobacco; no distinct changes were detected in the expression of any NbPIN members between wild-type, 35S:MdMIR156a6, and 35S:MIM156 tobacco plants (**Figure 7C**).

### *ARF* Genes Expression Did Not Change with miR156 Levels

The MxARF4, MxARF7, MxARF14, MxARF15, MxARF16, MxARF17, MxARF19, MxARF20, and MxARF28 genes showed higher expression in Mx-A than in Mx-R cuttings with or without IBA treatment during the adventitious root formation process (**Supplementary Figure 6**). In contrast, expression of MxARF3 and MxARF8 was higher in Mx-R than in Mx-A cuttings (**Supplementary Figure 6**). MxARF21, MxARF23 and MxARF25 were down regulated in plants treated with IBA in Mx-R but not Mx-A cuttings (**Figure 8A**). However, no distinct changes were detected in 16 putative NbARF gene family members between wild-type, 35S:MdMIR156a6, and 35S:MIM156 tobacco plants during the adventitious rooting process (**Figure 8B**). This indicates that miR156 may not be associated with the expression of ARF transcription factors during the regulation of adventitious root formation.

# *RTCS*-Like Gene Expression Varied with miR156 Levels

In M. xiaojinensis, the MxRTCS-like gene was up-regulated 6–24 h after IBA treatment in both Mx-A and Mx-R cuttings, but the maximum induction of MxRTCS was observed in Mx-R cuttings at 120–168 h, which was 4-fold higher than that measured in Mx-A cuttings (**Figure 9**). Similarly, the expression of NbRTCS was also significantly induced 24–72 h after subculture in hormone-free medium in 35S:MdMIR156a6 plants, but was delayed 72 h subculture in hormone-free medium in wild-type. However, the expression of NbRTCS did not change throughout the experiment in 35S:MIM156 plants (**Figure 9**).

# Response in *MxSPL* Gene Expression to miR156 Levels

Of the 13 putative miR156-regulated SPL gene family members in apple genome, nine were actively expressed in both Mx-J and Mx-A plant cuttings, but did not include MxSPL3, MxSPL10, MxSPL11, and MxSPL12. To investigate which M. xiaojinensis SPL gene family member was involved in adventitious root formation, mutants of SPL genes, indicated here as resistant SPLs (rSPLs), were designed to no longer be targeted by miR156 (**Supplementary Figure 8**; Schwab et al., 2005). MxrSPL4a&4b, MxrSPL18, MxrSPL19, MxrSPL20, MxrSPL21&22, MxrSPL24, and MxrSPL26, driven by CaMV 35S promoter, were transformed into 35S:MdMIR156a6 transgenic tobacco plants (**Supplementary Figure 9**). In independent bivalent transgenic lines, 35S:rSPL4a&4b/35S:MdMIR156a6, 35S:rSPL18/35S:MdMIR156a6, 35S:rSPL19/35S:MdMIR156a6, and 35S:rSPL24/35S:MdMIR156a6 the adventitious rooting rate and adventitious root number did not significantly differ from that of plants expressing 35S:MdMIR156a6 (**Supplementary Figure 10**), indicating that MxSPL4a&4b, MxSPL18, MxSPL19, and MxSPL24 are not involved in adventitious rooting. In contrast, the bivalent transformants 35S:rSPL20/35S:MdMIR156a6, 35S:rSPL21&22/35S:MdMIR156a6, and 35S:rSPL26/35S: MdMIR156a6 exhibited reduced adventitious rooting ability (**Figure 10A**). 35S:rSPL20/35S:MdMIR156a6 and 35S:rSPL21&22/35S:MdMIR156a6 produced significantly fewer adventitious roots than 35S:MdMIR156a6 plants, but still more than wild-type plants (**Figure 10B**). The adventitious root number in 35S:rSPL26/35S:MdMIR156a6 plants was significantly fewer than that in 35S:MdMIR156a6 plants (**Figure 10B**). In addition, the adventitious root development rate in 35S:rSPL20/35S:MdMIR156a6, 35S:rSPL21&22/35S:MdMIR156a6, and 35S:rSPL26 /35S:MdMIR156a6 plants were similar as that measured in wildtype, but slower than that observed in 35S:MdMIR156a6 plants 7 d after culture on MS medium (**Figure 10C**). These results

adventitious root primordia. Scale bars = 200 µm.

indicate that MxSPL20, MxSPL21&22, and MxSPL26 not only affected adventitious root number but also delayed rooting rate.

The MxSPL20, MxSPL21&22, and MxSPL26 expression profile was detected at 0, 6, 12, 24, 72, 120, and 168 h after IBA treatment (**Figure 11**). There were no obviously differences in the expression of MxSPL20 and MxSPL21&22 between Mx-A and Mx-R cuttings; however, the expression of MxSPL26 was significantly higher in Mx-A than in Mx-R cuttings, indicating MxSPL26 could be a key SPL family member involved in adventitious rooting of leafy cuttings.

#### DISCUSSION

Although, it appears that juvenile phase softwood cuttings are much easier to root than the adult ones, there is little data detailing the mechanism of these differences. The expression level of miR156 was decreased during juvenile to adult phase change (Du et al., 2015; Ji et al., 2016). However, the effect of miR156 on adventitious root formation barely rated a mention in previous reports (Zhang et al., 2011; Feng et al., 2016; Massoumi et al., 2017). Although, miR156-targeted SPLs are known to control varied physiological and developmental processes (Wang et al., 2008; Shikata et al., 2009; Yu et al., 2010, 2015; Gou et al., 2011), with which the miR156-targeted SPL gene member is associated with adventitious rooting and how miR156 interacts with other rooting regulatory factors such as IAA are so far not fully understood to date. In the present study, we found that the high miR156 expression was required for adventitious roots formation in an apple rootstock, M. xiaojinensis. We also confirmed the involvement of MxSPL26 in inhibiting adventitious root formation.

### Auxin and High miR156 Expression Level Are Both Necessary for Adventitious Rooting

For some rooting recalcitrant woody plants, juvenility is necessary for efficient adventitious rooting. In general, cuttings from juvenile nodes root more readily than cuttings from adult nodes in woody plants (Greenwood, 1995). Indeed, miR156 expression levels were significantly higher in Mx-J and Mx-R cuttings than Mx-A cuttings (**Figure 2**), and the rooting ability of Mx-R and Mx-J cuttings were consistently higher than that of Mx-A cuttings (**Figure 1**). A significant decrease in miR156 expression was observed in shoots taken from 1.4 m trunk above ground compared to shoots taken from below 1.4 m in M. xiaojinesis seedlings (Ji et al., 2016). In our previous data, the micro-shoots from adult phase M. xiaojinnesis explants were

same lowercase letter indicate no significant differences.

rejuvenated successfully after 15 passages of in vitro subculture, marked by the elevated expression of miR156 expression in leaves of the micro-shoots, and coupled with recovered adventitious rooting ability and leaf lobes (Xiao et al., 2014).

When miR156 level was manipulated via transformation with a 35S:MIR156 construct in tomato, tobacco, or Arabidopsis, the adventitious rooting increased (Zhang et al., 2011; Feng et al., 2016; Massoumi et al., 2017). In the present study, the miR156 expression level was manipulated in transgenic tobacco to analyze how miR156/SPL modules are involved in adventitious rooting. Consistent with the previous reports, the adventitious rooting capacity also varied with 35S:MIM156 and 35S:MdMIR156a6 overexpressing transgenic tobacco plants (**Figure 3**). However, the uncoupling of miR156 expression and adventitious root formation was reported in E. grandis (Levy et al., 2014); therefore, miR156 may be necessary but not sufficient for adventitious rooting in woody plants. Except for high expression of miR156, auxin was also integrant for adventitious root formation. In absence of IBA treatment, even Mx-J and Mx-R leafy cuttings exhibit adventitious root defects (**Figure 1**). In agreement to this result, the adventitious rooting ability was reduced in all tobacco lines upon treatment with the polar auxin transport inhibitor NPA (**Figure 6**).

### miR156 Affects Adventitious Root Formation Independently from *PIN* and *ARF* Genes Expression

Auxin and miR156 are both involved in adventitious root development (Zhang et al., 2011; Feng et al., 2016; Steffens and Rasmussen, 2016; Massoumi et al., 2017). However, the interaction of miR156 and auxin signaling pathway is unexplored during adventitious root development. In comparison to the wild-type samples, NbPINs and NbARFs expressions was not substantially modulated in 35S:MdMIR156a6 and 35S:MIM156 plants (**Figures 7**, **8**), indicating that miR156 promoted adventitious root formation independently from changing PIN and ARF genes expression. These results are supported by the transcriptome data, which show that PIN and ARF genes in Medicago sativa are not significantly differentially expressed between miR156 overexpression and wild-type plants (Gao et al., 2016). Similarly, the auxin response was not changed in either Pro35S:MIR156 or Pro35S:MIM156 Arabidopsis, indicating that miR156 does not modulate the auxin response during regulating shoot regeneration (Zhang T. Q. et al., 2015). Collectively, these findings suggest that miR156 does not modulate the early auxin response.

Conversely, auxin can induce the expressions of two MIR156 genes and two SPL genes during lateral root development in transgenic Arabidopsis (Yu et al., 2015). GUS staining revealed that MIR156B was specifically expressed in primary and lateral root primordia, MIR156D was especially active in primary and lateral root tip, and two SPL genes were also highly or specifically expressed in root (Yu et al., 2015). Hence, auxin inducing MIR156 and SPL expression may be tissue/organ specific during lateral root development. Our results revealed that the expression level of MxSPL26 or MxSPL20 or MxSPL21&22 was not obviously affected under IBA treatment (**Figure 11**). Similarly, miR156 expression was not affected by IBA application during the induction of adventitious root development in both juvenile and adult E. grandis stem cuttings (Levy et al., 2014). Thus, miR156/SPL modules were not downstream targets of auxin during adventitious root formation.

Rtcs and Rtcl in maize are orthologs of CRL1 in rice (O. sativa) and of AtLBD29 in A. thaliana, and all of these genes are involved in the initiation of adventitious root formation and are targets of the ARF gene family (Inukai et al., 2005; Taramino et al., 2007; Liu J. C. et al., 2014). Although MxRTCS-like gene expression was induced in both Mx-A and Mx-R cuttings after IBA treatment, the distinct fold-change in expression occurred only in Mx-R cuttings (**Figure 9**). In agreement with these finding, the NbRTCS expression was significantly induced in 35S:MdMIR156a6 transgenic tobacco plants, but no substantial changes were detected in 35S:MIM156 transformants (**Figure 9**). These data suggest that high expression level of miR156 is required for auxin inducing expression of RTCS-like during adventitious root formation in M. xiaojinesis and transgenic tobacco. Therefore, we speculate that decline of miR156 expression in adult phase leafy cuttings may inhibit transcript abundance of RTCS-like gene induced by auxin, thereby reducing the adventitious rooting capacity (**Figure 12**). The histological features showed that Mx-A leafy cuttings and 35S:MIM156 transgenic tobacco plants exhibited defect in adventitious root primordia initiation (**Figures 1E**,**4**), which provided an evidence for above mentioned speculation. In agreement with our observations, maize mutant rtcs and rice mutant rtcl both failed in adventitious root primordia initiation (Inukai et al., 2005; Muthreich et al., 2013). However, the molecular mechanism underlying miR156 modulate auxin induced RTCS-like gene expression remains unknown.

Although, the expressions of PIN and ARF genes were independently from the miR156 expression level, the response of MxPIN1 and MxPIN10 gene expression to auxin treatment

differed in Mx-R and Mx-A samples. The different responses of PIN family members to auxin treatment between Mx-R and Mx-A cuttings may be caused by other developmental signals, such as glutathione content. In apple seedlings, the glutathione content and glutathione/glutathione disulfide ratio were much higher in the juvenile phase than in the adult phase, and modulating glutathione content caused concomitant changes in miR156 expression levels (Du et al., 2015). It has been proved that the reduction of glutathione availability by L-buthionine-(S,R)-sulfoximine (BSO) treatment reduced the

from Mx-A and Mx-R during adventitious root formation after IBA treatment. Bars show *SD* from three biological replicates. (C) *NbPIN* genes expression patterns in WT, *35S:MdMIR156a6,* and *35S:MIM156* transgenic tobacco stems during the adventitious rooting process (original results show in Supplementary Figure 5).

FIGURE 8 | Expression profiles of *ARFs* during adventitious rooting in *Malus xiaojinensis* and transgenic *Nicotiana benthamiana*. All the results of semi-quantitative RT-PCR were quantified using the ImageJ software. Log-transformed values of the relative expression levels of *MxARF* family genes under IBA treatment compared to controls were used for hierarchical cluster analysis with MeV 4.8.1. The color scale represents relative expression levels with red denoting up-regulation and green denoting down-regulation. The relative expression levels of *NbARFs* genes compared to *NbEF1*α were used for hierarchical cluster analysis with MeV 4.8.1. Sampling times are indicated at top of the figure. (A) Expression profiles of *MxARF* genes in stem bark of Mx-A and Mx-R during adventitious root formation process after IBA treatment (original results show in Supplementary Figure 6). (B) *NbARF* genes expression patterns in WT, *35S:MdMIR156a6,* and *35S:MIM156* transgenic tobacco stems during the adventitious rooting process (original results show in Supplementary Figure 7).

*35S:rMxSPL*/*35S:MdMIR156a6* plants. Scale bars = 1 cm. (B,C) Quantitative analysis of rooting ability of tobacco stem cuttings. (B) Adventitious root numbers per cutting were counted after 14 d on MS medium. (C) Percent rooting was investigated after 7 and 14 d of growth on MS medium. Bars show *SD* with three biological replicates; *n* = 5 for each replicate. The statistical analysis was performed by Duncan's multiple range test at level *p* ≤ 0.05. The same lowercase letter indicate no significant differences.

FIGURE 11 | Expression profile of *MxSPL20, 21*&*22,* and *26* genes during adventitious root formation in *Malus xiaojinensis* leafy cuttings. RNA was extracted from Mx-A and Mx-R stem bark after IBA treatment at serial time-points. Relative expression was measured using quantitative real time PCR normalized to *MxEF1*α. Bars show *SD* from three biological replicates.

expression of PIN1, PIN2, and PIN3 in Arabidopsis (Bashandy et al., 2010).

In addition, MxARF4, MxARF7, MxARF14, MxARF15, MxARF16, MxARF17, MxARF19, MxARF20, and MxARF28 showed relative higher expression levels in Mx-A than in Mx-R cuttings during adventitious root formation (**Supplementary Figure 6**). Interestingly, the peptide sequences of these ARFs, except for ARF20, are enriched in serine, proline, and leucine in their middle regions (**Supplementary Figure 11**). Arabidopsis ARFs with this middle region are usually transcriptional repressors (Guilfoyle and Hagen, 2007). The differential expression of MxPIN and MxARF might partly cause Mx-A recalcitrant to root formation.

#### *MxSPL26* Is the Key Target Gene in miR156 Regulation of Adventitious Rooting in *M. xiaojinensis*

MiR156 and its targeted SPL genes were demonstrated to control plant phase transition and related traits such as cell size and number, trichome development, anthocyanin synthesis, leafy morphology, flowering time, and lateral root development (Wang et al., 2008; Shikata et al., 2009; Yu et al., 2010, 2015; Gou et al., 2011). Yet, the function of SPL genes involving in adventitious rooting are largely unexplored. It was reported that hypocotyls of spl2/9/11/13/15 mutants, as well as plants transformed with 35S:MIR156A, produced the same number of adventitious roots as wild-type plants (Xu et al., 2016). However, it is not clear which SPL gene involved in regulating adventitious rooting in Arabidopsis, because the expression level of miR156 targeted SPLs had no significantly difference among spl2/9/11/13/15 mutants, 35S:MIR156A plants and wild-type young seedlings (Xu et al., 2016). Here, MxSPL26 was experimentally confirmed to be involved in inhibiting adventitious root formation most possibly in function redundancy with SPL20 and SPL21&22, but SPL26 seemed to play a major role. The expression of MxSPL26 was constantly higher than that of MxSPL20 and MxSPL21&22 in M. xiaojinensis cuttings. MxSPL26 relative expression exhibited a 5- to 10-fold increase in stem bark from Mx-A cuttings than in that from Mx-R after 72∼120 h from plugging into the media, but no obvious differences in the expression of MxSPL20 and MxSPL21&22 genes were found between Mx-A and Mx-R (**Figure 11**). The number of roots per plant was significantly reduced in all the 35S:rMxSPLs/35S:MdMIR156a6 lines but MxSPL26 resistant form transgenic tobacco lines produced the least number of adventitious roots compared with rMxSPL20, rMxSPL21&22 and 35S:MdMIR156a6 transgenic lines (**Figure 10** and **Supplementary Figure 10**). According to the phylogenetic tree, MxSPL20 and MxSPL21&22 exhibited close evolutionary, but quite far phylogenetic relationship from MxSPL26 (**Supplementary Figure 12**). Notably, although SPL3, SPL9, and SPL10 were all involved in repressing Arabidopsis lateral root development, SPL10 seemed to play a major role (Yu et al., 2015). Similarly, AtSPL3 and AtSPL9 exhibited close evolutionary, but quite far phylogenetic relationship from AtSPL10 (**Supplementary Figure 12**).

In conclusion, in semi-lignified leafy cuttings of M. xiaojinensis, a relatively higher expression of the miR156 is necessary for adventitious root formation. MiR156 functions via its target gene MxSPL26 and rooting-related genes such as MxRTCS-like; but acts independently of MxPIN or MxARF family members in response to auxin. These results provided a potential strategy for the improvement of the adventitious rooting ability of perennial woody plants via manipulating miR156 and SPL gene expression.

### AUTHOR CONTRIBUTIONS

XZ, YW, TW, XzX, and ZH prepared the plant materials and designed the experiments. XzX, XL, and XH conducted the experiments. XzX took the photographs. XzX and XZ analyzed the data and XzX wrote the manuscript. All authors read and approved the manuscript.

### FUNDING

This work was funded by Special Fund for Agro-scientific Research in the Public Interest (201203075); the Modern Agricultural Industry Technology System (CARS-28); the National Natural Science Foundation of China (NSFC) (Grant no. 31372020; 31672107); Key Laboratory of Biology and Genetic Improvement of Horticulture Crop (Nutrition and Physiology), Ministry of Agriculture, and Beijing Collaborative innovation center for eco-environmental improvement with forestry and fruit trees.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2017. 01059/full#supplementary-material

Supplementary Figure 1 | Expression profiles of nine miR156 precursors were analyzed using semi-quantitative RT-PCR in *Malus xiaojinensis* stem bark.

Supplementary Figure 2 | Histological features of Mx-A, Mx-J, and Mx-R leafy cuttings of *Malus xiaojinensis* before IBA treatment. Cross sections of the stems were stained with toluidine blue. Scale bars = 200 µm.

# REFERENCES

Abu-Abied, M., Szwerdszarf, D., Mordehaev, I., Levy, A., Rogovoy, O., Belausov, E., et al. (2012). Microarray analysis revealed upregulation of nitrate reductase in juvenile cuttings of Eucalyptus grandis, which correlated with increased Supplementary Figure 3 | NPA (1-N-naphthylphthalamic acid) concentration selected for wild type tobacco plant treatment. Tobacco plants were transferred to MS medium with 5, 20, or 80 µM NPA for 14 days. (A) The phenotype and (B) percent rooting were evaluated. Bars show *SD* from three biological replicates; *n* = 10 individuals in each replicate. Scale bars = 15 mm.

Supplementary Figure 4 | Semi-quantitative RT-PCR analysis of the expression dynamics of *MxPINs* in stem barks from Mx-A and Mx-R during adventitious root formation in *Malus xiaojinensis* under IBA treatment. *MxEF1*α was used as an internal control. The upper and lower bands represent treatment and control, respectively.

Supplementary Figure 5 | *NbPIN* genes expression pattern in tobacco stems from WT, *35S:MdMIR156a6,* and *35S:MIM156* transgenic lines growing on MS medium during the adventitious rooting process. *NbEF1*α was used as an internal control.

Supplementary Figure 6 | Semi-quantitative RT-PCR analysis of the expression dynamics of *MxARFs* in stem barks from Mx-A and Mx-R during adventitious root formation in *Malus xiaojinensis* under IBA treatment. *MxEF1*α was used as an internal control. The upper and lower bands represent treatment and control, respectively.

Supplementary Figure 7 | *NbARF* genes expression pattern in tobacco stems from WT, *35S:MdMIR156a6,* and *35S:MIM156* transgenic lines growing on MS medium during the adventitious rooting process. *NbEF1*α was used as an internal control.

Supplementary Figure 8 | Diagram of the miR156 target sites of the WT and modified version of *MxSPLs*. Capital letters indicate the amino acid sequences in *Malus xiaojinesis*.

Supplementary Figure 9 | Molecular characterization of transgenic tobacco plants. (A) Semi-quantitative PCR analysis of *rMxSPLs* DNA and mRNA levels and (B) miR156 expression levels in 10-day old seedling leaves.

Supplementary Figure 10 | Role of *MxSPL4a*&*4b, 18, 19,* and *24* genes during adventitious root formation in transgenic *Nicotiana benthamiana*. (A) Adventitious root formation in *35S:rMxSPL*/*35S:MdMIR156a6* plants. Scale bars = 1 cm. (B,C) Quantitative analysis of the rooting ability in tobacco stem cuttings. (B) Percent rooting was investigated after 7 and 14 d on MS medium. (C) Adventitious root numbers per cutting were counted after 14 d on MS medium. Bars show *SD* from three biological replicates; *n* = 5 plants in each individual replicate. The statistical analysis was performed by Duncan's multiple range test at level *p* ≤ 0.05; means with different letters are significantly different from each other.

Supplementary Figure 11 | The MxARF family of transcription factors in *Malus domestic*. MdARF2, 11, 12, 20, 30, 33, and 36–39 have an activation domain (AD) that is enriched in glutamine (Q), serine (S), and leucine (L). The remainder of the ARFs consist of transcriptional repressors with a repression domain (RD) that is enriched in serine (S) and in some cases proline (P). All ARFs contain a conserved DNA binding domain (DBD).

Supplementary Figure 12 | Phylogenetic analysis of miR156-targeted SPL between apple and *Arabidopsis*. Phylogenetic tree was constructed with SBP domain protein sequences. Phylogenetic tree was constructed using MEGA 4.0 software with the neighbor-joining (NJ) method and the bootstrap test replicated 1,000 times.

Supplementary Table 1 | Primers for constructing vector.

Supplementary Table 2 | PCR primers of genes in *Malus xiaojinesis*.

Supplementary Table 3 | PCR primers of genes in *Nicotiana benthamiana.*

nitric oxide production and adventitious root formation. Plant J. 71, 787–799. doi: 10.1111/j.1365-313X.2012.05032.x

Abu-Abied, M., Szwerdszarf, D., Mordehaev, I., Yaniv, Y., Levinkron, S., Rubinstein, M., et al. (2014). Gene expression profiling in juvenile and mature cuttings of Eucalyptus grandis reveals the importance of microtubule remodeling during adventitious root formation. BMC Genomics 15:826. doi: 10.1186/1471-2164-15-826


association with shoot maturation in the reproductive phase. Plant Cell Physiol. 50, 2133–2145. doi: 10.1093/pcp/pcp148


**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 Xu, Li, Hu, Wu, Wang, Xu, Zhang and Han. 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.

# Dancing with Hormones: A Current Perspective of Nitrate Signaling and Regulation in Arabidopsis

#### Peizhu Guan\*

Section of Cell and Developmental Biology, Division of Biological Sciences, University of California, San Diego, La Jolla, CA, United States

In nature and agriculture, nitrate availability is a main environmental cue for plant growth, development and stress responses. Nitrate signaling and regulation are hence at the center of communications between plant intrinsic programs and the environment. It is also well known that endogenous phytohormones play numerous critical roles in integrating extrinsic cues and intrinsic responses, regulating and refining almost all aspects of plant growth, development and stress responses. Therefore, interaction between nitrate and phytohormones, such as auxins, cytokinins, abscisic acid, gibberellins, and ethylene, is prevalent. The growing evidence indicates that biosynthesis, de-conjugation, transport, and signaling of hormones are partly controlled by nitrate signaling. Recent advances with nitrate signaling and transcriptional regulation in Arabidopsis give rise to new paradigms. Given the comprehensive nitrate transport, sensing, signaling and regulations at the level of the cell and organism, nitrate itself is a local and long-distance signal molecule, conveying N status at the whole-plant level. A direct molecular link between nitrate signaling and cell cycle progression was revealed with TEOSINTE BRANCHED1/CYCLOIDEA/PROLIFERATING CELL FACTOR1-20 (TCP20) – NIN-LIKE PROTEIN 6/7 (NLP6/7) regulatory nexus. NLPs are key regulators of nitrogen responses in plants. TCPs function as the main regulators of plant morphology and architecture, with the emerging role as integrators of plant developmental responses to the environment. By analogy with auxin being proposed as a plant morphogen, nitrate may be an environmental morphogen. The morphogen-gradient-dependent and cell-autonomous mechanisms of nitrate signaling and regulation are an integral part of cell growth and cell identification. This is especially true in root meristem growth that is regulated by intertwined nitrate, phytohormones, and glucose-TOR signaling pathways. Furthermore, the nitrate transcriptional hierarchy is emerging. Nitrate regulators in primary nitrate signaling can individually and combinatorially control downstream transcriptional networks and hormonal pathways for signal propagation and amplification. Under the new paradigms, nitrate-induced hormone metabolism and signaling deserve fresh examination. The close interplay and convergent regulation of nitrate and hormonal signaling at morphological, physiological, and molecular levels have significant effects on important agronomic traits, especially nutrient-dependent adaptive root system growth and architecture.

#### Edited by:

Chi-Kuang Wen, Shanghai Institutes for Biological Sciences (CAS), China

#### Reviewed by:

Jeremy Dale Murray, John Innes Centre (BBSRC), United Kingdom Ertao Wang, Institute of Plant Physiology and Ecology, Shanghai Institutes for Biological Sciences (CAS), China

> \*Correspondence: Peizhu Guan peguan@ucsd.edu

#### Specialty section:

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

Received: 30 June 2017 Accepted: 15 September 2017 Published: 28 September 2017

#### Citation:

Guan P (2017) Dancing with Hormones: A Current Perspective of Nitrate Signaling and Regulation in Arabidopsis. Front. Plant Sci. 8:1697. doi: 10.3389/fpls.2017.01697

Keywords: nitrate signaling, hormones, TOR signaling, cell cycle, root growth

**69**

# INTRODUCTION

fpls-08-01697 September 26, 2017 Time: 17:47 # 2

As a constituent of amino acids and nucleotides, nitrogen (N) is an essential building block for all forms of life. Not surprisingly, the mineral nutrient needed in greatest abundance by plants is N (Crawford, 1995). N availability is crucial for plant anabolism and catabolism. Despite the abundance of N (78%) in the atmosphere, the availability of fixed N in Earth's crust is scarce to such an extent that N is the quantitatively most limiting nutrient for plants (Vance, 2001; Miller and Cramer, 2004). In nature and agricultural systems, plants take up N mainly from soils in two forms, nitrate and ammonium, by roots during their postembryonic growth. Nitrate is the predominant form of N in aerobic soils where nitrification occurs rapidly (Crawford and Forde, 2002). As most soils on Earth are aerobic, nitrate is a primary N source and hence an essential nutrient for most plants.

Plants are sessile organisms that always face spatiotemporal fluctuations of nitrate concentrations in soil solution by up to four orders of magnitude due to leaching and microbial activity (Crawford, 1995; Vance, 2001; Miller and Cramer, 2004). In interacting with the environment, plants have evolved elaborate adaptive sensing, signaling and regulatory network in response to nitrate availability for survival, fitness and reproduction (Crawford, 1995; Wang et al., 2004, 2012; Remans et al., 2006a,b; Ho et al., 2009; Ruffel et al., 2011; Marchive et al., 2013; Guan et al., 2014, 2017; Vidal et al., 2015; Bellegarde et al., 2017; Liu et al., 2017). Nitrate is hence an essential nutrient as well as a crucial signal for plant growth, development, and stress responses.

Furthermore, nitrate and hormonal signaling and their interaction are of fundamental importance, underlying a plethora of plant physiological, morphological, and developmental processes in plants (Krouk et al., 2011; Nacry et al., 2013; Krapp, 2015; Krouk, 2016; O'Brien et al., 2016; Bellegarde et al., 2017). Much of our understanding of the process has been achieved so far by the molecular genetic studies using Arabidopsis thaliana as a model. The accumulating evidence indicates that biosynthesis, de-conjugation, degradation, transport, and signaling of hormones are partly controlled by nitrate signaling, so that the environmental and internal signaling pathways are seamlessly integrated.

Much of the literature on the interaction between nitrate and hormonal signaling pathways has been focused on hormonal control of nitrate metabolism and signaling (Kiba et al., 2010), while this review focuses more on the other side of the coin – nitrate signaling control of hormone metabolism and signaling (Krouk, 2016). The latest findings in nitrate research revealed novel molecular links between N and plant development (Ristova et al., 2016; Guan et al., 2017; Liu et al., 2017) shed new light on nitrate-hormone interconnections. In the context of agronomy, nitrate- and hormone-regulated lateral root (LR) growth and development are among the main determinants of root plasticity in response to nitrate availability and of nitrogen use efficiency (NUE) in crops. Therefore, the interplay and convergent regulation of nitrate and hormones in LR, which is highlighted in this review, is of agronomic importance.

### NITRATE SIGNALING, UNCOUPLED FROM NITRATE METABOLISM, ACTS AT LOCAL AND WHOLE-PLANT LEVELS

Nitrate is taken up by roots then transported into root and shoot cells mainly via the NITRATE TRANSPORTER 1 (NRT1) and NITRATE TRANSPORTER 2 (NRT2) family of nitrate transporters (Wang et al., 2012). Once inside the cells, nitrate is reduced to nitrite by NITRATE REDUCTASE (NR) in the cytosol. In A. thaliana, two NR enzymes, NIA1 and NIA2, are responsible for 10 and 90% of the total NR activity in seedlings, respectively (Cheng et al., 1988; Wilkinson and Crawford, 1991, 1993). Nitrite is then reduced to ammonium by NITRITE REDUCTASE (NiR) in plastids, where ammonium is in turn assimilated into glutamine (Gln) (Crawford, 1995). Notably, with external nitrate concentration increases, nitrate assimilation into amino acids in higher plants is increasingly achieved in shoots, which become the main sites of NR activity (Andrews, 1986). In addition to amino acids production, nitrate metabolism supports plant use of light, CO<sup>2</sup> and water to produce sugars and organic acids. In spite of its importance, nitrate assimilation is energetically costly, demanding intensive use of adenosine triphosphate (ATP), reducing equivalents, and C skeletons (Nunes-Nesi et al., 2010). Therefore, nitrate assimilation is subject to restraint or stimulation by resource availability in the environment and the demands of plant growth and development.

Fundamentally, nitrate signaling is uncoupled from, but executes tight control over, nitrate metabolism (Wang et al., 2000, 2003, 2004, 2007, 2009; Scheible et al., 2004; Muños et al., 2004; Ho et al., 2009; Nunes-Nesi et al., 2010; Konishi and Yanagisawa, 2013; Marchive et al., 2013; Bouguyon et al., 2015; Guan et al., 2017; Liu et al., 2017). Most plants have been wired to perceive nitrate, but not its downstream metabolites, e.g., ammonium and Gln, as the principal source of N offered by the environment. Among the earliest and most convincing evidence is that revealed by transcriptome analysis in NR-null (nia1 nia2) mutants, numerous genes, including the key nitrate assimilatory genes, directly respond to nitrate independent of nitrate reduction (Wang et al., 2004). Hence, nitrate is a signal molecule of paramount importance, from stimulating germination, to sustaining substantial postembryonic growth, to controlling developmental phase transitions (Alboresi et al., 2005; Chopin et al., 2007; Fan et al., 2009; Nunes-Nesi et al., 2010; Vidal et al., 2014b; Yan et al., 2016; Guan et al., 2017; Liu et al., 2017). In response to soil nitrate availability, plants reprogram genomewide short-term and long-term gene expression at the wholeplant level and promote adaptive regulation of organogenesis, involving root system architecture, root and shoot growth, leaf expansion, flowering time, stomata opening, defense responses, etc. (Forde, 2002; Walch-Liu et al., 2005; Krouk et al., 2010a; Castro-Marín et al., 2011; Ruffel et al., 2011; Fagard et al., 2014; Guan et al., 2014, 2017; Liu et al., 2017).

When exposed to nitrate, the first, and also one of the foremost nitrate responses in plants that has been extensively studied is the primary nitrate response (PNR) (Redinbaugh and Campbell, 1991; Wang et al., 2000, 2003, 2004, 2007; Scheible et al., 2004; Medici and Krouk, 2014; Ristova et al., 2016). The PNR is rapid, independent of de novo protein synthesis, and responsive to nitrate concentrations as low as 100 nM in roots and 250 µM in shoots of pre-starved Arabidopsis seedlings. The PNR affects the expression of 1,596 genes at the significance level in wild-type (WT) plants. Among those genes, 595 genes in both roots and shoots directly responded to nitrate, which is confirmed in the NR-null mutant (nia1 nia2) (Wang et al., 2004).

Indeed, the nitrate response is a whole-plant response and nitrate itself can function as both a local and long-distance systemic signal (Wang et al., 2004; Ruffel et al., 2011). The phenomenon was initially revealed by microarray analysis of nitrate-regulated gene expression in roots and shoots of the seedlings that were grown hydroponically. When treated with 0.25 mM nitrate for only 20 min, the roots have a much broader response than shoots in terms of the number of genes being affected (Wang et al., 2000). However, with sufficient nitrate concentrations and sufficient time (5 mM nitrate for 2 h) for nitrate transport facilitation (but not for reduction), the shoot genes in the NR-null mutant can be as responsive to nitrate as root genes, although the two groups of genes are still organ-specific (Wang et al., 2004). The PNR genes in roots and shoots are selectively targeted, with an overall concentration on energy and metabolism, including glycolysis and gluconeogenesis, amino acid metabolism, nitrogen and sulfur utilization, and transport facilitation (Wang et al., 2004); nevertheless, there are also a large group of genes for signaling and regulatory components intimately related to the twocomponent systems (TCS), calcium and sugar transport, auxin, cytokinins, and abscisic acid (ABA) metabolism and signaling, and so on (Wang et al., 2000, 2003, 2004; Ristova et al., 2016; Liu et al., 2017).

Nitrate signaling and regulation underlie the genome-wide expression reprogramming in nitrate responses, which leads to activation and adaptation of N-regulated metabolism and development. The process involves membrane and cytosol sensing, signal transduction, transcription factors (TFs), the interactions of TFs, and nitrate-responsive DNA regulatory elements (Ho et al., 2009; Wang et al., 2010; Guan et al., 2014, 2017; Liu et al., 2017). On the other hand, the transcriptome analysis revealed that numerous pathways and processes, particularly hormone signaling pathways, have interaction with and depend on nitrate signaling (Krouk, 2016; O'Brien et al., 2016; Ristova et al., 2016) for N status before making collective decisions in growth, development and stress responses.

#### NITRATE TRANSPORT, SIGNALING AND REGULATION AT A GLANCE

In plants, the first identified and characterized nitrate transporter is known as chlorate resistant 1 (CHL1) or AtNRT1.1 or AtNPF6.3 (Tsay et al., 1993). It is also the first plant member of NITRATE TRANSPORTER 1/PEPTIDE TRANSPORTER (NRT1/PTR) Family (also named NPF) discovered. The NRT1/PTR Family (NPF) comprises of membrane proteins ubiquitously found across all major kingdoms of life and sharing sequence homology. In bacteria, fungi, animals and plants, the family members were found to transport dipeptides (Leran et al., 2014; von Wittgenstein et al., 2014). In higher plants, there are at least four families of nitrate transporters. Besides NRT1/PTRs, the other three families are: NITRATE TRANSPORTER 2 (NRT2), CHLORIDE CHANNEL (CLC) a/b, and SLOW ANION CHANNEL-ASSOCIATED 1 HOMOLOG 3 (SLAH3) (Wang et al., 2012; Krapp et al., 2014).

The nitrate transporters contribute to numerous physiological functions involved in different stages and processes of nitrate distribution, assimilation, signaling, and osmotic regulation. They are individually critical, such as NRT1.1 (CHL1/NPF6.3), NRT1.2 (NPF4.6/AIT1), NRT2.1, NRT2.2, NRT2.4, and NRT2.5 in nitrate uptake from soil (Tsay et al., 1993; Huang et al., 1999; Cerezo et al., 2001; Kiba et al., 2012; Lezhneva et al., 2014); NAXT1 (NPF2.7) in nitrate efflux (Segonzac et al., 2007); NRT1.5 (NPF7.3), NRT1.8 (NPF7.2), and NRT1.9 (NPF2.9) in root-to-shoot xylem translocation, a primary route of longdistance nitrate transport driven by transpiration, which is accompanied by shoot-to-root xylem and phloem transport of nitrate (Lin et al., 2008; Li et al., 2010; Wang and Tsay, 2011); NRT1.7 (NPF2.13), NRT1.9 (NPF2.9), NRT2.4 and NRT2.5 in source-to-sink phloem remobilization, a secondary route of long-distance nitrate transport driven by osmotic gradients in both roots and shoots (Fan et al., 2009; Wang and Tsay, 2011; Kiba et al., 2012; Lezhneva et al., 2014); NRT1.4 (NPF6.2) in nitrate petiole storage (Chiu et al., 2004); CLCa/b in nitrate accumulation in vacuoles (De Angeli et al., 2006; von der Fecht-Bartenbach et al., 2010); NRT1.6 (NPF2.12) and NRT2.7 in nitrate accumulation in seeds (Chopin et al., 2007; Almagro et al., 2008); NRT1.1 (CHL1/NPF6.3) and SLAH3 in stomatal closure and opening (Guo et al., 2003; Geiger et al., 2011); NRT1.11 (NPF1.2) and NRT1.12 (NPF1.1) in xylem-to-phloem transfer for redistributing nitrate (Hsu and Tsay, 2013); NPF2.3 in nitrate translocation to shoots for acclimation to salt stress (Taochy et al., 2015); and NPF5.5 in embryo N accumulation (Leran et al., 2015) (**Figure 1**). All transport proteins are localized at the plasma membrane, except that NRT2.7 and CLCa/b are localized at the tonoplast (Wang et al., 2012; O'Brien et al., 2016).

As a result of dramatic fluctuations of nitrate concentrations in soil, plants have evolved two uptake systems: low-affinity transport system (LATS) for high external nitrate concentration (>0.5 mM) and high-affinity transport system (HATS) for low nitrate concentration (<0.5 mM), into which most nitrate transporters in roots and shoots can be categorized (Crawford and Glass, 1998; Forde, 2000; Miller et al., 2007). All known NPF transporters, except NRT1.1 (CHL1/NPF6.3), solely belong to LATS. Even though a majority of low- and high-affinity transporters are inducible, the two exceptions are NRT2.5 in HATS and NRT1.2 (NPF4.6/AIT1) in LATS, which are constitutive nitrate transporters (Huang et al., 1999; Lezhneva et al., 2014; Kotur and Glass, 2015). This gave rise to the four subsystems: the constitutive high-affinity system (cHATS), the

inducible high-affinity system (iHATS), the constitutive lowaffinity system (cLATS), and the inducible low-affinity system (iLATS) (Crawford and Glass, 1998; Forde, 2000; Tsay et al., 2007). Much attention has been given to the nitrate transporters that play crucial roles of mineral uptake in roots, the principal nutrient absorbing organs. In Arabidopsis roots, LATS involves NRT1.1 (CHL1/NPF6.3) and NRT1.2 (NPF4.6/AIT1) (Tsay et al., 1993; Huang et al., 1999), and HATS involves NRT1.1 (CHL1/NPF6.3), NRT2.1, NRT2.2, NRT2.4, and NRT2.5 (Wang et al., 1998; Liu et al., 1999; Cerezo et al., 2001; Li et al., 2007; Kiba et al., 2012; Lezhneva et al., 2014; Kotur and Glass, 2015). The NRT1s and NRT2s are both proton-coupled transporters. The interaction with NAR2 is critical for transport capacity of most high-affinity NRT2s in plants (Kotur et al., 2012; Kotur and Glass, 2015). When root epidermal cells are exposed to nitrate, the H+-ATPase in the plasma membrane pumps protons out of the cell, producing pH and electrical (19) gradients, which potentially provides required energy to both LATS and HATS for co-transporting two or more protons per nitrate into the cell, a process also involving membrane depolarization. Both nitrate influx and efflux could be mediated by the proton-coupled mechanism (Crawford, 1995; Forde, 2000; Tsay et al., 2007; Wang et al., 2012).

Among the nitrate transporters so far characterized in Arabidopsis, NRT1.1 (CHL1/NPF6.3) is the only dual-affinity transporter (Wang et al., 1998; Liu et al., 1999; Liu and Tsay, 2003), although dual-affinity transport activity was also found in the potassium transporter AtKUP (Fu and Luan, 1998; Kim et al., 1998) and the nitrate transporter MtNRT1.3 (Morère-Le Paven et al., 2011). Moreover, NRT1.1 (CHL1/NPF6.3) mediates the expression of NRT2.1 and NRT3.1/ NAR2.1, depending on nitrate/ammonium concentrations. The process is the critical regulation of HATS, which is also under the feedback repression by N metabolites (Muños et al., 2004; Krouk et al., 2006). Tightly regulated by nitrate signaling, NRT1.1 (CHL1/NPF6.3) and NRT2.1 are most transcriptionally abundant (Wang et al., 2012). Beyond being transporters, they are also deeply involved in sensing and activating downstream gene expression, including the PNR and nitrate-regulated root development (Little et al., 2005; Remans et al., 2006a,b; Ho et al., 2009; Gojon et al., 2011).

Nitrate uptake, sensing and signaling are regulated by multiple mechanisms, which mainly include nitrate availability, feedback repression by N status, stimulation by photosynthesis, and hormone signaling (Forde, 2000; Wang et al., 2000, 2003, 2004, 2007; Nacry et al., 2013; Krouk, 2016; O'Brien et al., 2016). Significantly, primary nitrate signaling in response to nitrate availability is amplified and propagated overriding feedback constraints from downstream N metabolites and low sugars (Wang et al., 2004, 2007; Nunes-Nesi et al., 2010; Guan et al., 2017; Liu et al., 2017). The homeostasis of nitrate concentration, calcium, pH, redox, and phosphate in the cytosol is delicately regulated and maintained in all cells, for which nitrate availability is a critical determinant. For instance, the steady-state cytosolic nitrate concentrations in barley root cells were recorded between 3 and 5 mM, which corresponds with a potential "optimal" range of exogenous nitrate concentrations (1–10 mM) for plants (Miller and Smith, 1996, 2008), and cytosolic pH varies from 7.3 to 8 in plants (Martinière et al., 2013). Among the primary signaling roles of nitrate is that the disruption of cytosolic ionic environment resulting from nitrate availability/unavailability and nitrate concentrations outside of the potential "optimal" range can trigger multiple downstream cascades of N-regulated events (Champigny and Foyer, 1992; Fan et al., 2016; Liu et al., 2017; Xuan et al., 2017). The nitrate-induced disruptions are rapidly captured by the evolutionarily conserved calcium signaling. The patterns of calcium level increase are elicited by nitrate in a context-dependent manner and calcium is specifically involved in the nitrate response and signaling as a modulator and/or a second messenger (Wang et al., 2000, 2003, 2004; Riveras et al., 2015; Liu et al., 2017).

In response to low exogenous nitrate concentration (<1 mM), NRT1.1T101 phosphorylation is switched on, involving the CBLinteracting protein kinase CIPK23, a plant-specific calcium sensor (Ho et al., 2009; Hu et al., 2009). Through its dual-affinity binding and a phosphorylation-controlled dimerization switch between the two affinities, NRT1.1 (CHL1/NPF6.3) functions as a nitrate membrane sensor required for the PNR and other nitrate responses, independent of its uptake function as a transporter (Remans et al., 2006b; Ho et al., 2009; Wang et al., 2009). NRT1.1 (CHL1/NPF6.3) is regarded as the first transceptor discovered in plants by analogy with yeast nutrient transceptors (Ho et al., 2009; Gojon et al., 2011). It is a dose-dependent master controller of multiple signaling mechanisms capable of

responding to a wide range of soil nitrate levels (Ho et al., 2009; Krouk et al., 2010a,b; Bouguyon et al., 2015; Leran et al., 2015). Nevertheless, prolonged N starvation rendered the nitrate response NRT1.1 (CHL1/NPF6.3) independent, suggesting alternative or redundant nitrate membrane sensing systems must be present (Wang et al., 2009). NRT2.1 has been proposed as such a candidate, which shows an uncoupled dual (uptake/signaling) function in root growth in response to nitrate availability (Remans et al., 2006a).

Although not capable of evoking the PNR per se, subgroup III calcium-dependent protein kinases (CPKs), CPK10, CPK30, and CPK32 are also required for rapid nitrate-induced cellular and metabolic responses, and nitrate-regulated root and shoot growth (Liu et al., 2017). In response to nitrate availability, Ca2+-sensor CPKs translocate to the nucleus, where the phosphorylation of NLP7S205 by CPK10 is responsible for nitrate-stimulated nuclear retention of NLP7 (Liu et al., 2017). The widely overlapped transcriptomic and phenotypic defects in icpk and nlp7 mutants further substantiate the existence of the nitrate–CPK–NLP signaling-regulatory pathway potentially activating the downstream nitrate transcriptional network for signal amplification (Liu et al., 2017). Nevertheless, multiple sensing and signaling mechanisms with redundancy are synergically required for the context-dependent broad-ranged nitrate responses (Wang et al., 2009; Guan et al., 2017; Liu et al., 2017).

Furthermore, nitrate-responsive DNA regulatory elements (Girin et al., 2007; Konishi and Yanagisawa, 2010; Wang et al., 2010) and transcriptional regulators, including ARABIDOPSIS NITRATE REGULATED 1 (ANR1), NLP6, NLP7, LOB DOMAIN-CONTAINING PROTEIN 37/38/39 (LBD37/LBD38/LBD39), SQUAMOSA PROMOTER BINDING PROTEINLIKE 9 (SPL9), HIGH NITROGEN INSENSITIVE 9 (HNI9), NAC DOMAIN-CONTAINING PROTEIN 4 (NAC4), BASIC LEUCINE-ZIPPER 1 (bZIP1), TGACG MOTIF-BINDING FACTOR 1/4 (TGA1/TGA4), TCP20, HYPERSENSITIVE TO LOW PI-ELICITED PRIMARY ROOT SHORTENING 1 (HRS1), NITRATE REGULATORY GENE2 (NRG2), BRIC-A-BRAC/TRAMTRACK/BROAD-COMPLEX 1/2 (BT1/BT2), and NLP8; (Zhang and Forde, 1998; Castaings et al., 2009; Rubin et al., 2009; Krouk et al., 2010c; Widiez et al., 2011; Konishi and Yanagisawa, 2013; Marchive et al., 2013; Alvarez et al., 2014; Guan et al., 2014, 2017; Para et al., 2014; Vidal et al., 2014b; Medici et al., 2015; Araus et al., 2016; Xu et al., 2016; Yan et al., 2016) have been identified. Some TFs, e.g., ANR1, LBD37/38/39, SPL9, NAC4, bZIP1, TGA1/TGA4, and HRS1, are nitrate-responsive while NLP6, NLP7, HNI9, TCP20, NRG2, BT1/BT2, and NLP8 are not. Among the key regulators, direct interactions of NLP6, NLP7, TGA1, bZIP1, TCP20, HRS1, and NLP8 with target gene promoters have been verified. Intriguingly, multiple transcriptional mechanisms are involved in nitrate responses. bZIP1, following a hit-and-run transcriptional model, transiently bind to its target gene promoters to enable a rapid and dynamic N-signal propagation (Para et al., 2014). The propagation of nitrate signaling into metabolism and stress response are observed in the clusters of genes potentially targeted by TGA1/TGA4 (Alvarez et al., 2014). The upregulation of gene expression of TGA1 is specifically dependent on a phospholipase C (PLC)-calcium signaling pathway downstream of NRT1.1 (CHL1/NPF6.3) (Riveras et al., 2015). In response to nitrate, NLP7 was found to be retained in nucleus via a CPK-dependent phosphorylation to promote a genome-wide gene expression regulation, giving rise to the concept of primary nitrate signaling (Castaings et al., 2009; Marchive et al., 2013; Liu et al., 2017).

The primary nitrate signaling controlled by non-nitrateresponsive TFs that are not regulated by nitrate at the transcriptional level seems to regulate the proper level of expression of downstream nitrate-responsive TFs (**Figure 2**) (Marchive et al., 2013; Guan et al., 2014, 2017; Xu et al., 2016; Bellegarde et al., 2017; Liu et al., 2017). Moreover, several nonnitrate-responsive regulators, NLP6, NLP7, TCP20, NRG2, and BT1/BT2 were shown to regulate the expression of the sentinel PNR genes, such as NRT1.1 (CHL1/NPF6.3), NRT2.1 and NIA1 at different nitrate concentrations (Guan et al., 2014, 2017; Araus et al., 2016; Xu et al., 2016). In contrast with NLP6, NLP7, and TCP20, the specific molecular function of NRG2 is not known; nevertheless, NRG2 could be also involved in primary nitrate signaling based on its regulatory roles in the PNR and their close ties to nitrate levels (Xu et al., 2016). It is well known that the protein–protein interactions define the specificity of signal transduction and transcriptional regulation (Lamb and McKnight, 1991; Pawson and Nash, 2000). The interaction between NLP7 and NRG2 was reported; however, its function in N signaling and regulation remains to be investigated (Xu et al., 2016). Most recently, a central regulatory nexus in response to nitrate availability, involving TCP20-NLP6/7 interactions, was identified (Guan et al., 2017). Centered on this regulatory nexus, a primitive transcriptional regulatory hierarchy is emerging (**Figure 2**) (Bellegarde et al., 2017; Guan et al., 2017; Liu et al., 2017).

TCP20 and NLP6/7 proteins are constitutively and ubiquitously expressed in plants (Winter et al., 2007; Castaings et al., 2009; Hervé et al., 2009; Danisman et al., 2012; Marchive et al., 2013; Chardin et al., 2014). TCP20 and NLP6/7 belong to two ancient gene families, the protein sequences of which contain multiple, deeply conserved motifs in plants (Cubas et al., 1999; Schauser et al., 2005; Martín-Trillo and Cubas, 2010). The NIN-like protein gene NLP7 was identified through its sequence similarity to the nitrate regulatory gene NIT2 in Chlamydomonas (Camargo et al., 2007; Castaings et al., 2009). NLPs and RWP-RK domain proteins, whose founding members are the nodulation-specific NIN proteins, constitute the RWP-RK family. The origin of NLPs predates the monocot/eudicot divide. The RWP-RK proteins are key regulators of N responses in plants (Schauser et al., 2005; Chardin et al., 2014). With their origin predating the emergence of land plants, TCPs are plant-specific TFs that function as the main regulators of plant morphology and architecture, mainly because of direct transcriptional control of cell cycle genes and regulation of hormone activity by class I TCPs, (Martín-Trillo and Cubas, 2010; Manassero et al., 2013; Nicolas and Cubas, 2016). Intriguingly, distinct but overlapping binding sites between the classes I and II TCPs indicate either coordinate or competitive regulation of transcription. For example, TCP20 and TCP9 (class I) and TCP4 (class II) were found to act antagonistically on jasmonic acid (JA) metabolism

initially confirmed pathways.

and leaf development (Danisman et al., 2012). The delicate balance of transcriptional regulation between the two classes of TCPs in the distal meristem boundary zone where cell division transitions into expansion and differentiation was proposed as a key control of organ growth rate, ultimately shaping organs (Li et al., 2005). In inflorescence shoot apex, gibberellin (GA) regulated DELLA-TCP interactions control plant height (Daviere et al., 2014).

TCP20 and NLP6/NLP7 bind to adjacent sites in the upstream promoter region of the NR gene, NIA1, and physically interact under continuous nitrate and N-starvation conditions. The subcellular localization and nuclear accumulation of single NLP6/7 and TCP20-NLP6/7 complexes depend on nitrate availability (Guan et al., 2017). The regulatory interactions could perceive nitrate availability via multiple upstream sensing and signal transduction in the cell membrane and cytosol, such as by the nitrate transceptor NRT1.1 (CHL1/NPF6.3), nitrate–CPK–NLP pathway (Liu et al., 2017). In the presence of nitrate, both NLP6 and NLP7 are retained in the nucleus. The nitrate-dependent nuclear retention in response to nitrate availability occurs within minutes. NLP6 and NLP7 thereby function as two partially redundant master regulators for

rapid nitrate signaling and responses and growth (Marchive et al., 2013; Guan et al., 2017). The severe growth defects in nlp6 nlp7 double mutants with nitrate as the sole N source, and in NR-null (nia1 nia2) mutants are comparable. Putative binding sites of NLP6 and NLP7 were found in the CYCB1;1 promoter region. The defective CYCB1;1 expression was also observed in nlp6 nlp7 double mutants, which, however, did not satisfy the significance test when total roots were measured, but would likely pass the test if measuring only root tips. Under N starvation, TCP20-NLP6/7 heterodimers accumulate in the nucleus. The transcriptional complexes not only bind to and upregulate sentinel nitrate-responsive genes for transport, assimilation and signaling, but also bind to and downregulate CYCB1;1, a division marker of apical meristems, for the control of G2/M transition in cell cycle progression, supporting root apical meristem (RAM) growth (Li et al., 2005; Guan et al., 2014, 2017). The direct molecular link between nitrate availability and G2/M cell cycle progression in the RAM is crucial for plant adaptive postembryonic development that depends on meristems. Genome-wide transcriptional profiling further revealed that potential NLP7 targeted genes include other cell cycle genes, such as MCM2/3, ETG1, CDC20.2, and ORC (Liu et al., 2017). Therefore, TCP20-NLP6/7 complexes have a much wider presence in cell cycle regulation (**Figure 2**).

### NITRATE, HORMONES, AND GLUCOSE-TOR IN ROOT APICAL MERISTEM GROWTH

TCP20 regulates mitotic cyclin gene CYCB1;1 and putatively ribosomal protein genes by binding to the GCCCR motif in their promoters in vitro and in vivo, which was proposed to be a mechanism for regulation of the cell cycle and cell growth at a synchronized rate (Li et al., 2005; Guan et al., 2017). Under N starvation, total RNA, a majority of which is ribosomal RNA (rRNA), recorded in seedling roots of tcp20, nlp6, nlp7, tcp20 nlp6, tcp20 nlp7, or nlp6 nlp7 mutants was only approximately 50% percent of the total RNA recorded in seedling roots of WT, when all were measured on same total fresh root weight, suggesting defective ribosome biogenesis and cell growth (Guan et al., 2017, unpublished results). It implies that these proteins may function as a complex in regulating ribosomal protein genes to stimulate ribosome biogenesis. The distinct nitrate-dependent TCP20-NLP6/7 interactions and their regulation under two conditions also support that nitrate signaling is an integral part of the synchronized the cell cycle and cell growth in meristem, likely via controlling cell division.

A master regulator of cytoplasmic growth is TARGET OF RAPAMYCIN (TOR), which is at the interface of growth and nutrient availability in unicellular organisms and through acquiring additional roles, becomes a central controller of organism growth, and energy and nutrient homeostasis in multicellular organisms (Zoncu et al., 2011; Sablowski and Dornelas, 2014). (m)TORC1 in both mammalian and yeast cells critically regulates and maintain the robust transcription of genes involved in ribosome biogenesis along with translation initiation and nutrient import under favorable growth conditions (Powers and Walter, 1999; Mayer et al., 2004; Xiao and Grove, 2009). That is in alignment with the roles of single NLP6 and NLP7 proteins in the presence of nitrate (Guan et al., 2017). TOR signaling is suppressed under stress conditions, leading to cell cycle arrest so as to prevent uncontrolled cell growth. However, in the face of constant nutrient stress in nature, the sustained cell growth in RAMs via transcriptionally repression of CYCB1;1 by TCP20-NLP6/7 complexes are crucial for plants as sessile organisms. It could enable continuous nutrient acquisition in soil and adequate remobilization within plants, which are key factors for survival by maximizing NUE (Masclaux-Daubresse et al., 2010). Intriguingly, sustained RAM growth was also physiologically observed under phosphate deficiency, which is at the expense of photosynthesis (Kang et al., 2014). Moreover, during cell expansion that predominates in post-mitotic cells, the TCP20-NLP6/7 also play critical regulatory roles in cell wall biogenesis and modification (Hervé et al., 2009; Danisman et al., 2012; Karve et al., 2016). The growth of organs and whole plants depends on system-wide synchronized coordination of nutrient availability, cell growth and cell-cycle progression, for which the functions of TCP20-NLP6/7 interactions are central.

Interestingly, the TCP20-NLP6/7 regulatory nexus employs the type I/II Phox and Bem1p (PB1) domains of NLP6/7, a protein-interaction module conserved in animals, fungi, amoebas, and plants (Sumimoto et al., 2007; Chardin et al., 2014; Guan et al., 2017). In animals, PB1 domains are employed for activation of mTOR1 by amino acids and organizing growth factors (Linares et al., 2015). In plants, the type I/II PB1 domains are also employed in the homo- and hetero-oligomerization of auxin response factor (ARF) TFs and auxin/indole 3 acetic acid (Aux/IAA) repressor proteins (Boer et al., 2014; Korasick et al., 2014; Guilfoyle, 2015). This indicates that the TCP20-NLP6/7 interactions are part of a more general pattern used for nutrient–growth signaling, cellular homeostasis, and morphogenetic signaling in both plants and animals (Guan et al., 2017). However, TCP20-NLP6/7 regulatory nexus compares with recently discovered plant glucose-TOR signaling in that both of them are not well framed in conventional non-transcriptional mechanisms of mammalian TOR, which indirectly modulate limited messenger RNAs and target genes via 4E-BP1 and S6K1 phosphorylation (Xiong et al., 2013; Guan et al., 2017). Instead, they are two central transcriptional machineries controlling a broad range of nutrient–growth gene expression at the wholeplant level (Xiong et al., 2013; Guan et al., 2017).

For plant growth and development, nitrate, sugars, and the phytohormones, particularly auxin and CKs, are of vital importance. They are integral parts of the regulation of the dynamic balances of cell division and cell differentiation, which controls organ shape and size. Especially in root growth, they are intricately coordinated in controlling the balance between the cell cycle and cell growth (Xiong et al., 2013; Sablowski and Dornelas, 2014; Barrada et al., 2015; Guan et al., 2017). CK and auxin have long been implicated in regulating the components of the cell cycle (Himanen et al., 2002; Perrot-Rechenmann, 2010; Schaller et al., 2014). In Arabidopsis, RAM growth is under the antagonistic effects of auxin and CK, which mediate cell division

at the apical meristem and cell differentiation at the transition zone, respectively (Ioio et al., 2007). For LR formation consisting of pericycle activation and meristem establishment, auxin is a dominant regulator (Himanen et al., 2002; Fukaki and Tasaka, 2009). CK was also reported to repress LR initiation and promote LR elongation (Rani Debi et al., 2005; Laplaze et al., 2007). Recently, the glucose-TOR signaling pathway was reported to control the G1/S transition by an unconventional mechanism of transcriptional regulation. TOR kinase directly phosphorylates and activates E2Fa, which in turn transcriptionally activates S-phase genes in response to glucose and sucrose signaling, which is independent of S6K, RBR or translational control (Xiong et al., 2013). The glucose-TOR signaling for the glycolysismitochondrial energy relays is indispensable for RAM growth.

Intriguingly, the induction level of primary auxin and CK marker genes and spatial expression of patterning genes were intact in the presence of rapamycin or antimycin A (AMA) in WT (in both cases TOR activity is inhibited) and in tor seedlings. Upon glucose starvation, neither RAM cell number nor RAM length were significantly reduced, which is overtly opposite to what was observed in RAM upon N starvation, where arrest at the G2/M transition occurred (Xiong et al., 2013; Guan et al., 2017). Under N starvation, significant reductions of LR number per plant were displayed across the mutant lines, nlp6, nlp7, tcp20 nlp6, tcp20 nlp7, and nlp6 nlp7 (Guan et al., 2017, unpublished results). It is consistent with that in Arabidopsis, the initial xylem pole pericycle cell divisions during first LR initiation event are accompanied with regulation of G2/M transition (Himanen et al., 2004; Malamy, 2005). Auxin and CK signaling and stem cell niche maintenance seems not to rely on sugar signaling and metabolism pathways (Xiong et al., 2013). The accumulating evidence as reviewed here suggests that nitrate signaling and metabolism is crucial for hormone signaling and maintenance of stem cell niche integrity.

TORC1 and TORC2 complexes, and a large part of the evolutionary "core" of TOR pathway likely originated in or before the last eukaryotic common ancestor (LECA) that gave rise to all currently known living eukaryotic species (van Dam et al., 2011). Although the two TOR complexes are found in other major lineages of eukaryotes, plants possess only TORC1 (van Dam et al., 2011). In plant TOR signaling as has been uncovered so far, the integration of N status with cell growth and the cell cycle, which is the main component in TOR1 pathways of both yeast and mammals and required by virtually all eukaryotic cells, is still missing. It is mainly because the upstream regulators that directly sense N availability and organize growth factors are unknown. Plants lack orthologs of small guanosine 59-triphosphatases (GTPases): Ras homolog enriched in brain (RHEB), and Rag guanosine 59-triphosphatases (RAGs) (Xiong and Sheen, 2014). TOR signaling is highly conserved; however, it is adequately flexible to include new signals and mechanisms in response to environmental challenges in the evolution of animals and plants (van Dam et al., 2011). For example, insulin signaling is an animal-specific addition to the pathway to use sugar (glucose) for cellular growth in a multicellular environment. Recently, the plant-specific small GTPase Rho-related protein 2 (ROP2) (Li et al., 2001) was demonstrated to transduce lightauxin signal to activate TOR by direct interaction (Li et al., 2017). Interestingly, TOR kinase can also be activated by nitrate and amino acids via S6K1 T449 phosphorylation by unknown mechanisms in Arabidopsis seedlings (Xiong and Sheen, 2015). Could TCP20-NLP6/7 complexes function in parallel with plant TOR or upstream of plant TOR for N signaling?

The glucose-TOR signaling imposes an overall limit on plant organ growth, specifically on organ size (Sablowski and Dornelas, 2014). By contrast, nitrate and its close interplay with hormones have a determining effect on patterning tissues and shaping organs (Crawford, 1995; Stitt, 1999; Forde, 2002; Miller et al., 2008; Ruffel et al., 2011; Nacry et al., 2013; Guan et al., 2014, 2017; Mounier et al., 2014; Sablowski and Dornelas, 2014; Krouk, 2016; Ristova et al., 2016). This is a strong indication of interaction and integration of nitrate and hormonal signaling pathways in plant growth and development. One thing is clear: RAM growth that takes place underground is particularly under convergent regulation of nitrate and hormone signaling. Notably, the distinct mechanism and coordination between hormone and glucose-TOR energy signaling are involved in regulation of shoot apical meristem (SAM). Light, a main aboveground environmental cue, is required for producing auxin, which in turn can activate downstream ROP2-TOR-E2Fa/b signaling pathway and promoting SAM growth (Li et al., 2017). The involvement of nitrate signaling in SAM growth should also be investigated in the integrated context.

#### NITRATE IN AUXIN BIOSYNTHESIS, TRANSPORT, SIGNALING, AND RESPONSES

Auxins are a class of essential phytohormones involved in tailoring plant growth and morphology to environmental conditions (Vanneste and Friml, 2009; Overvoorde et al., 2010). As the main endogenous auxin in most plants, indole-3 acetic acid (IAA) is the most potent native auxin, regulating almost every aspect of plant life, i.e., growth, development, and biotic and abiotic stress responses (Woodward and Bartel, 2005; Zhao, 2012). IAA biosynthesis is defined by a two-step complete pathway where indole-3-pyruvate (IPA) is converted from tryptophan (Trp) by the TAA family of amino transferases, before IPA being converted to IAA by the YUC family of flavin monooxygenases (Zhao, 2012). It is generally accepted that auxin regulation of plant morphogenesis relies on its tissue-specific concentration gradients collectively formed by the processes of auxin biosynthesis, conjugation, degradation and transport (Normanly, 2010). Recent studies also showed that localized auxin biosynthesis is indispensable in many developmental processes including embryogenesis, seedling growth, root development, vascular patterning, phyllotaxis and flower development (Cheng et al., 2006, 2007; Overvoorde et al., 2010; Pinon et al., 2013; Chen et al., 2014). In Arabidopsis roots, defective localized auxin biosynthesis cannot be replenished by auxin transported from shoots, indicating that shoot-derived

auxin alone is not sufficient for supporting root elongation and root gravitropic responses (Chen et al., 2014).

Notably, N availability directly regulates TAA1 and its close homologs TAR1 and TAR2 in the first step of IAA biosynthesis (Ma et al., 2014). In Arabidopsis, the expression levels of the three genes in roots and shoots under high N conditions (3 mM NH4NO3) were compared with those under low N conditions (0.1 mM NH4NO3) after 7 days treatment (Ma et al., 2014). The expression of TAR2 was significantly induced by low N in roots, where the expression of TAA1 was moderately induced. The expression of TAA1 and TAR1 were both repressed in shoots. TAR2 was expressed in the root pericycle and vasculature of root maturation zone near the root tip. The tar2 mutants showed repressed auxin accumulation in LR primordia and reduced LR primordia emergence and numbers under low N conditions (Ma et al., 2014).

However, with ammonium in the media, it was difficult to identify which N source, nitrate or ammonium, or both of them, could be responsible for the gene expression regulation. The recent genome-wide transcriptional profiling showed that TAR2 and PIN-FORMED PROTEIN 7 (PIN7) are among the top NLP7 activated genes, together with nitrate assimilation genes such as NiR, NIA1, FNR2, and NRT2.1 (Marchive et al., 2013; O'Malley et al., 2016; Liu et al., 2017). Plasma membrane-localized PIN7 is a main auxin efflux carrier protein (Friml et al., 2003). This evidence substantiates that the auxin biosynthesis and transport is transcriptionally regulated by nitrate in roots. It further suggests that the TCP20-NLP6/7 complexes function upstream of auxin-ROP2-TOR-E2Fa/b signaling pathway (**Figure 2**).

Auxin transport has been postulated to be a major factor determining intercellular and intracellular distributions of IAA. In plant cells, transporters and their asymmetrical localization are required for suggested directional efflux of anionic auxins and formation of polar flow. Therefore, NLP7-regulated auxin efflux via PIN7 could contribute to regional auxin gradient and local maxima to establish and maintain a root primordium and determine LR numbers (Overvoorde et al., 2010). Auxin was also shown to be transported away from the LR primordium by NRT1.1 (CHL1/NPF6.3) at low nitrate conditions (<0.5 mM), therefore preventing the growth of pre-emerged LR primordia and young LRs; when nitrate being plentifully supplied, auxin accumulated in the LR primordium to promote growth as a result of repressed auxin transport activities of NRT1.1 (CHL1/NPF6.3) (Krouk et al., 2010b; Mounier et al., 2014). The phosphorylated form of NRT1.1 (CHL1/NPF6.3) is predominantly active in auxin transport among the point mutations in NRT1.1 (CHL1/NPF6.3) being tested for auxin influx activity in Xenopus oocytes. It is also responsible for modulation of auxin gradient in LR primordium (Bouguyon et al., 2015).

Transport of other plant hormones across plasma membranes also requires transporter proteins that are spatiotemporally regulated during development instead of occurring simply by diffusion (Saito et al., 2015). Interestingly, more new substrates, such as ABA, GAs, jasmonoyl-L-isoleucine, and glucosinolates, were recently found to be transported by NRT1/PTRs or NPFs (Kanno et al., 2012; Nour-Eldin et al., 2012; von Wittgenstein et al., 2014; Saito et al., 2015; Chiba et al., 2015; Tal et al., 2016). In addition to being nitrate transporters, the capability of transporting hormones was suggested to be another critical feature of this family in plants. However, no long-distance transport of any hormones, i.e., loading/unloading of them into/out of xylem/phloem vessels, has been demonstrated. All the NRTs-dependent transports so far reported only involve local redistribution of the hormones (Kanno et al., 2012; Tal et al., 2016).

In Arabidopsis roots, a miR393/AFB3 regulatory module was identified as nitrate-responsive, which integrates nitrate and auxin signaling in modulating both primary and LR growth (Vidal et al., 2010). miR393 was the only N-responding sRNA identified in 454 sequencing and it specifically responded to nitrate not sucrose. The auxin receptor genes, TIR1, AFB1, AFB2, and AFB3, are regulated by miR393. Among them, a strong induction of AFB3, also the only induction, by nitrate, was observed. The AFB3 induction peaked at 1 h after nitrate (5 mM KNO3) exposure, and the nitrate induction of miR393 peaked at 2 h, strikingly coinciding with the fast declining of the already peaked expression of AFB3. This miR393-dependent repression was not observed in NR-null mutants, correlating with the absence of miR393 expression. Further evidence support that nitrate signal alone is responsible for the transcriptional induction of AFB3 in root tips, which can be subsequently posttranscriptionally repressed by miR393 induced by unidentified N metabolite(s) downstream of nitrate reduction. Such a mechanism agrees with the type I incoherent feed-forward loop (FFL) motif featured in transcriptional controls in yeast, bacteria, and mammals (Shen-Orr et al., 2002; Mangan and Alon, 2003; Tsang et al., 2007; Vidal et al., 2010). Accompanied with it, accumulation of auxin and the regulatory of many auxinresponsive and auxin-related genes involved in multi-level of auxin signaling and responses in the root tips and pericycle cells were also observed. The nitrate-regulated miR393/AFB3 module is capable to integrate nitrate (5 mM) signal into auxindependent root growth. Shorter primary root due to inhibited root elongation and more dense LRs due to higher rate of LR initiation and emergence were formed in response to nitrate availability in soil (Vidal et al., 2010).

Auxin binds to TIR1/AFB receptors, members of the SCFTIR1/AFB E3 ubiquitin ligase complex. It promotes the recognition and degradation of the Aux/IAA repressors via by polyubiquitination, which free the inhibition of the auxin response factors (ARFs) that allows auxin-responsive transcription (Chapman and Estelle, 2009). The activation of AFB3 is not the cause but one of the consequences of nitrate response (Vidal et al., 2013). AFB3-dependent auxin signaling, including perception and response, and its regulation of root growth is downstream of nitrate signaling in response to nitrate availability, independent of nitrate transport and metabolism. Specifically, NAM/ATAF/CUC TF, NAC4 and its targeted TF gene OBP4, functions as a downstream branch of nitrate-AFB3. The nitrate-AFB3-NAC4-OBP4 signaling, with all their proteins found expressed in root pericycle cells, is required for nitratedependent LR initiation and emergence. The NAC4-OBP4 part of the pathway is possibly regulated by AUX/IAA proteins, such as IAA14. These observations suggest convergent regulation

between nitrate and auxin signaling pathways on LR growth (Vidal et al., 2013).

With three different NRT1.1 (CHL1/NPF6.3) mutants: chl1-5, chl1-9, and NRT1.1T101D mutants, the role of nitrate membrane sensor/transporter, NRT1.1 (CHL1/NPF6.3) in regulating nitrate response of AFB3 and NAC4 were tested (Vidal et al., 2014a). Specifically, chl1-5 is a deletion mutant without uptake and sensing function; chl1-9 is defective in both high- and low-affinity nitrate uptake but not in nitrate signaling, and NRT1.1T101D mutant mimics a constitutively phosphorylated transporter, with only the high-affinity mode (Ho et al., 2009). Interestingly, not like sentinel PNR genes, such as NRT2.1, NIA1, and NIA2, which were tightly controlled by the signaling functions of NRT1.1 (CHL1/NPF6.3), only the transport function of the NRT1.1 (CHL1/NPF6.3), not those of NRT2.1, NRT1.2 (NPF4.6/AIT1), and NRT2.2, matters on the expression levels of AFB3 and NAC4. Moreover, the nitrate induction of AFB3 and NAC4 was independent of affinity mode of NRT1.1 (CHL1/NPF6.3). Notably, the NRT1.1T101D was demonstrated to modulate auxin gradient in LRP as WT NRT1.1 (CHL1/NPF6.3) in the absence of nitrate, which excludes the possible involvement of auxin transport of NRT1.1 (CHL1/NPF6.3) in the nitrate induction of AFB3 and NAC4 (under 5 mM nitrate treatment). It was suggested that an unidentified signaling pathway independent from the signaling via NRT1.1 (CHL1/NPF6.3) phosphorylation was triggered by NRT1.1 (CHL1/NPF6.3) transport of nitrate (Vidal et al., 2014a). It further substantiates that nitrate responses, which include the nitrate-AFB3-NAC4-OBP4 auxin perception, signaling and response, are established via multiple signaling mechanisms and their coordination (Liu et al., 2017).

#### NITRATE IN CYTOKININ BIOSYNTHESIS, SIGNALING, AND RESPONSES

Involved in many phases of plant growth and development, cytokinins (CKs) are a class of phytohormones known for promoting cell division and differentiation (Mok and Mok, 2001). CKs can interact with auxins either synergistically or antagonistically and also promote the production of ethylene (Mok and Mok, 2001). Since CKs are translocated at cellular and whole-plant levels, CK root-shoot communication is proposed as a model of systemic signaling for nutrient status (Sakakibara, 2006; Ruffel et al., 2011). CK activity in plants is tightly related to nitrate availability. Nitrate, not its downstream N metabolites, has been known to induce rapid de novo CK synthesis and accumulation in the roots of barley, maize, and Arabidopsis (Takei et al., 2004). CK biosynthesis can also occur in other tissues where the adenosine phosphate-isopentenyltransferases (IPTs) are expressed. IPTs are key enzymes that catalyze the first and ratelimiting step of CK biosynthesis, i.e., prenylation of adenosine 5 <sup>0</sup> phosphates, such as ATP and ADP, at the N<sup>6</sup> -terminus with dimethylallyl diphosphate (DMAPP) (Sakakibara, 2005). IPT expression is ubiquitous and peaks in proliferating tissues. In Arabidopsis, IPT3 is regulated by N in a nitrate-specific manner. The expression of IPT3 and several Arabidopsis response regulators 3, 5, 6 (ARR3, 5, 6), is induced by nitrate during the PNR (Wang et al., 2004). IPT3 was strongly induced in roots and weakly induced in shoots in both WT and NR-null mutant plants during the PNR, partly mediated by NRT1.1 (CHL1/NPF6.3) (Wang et al., 2004, 2009). When nitrate (10 mM KNO3) was re-supplied to nitrogen-limited seedlings, the kinetics of IPT3 and NIA1 transcripts that were rapidly accumulated within 1 h. resembled each other (Takei et al., 2004). NIA1 is among the most induced genes in the PNR (Wang et al., 2000); therefore, nitrate has a tight control over CK biosynthesis via activation of IPT3. All are consistent with the idea that IPT3 is the main determinant of short-term nitrate-dependent CK biosynthesis, particularly in roots, in response to the rapid change of nitrate availability in soil (Takei et al., 2004). More recent transcript profiling of CK metabolism and signaling genes further revealed that besides IPT3, high nitrate upregulates the transcripts of CYP735A2, which is responsible for the production of trans-zeatin-type (tZtype) CK in roots, while it downregulates that of LOG5. The type-A ARR genes ARR3, ARR5, and ARR7, like the CK metabolism genes, were shown to respond to nitrate but not to ammonium (Ramireddy et al., 2014; Liu et al., 2017). Also induced by nitrate are CYTOKININ RESPONSE FACTORS (CRFs) (Rashotte et al., 2006; Liu et al., 2017), which is known to be transcriptionally induced by CK and whose disruption affects the basal expression of a significant number of CK-regulated genes, including the type-A ARRs. CRFs are implicated in promoting root and shoot growth and leaf senescence (Raines et al., 2016).

Among the most highly expressed IPTs, IPT3 is mainly expressed in phloem tissue throughout the whole plant, specifically found in phloem companion cells, and IPT5 is in the LR primordium and pericycle, which are consistent with where CK biosynthesis is suggested to occur. The spatial differentiation of IPTs transcript also suggests that in terms of CK production, IPT5 and IPT3 could contribute most in roots, while IPT3 is most dominant in shoots. IPT5 was not responsive to either nitrate or ammonium under a short-term hour-long treatment, however, was demonstrated to be a "long-term" or systemic N statusresponsive gene, with its transcript abundance being responsive to both nitrate and ammonium media at different concentrations over a long course of observation, 11 days. By contrast, IPT3 expression pattern (strongly in roots and weakly in shoots) is quickly induced during the PNR, a frequently occurring wholeplant nitrate response due to dramatic fluctuations of nitrate in soil (Wang et al., 2004). The close regulation of CYP735A2 and IPT3 by nitrate could be a major factor shaping nitrate-dependent spatiotemporal CK distribution in plants and regulating root system architecture in response to a variety of abiotic stresses (Ramireddy et al., 2014).

Compared to the rapid nitrate response of IPT3 (within 1 h), CK signaling has a relatively delayed (within 4 h) feedback control over most the IPTs, IPT1,3,5,7 by downregulating them in roots, where IPT5 and IPT7 can be upregulated concurrently by auxin (Miyawaki et al., 2004). IPT7 is expressed in root stele and phloem companion cells. To add another layer of dynamic complexity of interactions, CK and auxin also exert feedback controls over nitrate uptake and assimilation (Guo et al., 2002; Krouk, 2016). In this context, nitrate and two hormonal mediators, CK and its antagonistic partner, auxin, act

in concert to modulate CK biosynthesis in root development. The dual nitrate-CK response system, employing IPT3 and IPT5, along with CK/auxin feedback regulations on IPTs could have a critical role in mediating root foraging for nitrate, a classic plant response to nitrate availability. To compete for nutrients in diverse soil microenvironments, plants have evolved the unique capability to proliferate LRs preferentially in nutrient-rich zones, called "root foraging" (Drew et al., 1973). Root foraging for nitrate involves both local and systemic signaling (Ruffel et al., 2011; Guan et al., 2014; Mounier et al., 2014). Besides their effects on localized CK and auxin biosynthesis, the concerted nitrate-CK-auxin regulation could also be an integral part of N systemic signaling that coordinates nutritional requirements among different organs and at different developmental stages. Notably, ammonium and downstream N metabolites are unlikely to be major players in systemic N signaling (Howitt and Udvardi, 1999; Forde, 2002; Bellegarde et al., 2017).

Using WT, NR-null and ipt3,5,7 mutants in split-root experiments, nitrate signaling was demonstrated to act both locally and systemically to integrate N supply and demand. The systemic N signaling also involves a nitrate-CK relay, where IPT3, IPT5 and IPT7 play a central role. The nitrate-CK relay is suggested to be necessary for shaping root foraging (Ruffel et al., 2011). The study also suggested that there is an additional systemic signaling pathway also required. Using decapitation experiments, the concept of shoot-root CK-dependent feedback specifically for N demand was proposed. However, questions remain. Is decapitation a definitive way to confirm CK's independence from nitrate in the systemic N demand signaling? In analogy to CK's root-shoot-root signaling/relay mechanism, a similar model was proposed for small peptides in root foraging (Okamoto et al., 2013; Tabata et al., 2014; Ohkubo et al., 2017). Clavata3/ESR (CLE)-related peptide signal and N starvationtriggered C-terminally Encoded Peptide (CEP) were identified as "satiety" and "hunger" signals. In root-to-shoot route, CLE and CEP were first derived from roots, then transmitted to shoots where being perceived by leucine-rich repeat receptor-like kinase (LRR-RLK) receptors HAR1 and CEPRs, respectively. In the following shoot-to-root route, CEP Downstream1 (CEPD1) and CEPD2, two phloem-specific polypeptides, are regulated by CEPRs and then transmitted to roots, where NRT2.1 is thereby activated (Ohkubo et al., 2017). Notably, the two hormonedependent systemic signaling pathways could be necessary but not sufficient for root foraging independent of local and systemic signaling by nitrate (Ruffel et al., 2011, 2016; Guan et al., 2014; Mounier et al., 2014).

Furthermore, there is also intriguing spatiotemporal regulation of CK signaling by nitrate in the context of root foraging. The expression of primary CK-response genes and negative regulators of CK signaling, type-A ARRs(3,5-9) was globally up-regulated by nitrate in roots and shoots at a later time (2 h, 8 h, and 2 days) compared with the much quicker expression of IPT (within 1 h) (Ruffel et al., 2011). The response levels of ARRs in NR-null roots and shoots are very comparable, in some cases even lower in roots, which is strikingly opposite to those of IPTs (Wang et al., 2004). Since both CK biosynthesis after IPT induction and induction of ARRs by the produced CK are rapid (Wang et al., 2002), the much later (>8 h) regulations of ARRs by nitrate (Ruffel et al., 2011) suggest additional nitrateregulated mechanism(s) are likely to be involved rather than CK biosynthesis-dependent replay.

Another branch of the evidence that deserves our attention is that TCP20 as a cell-autonomous systemic nitrate regulator is clearly required for the nitrate foraging by roots (Guan et al., 2014). tcp20 mutants strongly suppress the preferential growth of LRs by equalizing growth across heterogeneous nitrate environments, mainly through increasing the LR growth on lownitrate media of split-root plates as if the plants were impervious to any systemic signal. An earlier study showed that the main class of AtTCP20::EAR-repressed genes include ARR4,6,7 and AUX/IAA13,16,27 that repress ARFs in auxin signaling (Hervé et al., 2009). All the genes possess at least one class I TCP binding motif in their promoters. In tcp20 mutants, the foragingdefective LR growth in high/low nitrate media was largely due to much shorter/longer LR length but not to the less/more number of LRs (Guan et al., 2014). The RAM growth is indeed under the balanced control of nitrate-CK-auxin signaling as also previously discussed. The concerted nitrate-CK-auxin signaling could also have TCP20 as a mediator between CK and auxin for the regulation of root foraging.

# NITRATE IN ABSCISIC ACID DECONJUGATION, DEGRADATION, TRANSPORT, AND SIGNALING

Nitrate sensing, signaling and regulation, and their interaction with hormones are very dose-dependent. Beyond the optimal range (1–10 mM) corresponding to steady-state cytosolic nitrate concentrations (4–6 mM) (Miller and Smith, 2008), additional interaction between nitrate and hormones occurs. Transferring Arabidopsis seedlings between media with different nitrate concentrations has been used to mimic plant responses to a variety of nitrate availability in soil. The reversible oscillating responses have been thereby observed in nitrate-dependent hormone biosynthesis and accumulation, and LR growth (De Smet et al., 2003; Tian et al., 2009).

Abscisic acid has been long regarded as a stress hormone crucial to plant abiotic and biotic stress responses (Zhu, 2002; Kiba et al., 2012). In the face of high nitrate concentrations (approximately >10 mM), nitrate and ABA are close partners, especially in the control of LR growth (Signora et al., 2001; De Smet et al., 2003). Nitrate serves as an osmolyte; therefore, the changes of nitrate availability and concentration alter osmotic potential of plant cells. Repression of LRs in Arabidopsis due to very high nitrate (30 mM) resembled the repressed growth of LR treated by 30 mM KCl or 60 mM mannitol. All the treatments, including high nitrate (>30 mM), could impose osmotic stress (Deak and Malamy, 2005). Exogenous ABA also inhibits LR development, mimicking high nitrate repression of LR (De Smet et al., 2003). Furthermore, ABA synthesis and ABA-sensing mutants displayed significantly reduced inhibitory effects by high nitrate concentrations (>10 mM) (Signora et al., 2001). The ABA-induced growth arrest occurred right after LR

emergence and before the activation of the LR meristem, which is due to ABA suppression of the transcription of two cell cyclerelated genes, CYCD3;1 and CDKB1;1. It serves as a checkpoint that, however, is reversible (De Smet et al., 2003).

The accumulation of ABA was detected mainly in the endodermis and quiescent center of Arabidopsis root tips, similar to the expression pattern of SCARECROW, and to a lesser extent in the vascular cylinder (Ondzighi-Assoume et al., 2016). A threefold increase of ABA level in root tips was observed in seedlings being transferred from the medium containing 20 mM nitrate to that containing 30 mM nitrate. It was accompanied by the increased activity of the endoplasmic reticulum-localized, ABA-GE-deconjugating enzyme β-GLUCOSIDASE1, but not with de novo ABA biosynthesis. High nitrate thereby stimulates release of bioactive ABA from the inactive storage form, ABAglucose ester (ABA-GE) (Ondzighi-Assoume et al., 2016). In parallel with this, osmotic stress causes accumulation of the endogenous ABA; therefore, ABA has been regarded as a mediator of responses of osmotic stress imposed by drought and high salt (Zhu, 2002). All suggests that both ABA and high nitrate share a single pathway which is likely part of general osmotic stress responses, during a specific LR development stage in Arabidopsis. And notably the high nitrate-ABA pathway is independent of auxin (De Smet et al., 2003).

ABA-IMPORTING TRANSPORTER (AIT) 1, also characterized as the constitutive, low-affinity nitrate transporter NRT1.2 (NPF4.6), mediates cellular ABA uptakes during seed germination and post-germination growth of Arabidopsis (Kanno et al., 2012). In response to drought stress, plants synthesize ABA to trigger closing of stomatal pores. NRT1.2 (NPF4.6/AIT1) is suggested to be involved in regulation of stomatal aperture in inflorescence stems via transporting ABA synthesized in vascular tissues to guard cells. Being an osmolyte, nitrate is also known for controlling gas exchange by stomates. In the presence of nitrate, NRT1.1 (CHL1/NPF6.3) is required in nitrate induced depolarization and nitrate accumulation in guard cells during stomatal opening. Its mutants showed reduced stomatal opening and transpiration rates in the light or when deprived of CO<sup>2</sup> in the dark, leading to drought resistance (Guo et al., 2003). The two nitrate transporters seem to be able to work in a "coordinated" manner to regulate the stomatal functions in response to drought stress or nitrate, whichever signal becomes dominant. An intriguing coordination between the two transporters occurred when the induction of NRT1.1 (CHL1/NPF6.3) by nitrate caused a transient repression of NRT1.2 (NPF4.6/AIT1) (Huang et al., 1999). With the exception of this temporary coupled reaction of NRT1.1 (CHL1/NPF6.3) and NRT1.2 (NPF4.6/AIT1) in response to nitrate induction, NRT1.2 (NPF4.6/AIT1) is constitutively expressed before and after nitrate exposure (Huang et al., 1999).

The mechanism of interaction between nitrate and ABA signaling, which could be behind such coordination is further understood in roots. In another study, ABA insensitive2 (ABI2; an ABA inactivated PP2C) has been identified as a potential interacting protein of the CBL1-CIPK23 complex, which like CBL9-CIPK23 has inhibitory effects on nitrate transport of NRT1.1 (CHL1/NPF6.3), under >30 mM nitrate (Leran et al., 2015). The CBL9-CIPK23 complex is known to be responsible for the phosphorylation of the NRT1.1 (CHL1/NPF6.3), resulting in switching to high-affinity transport mode in response to low nitrate availability (<1 mM) (Ho et al., 2009). ABI2 negatively regulates the full activation of CBL1-CIPK23 toward their targeted proteins by substantially reducing CIPK autophosphorylation and CIPK-dependent phosphorylation of the Ca2+-sensor moiety in the associated CBL (Hashimoto et al., 2012). During drought and osmotic stresses, stress-induced ABA could inactivate ABI2 by RCAR/PYL/PYR interaction, which enhances phosphorylation of NRT1.1 (CHL1/NPF6.3) and phosphorylation of AKT1 by CBL1-CIPK23. The similar phenotypes associated with nitrate transport and signaling were observed in both chl1 and abi2-2 mutants. The mechanism could allow plants to rechannel their energy and resource from nitrate assimilation to stress response via reducing nitrate uptake in favor of uptake of potassium ions, which is critical in abiotic and biotic stress responses (Wang et al., 2013). It suggests that ABA-dependent stress signals could be required to be conveyed to and processed through the nitrate transceptor, NRT1.1 (CHL1/NPF6.3), so that the abiotic stress response is likely a collective decision made in conjunction with nitrate signaling. The conclusion is also supported by the results of an earlier study that in Arabidopsis guard cells, ABI1 and ABI2 protein phosphatases are downstream of NR-mediated nitric oxide (NO) in the ABA signal-transduction cascade (Desikan et al., 2002). The NO synthesis regulated by nitrate signaling is required for ABA-induced stomatal closure (Desikan et al., 2002).

Recently, a direct molecular link between nitrate signaling and ABA degradation in seed germination was revealed (Yan et al., 2016). The conserved nitrate regulator, NLP8, was found to regulate ABA catabolism and activate the expression of CYP707A2, which is indispensable for nitrate-induced seed germination. This activation appears to occur directly, through NLP8 binding to the promoter of CYP707A2, which encodes ABA 8<sup>0</sup> -hydroxylase, a key ABA catabolic enzyme (Kushiro et al., 2004; Yan et al., 2016). ABA negatively regulates the germination process. Hence, seed germination after the onset of imbibition can be triggered in a timely fashion upon reduced level of ABA. Notably, CYP707A2 has been shown to be a hub processing environmental signaling, i.e., nitrate, light, and temperature, during germination (Footitt et al., 2011, 2013).

#### NITRATE IN ETHYLENE BIOSYNTHESIS AND SIGNALING

With the chemically simplest form among phytohormones, ethylene is a gaseous signal molecule and potent regulator of developmental adaptations (Ecker, 1995; Bleecker and Kende, 2000). The production of ethylene is regulated by internal signals during developmental phases, including seed germination, root growth, fruit ripening, organ senescence, etc., and also in response to biotic and abiotic stresses (Wang et al., 2002). Compared with the significant ABA accumulation in roots of the seedlings that were transferred from low nitrate (20 mM) to high nitrate (30 mM), transferring seedlings from low nitrate (0.1 mM) to high nitrate (10 mM) caused a rapid

burst of ethylene production in roots (Tian et al., 2009). Both of them contribute to the inhibitory effects of LR growth exerted by transferring to high nitrate conditions. Strikingly, the LR growth inhibition and the elicited ethylene evolution can be reversed by transferring the seedlings back to the low nitrate (0.1 mM), similar to the reversible arrest observed in the case of ABA (De Smet et al., 2003). Ethylene is synthesized from methionine through S-adenosyl-L-methionine and 1-aminocyclopropane-1-carboxylic acid (ACC), which are catalyzed by ACC synthase (ACS) and ACC oxidase (ACO) (Kende, 1993). The nitrate-dependent ethylene evolution and accumulation were accompanied by transient but significant increase of ACS and ACO, which are transcriptionally induced by the transferring to high nitrate (10 mM). The inactivation of ACS and ACO by their antagonists alleviated LR growth defects. Nitrate-induced ethylene inhibited the growth of immature LRs, which is at a later development stage compared to ABA-induced LR inhibition.

Employing the combinations of Chl1-5 and nrt2.1-1 mutants and ethylene-insensitive mutants, etr1-3 and ein2-1, ethylene was demonstrated as an important modulator in the regulation of nitrate-dependent expression of the two main transporters, NRT1.1 (CHL1/NPF6.3) and NRT2.1 (Tian et al., 2009). Notably, in the comparable range of low nitrate (0.5 mM), NRT2.1 promotes initiation of LR primordia, which is likely a different mechanism occurring at an early stage of LR development (Remans et al., 2006b). The ethylene-dependent regulation, or nitrate signaling relay (Krouk, 2016) observed here when seeding roots being challenged by high nitrate conditions, could be centered on NRT1.1 (CHL1/NPF6.3) whose expression is much more strongly affected. This signaling relay via ethylene might be part of the mechanism where NRT1.1 (CHL1/NPF6.3) mediates the repression of NRT2.1 between high-affinity transport mode and low-affinity transport mode (Muños et al., 2004; Krouk et al., 2006). Interestingly, transferring seedlings from high nitrate (10 mM) to low nitrate (0.1 mM) also caused a rapid burst of ethylene production measured on a whole-plant basis. NRT2.1 whose repression is relieved by NRT1.1 (CHL1/NPF6.3) under low nitrate concentration (<0.5 mM) was singled out to be responsible for stimulating ethylene production (Zheng et al., 2013).

#### NITRATE IN GIBBERELLIN BIOSYNTHESIS, TRANSPORT, AND SIGNALING

Gibberellins are tetracyclic diterpenoid hormones. GAs are key endogenous regulators involved in seed germination, root and shoot elongation, flowering, and fruit patterning (Daviere et al., 2014; Tal et al., 2016). Much of the lead role of nitrate in dancing with hormones has been revealed in roots. Nevertheless, nitrate-hormone interaction certainly takes place in whole plants. For example, in the transition from vegetative growth to reproduction, earlier flowering was favored at low nitrate growth conditions rather than at high nitrate conditions. The major repressor of flowering in Arabidopsis, FLOWERING LOCUS C (FLC), is repressed and activators of flowering, FLOWERING LOCUS T (FT), LEAFY (LFY), and APETALA1 (AP1), are induced in low-nitrate conditions. Interacting with photoperiodand temperature- and GA-signaling pathways, nitrate regulates floral induction by communicating nutrient availability (Castro-Marín et al., 2011; Kant et al., 2011; Liu et al., 2013). The low nitrate (1 mM) was shown to transcriptionally induce expression of GA1, the main GA biosynthesis gene, therefore promoting bioactive GAs in various tissues of flowering plants. Along with GA1, nitrate also induced the expression of SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1), an integrator of the GA-dependent flowering pathway which coordinates all the endogenous pathways: GA, vernalization, autonomous, and photoperiod (Liu et al., 2013). The transcriptome of pre-starved Arabidopsis seedlings in response to nitrate re-addition (3 or 5 mM KNO3) also showed the repressed expression of GID1B, GA receptor and the induced expression of GATA, NITRATE-INDUCIBLE, CARBON-METABOLISM INVOLVED (GNC) and GNC-LIKE/CYTOKININ-RESPONSIVE GATA FACTOR1 (GNL/CGA1) TFs, negative regulators of GA signaling downstream from DELLA proteins and PHYTOCHROME-INTERACTING FACTORS (PIFs) (Wang et al., 2003, 2004; Scheible et al., 2004; Richter et al., 2010).

NPF3.1 expression in endodermis were found to be transcriptionally repressed by GA and promoted by ABA (Tal et al., 2016). NPF3.1 is a plasma membrane localized protein mediating nitrate and nitrite uptake (Sugiura et al., 2007; Leran et al., 2014). In addition, the experiments in X. oocytes showed that NPF3.1 is an active GA importer and it is also capable of transporting ABA. Another NPF protein, AIT3/NPF4.1, was found earlier to have ABA and GA transport activities (Kanno et al., 2012). NPF3.1 as such, is involved in two antagonistic hormone signaling in endodermal cells controlling root meristem size (Tal et al., 2016). Such an intimate interplay between nitrate, GA and ABA were also observed in determining seed dormancy and germination times (Alboresi et al., 2005; Chopin et al., 2007; Yan et al., 2016). GTR1/NPF2.10 was also proposed as a multifunctional transporter employed by the structurally distinct compounds glucosinolates, JA-Ile and GA, which promote stamen development via mediating bioactive GA transport (Saito et al., 2015).

# CONCLUSION

Decades of nitrate research have given rise to new paradigms. By analogy with molecular oxygen (O2) being an environmental morphogen in embryonic development and stem cell function in animals (Simon and Keith, 2008), and auxin being proposed as a plant morphogen (Bhalerao and Bennett, 2003; Esmon et al., 2006), nitrate could be a potent environmental morphogen in plants given the comprehensive nitrate transport, sensing, signaling and regulations at the level of the cell and organism (Wang et al., 2004, 2012; Bellegarde et al., 2017). Remarkably, less than 0.1% seed N is from nitrate (Chopin et al., 2007), which doesn't prevent nitrate from being a crucial signal and creating specific niches (Yan et al., 2016); and phloem-based

transport of nitrate that represents only 1–10% of total N in the phloem sap, has a more morphogenetic role than a nutritional role in sink organs (Fan et al., 2009; Kiba et al., 2012; Hsu and Tsay, 2013; Bellegarde et al., 2017). The differential responses to extracellular nitrate availability while maintaining cellular homeostasis, especially steady-state ionic environment, are among the main morphogenetic effects in determining cell growth and identity in plants. At a more detailed level, the intercellular and intracellular gradients of nitrate could be responsible for diversified patterns of transient and sustained promotion or repression of expression levels of specific subsets of nitrate-responsive genes, which were observed in a variety of nitrate responses, including the PNR. Behind the whole-plant responses is that the changes of N status of the shoot could be potentially communicated to the root via the peaks and valleys of nitrate gradient and xylem- and phloem-based transport, serving for long-distance signaling (Bellegarde et al., 2017).

Why does nitrate signaling critically regulate so many types of phytohormones at so many levels? The localized hormone biosynthesis, deconjugation and degradation seem to be the primary connection between nitrate and hormones, for which solid molecular evidence has been increasingly found. Being an environmental cue and once being absorbed into plants, also becoming an environmental morphogen, nitrate transcriptionally regulates the metabolism and signaling of hormones at the whole plant level. Indeed, this regulatory process is highly context-dependent and spatiotemporal. Conversely, the hormones from nitrate-induced production, deconjugation and degradation, could act as mediators and/or modulators in N-dependent signaling and regulation and provide feedback controls at the regional level. Notably, numerous hormone signaling and regulatory components were found to be transcriptionally activated by nitrate in certain contexts, suggesting that the signaling of the two morphogens is intertwined at multiple levels, if not all levels. The genome-wide transcriptome further revealed that the molecular conductors at the top-level of nitrate regulatory hierarchy could exert direct controls over hormonal pathways (Hervé et al., 2009; Danisman et al., 2012; Marchive et al., 2013; Guan et al., 2017; Liu et al., 2017), so that a variety of hormones are employed in propagation and amplification of nitrate signaling (**Figure 2**).

Intriguingly, cell-autonomous regulation by N in determining cell growth and fate is strongly indicated by TCP20-NLP6/7 regulatory nexus, which is involved in sensing nutrient status and transcriptional control of G2/M transition in cell cycle progression (Guan et al., 2017). Growing evidence suggests that there could exist PB1 domain-mediated interactions between

#### REFERENCES


nitrate and auxin signaling regulators upstream of TOR signaling, which are central in nutrient–growth process in plants (Xiong et al., 2013; Guan et al., 2017; Li et al., 2017). Moreover, the analogy between regulatory roles of TCP20 in response to nitrate availability and of TCP21/CHE in the circadian oscillator suggest a general TCP-dependent cell-autonomous mechanism for plant responses to variations in environmental cues, i.e., nutrients, light, and temperature (Pruneda-Paz et al., 2009; Guan et al., 2017). The intertwined coordination of cell autonomous and morphogen-gradient-dependent mechanisms is deeply conserved in eukaryotes, being well observed in the amoeba, Dictyostelium discoideum (Clay et al., 1995).

Hormones have long been regarded to provide an indispensable link between N and plant growth and development. However, among the most deeply conserved in plants, nitrate signaling and regulation with a highly organized transcriptional hierarchy are as crucial as hormonal signaling and regulation in growth, development and stress responses. The novel model that underlies substantial plant development and adaptive responses could involve other TCPs because of functional redundancy between TCP20 and its homologs. The classes I and II TCPs exert either coordinate or competitive regulation of transcription that could be essential for defining growth rate and organ development (Li et al., 2005). The interaction between TCPs and hormone biosynthesis, transport, signaling and responses in growth, development and defense has been increasingly reported (Nicolas and Cubas, 2016). With TCPs in the picture, much extended interplay and convergent regulation between nitrate and hormone signaling will be expected.

In the past half century, N fertilizer is the main contributor to global crop production increases that support two billion more people on Earth. Nevertheless, NUE has become a major constraint on agricultural productivity and environmental sustainability worldwide. Understanding nitrate signaling and regulation and their interaction with hormones is central to meet the global challenges, which demands extensive research under the new paradigms.

#### AUTHOR CONTRIBUTIONS

The author confirms being the sole contributor of this work and approved it for publication.

# ACKNOWLEDGMENT

I thank Dr. Nigel Crawford for discussions.



constitutive component of low-affinity uptake. Plant Cell 11, 1381–1392. doi: 10.1105/tpc.11.8.1381






Zoncu, R., Efeyan, A., and Sabatini, D. M. (2011). mTOR: from growth signal integration to cancer, diabetes and ageing. Nat. Rev. Mol. Cell Biol. 12, 21–35. doi: 10.1038/nrm302

**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.

The reviewer EW and handling Editor declared their shared affiliation.

Copyright © 2017 Guan. 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.

# FAR-RED INSENSITIVE 219/JAR1 Contributes to Shade Avoidance Responses of Arabidopsis Seedlings by Modulating Key Shade Signaling Components

#### Swadhin Swain† , Han-Wei Jiang and Hsu-Liang Hsieh\*

Institute of Plant Biology, College of Life Science, National Taiwan University, Taipei, Taiwan

#### Edited by:

Chi-Kuang Wen, Institute of Plant Physiology and Ecology, Shanghai Institutes for Biological Sciences (CAS), China

#### Reviewed by:

Lin Li, Fudan University, China Rongcheng Lin, Institute of Botany (CAS), China Xin Zhou, Shanghai Normal University, China

#### \*Correspondence:

Hsu-Liang Hsieh hlhsieh@ntu.edu.tw

#### †Present address:

Swadhin Swain, Department of Microbiology and Plant Biology, The University of Oklahoma, Norman, OK, United States

#### Specialty section:

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

Received: 15 August 2017 Accepted: 20 October 2017 Published: 02 November 2017

#### Citation:

Swain S, Jiang H-W and Hsieh H-L (2017) FAR-RED INSENSITIVE 219/JAR1 Contributes to Shade Avoidance Responses of Arabidopsis Seedlings by Modulating Key Shade Signaling Components. Front. Plant Sci. 8:1901. doi: 10.3389/fpls.2017.01901 To receive an ample amount of light, plants use elongation growth in response to vegetation shade. The combined interaction of light and hormones, including jasmonic acid (JA) signaling controls this elongation. However, the detailed molecular mechanisms underlying the response are still emerging. FAR-RED INSENSITIVE 219/JASMONATE RESISTANCE 1 (FIN219/JAR1), a cytoplasmic localized JA-conjugating enzyme, integrates far-red light and JA signaling. Here, we report that FIN219/JAR1 negatively regulates shade-induced hypocotyl elongation and gene expression in Arabidopsis seedlings in response to shade. In turn, simulated shade reduces FIN219 protein accumulation. Analysis of phyA 211 fin219-2 double mutants indicated that FIN219 and phyA are synergistic in regulating shade-induced hypocotyl elongation and gene expression. Moreover, FIN219 differentially affected the expression of the shadesignaling bHLH factors PIF5 and PAR1, thereby increasing the expression of the auxin-response genes IAA29 and SAUR68 on exposure to shade. Furthermore, the protein level of CONSTITUTIVE PHOTOMORPHOGENIC 1 (COP1) was affected in both fin219 mutants and overexpression lines as compared with the wild type under shade. Intriguingly, ectopic expression of FIN219 inhibited the nuclear accumulation of COP1 in response to shade. Further co-immunoprecipitation studies revealed that FIN219 interacted with COP1 and phyA under shade. Therefore, FIN219/JAR1 may play a vital role in modulating the Arabidopsis response to simulated shade via multiple layers of molecular mechanisms.

Keywords: shade avoidance response, FIN219/JAR1, jasmonates, shade-induced hypocotyl elongation, Arabidopsis, shade signaling

# INTRODUCTION

Light carries most of the essential information needed for plant growth and development (Von Arnim and Deng, 1996; Fankhauser and Chory, 1997; Chory, 2010). Photoreceptors such as phytochromes, cryptochromes, and UVR8 perceive light as a developmental signal (Li et al., 2011; Liu et al., 2011; Tilbrook et al., 2013). Vegetation shade reduces plants' access to sufficient light. Shade-intolerant plants such as Arabidopsis trigger a suite of responses, collectively called shade

**89**

avoidance syndrome (SAS), including elongation of hypocotyls, stems and petioles, hyponasty (upward bending of leaves) and early flowering (Franklin, 2008; Casal, 2012). SAS is a default developmental program and suppressed under normal light by phyB along with phyD and phyE (Franklin and Quail, 2010; Casal, 2013).

Phytochromes exist in two forms: the red light (666 nm) absorbing P<sup>r</sup> form, and far-red light (730 nm)-absorbing Pfr form (Li et al., 2011). However, monochromatic red (R) or far-red (FR) light usually does not convert 100% of P<sup>r</sup> to Pfr or vice versa. Thus, a dynamic equilibrium is established between the two phy forms (P<sup>r</sup> and Pfr) depending on the quality of light (Possart et al., 2014). Under white light (high R:FR), the active phyB, Pfr, migrates to the nucleus and interacts with various PIFs and drives their degradation (Lorrain et al., 2008). The onset of low R:FR ratio shifts the steady-state equilibrium toward the inactive Pr form (Franklin and Quail, 2010). As a result, PIFs become stabilized and more abundant (Lorrain et al., 2008). Genomewide analysis revealed that PIF4 and PIF5 preferentially bind to auxin biosynthetic and signaling gene promoters (Hornitschek et al., 2012; Leivar et al., 2012b) and activate their expression under low R:FR light. PIF7 also functions as a major regulator for shade-induced hypocotyl elongation and is dephosphorylated in response to shade (Li et al., 2012). HFR1 physically interacts with PIF4 and PIF5 and inhibits their binding to the target promoters, thus negatively regulating shade responses (Hornitschek et al., 2009).

CONSTITUTIVE PHOTOMORPHOGENIC 1 (COP1) is a repressor of photomorphogenesis and encodes an RING-finger E3 ubiquitin ligase. Under darkness, COP1 accumulates in the nucleus and targets degradation of positive regulators of photomorphogenesis such as HY5, HYH, LAF1, and HFR1 (Deng et al., 1991; Ballesteros et al., 2001; Lorrain et al., 2006). As well, COP1 is required for activating shade signaling. Mutants of COP1 can barely sense vegetation shade, whereas overexpression of COP1 leads to a constitutive shade avoidance phenotype (Pacín et al., 2013). The molecular mechanisms of COP1-mediated shade responses are emerging. COP1 rapidly accumulates in the nucleus under low R:FR, and this nuclear accumulation is essential for full shade responses (Pacín et al., 2013). Further evidence indicates that COP1 enhances HFR1 degradation under shade, leading to an increase of PIF-mediated gene expression, auxin levels, and thus stem growth (Pacín et al., 2016). COP1 and SPA1 together participate in hypocotyl and leaf petiole elongation in response to low R:FR by targeting the HFR1degradation by 26S proteasome (Rolauffs et al., 2012). The cop1 bbx21 bbx22 triple mutant is responsive to the shade, which suggests that the shade insensitivity phenotype of cop1 is mediated through both BBX21 and BBX22 (Crocco et al., 2010). However, how COP1 regulates SAS at the molecular level needs further investigation.

Phytohormones play important roles in light control of plant development (Lau and Deng, 2010; Kurepin and Pharis, 2014). The interplay of jasmonates (JAs) and light signaling is accepted (Kazan and Manners, 2011; Hsieh and Okamoto, 2014). Although low R:FR light perturbs JA signaling (Moreno et al., 2009; Robson et al., 2010; Cerrudo et al., 2012; De Wit et al., 2013; Chico et al., 2014; Leone et al., 2014), modulation of shade signaling by JA or its signaling components is still under examination. Recent evidence has revealed the interplay between JAs and shade, which is critical for growth-defense balance. This process may involve resource allocation between growth and immunity (Yang et al., 2012; Ballaré, 2014; Mazza and Ballaré, 2015).

The jar1 mutants show reduced sensitivity to exogenous JA and enhanced susceptibility against soil fungus, Pythium irregulare (Staswick et al., 1992, 1998). Later, jar1-1 was mapped to the same locus as fin219-1 (Staswick et al., 2002), a suppressor of cop1 mutant under darkness (Hsieh et al., 2000). FIN219 (GH3.11) belongs to a GH3 family of proteins, and its expression is rapidly induced by auxin (Hsieh et al., 2000). FIN219/JAR1 encodes JA-amino-synthetase, which conjugates isoleucine (Ile) with jasmonic acid (JA) to form the bioactive jasmonoyl-Lisoleucine (JA-Ile) (Staswick and Tiryaki, 2004). JA-Ile is an oxylipin that profoundly affects plant developmental and stress responses (Bari and Jones, 2009; Browse, 2009).

The fin219/jar1 mutant exhibits an insensitive long-hypocotyl phenotype under continuous far-red (cFR) light, which indicates its role in phyA-mediated signaling (Hsieh et al., 2000; Chen et al., 2007). Moreover, FIN219/JAR1 physically interacts with COP1 under darkness and cFR light and negatively regulates COP1 under FR light. Artificial induction of FIN219 protein accumulation via overexpression inhibited COP1 nuclear localization and hence stabilized HY5 protein (Wang et al., 2011). In addition, FIN219 might contribute to signaling of other hormones such as auxin, ethylene, gibberellin and abscisic acid by regulating a large number of bHLH transcription factors (TFs) (Chen et al., 2015).

Under low R:FR light, both phyA and fin219 mutant seedlings show enhanced hypocotyl elongation as compared with the wild type (Johnson et al., 1994; Yanovsky et al., 1995; Robson et al., 2010). Although the phyA phenotype is attributed to a FR lightmediated high irradiance response, the molecular mechanism underlying fin219-mediated shade avoidance phenotype is still under debate.

In the present study, we examined the functional significance of FIN219 under simulated shade in Arabidopsis seedlings. With a combination of genetic and biochemical experiments, we demonstrate that FIN219 negatively regulates shade avoidance responses by modulating key shade signaling components. Moreover, FIN219-mediated shade avoidance responses were independent of phyA-mediated high irradiance responses.

#### MATERIALS AND METHODS

#### Plant Materials and Growth Conditions

The fin219-2 (SALK\_059774), fin219-1, jar 1-1, phyA 211, cop1-4, cop1-6, cop1-4 fin219-2, and cop1-6 fin219-2 mutants and glucocorticoid-inducible FIN219 transgenic line (pGR:FIN219; PGR219) were described previously (Hsieh et al., 2000; Chen et al., 2007; Wang et al., 2011). pGR219 seedlings were harvested in GM plates with 1 µM dexamethasone (Dex) to induce FIN219 expression. The phyA 211 fin219-2 double mutant was generated by crossing phyA 211 with the fin219-2 mutant and selecting homozygous plants in an F2 generation by T-DNA-specific primers. All mutants are in a Col-0 background. Seeds of Arabidopsis thaliana were surface-sterilized and plated on growth medium (1/2 strength Murashige and Skoog medium, Duchefa Biochemie; 0.3% sucrose, 0.5% MES, 0.5% agar). After 3 days of incubation at 4◦C, seed plates were kept at 22◦C, 150 µmol m−<sup>2</sup> s <sup>−</sup><sup>1</sup> fluorescence white light for 16 h for germination, then transferred to an LED growth chamber (HIPONT, 721FTEC, Taiwan). Germinating seedlings were grown in continuous white light (high R:FR) for 2 days, then kept in continuous white light or transferred to continuous simulated shade (low R:FR) for an additional 4 days.

#### Light Measurements

fpls-08-01901 October 31, 2017 Time: 18:29 # 3

The HIPONT, 721FTEC (Taiwan) growth chamber equipped with monochromatic LED light sources for white light (455 nm), red light (650 nm), and far-red light (730 nm) (Supplementary Figure S4), was used for shade avoidance study. In our study, white light and simulated shade represented high red:far-red light ratio (R5.21 µmol m−<sup>2</sup> s −1 /FR2.25 µmol m−<sup>2</sup> s <sup>−</sup><sup>1</sup> ∼ 2.3) and low red:far-red ratio (R5.21 µmol m−<sup>2</sup> s −1 /80.02 µmol m−<sup>2</sup> s <sup>−</sup><sup>1</sup> ∼ 0.06), respectively. Photon fluencies were estimated with use of a Li-Cor (LI-250A, LiCor Corp., Lincoln, NE, United States) and spectral distribution was measured by an Ocean Optics spectrum meter (USB2000, Florida, United States). Photosynthetically active radiation (white + red light) was kept constant at 70 µmol m−<sup>2</sup> s −1 .

#### Hypocotyl Length Measurements and Western Blot Quantification

Digital images were analyzed by the US National Institutes of Health ImageJ software (Bethesda, MD, United States<sup>1</sup> ) to measure lengths of hypocotyl and to quantify Western blot band intensities.

### RNA Extraction, cDNA Synthesis, and Quantitative Real-Time PCR

RNA extraction and cDNA synthesis were as described previously (Chen et al., 2015). Samples of 100 mg were ground in 0.5 ml Buffer A (1MTris-HCl, pH 7.3, 5 mMEDTA, pH 8.0, 1% SDS), then extracted twice with an equal volume of phenol and once with chloroform:isoamylalcohol (24:1).

<sup>1</sup>http://rsb.info.nih.gov/

The supernatant was precipitated with LiCl (final working concentration 3 M) and incubated at −20◦C overnight. After centrifugation, the pellet was dissolved completely in 0.5 ml 2% potassium acetate, then precipitated again with isopropanol. Total RNA (2 µg) was treated with DNase to prevent genomic DNA contamination, then used as a template for cDNA synthesis with the ABI cDNA transcription kit (#4368814). Real-time PCR involved CFX96 Touch Real-time PCR Detection System (Bio-Rad, United States). Gene-specific primers (Supplemental Table S1) were used for analyzing mRNA levels of ACT2 (AT3G18780), PIL1 (AT2G46970), ATHB2 (AT4G16780), FIN219 (AT2G46370), HFR1 (AT1G02340), PAR1 (AT2G42870) and PIF5 (AT3G59060), IAA29 (AT4G32280) and SAUR68 (AT1G29490) by qPCR. ACT2 was used as internal control to normalize the expression levels, then standardized to the wild-type level under white light.

#### Protein Extraction and Protein Gel Blot Analysis

Total protein was extracted with extraction buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM MgCl2, 0.1% NP-40, 1 mM PMSF and 1X protease inhibitor) as described (Hsieh et al., 2000). Total protein, 150 µg, was loaded in each lane and separated on 8% SDS-PAGE and transferred to PVDF membrane (Millipore). Protein gel blot analyses involved standard methods (Sambrook and Russell, 2001) with FIN219, PHYA and α-tubulin monoclonal, and COP1 polyclonal antibodies.

# Protoplast Isolation and Subcellular Localization Study

Arabidopsis mesophyll protoplast isolation and transfection were performed as described (Yoo et al., 2007). In brief, 4-week-old well-expanded leaves were peeled (Wu et al., 2009) and incubated in enzyme solution (20 mM MES, 1.5% cellulase R10, 0.4% macerozyme R10, 0.4 M mannitol, 20 mM KCl, 10 mM CaCl2, 5 mM b-mercaptoethanol, and 0.1% bovine serum albumin, pH 5.7) for 1 h. Protoplasts were collected by centrifugation at 100 × g and washed twice with W5 solution (2 mM MES, 154 mM NaCl, 125 mM CaCl2, and 5 mM KCl, pH 5.7). Protoplasts were resuspended in W5 solution and incubated on ice for at least 30 min, then washed with MMG solution (4 mM MES, 0.4 M mannitol, and 15 mM MgCl2, pH 5.7) and resuspended at 2 × 10<sup>5</sup> cells mL−<sup>1</sup> in MMG solution. For protoplast transfection, 200 µL protoplasts was mixed with 10 µg (∼20 µL) DNA and 220 µL PEG solution (40% polyethylene glycol 4000, 0.2 M mannitol, and 100 mM CaCl2), then transfected protoplasts were washed twice, resuspended in W5 solution with 1 µM Dex and incubated overnight in the dark. On the next day, the protoplasts were treated with white light and shade light. The GFP fusion construct p35S:GFP-COP1 was used for transfection. The nuclei of protoplasts were stained with 4<sup>0</sup> ,6-diamidino-2-phenylindole. Transformed protoplasts were visualized by confocal laser scanning microscopy (Leica TCS SP5 Confocal Spectral Microscope) and image processing involved LAS AF lite software from Leica.

FIGURE 3 | FIN219 and phyA have a synergistic effect on shade-mediated hypocotyl elongation and gene expression. (A) Seedlings of fin219-2, phyA 211 and the double mutant phyA 211 fin219-2 were grown under white light (high R:FR) or simulated shade (low R:FR) for 4 days. (B) Quantification of hypocotyl lengths of seedlings shown in A. Data are mean ± SE (n = 25). (C) qRT-PCR of PIL1 and ATHB2 expression in seedlings grown under white light or simulated shade for 4 days standardized to the wild type under white light. Data are mean ± SE from three biological replicates. Different lowercase letters represent significant differences by ANOVA at P < 0.05.

# Co-immunoprecipitation Analysis

Co-immunoprecipitation analysis was performed as described (Chen et al., 2007). The seedlings were ground with grinding buffer (50 mM Tris-HCl, pH 5.7, 150 mM NaCl, 10 mM MgCl2, 0.1% bovine serum albumin, 0.1% Nonidet P-40, 1 mM PMSF, 2X Protease Inhibitor Cocktail [Sigma], and 50 µM MG132). A total of 2 mg of proteins was mixed with beads and incubated at 4◦C for 4 h, then washed three times with same grinding buffer. Pellets were analyzed by standard SDS-PAGE and subjected to protein gel blot analysis.

# RESULTS

# FIN219 Negatively Regulates Hypocotyl Elongation and Gene Expression under Simulated Shade

Jasmonic acid biosynthetic and signaling mutants display enhanced shade-induced hypocotyl elongation (Robson et al., 2010). We previously demonstrated that FIN219/JAR1 integrates phyA-mediated FR light and JA signaling (Wang et al., 2011; Chen et al., 2015). To further understand the role of FIN219 in the shade avoidance response, we examined hypocotyl elongation and gene expression responses under white light (high R:FR) and simulated shade (low R:FR). Seven-day-old seedlings of fin219 mutants [fin219-2 (Wang et al., 2011), jar1-1 (Staswick et al., 1998), and fin219-1 (Hsieh et al., 2000)], and overexpression of FIN219 in a fin219-2 genetic background driven by glucocorticoid-inducible promoter [pGR:FIN219, also pGR219 (Wang et al., 2011)] were compared with the corresponding wild type. Under white light, fin219 mutants showed similar hypocotyl length to that of the wild type. However, under simulated shade, the hypocotyl was longer for all fin219 mutants than the wild type [129.7% (fin219-2), 180.2% (jar1-1) and 197.6% (fin219-1), p < 0.01] (**Figures 1A,B**). The FIN29 overexpression line, pGR219, induced by Dex had a short-hypocotyl phenotype under white light. Although pGR219 responded to low R:FR light, hypocotyl elongation was lower than that of the wild type (**Figures 1A,B**). Furthermore, we quantified the expression of shade-inducible marker genes PIL1 and ATHB2. Under shade, the expression of these genes was induced in the wild type. Under white light, their expression in the fin219 mutants and pGR219 was largely similar to that of the wild type. However, under simulated shade, the expression was enhanced in the fin219 mutants but significantly reduced in pGR219 (**Figures 1C,D**). FIN219 may inhibit shade-induced hypocotyl elongation and gene expression.

# Simulated Shade Reduces FIN219 Protein Level and FIN219 Affects phyA and COP1 Protein Accumulation

The onset of shade avoidance is coupled with transcriptional and translational reprogramming. FIN219 (AtGH11.3) is a GH3 family protein, which is rapidly upregulated in response to auxin (Hsieh et al., 2000). Recently, GH3.3 was found induced by simulated shade (Crocco et al., 2015). We used quantitative

real-time PCR (qRT-PCR) to determine the effect of simulated shade on FIN219 transcript levels. Simulated shade did not alter FIN219 expression in the wild type and fin219 mutants; however, FIN219 mRNA level was induced in pGR219 (Supplementary Figure S1). Furthermore, we used protein gel blot analysis to demonstrate how shade affects FIN219 protein levels. FIN219 protein level was significantly reduced under simulated shade as compared with white light both in the wild type and pGR219 (**Figures 2A–C**). Under white light, FIN219 protein level was lower in the fin219 mutants (the null mutant fin219-2 and jar1-1) than in the wild type. Simulated shade further reduced FIN219 level in these mutants (**Figure 2B**). Thus, shade light may reduce FIN219 protein level, and overaccumulation of FIN219 protein in pGR219 (**Figure 2C**) may inhibit shade responses (**Figures 1A–C**).

We further determined the effect of FIN219 on PHYA and COP1 protein levels under simulated shade. In agreement with previous reports (Martínez-García et al., 2014), under shade light, PHYA protein level was increased in the wild type and slightly reduced in fin219 mutants (**Figures 2A,B**). However, the pGR219 line showed spontaneous accumulation of high levels of PHYA under white light, with no enhanced accumulation under simulated shade (**Figure 2C**). Thus, pGR219 line may be insensitive to shade light.

CONSTITUTIVE PHOTOMORPHOGENIC 1 is essential for hypocotyl and petiole elongation under shade light (Rolauffs et al., 2012). COP1 level does not change with exposure to shade, and COP1 nuclear accumulation is required for full shade responses (Pacín et al., 2013). To determine how FIN219 affects COP1 level, we measured COP1 protein levels. As expected, COP1 level did not change largely under shade light; however, under white light and shade, COP1 level was greater in fin219 mutants and pGR219 than the wild type (**Figures 2B,C**). FIN219 altered COP1 may in turn affect the shade responses.

#### fin219-2-Mediated Shade Responses Are Independent of phyA-Mediated High-Irradiance Response

The phyA mutant exhibits a long hypocotyl phenotype under continuous FR light or continuous white light supplemented with FR light. This response of phyA is due to the FR high-irradiance response (HIR) (Johnson et al., 1994; Yanovsky et al., 1995; Martínez-García et al., 2014). FIN219 is a component of phyAmediated FR light signaling (Hsieh et al., 2000). FIN219-mediated shade responses may be due to phyA-mediated HIR signaling. To clarify this possibility, we analyzed phenotypic responses of the phyA211 fin219-2 to shade (Supplementary Figures S2A,B). Shade-induced hypocotyl elongation and gene expression were measured in fin219-2 and phyA 211 single mutants and phyA211 fin219-2 double mutants under white light and simulated shade. Under simulated shade, hypocotyl elongation and the expression of shade-induced marker genes such as PIL1 and ATHB2 was greater in the double mutant than each of the single mutants (**Figures 3A–C**), which suggests that FIN219-mediated shade response is independent of phyA-mediated FR-HIR.

We showed reduced FIN219 level under shade light. However, in a dark-to-light transition experiment, FIN219 level in the wild type was reduced under light as compared with under dark, with a significant increase in FIN219 level under white

and shade light (**Figure 4A**). Intriguingly, FIN219 level was lower in phyA mutants than the wild type under all light conditions (**Figure 4A** and Supplementary Figure 2B) except dark (**Figure 4A**), which implies that phyA positively regulates FIN219 under short-term shade. Similarly, in contrast to light-grown seedlings (**Figures 2B,C**), in dark-grown seedlings transferred to white light for 5 h, PHYA level was lower in the fin219-2 mutant than the wild type, and under shade light, it was comparable in fin219-2 and wild type (**Figure 4A**). However, under all light conditions examined, PHYA protein level was greater in pGR219 than the wild type (**Figure 4B**). FIN219 and phyA may regulate each other depending on the genotype and light condition, and output of this regulation may depend on a specific light condition.

#### FIN219 Differentially Alters the Expression of PIF5 and PAR1, for Altered Auxin Responses

Several groups of TFs are involved in shade light signaling. In our recent study, we found that FIN219 affects a number of bHLH TFs (Chen et al., 2015). While checking the expression of shade signaling genes, we found that the bHLH TFs PIF5 and PAR1 were affected by FIN219. As expected, PIF5 expression in the wild type did not differ under simulated shade and white light (**Figure 5A**). Disruption of FIN219 expression by mutation (fin219-2 and fin219-1) significantly induced PIF5 expression under simulated shade, which was decreased in pGR219 regardless of light condition (**Figure 5A**). Similarly, the effect of FIN219 on PAR1 expression was significantly pronounced under both white and shade light. Under white light, PAR1 expression was greater in fin219 mutants and pGR219, whereas under shade light, PAR1 expression was reduced in fin219 mutants and remained unchanged in pGR219 (**Figure 5C**). Both PIF5 and PAR1 affect auxin biosynthesis and signaling under shade light (Roig-Villanova et al., 2007; Hornitschek et al., 2012). To check how their altered expression was translated into downstream effects, we evaluate the expression patterns of auxin-response genes IAA29 and PAR1 target SAUR68. The expression of both IAA29 and SAUR68 was enhanced in fin219 mutants and reduced in pGR219 under shade (**Figures 5B,D**). In addition, PIF5 showed high induction and PAR1 reduced

expression in the phyA 211 fin219-2 double mutant under shade (**Figures 6A,C**). Moreover, the expression of both IAA29 and SAUR68 was synergistically increased in phyA 211fin219-2 under simulated shade (**Figures 6B,D**). Thus, FIN219 modulates auxin homeostasis under simulated shade through PIF5 and PAR1.

#### FIN219 Reduces Nuclear Accumulation of COP1 through Direct Physical Interaction under Shade

The cop1 mutants show very short hypocotyls under high and low R:FR conditions. The cop1-4 mutant still shows a residual shade response (Pacín et al., 2013), which is not observed in cop1-6 (Supplementary Figures S3A,B). To demonstrate how fin219 mutation affects cop1 mutants under simulated shade, we assessed the double mutants cop1-4 fin219-2 and cop1-6 fin219- 2 in response to simulated shade. Shade-induced hypocotyl elongation was significantly enhanced although slightly in cop1- 4 fin219-2 as compared with cop1-4, with no difference between cop1-6 fin219-2 and cop1-6 (Supplementary Figures S3A,B).

Enhanced COP1 nuclear accumulation is essential for shade responses (Pacín et al., 2013). Previous studies showed that FIN219 interacted with COP1 under dark and FR light. Moreover, FIN219 overexpression can exclude COP1 from the nucleus to the cytoplasm even in the dark (Wang et al., 2011). To further investigate the effect of FIN219 on COP1 nuclear localization under white and shade light, we performed protoplast transient assays using Arabidopsis mesophyll protoplasts from the wild type, fin219-2 and pGR219 and the construct p35S: GFP-COP1 in a binary vector. Under the dark, GFP-COP1 localized in the nucleus in wild-type protoplasts and fin219-2 (**Figures 7Ba–i,E**) but mostly in the cytoplasm in pGR219 protoplasts (**Figures 7Bk–n,E**). Under white light, GFP-COP1 signal is mostly abundant in the cytoplasm in wild type, fin219-2 and pGR219 protoplasts (**Figures 7Ca–d,f–i,k–n,E**). Upon exposure to simulated shade for 1 h, GFP-COP1 signals mainly concentrated in the nucleus in wild type and fin219-2 (**Figures 7Da–d,f–i,E**). However, overexpression of FIN219 in pGR219 inhibited COP1 nuclear accumulation (**Figures 7Dk– n,E**). Therefore, FIN219 overexpression can induce COP1 accumulation in the cytoplasm in response to simulated shade.

Furthermore, to determine the possibility of physical interaction between FIN219 and COP1 under shade, we used co-immunoprecipitation with wild type, fin219-2 and pGR219 seedlings grown under simulated shade. FIN219 directly interacted with COP1 and phyA (**Figure 8**). Interaction of both molecules was stronger in the FIN219 overexpression line pGR219, which suggests that FIN219 abundance may affect the degree of interactions (**Figure 8**). Hence, FIN219 can regulate the subcellular localization of COP1 under white and simulated shade, and lack of FIN219 regulation of COP1 in fin219 mutants compared to the wild type makes it more sensitive to shade light.

#### DISCUSSION

Hypocotyl elongation is a key adaptation of Arabidopsis seedlings to avoid shade light. Here, we report that FIN219/JAR1, a JA-conjugating enzyme, plays a negative role in regulating shade responses such as hypocotyl elongation and expression of shade components. The double mutant phyA 211fin219-2 had a synergistic effect in response to simulated shade, which suggests that phyA and FIN219/JAR1 work in parallel pathways to regulate shade signaling, which is consistent with the additive levels of TFs PIL1 and ATHB2 and auxin-response genes IAA29 and SAUR68 under the same condition. Moreover, FIN219/JAR1 and PHYA levels were down- and upregulated by shade, respectively. Intriguingly, PHYA and COP1 levels were affected in the fin219 mutants under simulated shade. In particular, FIN219/JAR1 overexpression under shade could change COP1 subcellular localization from the nucleus to the cytoplasm. Further Co-IP studies under shade revealed that FIN219/JAR1, phyA and COP1 interacted with each other. These data indicate that FIN219/JAR1 plays a vital role in regulating shade responses likely by modulating the expression and subcellular location of shade components.

FIN219/JAR1 is a JA-conjugating enzyme, responsible for the formation of JA-Ile and involved in the regulation of plant development and defense responses likely via crosstalk with different hormones and TF-mediated signaling pathways (Chen et al., 2015). Mutation in the FIN219/JAR1 locus resulted in enhanced shade-induced hypocotyl elongation with different degrees (**Figures 1A,B**) (Robson et al., 2010), whereas its overexpression led reduced elongation responses (**Figures 1A,B**), which suggests a negative role of FIN219 in shade signaling. Since FIN219 is an auxin and JA inducible gene and its function in JA signaling is well known, how it modulates other hormone signaling pathways is largely unknown. Our previous studies also indicated that FIN219 could crosstalk with other different hormones (Chen et al., 2015). The fin219-2 is a T-DNA insertion knockout mutant (Wang et al., 2011), whereas jar1-1 is an EMS mutant (Staswick et al., 2002) and fin219-1 is a mis-regulated mutant with changes of methylation status in the promoter of its gene (Hsieh et al., 2000). Thus, jar1-1 and fin219-1 mutants contain reduced levels of proteins compared to wild type. It could be possible that in these mutant lines (jar1-1 and fin219-1), FIN219 perturbs other hormone signaling pathways and makes plants more sensitive to shade. This is our hypothesis that needs further investigation to clarify. In addition, COP1 is necessary

for hypocotyl and petiole elongation under shade light (Rolauffs et al., 2012). Furthermore, B-box-containing 21 (BBX21) acts downstream of COP1 to negatively regulate the shade response (Crocco et al., 2010). FIN219/JAR1 appears to affect COP1 levels under shade (**Figures 2B,C**). Intriguingly, BBX21 negatively regulated FIN219 transcript levels under canopy shade (Crocco et al., 2010), so FIN219-COP1-BBX21 may form a regulatory feedback loop in response to shade environments. Recent studies revealed that phyB and PIFs form a mutually negative feedback loop under continuous red but not shade light (Leivar et al., 2012a). Thus, a local loop regulation initiated by FIN219-COP1- BBX21, leading to reduced levels of active JA-Ile, may play a vital role in fine-tuning shade light signaling.

FIN219/JAR1 acts as a major enzyme for the formation of physiologically active JA-Ile to regulate plant growth and defense responses (Staswick et al., 2002; Wang et al., 2011). Its protein levels should be strictly modulated in response to developmental and environmental cues. So far, multiple photoreceptors, including phyA, and different hormones such as auxin and JAs can regulate FIN219 levels. Current work indicated that pGR219 with FIN219 overexpression has always severely stunted phenotype (**Figure 1**). The expression levels of IAA29 and SAUR68 in pGR219 are always similar with Col-0 under white light (**Figures 5B,D**). Moreover, PIF5 and PAR1 have opposing effects on shade responses, which suggests that FIN219 may trigger major effects on shade-responsive genes mainly under shade conditions. Besides, PIF5 positively regulates and PAR1 negatively affects shade responses. In pGR219, the PIF5 expression is less and PAR1 expression is greater than wild type (**Figures 5A,C**). However, these gene expressions do not alter with response to shade in pGR219. In addition, our previous studies indicated that FIN219 overexpression could exclude COP1 from the nucleus to the cytoplasm even in the dark without greatly altering COP1 level (Wang et al., 2011). Shade also resulted in COP1 accumulation in the nucleus (**Figures 7D,E**). In contrast, pGR219 under shade showed more COP1 accumulation in the cytoplasm and substantial reduction of COP1 in the nucleus compared to wild-type Col-0 under the same condition, which suggests that the accuracy of COP1 subcellular location is critical for shade responses in addition to expression of shade components. Thus, the PGR219 with pronounced short-hypocotyl phenotype under white light and shade conditions may have substantial effects on plant growth and development in addition to light signaling.

Shade light illumination affects a number of genes, including TFs (Sessa et al., 2005; Roig-Villanova et al., 2006). PIF-related TFs are essential for shade-induced responses (Lorrain et al., 2008; Casal, 2013). As expected, wild-type PIF5 transcript levels did not change greatly under shade light. However, FIN219 negatively regulated PIL1 and PIF5 gene expression under low R:FR light (**Figures 1C**, **5A**). As well, the expression of several TFs such as HFR1, ATHB2, PIL1, and PAR1/2 were rapidly induced under shade light (Carabelli et al., 1996; Salter et al., 2003; Sessa et al., 2005; Roig-Villanova et al., 2006). HFR1 and PAR1/2 negatively regulate shade responses (Roig-Villanova et al., 2007; Hornitschek et al., 2009), whereas ATHB2 and PIL1 are positive (Steindler et al., 1999; Salter et al., 2003) and negative regulators, respectively (Li et al., 2014). PAR1 expression was positively regulated by FIN219 under shade (**Figure 5C**). Therefore, FIN219 may regulate the expression of early shade-induced genes positively and negatively. In addition, PIF5 and PAR1 modulate auxin biosynthesis and signaling (Roig-Villanova et al., 2007; Hornitschek et al., 2012). In agreement, the expression of auxin-response genes IAA29 and PAR1 target SAUR68 is associated with PIF5 and PAR1 expression, respectively. As well, the PIF5 and PAR1 expression was higher in the double mutant phyA211 fin219-2 than in each single mutant under shade (**Figures 6A,C**), so FIN219 may act independently of phyAmediated signaling to modulate auxin signaling in response to shade.

In addition, PAR1 transcripts are repressed by PHYA and PHYB overexpression under simulated shade (Roig-Villanova et al., 2006). Its transcripts are also suppressed by phyA and phyB under FR and R light, respectively. The PHYA protein level in fin219 mutants was similar to that in wild type (**Figure 2B**) and it was greater in PGR219 under white light (**Figure 2C**), which suggests that PAR1 transcripts would be less than wild type under white light. However, in **Figure 5C**, PAR1 expression was even greater in both fin219 mutants and PGR219 than in wild type under white light, which implies that PAR1 expression patterns

gene expression and hypocotyl elongation. Regular arrow, positive effect; inverted T, negative effect. Dashed lines represent an indirect regulation.

in fin219 mutants and PGR219 may involve hormone effects such as JAs and auxin. The detailed mechanisms remain elusive.

Phytochromes were shown to play vital roles in shade signaling (Devlin et al., 1999; Franklin and Whitelam, 2005; Casal, 2012). phyB is stable in light and functions as a major photoreceptor in suppression of shade-mediated hypocotyl elongation. Under a high R:FR ratio, the phenotype of the phyB mutant was similar to the wild type under shade. Furthermore, phyD and phyE were redundant to phyB in repressing shade responses, whereas phyC did not play a role in shade responses (Franklin and Quail, 2010). Functional roles of phyA involved in the repression of shade responses remain largely unknown. Here, we found that PHYA levels were increased by simulated shade (low R:FR ∼ 0.05) (**Figures 2**, **4**), which agrees with the report by Martínez-García et al. (2014). An increase in PHYA level on exposure to shade even for 5 h (**Figure 4**) is likely related to the phyA function in fine-tuning the chlorophyll biosynthesis in response to partial shading (Brouwer et al., 2014). Moreover, phyA appears to positively regulate FIN219 levels in short-term exposure to shade (**Figure 4A**), rather than long-term exposure to shade (Supplementary Figure S2B). In turn, increased FIN219 levels leading to enhanced JA-Ile levels resulted in reduced the chlorophyll content under shade. This speculation is consistent with JAs being able to decrease chlorophyll content and reduce photosynthesis in light (He et al., 2002; Zhai et al., 2007). In contrast, with longer exposure to shade, FIN219 level was reduced in the wild type (**Figure 2** and Supplementary Figure S2B) and positively regulated PHYA level (**Figure 2B**). Thus, reduced FIN219 level under shade may involve other mechanisms, and both FIN219 and phyA may regulate each other in response to low R:FR light.

CONSTITUTIVE PHOTOMORPHOGENIC 1 is a negative regulator of photomorphogenesis and encodes an E3 ubiquitin ligase. Under the dark, COP1 accumulates in the nucleus and results in the degradation of positive regulators such as HY5 of photomorphogenesis. Upon light exposure, COP1 is inactivated by migrating to the cytoplasm and other unknown mechanisms. Shade light can trigger COP1 accumulation rapidly in the nucleus, and it positively modulates the shade responses (Pacín et al., 2013). Moreover, COP1 physically interacts with FIN219 under the dark and continuous FR light and restricts its nuclear accumulation on overexpression of FIN219 (Wang et al., 2011). We found COP1 mainly accumulated in the cytoplasm and unable to migrate to the nucleus in pGR219 under the conditions examined, including shade for 1 h (**Figures 7B–E**), which significantly affects hypocotyl elongation (**Figures 1A,B**). Thus, abundant FIN219 protein affects the total content of active COP1 protein in the nucleus. Ethylene promotes hypocotyl elongation in the light by triggering COP1 nuclear accumulation, which enhances HY5 degradation (Yu et al., 2013). FIN219 levels

responsible for the formation of JA-Ile, an active form of JA, are likely tightly regulated in response to low R:FR to modulate the subcellular localization of COP1.

FIN219/JAR1 is a cytoplasmic protein in FR light and remains in the same location even under shade (Hsieh et al., 2000; **Figure 7**). Co-IP studies indicated that FIN219/JAR1 dosedependently interacted with phyA and COP1 proteins under shade (**Figure 8**). FIN219 overexpression excluded COP1 from the nucleus to the cytoplasm under FR light (Wang et al., 2011). As well, ectopic expression of FIN219 under low R:FR light changed the COP1 subcellular location from the nucleus to the cytoplasm because COP1 was localized in the nucleus under shade light. However, FIN219 level in the wild type was reduced by shade light (**Figure 2**) and was increased by 5-h shade as compared with white light (**Figure 4**). FIN219 and COP1 interaction likely occurred in the cytoplasm at the early stage of shade light exposure. This speculation is consistent with transient expression at 1-h shade that resulted in more accumulation of GFP-COP1 in the cytoplasm by ectopic expression of FIN219 (pGR219) (**Figure 7**).

In addition, the levels of PHYA were increased and stabilized by continuous and transient shade light (**Figures 2**, **4**), which leads to suppression of hypocotyl elongation, with an antagonistic effect on phyB deactivation by low R:FR light. PhyA and phyB likely function dynamically to modulate hypocotyl elongation in response to changes in R:FR ratios in natural environments. In addition, FIN219 and phyA positively regulated each other under shade (**Figures 2**, **4**, **8**). phyA is activated by FR light and migrates to the nucleus. Whether FIN219 affects phyA subcellular localization, leading to their interaction in the cytoplasm upon exposure to shade light, remains to be elucidated.

FIN219/JAR1 is a JA-conjugating enzyme and interacts with multiple partners, including COP1 (Chen et al., 2007; Wang et al., 2011; **Figure 8**). These data suggest that FIN219 may have a dual function with an enzymatic and protein–protein interaction activities. Collectively, our data show that FIN219/JAR1, a JAconjugating enzyme, functions as a negative regulator in shade signaling and may work with phyA and COP1 in response to

#### REFERENCES


shade. Alternatively, low R:FR ratio may inactivate phyB by reducing its abundance in the nucleus and decrease FIN219/JAR1 levels, thereby leading to increased PIF5 level and reduced PAR1 level and increased COP1 accumulation in the nucleus. These events activate downstream genes such as IAA29 and SAUR68 and shade responses, including hypocotyl elongation (**Figure 9**).

#### AUTHOR CONTRIBUTIONS

SS and H-LH designed the experiments; SS and H-WJ performed experiments; SS and H-LH did data analyses; SS and H-LH wrote the manuscript.

# FUNDING

This work was supported by the National Science Council, Taiwan (NSC 101-2311-B-002-002-MY3) and Ministry of Science and Technology (MOST 104-2311-B-002-035-MY3).

# ACKNOWLEDGMENTS

We are grateful to the Arabidopsis Biological Research Center (Ohio State University, Columbus) for the fin219-2 and jar1-1 seeds. We thank the Technology Commons staff in the College of Life Science, National Taiwan University, for technical assistance with confocal microscopy and real-time PCR analyses. SS is a postdoctoral fellow supported by the Excellence Research Program of National Taiwan University (104R4000).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2017.01901/ full#supplementary-material

Casal, J. J. (2012). Shade avoidance. Arabidopsis Book 10:e0157. doi: 10.1199/tab. 0157


by shade involves differential regulation of protein stability of MYC transcription factors and their JAZ repressors in Arabidopsis. Plant Cell 26, 1967–1980. doi: 10.1105/tpc.114.125047



**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 Swain, Jiang and Hsieh. 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.

# Hormonal Regulation in Shade Avoidance

#### Chuanwei Yang and Lin Li\*

State Key Laboratory of Genetic Engineering, Institute of Plant Biology, School of Life Sciences, Fudan University, Shanghai, China

At high vegetation density, shade-intolerant plants sense a reduction in the red (660 nm) to far-red (730 nm) light ratio (R/FR) in addition to a general reduction in light intensity. These light signals trigger a spectrum of morphological changes manifested by growth of stem-like tissue (hypocotyl, petiole, etc.) instead of harvestable organs (leaves, fruits, seeds, etc.)—namely, shade avoidance syndrome (SAS). Common phenotypical changes related to SAS are changes in leaf hyponasty, an increase in hypocotyl and internode elongation and extended petioles. Prolonged shade exposure leads to early flowering, less branching, increased susceptibility to insect herbivory, and decreased seed yield. Thus, shade avoidance significantly impacts on agronomic traits. Many genetic and molecular studies have revealed that phytochromes, cryptochromes and UVR8 (UV-B photoreceptor protein) monitor the changes in light intensity under shade and regulate the stability or activity of phytochrome-interacting factors (PIFs). PIF-governed modulation of the expression of auxin biosynthesis, transporter and signaling genes is the major driver for shade-induced hypocotyl elongation. Besides auxin, gibberellins, brassinosteroids, and ethylene are also required for shade-induced hypocotyl or petiole elongation growth. In leaves, accumulated auxin stimulates cytokinin oxidase expression to break down cytokinins and inhibit leaf growth. In the young buds, shade light promotes the accumulation of abscisic acid to repress branching. Shade light also represses jasmonate- and salicylic acid-induced defense responses to balance resource allocation between growth and defense. Here we will summarize recent findings relating to such hormonal regulation in SAS in Arabidopsis thaliana, Brassica rapa, and certain crops.

#### Edited by:

Yunde Zhao, University of California, San Diego, United States

#### Reviewed by:

Yi Tao, Xiamen University, China Rongcheng Lin, Institute of Botany (CAS), China Christian Fankhauser, University of Lausanne, Switzerland

> \*Correspondence: Lin Li linli@fudan.edu.cn

#### Specialty section:

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

Received: 26 June 2017 Accepted: 21 August 2017 Published: 04 September 2017

#### Citation:

Yang C and Li L (2017) Hormonal Regulation in Shade Avoidance. Front. Plant Sci. 8:1527. doi: 10.3389/fpls.2017.01527 Keywords: shade avoidance syndrome, light signaling, PIFs, hormone regulation, crosstalk

#### INTRODUCTION

Over the past few decades, a substantial body of studies has focused on understanding how plants sense the proximity of neighbors, how they respond at molecular levels, and how they adjust their morphological and physiological indexes. Many important light signaling components have been shown to regulate the shade avoidance responses—for example, PIFs (phytochrome interacting factors), HFR1 (long hypocotyl in far-red 1), PAR1/2 (phytochrome rapidly regulated 1/2) and COP1 (constitutive photomorphogenic 1). Meanwhile, various phytohormones are also involved and coordinated to shape shade-regulated plant architecture. Analyses of hormonal biosynthetic and signaling mutants, combined with studies of exogenous hormone applications, have implicated the roles of these phytohormones in multiple shade avoidance responses. In this review, we provide an overview of the current understanding of shade light and subsequent hormonal regulation.

#### SHADE SIGNAL AND PLANT PERCEPTION

fpls-08-01527 August 31, 2017 Time: 17:8 # 2

Light-quality signals are of paramount importance in detecting neighboring vegetation. Photosynthetic pigments in leaves absorb strongly in the range of photosynthetically active radiation (PAR) (400–700 nm) and UV radiation (280–400 nm), and reflect far-red wavelength (700–800 nm) (Casal, 2013). Thus, natural shade is a combination of the reduction in the red/far-red ratio (R/FR), the reduction in red plus far-red irradiance, the reduction in blue and UV irradiance, and the reduced blue/green ratio. To detect these spectral differences, plants use multiple light sensors, such as red and far-red light absorbing phytochromes, the blue/UV-A light sensing cryptochromes, and the UV-B photoreceptor protein (UVR8).

### A BRIEF ACCOUNT OF THE SHADE SIGNALING PATHWAY

Phytochromes exist in two photoconvertible forms: an inactive R-absorbing Pr form and an active FR-absorbing Pfr form. The steady-state ratio of Pr and Pfr forms depends on R/FR. The constitutive shade avoidance syndrome (SAS) phenotype of Arabidopsis phyB mutant plants indicates that phyB plays a dominant role in inhibiting SAS (Franklin and Quail, 2010). High R/FR establishes a high proportion of phyB Pfr, which interacts with the bHLH family of transcription factor PIFs and triggers the phosphorylation, ubiquitination and degradation of PIFs. In contrast, low R/FR drives Pfr-to-Pr conversion and releases the suppression of PIFs. Activated PIFs promote gene expression related to shade-induced growth. PIF7, PIF4 and PIF5 play central roles in this process (Lorrain et al., 2008; Li L. et al., 2012).

To prevent exaggerated shade-avoidance responses, shadeinduced HFR1 (Sessa et al., 2005; Hornitschek et al., 2009), PAR1/2 (Roig-Villanova et al., 2006; Galstyan et al., 2011; Bou-Torrent et al., 2014), and PIL1 (PIF3 like 1) (Li et al., 2014; Luo et al., 2014) are proposed as the negative regulators of PIFs. The bZIP transcription factor, elongated hypocotyl 5 (HY5), is also reported to form non-functional complexes with PIFs (Chen et al., 2013; Toledo-Ortiz et al., 2014). In addition to directly binding with PIFs, the Suppressor of phyA-105 (SPA)/COP1 E3 ubiquitin ligase complex indirectly enhances PIF activity by degrading HFR1 and HY5 to augment shade responses (Sheerin et al., 2015; Pacin et al., 2016). BBX (double B-box) 21 and BBX25 regulate shade response through the function in the COP1 signaling pathway (Crocco et al., 2010; Gangappa et al., 2013).

Cryptochromes (CRYs) are involved in repressing a low bluemediated SAS by regulating PIF abundance and activity (de Wit et al., 2016; Pedmale et al., 2016). PIF activity is enhanced directly through CRY inactivation and indirectly through relieved inhibition of COP1, which increases the degradation of negative regulators of PIF, including HFR1 and HY5 (de Wit et al., 2016).

UV-B-mediated inhibition of shade responses has been reported to occur through the degradation of PIF4/5 (Hayes et al., 2014).

In summary, downstream of photoreceptors, PIFs, as the key regulators, determine the massive transcriptional reprogramming upon perception of shade light, and also mediates the convergence between light and hormones.

# AUXIN, A PROMINENT PLAYER IN SHADE-INDUCED ELONGATION GROWTH

A forward genetic screen for impaired shade-induced hypocotyl elongation in Arabidopsis identified TAA1, an enzyme catalyzing the first step of an auxin biosynthetic pathway (Tao et al., 2008; Won et al., 2011). Later, a family of enzymes encoded by YUCCA (YUC) genes has been functionally positioned as the second and rate-limiting step of TAA1-dependent auxin biosynthesis (the indole-2-pyruvic acid pathway, or "IPA pathway"). The transcriptional regulation of YUCCA genes by PIF7 has been found to link photoperception with auxin biosynthesis (Li L. et al., 2012). The level of shade-stimulated free indole-3 acetic acid (IAA) is blunted in taa1, and pif7 mutants confirm that auxin production through the TAA1-YUC pathway is required to initiate the SAS in seedlings (Tao et al., 2008; Li L. et al., 2012; Procko et al., 2014). PIF4 and PIF5 are partially redundant, with PIF7 regulating the expression of YUCCA genes (Hornitschek et al., 2012). Correspondingly, the yuc2 yuc5 yuc8 yuc9 quadruple mutant displays the completely disrupted SAS (Nozue et al., 2015; Muller-Moule et al., 2016). Tissue-level measurement in Brassica rapa seedlings has suggested that auxin appears to be generated in the cotyledons and transported to the hypocotyl (Procko et al., 2014). Indeed, seedlings treated with the auxin transport inhibitor naphthylphalamic acid (NPA) totally abolish shade-induced hypocotyl elongation (Tao et al., 2008). Consistently, pin3-3 (PIN3, auxin transporter) exhibits an impaired shade-induced hypocotyl elongation (Keuskamp et al., 2010), and the mutation in SAV4 leads to defective basipetal auxin transport and shade responses (Ge et al., 2017), indicating that auxin redistribution is important for shade-avoidance reactions (Morelli and Ruberti, 2000).

Besides auxin biosynthesis and transport, auxin sensitivity is also enhanced under shade (Nozue et al., 2011; Hornitschek et al., 2012; Bou-Torrent et al., 2014). Auxin signaling components, such as AUX/IAAs (Auxin/indole-3-acetic acid), have been reported to modulate the SAS (Steindler et al., 1999; Procko et al., 2016).

In addition to Arabidopsis, the key role of auxin on the SAS has also been confirmed in crop species (Carriedo et al., 2016). Shade-induced changes in auxin level have been found in sunflower (Kurepin et al., 2007) and tomato (Kozuka et al., 2010). Expression quantitative trait locus (eQTL) analysis identified a group of auxin-related genes, which were down-regulated in shade-tolerant tomato lines and up-regulated in the shade responders, suggesting the role of auxin in the natural variation of the SAS (Bush et al., unpublished). In maize seedlings (Wang et al., 2016) and rice seedlings (Liu et al., 2016), the expression of auxin-responsive genes is also dramatically affected by shade treatment.

Considered together, it may be concluded that intact auxin biosynthesis, transportation and signaling are required for shadeinduced stem growth.

#### GIBBERELLIN, ANOTHER SHADE GROWTH-PROMOTING HORMONE

Shade treatment resulted in an increased gibberellin (GA) concentration in bean internode (Beall et al., 1996), cowpea (Vigna sinensis) epicotyls (Martínez-García et al., 2000), sunflower stem (Kurepin et al., 2007) and Arabidopsis seedling (Bou-Torrent et al., 2014). The shade-induced GA biosynthetic enzymes GA20ox1, GA20ox2, and GA3ox at least in part account for the increase in active GA (Hisamatsu et al., 2005; Yu et al., 2015).

Bioactive GA leads to proteasomal degradation of DELLA proteins (Harberd et al., 2009). Lacking direct DNA binding capability, DELLAs are direct interactors of PIFs. Their binding prevents PIF proteins from binding DNA and thus negatively regulates the expression of genes involved in cell elongation (de Lucas et al., 2008; Feng et al., 2008). Shadeinduced breakdown of DELLA proteins due to increased gibberellin biosynthesis releases the suppression of PIFs, and activates the transcription of target genes. The GAinsensitive gai gain-of-function mutant, which has a stable GAI (DELLA) protein, shows a reduced SAS (Djakovic-Petrovic et al., 2007), suggesting that DELLA proteins constrain the SAS.

It is noteworthy that proteins that physically interact with DELLA proteins may alleviate DELLA-mediated repression of PIF activity, such as BBX24. The shade-response defect in bbx24 mutants is rescued by a GA treatment (Crocco et al., 2015).

In addition to GA-induced seedling phenotypes, GA biosynthesis and signaling are also important for shade-induced flowering. Silencing GA20ox2 expression delays flowering of Arabidopsis exposed to a FR-enriched light condition (Hisamatsu and King, 2008).

### ETHYLENE, AN ORGAN-SPECIFIC REGULATOR OF THE SAS

Low R/FR can enhance the production of ethylene in widetype tobacco (Pierik et al., 2004). In Arabidopsis, shade-induced petiole elongation was absent in the ethylene-insensitive mutants ein2-1 and ein3-1eil1-3, indicating that ethylene is a positive regulator of shade-induced petiole elongation (Pierik et al., 2009). However, the ein3eil1 mutant retains a full shadeinduced hypocotyl response (Das et al., 2016). The controversy suggests that ethylene plays a role in organ-specific shade response.

A recent research shows that light activation of photoreceptor phyB results in rapid degradation of EIN3, a master transcription factor in the ethylene signaling pathway (Shi et al., 2016). The position of ethylene signaling components under shade is worthy of further investigation.

#### BRASSINOSTEROID, A DYNAMIC REGULATOR UNDER SHADE

The promotion of stem growth by shade light requires brassinosteroids (BRs) because the BR biosynthesis mutant dwarf1 (Luccioni et al., 2002) and rot3 (Kim et al., 1998) are unable to show the elongation of hypocotyl under shade, as with wild-type seedlings treated with the BR synthesis inhibitor brassinazole (Keuskamp et al., 2011). BR biosynthesis is also required for petiole growth under low R/FR (Kozuka et al., 2010). However, short-term (4 h) simulated shade treatments resulted in lower levels of the active BR, and longer periods (24 h) abolished the differences in BR levels in whole seedlings (Bou-Torrent et al., 2014), suggesting that simulated shade altered BR levels in a dynamic fashion.

Beside the level of hormones, the sensitivity of seedlings to hormones also has an important effect on shade-induced growth. BR signaling components BR-ENHANCED EXPRESSION (BEE) and BES1-INTERACTING MYC-LIKE (BIM) are positive regulators of SAS hypocotyl responses because bee123 and bim123 seedlings display hypocotyl elongation defects after detecting simulated shade (Cifuentes-Esquivel et al., 2013). Remarkably, DELLAs negatively regulate BR signaling by binding BZR1 and reducing the expression of BR-responsive genes (Bai et al., 2012; Gallego-Bartolome et al., 2012; Li Q.F. et al., 2012). The transcription factor BZR1 and PIF4 physically interact and synergistically regulate target genes (Oh et al., 2012; Kohnen et al., 2016). Given that the binding of DELLA and PIFs impair the DNA-binding ability of PIFs, the complex of DELLAs, BZR1, and PIFs may play a role in stem elongation, and possibly exerts a similar function in shade avoidance, but this needs further investigation (Casal, 2013; de Lucas and Prat, 2014). In concordance with these findings, BR-responsive genes are overrepresented in end-of-day FR-induced genes in both the leaf blade and petiole (Kozuka et al., 2010). Although the majority of the BR genomic response comprises genes annotated as auxin responsive, the regulation of BR and auxin on SAS responses might nevertheless occur in a non-redundant and non-synergistic manner, because the response to blue light depletion will be fully inhibited only when both hormones are blocked simultaneously (Keuskamp et al., 2011).

In particular, the BR response appears to be required for the full expression of the SAS phenotypes under low blue light (Keller et al., 2011; Keuskamp et al., 2011). The question as to how BR biosynthesis and signaling dynamically respond to low R/FR or low blue light is yet to be answered.

### CYTOKININ, ENSURING REALLOCATION OF PLANT RESOURCES

The role of cytokinins (CKs) in shade avoidance responses was discovered from the response of plants to vertical light

intensity gradients in leaf canopies (Pons et al., 2001). In shaded leaves, where stomatal conductance and transpiration rate are reduced, the low delivery rate of CKs leads to reduced photosynthetic capacity and ultimately senescence (Boonman and Pons, 2007).

Another role of CKs was found in the inhibition of leaf growth in shade. Low R/FR signal can induce hypocotyl elongation and also trigger a rapid arrest of leaf-primordia growth by the breakdown of auxin-induced CKs through the action of AtCKX6 (cytokinin oxidase) in the incipient vein cells of developing primordia (Carabelli et al., 2007). In addition, the CK receptor AHK3 has been reported to mediate the root-to-hypocotyl ratio response under shade conditions (Novak et al., 2015).

The reduction of bioactive CKs triggers a reduced photosynthetic capacity and a transient arrest of leaf development, ensuring that energy resources are indeed redirected into extension growth in shade.

#### JASMONIC ACID, SHADE-REDUCED HORMONE RELATED TO DEFENSE

Plants often display a weak defense in insect and pathogen infection under shade conditions or FR-enriched conditions (Cerrudo et al., 2012; de Wit et al., 2013; Ballare, 2014). Shade has been shown to reduce herbivory-induced jasmonic acid (JA) accumulation (Agrawal et al., 2012), and FR-exposed plants suffer more insect herbivory than wild-type plants (Moreno et al., 2009), suggesting that shade can down-regulate the JA pathway to control plant immunity.

The JAZ-DELLA pathway is an important modulator of plant immunity under shade conditions (Moreno and Ballare, 2014). DELLA proteins positively regulate JA signaling by interacting with JAZs, and this interaction weakens the ability of JAZs to repress MYC2 (Hou et al., 2010; Yang et al., 2012). As described previously, DELLA proteins negatively regulate growth-related genes by binding PIFs (de Lucas et al., 2008; Feng et al., 2008). JAZ10 is required for the inhibitory effect of shade on JA responses (Leone et al., 2014). Therefore, shade conditions induce GA synthesis and the degradation of DELLA proteins, consequently increasing PIF-dependent growth and impairing JAZ-dependent defense. Canopy shade represses JA-mediated defenses via shade-induced stabilization of JAZ proteins and triggers inactivation of MYC2, MYC3, and MYC4 proteins (Chico et al., 2014). By contrast, regulation of the protein stability of MYCs and JAZs by shade facilitates reallocation of resources from defense to growth. The mutants deficient in JA biosynthesis and signaling display exaggerated shade-induced hypocotyl responses to a low R/FR ratio (Robson et al., 2010). Moreover, several FR light induced gene expressions are dependent on CORONATINE INSENSITIVE1 (COI1), a central component of JA signaling (Robson et al., 2010).

Canopy light cues affect emission of constitutive and methyl JA-induced volatile organic compounds, which can be detected by herbivorous insects (Kegge et al., 2013). A recent study found that in tomato (Solanum lycopersicum) phyB inactivation led the plants to produce a blend of JA-induced monoterpenes that increased their attractiveness to the predatory mirid bug Macrolophus pygmaeus (Cortes et al., 2016; Ballare and Pierik, 2017).

Certain transcription factors in the JA signaling pathway also participate in the regulation of SAS; for example, PHYTOCHROME AND FLOWERING TIME 1 (PFT1), a subunit of Mediator, is required for both JA-dependent defense gene expression and shade-induced early flowering (Cerdan and Chory, 2003; Cevik et al., 2012; Inigo et al., 2012). These factors could be the additional linkers of light signal and JA-mediated defenses.

#### SALICYLIC ACID, ANOTHER SHADE-REDUCED HORMONE

Salicylic acid (SA)-dependent disease resistance is also reduced under shade, which is considered as the early warning signal for plant competition (de Wit et al., 2013). Reduced SA synthesis (Griebel and Zeier, 2008) and response (de Wit et al., 2013) have been correlated with phyB inactivation. Under a low R/FR ratio, the phosphorylation level of the SA-signaling component NONEXPRESSOR of PATHOGENESIS-RELATED GENE 1 (NPR1) is reduced, which partly explains why shade reduces SAdependent disease resistance. A more detailed explanation of the mechanism that exists between shade avoidance responses and SA is required.

#### ABSCISIC ACID, REPRESSING BRANCHING UNDER SHADE

Abscisic acid (ABA) is commonly known as the "stress hormone" that responds to a variety of environmental stresses including both biotic and abiotic stress. Shade conditions increase ABA levels in sunflower (Helianthus annuus) (Kurepin et al., 2007) and tomato leaves (Cagnola et al., 2012). Shade increases the endogenous ABA level probably by enhancing the transcript levels of ABA biosynthetic gene NINE-CIS-EPOXYCAROTENOID DIOXYGENASE 3 (NCED3) and NCED5, particularly in hypocotyls (Kohnen et al., 2016). Several ABA signaling genes (ABF3, AFP1, AFP3, and GBF3) are up-regulated by a neighbor signal (Sellaro et al., 2017).

Shade light exerts a strong influence on branch development (Finlayson et al., 2010; Su et al., 2011). One recent study suggested that shade represses branching in bud n-2 by accumulation of ABA (Reddy et al., 2013). The genes involved in ABA biosynthesis and signal transduction showed varied gene expression patterns in responsive buds with increasing R/FR treatment. ABA biosynthesis mutants (nced3-2 and aba2-1) exhibited enhanced branching capacity under low R/FR.

However, ABA was not involved in shade-induced petiole elongation (Pierik et al., 2011), suggesting that the roles of ABA in the SAS may be organ specific.

#### STRIGOLACTONE, AN UNCLEAR ROLE IN THE SAS

Most shade-avoiding plants display reduced branching and enhanced apical growth, which helps them to compete for incident light. Strigolactone (SL) is one of the hormones that control lateral shoot growth. In Arabidopsis, BRC1 (BRANCHED1) is up-regulated in the axillary buds of plants grown at high density and is required for shade-mediated branch suppression (Aguilar-Martinez et al., 2007; Gonzalez-Grandio et al., 2013). In sorghum, inhibition of outgrowth in a phyB mutant and by FR treatment is correlated with an increase in the transcript levels of the SL-signaling gene SbMAX2 in buds (Kebrom et al., 2010). The involvement of SL in SAS has been observed, but more detailed studies of this mechanism are required.

Besides branching, Arabidopsis max2 mutants show longer hypocotyls under red, far-red and blue light than wild–type plants (Shen et al., 2012; Jia et al., 2014). The double mutant pif1max2

#### REFERENCES


shows a similar hypocotyl length to max2, which indicates that MAX2 is epistatic to PIF1 (Shen et al., 2012). MAX2 plays a role in the light signaling pathway, but further investigation of the mechanisms involved is needed.

# KARRIKINS, A POSSIBLE WAY TO ATTENUATE THE SAS

Studies have shown that karrikins enhance the sensitivity of seedlings to light (Waters and Smith, 2013). Since karrikins can inhibit elongation of the hypocotyl and increase the chlorophyll content (Nelson et al., 2010), they may be an efficient solution to attenuating plant SAS during the seedling stage (Meng et al., 2016).

# FINAL REMARKS

This review focused on understanding the interaction between phytohormones and the SAS (**Figure 1**). The regulations of these phytohormones on the SAS described here might vary according to tissue type (Kohnen et al., 2016), stage of development (Roig-Villanova and Martinez-Garcia, 2016) and species (Liu et al., 2016). In this regard, further research into the spatial and temporal regulation of phytohormones is necessary for a mechanistical understanding of the SAS. Moreover, crosstalk among hormones under shade conditions is also worthy of further investigation.

# AUTHOR CONTRIBUTIONS

CY and LL designed and wrote the manuscript.

# FUNDING

This work was supported by the National Key Research and Development Program of China SQ2017YFJC040047-01 and the National Natural Science Foundation of China Grants 31470374 and 31500973.

### ACKNOWLEDGMENTS

We are very thankful to Christian Fankhauser (University of Lausanne, Switzerland) for his critical comments on the manuscript.

within axillary buds. Plant Cell 19, 458–472. doi: 10.1105/tpc.106. 048934




of growth-promoting bHLH transcription factors. Plant J. 53, 312–323. doi: 10.1111/j.1365-313X.2007.03341.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 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.

# Phytohormone and Light Regulation of Chlorophyll Degradation

Xiaoyu Zhu1, 2†, Junyi Chen1, 2†, Kai Qiu1, 2† and Benke Kuai 1, 2 \*

*<sup>1</sup> State Key Laboratory of Genetic Engineering and Fudan Center for Genetic Diversity and Designing Agriculture, School of Life Sciences, Fudan University, Shanghai, China, <sup>2</sup> Ministry of Education, Key Laboratory for Biodiversity Science and Ecological Engineering, Institute of Biodiversity Science, Fudan University, Shanghai, China*

Degreening, due to the net loss of chlorophyll (Chl), is the most prominent symptom during the processes of leaf senescence, fruit ripening, and seed maturation. Over the last decade or so, extensive identifications of *Chl catabolic genes* (*CCGs*) have led to the revelation of the biochemical pathway of Chl degradation. As such, exploration of the regulatory mechanism of the degreening process is greatly facilitated. During the past few years, substantial progress has been made in elucidating the regulation of Chl degradation, particularly via the mediation of major phytohormones' signaling. Intriguingly, ethylene and abscisic acid's signaling have been demonstrated to interweave with light signaling in mediating the regulation of Chl degradation. In this review, we briefly summarize this progress, with an effort on providing a framework for further investigation of multifaceted and hierarchical regulations of Chl degradation.

#### Edited by:

*Benoit Schoefs, University of Maine, France*

#### Reviewed by:

*Carole Anne Llewellyn, Swansea University, United Kingdom Avtar Krishan Handa, Purdue University, United States*

\*Correspondence:

*Benke Kuai bkkuai@fudan.edu.cn*

*† These authors have contributed equally to this work.*

#### Specialty section:

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

Received: *14 August 2017* Accepted: *23 October 2017* Published: *06 November 2017*

#### Citation:

*Zhu X, Chen J, Qiu K and Kuai B (2017) Phytohormone and Light Regulation of Chlorophyll Degradation. Front. Plant Sci. 8:1911. doi: 10.3389/fpls.2017.01911* Keywords: chlorophyll degradation, phytohormone, ethylene, abscisic acid, jasmonic acid, light

# INTRODUCTION

Chlorophyll (Chl) molecules are synthesized almost instantly upon light exposure of seedlings for harvesting light energy to drive photosynthesis in green organs, and during the processes of leaf senescence, fruit ripening, and seed maturation, they are degraded rapidly, a process called degreening, to facilitate nutrient remobilization and, in some cases, vitamin biosynthesis (Christ and Hörtensteiner, 2014; Vom Dorp et al., 2015). Chl degradation is in fact imperative to plant development for its detoxifying the photo-toxicity of Chl molecules once they are freed from their binding proteins (Hörtensteiner, 2006; Li et al., 2017). Over the last decade or so, the major biochemical pathway of Chl degradation has been revealed by cloning and function analysis of Chl catabolic genes (CCGs). Because of an important role of the pheophorbide a oxygenase (PAO) in Chl degradation, this pathway is designated as PAO pathway (Christ and Hörtensteiner, 2014; **Figure 1**).

In higher plants, there are two forms of Chl molecules, Chl a and Chl b. Chl a is the degradable form of Chls, and, during leaf senescence, Chl b is converted to Chl a by Chl b reductase [CBR, encoded by NON-YELLOW COLORING 1 (NYC1) and NYC1-LIKE (NOL)] and 7-hydroxymethyl Chl a reductase (HCAR) (Kusaba et al., 2007; Horie et al., 2009; Sato et al., 2009; Meguro et al., 2011). For Chl a degradation, Magnesium is initially removed to convert Chl a to pheophytin a (Phein a) by Magnesium-dechelatase, encoded by Mendel's green cotyledon genes, NON-YELLOWINGs/STAY-GREENs (NYEs/SGRs) (Armstead et al., 2007; Ren et al., 2007; Chen et al., 2016; Shimoda et al., 2016; Wu et al., 2016). Phein a is then hydrolyzed by pheophytinase (PPH) to produce pheophorbide a (Pheide a) and phytol (Morita et al., 2009; Schelbert et al., 2009; Ren et al., 2010). Remarkably, the green color of Chl catabolites is completely lost when the porphyrin ring of Pheide a is cleaved by PAO, resulting in oxidized red Chl catabolite (RCC), which

is subsequently catalyzed by red Chl catabolite reductase (RCCR) to generate primary fluorescent Chl catabolite (pFCC) (Wüthrich et al., 2000; Pružinská et al., 2003; Pruzinská et al., 2007; Tanaka et al., 2003; Yao and Greenberg, 2006). Finally, the pFCC is modified and transported into the vacuole, and isomerized to non-fluorescent products by acidic pH (Christ et al., 2012, 2013; Hauenstein et al., 2016).

Phytohormones and environmental factors have long been known to regulate Chl degradation (Lim et al., 2007); however, the molecular mechanisms involved in these regulations remains largely unknown. In last few years, the success in revealing the biochemical pathway of Chl degradation has led to a rapid progress in elucidation of the molecular mechanisms. Particularly, substantial progress has been made on elucidation of the regulatory roles of ethylene, abscisic acid (ABA), jasmonic acid (JA), and light signaling components on Chl degradation, and a number of regulatory factors of CCGs have been identified by using the methods of biochemistry, genetics, and bioinformatics (Delmas et al., 2013; Liang et al., 2014; Sakuraba et al., 2014, 2016; Song et al., 2014; Qiu et al., 2015; Zhang et al., 2015; Zhu et al., 2015; Gao et al., 2016; Ghandchi et al., 2016; Li et al., 2016; Oda-Yamamizo et al., 2016; Yin et al., 2016; Chen et al., 2017; Mao et al., 2017; **Table 1**). These advances provide some valuable insight into the complexity of the molecular mechanism of hormone- and light-regulated Chl degradation. Here, we review recent progress in this field and discuss important yet unresolved questions regarding the roles and mechanisms of phytohormones and environmental factors in Chl degradation regulation.

# THE MOLECULAR MECHANISM OF ETHYLENE SIGNALING-MEDIATED CHL DEGRADATION

Ethylene is an important phytohormone, regulating diverse aspects of plant growth and development, especially leaf degreening and fruit ripening (Burg, 1973; Grbic and Bleecker, 1995; Lim et al., 2007; Qiu et al., 2015; Yin et al., 2016). During leaf degreening, the expression of ethylene biosynthetic genes encoding 1-Aminocyclopropane-1-carboxylic acid (ACC) synthase (ACS) and ACC oxidase (ACO) were significantly up-regulated, and the endogenous ethylene level increased accordingly (van der Graaff et al., 2006; Breeze et al., 2011). ACO1 antisense tomato plants synthesized less ethylene and delayed leaf degreening (John et al., 1995). ACSs octuple mutant, producing ∼10% of ethylene in WT, significantly delayed leaf degreening in Arabidopsis (Tsuchisaka et al., 2009). Exogenous application of ethylene could induce leaf degreening, whereas treatment with ethylene inhibitors could delay leaf degreening (Serek et al., 1995; Jing et al., 2005). The leaves of etr1-1, the mutant of ethylene receptor gene ETR1, cannot respond to ethylene treatment and shows a stay-green leaf phenotype (Bleecker et al., 1988; Grbic and Bleecker, 1995; Chao et al., 1997). Consistently, ectopic expression of a mutant form of the Arabidopsis ethylene receptor gene ETR1-1 delayed leaf Chl degradation in Nicotiana tabacum (Yang et al., 2008). ETHYLENE INSENSITIVE 2 (EIN2) and its downstream target EIN3 are key components of ethylene signaling, and the mutants of both EIN2 and EIN3 exhibit a severe stay-green phenotype



*<sup>a</sup>Transient over-expression of AtERF17 and SlERF16, which are the homologs of CitERF13 in Arabidopsis and tomato, can lead to Chl degradation in Nicotiana tabacum leaves (Yin et al., 2016).*

*<sup>b</sup>The null mutant of AtNAP has a significant stay-green phenotype during leaf and silique senescence (Guo and Gan, 2006; Kou et al., 2012).*

*<sup>c</sup>The prematurely senile 1 (ps1-D) is a gain-of-function mutant of OsNAP (Liang et al., 2014).*

*<sup>d</sup>OsNAC2 is a rice ortholog of ORE1/ANAC092/AtNAC2 (Mao et al., 2017).*

*<sup>e</sup>Over-expression of OsMYC2 significantly promote Chl degradation during leaf senescence in rice (Uji et al., 2017).*

*<sup>f</sup> Transient over-expression of oilseed rape BnaNAC55 (Brassica napus L.) lead to a significant decrease in Chl content in Nicotiana benthamiana leaves (Niu et al., 2016). <sup>g</sup>SOC1 is a negative regulator of Chl degradation during leaf degreening and senescence (Chen et al., 2017).*

during leaf senescence (Chao et al., 1997; Oh et al., 1997). EIN3 positively regulates ORE1 and NAP, the two important regulatory genes of senescence, either directly or indirectly via negatively regulating miR164, which in turn cleaves the transcript of ORE1 (Kim et al., 2009, 2014; Li et al., 2013). These reports convincingly demonstrate that ethylene signaling regulates the pathway of Chl degradation.

Recently, Qiu et al. (2015) reported that the expression of NYC1, NYE1, and PAO was significantly induced by ethylene treatment in the leaves of Arabidopsis, whereas largely repressed in ein3 eil1 double mutant. The electrophoretic mobility shift assay (EMSA) and dual-luciferase reporter assay demonstrated that EIN3 protein could directly bind to the EBS (EIN3 binding site, AC/TGA/TAC/TCT) in the promoters of NYC1, NYE1, and PAO, and enhance their promoter activity in Arabidopsis protoplasts. Therefore, EIN3 is a positive regulator of ethylene-mediated Chl degradation. Moreover, ORE1, the direct target of EIN3, could bind to the promoters of NYE1, NYC1, NOL, and PAO, and positively regulate their expression. Intriguingly, EIN3 and ORE1 could promote NYE1 and NYC1 expression in an additive manner (Qiu et al., 2015). This progress indicates that EIN3 and EIL1 constitute a major regulatory node of ethylene-triggered degreening, with EIN3 either directly or indirectly regulating the expression of CCGs. Notably, Yin et al. (2016) recently revealed that CitERF13, an ethylene responsive factor, could bind to CitPPH promoter and positively regulate its expression during citrus fruit degreening (**Table 1**).

#### THE MOLECULAR MECHANISM OF ABA SIGNALING-MEDIATED CHL DEGRADATION

ABA can be induced by age-dependent senescence or environmental stresses, such as drought, heat, and salt, and the increase of endogenous ABA level or the exogenous application of ABA accelerates chlorosis and senescence of plant leaves (Raab et al., 2009; Yang et al., 2014; Takasaki et al., 2015; Liu et al., 2016). ABA has therefore long been recognized as a positive regulator of degreening during leaf senescence in plants. It was reported that ABA accelerates leaf degreening and senescence via an AtNAP-SAG113 (a PP2C family protein phosphatase) regulatory module that is involved in the regulation of the stomata movement (Zhang and Gan, 2012).

With an attempt of investigating the direct regulation of CCGs, Gao et al. (2016) initially identified ABF3 as a transcriptional regulator of NYE1 by yeast one-hybrid (Y1H) screening. Further in vitro and in vivo analyses indicated that ABF2/3/4 directly bind to the promoter of NYE1, and

up-regulate its transcription. Notably, ABF2/3/4 also bind to the promoters of NYE2, NYC1, and PAO, and up-regulate their transcription. Under ABA treatment, detached leaves of abf2 abf3 abf4 triple mutants exhibited an obvious staygreen phenotype, while those of ABF4-OE transgenic lines showed an accelerated yellowing phenotype (Gao et al., 2016). ABI5 and EEL, two ABA signaling-related transcription factors, were also found to positively regulate the transcription of NYE1 and NYC1 by binding to their promoters (Sakuraba et al., 2014). Similarly, ANAC016, a senescence-associated NAC transcription factor, directly bind to the promoter of NYE1 and up-regulate its transcription. Leaves of anac016 mutant showed a stay-green phenotype, while ANAC016-OX line displayed an early leaf yellowing phenotype. Interestingly, it indirectly activates ABSCISIC ALDEHYDE OXIDASE3 (AAO3), an ABA biosynthesis gene, via a mediation of NAP (Kim et al., 2013; Yang et al., 2014; Sakuraba et al., 2016). Liang et al. (2014) found that ABA-induced leaf yellowing and senescence were mediated by OsNAP in rice. Unlike AtNAP, OsNAP was specifically induced by ABA but not ethylene. OsNAP directly bind to the promoters of OsSGR, OsNYC1, OsNYC3 (PPH), and OsRCCR1, and up-regulated their transcription in rice. Recently, Mao et al. (2017) reported that OsNAC2 could directly bind to the promoters of OsSGR and OsNYC3, and activate their expression during ABA-induced leaf yellowing and senescence in rice.

ABA also regulates seed maturation. During the processes of seed maturation and embryo degreening, a B3 domain transcription factor ABI3 directly binds to the promoters of NYE1 and NYE2, and up-regulates their transcription, consequently promoting Chl degradation in embryos. Intriguingly, the role of ABI3 in Chl degradation is seedspecific, as the mutant of ABI3 (abi3-6) does not show a stay-green leaf phenotype in the dark (Delmas et al., 2013). This progress has shed a light on the complex molecular mechanism underlying ABA-regulated Chl degradation (**Table 1**).

#### THE MOLECULAR MECHANISM OF JA SIGNALING-MEDIATED CHL DEGRADATION

Jasmonic acid is a phytohormone essential for the regulation of multiple developmental processes, including leaf degreening and senescence (Wasternack and Hause, 2013). Ueda and Kato (1980) firstly found that methyl jasmonate (MeJA) could induce leaf degreening in oats. Subsequently, this phenomenon was confirmed in various plant species such as Arabidopsis, wheat, rice, and maize (Beltrano et al., 1998; He et al., 2002; Shan et al., 2011; Yan et al., 2012; Lee et al., 2015). Mutants defective for JA synthesis exhibited delayed leaf degreening phenotype (Castillo and León, 2008; Schommer et al., 2008; Yan et al., 2012). COI1-JAZ complex is the co-receptor of JA (Sheard et al., 2010), and the leaves of coi1 mutant exhibit a stay green phenotype upon MeJA treatment (He et al., 2002; Shan et al., 2011; Lee et al., 2015). MYC2/3/4 could interact with JAZ, acting as the transcriptional activators in JA signaling, whereas bHLH03/13/14/17 were identified as the transcriptional repressors, repressing JA responses. Both MYC2/3/4 and bHLH03/13/14/17 could bind to the promoter of SAG29, and activate or repress the expression of SAG29 during JA-induced leaf senescence (Qi et al., 2015).

In a study of identifying the transcriptional regulators of CCGs, Zhu et al. (2015) revealed MYC2 as a putative transregulator of PAO by using the Y1H screening. MYC2 and its two homologs, MYC3 and MYC4, were then demonstrated to directly bind to the G-box (CACGTG) in the promoters of PAO, NYC1, and NYE1, and up-regulate their expression during JA-induced Chl degradation. The leaves of myc2 myc3 myc4 triple mutant showed a stay-green phenotype, whereas those of MYC2/3/4 overexpression lines displayed an accelerated yellowing phenotype upon MeJA treatment. Intriguingly, ANAC019/055/072, the immediate targets of MYC2/3/4 (Bu et al., 2008; Zheng et al., 2012), could also directly up-regulate the expression of NYE1, NYE2, and NYC1. The triple mutant of anac019 anac055 anac072 showed a similar stay-green phenotype as myc2 myc3 myc4 upon MeJA treatment. Moreover, the MYC2 and ANAC019 could interact with each other, and synergistically enhance NYE1 expression in Arabidopsis protoplasts. These findings indicate a hierarchical and coordinated regulatory network during JA-induced Chl degradation (Zhu et al., 2015; **Table 1**).

#### THE MOLECULAR MECHANISM OF LIGHT SIGNALING IN REGULATING CHL DEGRADATION

Light is the vital environmental factor for plant growth and development. Dark treatment, a simple and effective way for light deprivation, is widely used for studying leaf senescence and degreening (Ren et al., 2007; Christ and Hörtensteiner, 2014). phyB is a red light receptor (Schäfer and Bowler, 2002), and seedlings or mature leaves of phyB mutant yellow faster, whereas PHYB-OX plants yellow slower than those of WT during dark incubation (Sakuraba et al., 2014). phyB represses PIF4 and PIF5 at the post-transcriptional level (Leivar et al., 2008; Shin et al., 2009). In the dark, leaves of pif4, pif5, and pif1 pif3 pif4 pif5 quadruple mutants all show stay-green phenotypes, while those of PIF4-OX and PIF5-OX lines show early-yellowing phenotypes (Sakuraba et al., 2014). ELF3 inhibits leaf degreening and senescence by repressing PIF4 and PIF5 at the transcriptional level (Nusinow et al., 2011; Sakuraba et al., 2014). After incubating in darkness, leaves of elf3 senesced faster and leaves of ELF3-OX senesced slower than those of WT (Sakuraba et al., 2014). These findings collectively suggest that red light signaling is involved in the regulation of leaf degreening and senescence, with PIF4 and PIF5 acting as key mediators.

Both PIF4 and PIF5 associate with the promoters of ABI5 and EEL, two bZIP family transcription factors, and up-regulate their transcription (Sakuraba et al., 2014). Interestingly, PIF4, PIF5, ABI5, and EEL, as well as EIN3, can all activate the expression of ORE1, which encodes an important senescencepromoting transcription factor, by directly binding to its promoter. Meanwhile, ABI5 and EEL could directly activate NYE1 and NYC1 by binding to their promoters (Sakuraba et al., 2014). It was further demonstrated that PIF4 directly bind to the promoter of NYE1, and PIF5 to the promoters of NYE1 and NYC1 to up-regulate their transcription (Song et al., 2014; Zhang et al., 2015). Under dark treatment, endogenous ethylene level was significantly reduced in the leaves of pif4 mutant, while elevated in those of PIF4-OX lines. When treated with ethylene, mutants of pif3, pif4, and pif5 showed stay-green phenotypes, suggesting that PIF3/4/5 play roles in leaf degreening mediated by ethylene signaling (Song et al., 2014).

Recently, in a study designed for exploring the transcriptional regulation of PPH, Chen et al. (2017) demonstrated that SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1), a flowering pathway integrator, associates with the promoter of PPH, and negatively regulates its transcription. Under dark treatment, leaves of soc1-6 mutant yellowed earlier, whereas those of iSOC1-OE lines partially stayed green, in comparison to their respective controls. Moreover, SOC1 also negatively regulates NYE1 and SAG113 at the transcriptional level during dark-induced leaf degreening and senescence. Notably, SOC1 is the only negative regulator of Chl degradation identified so far (**Table 1**).

#### CONCLUSION AND PERSPECTIVES

Chl degradation is an active and progressive process which is regulated by diverse developmental and environmental clues, and mainly mediated by phytohormones' signaling. In Arabidopsis, ethylene signaling promotes leaf degreening through the transcriptional regulation of major CCGs by both EIN3 and ORE1, while in citrus fruits by CitERF13 (Qiu et al., 2015; Yin et al., 2016). The severe stay-green phenotype of the mutants of both EIN3/EIL1 and ORE1 implies that ethylene signaling is likely the major signaling pathway in regulating degreening during developmental leaf senescence (Kim et al., 2009; Li et al., 2013). ABA signaling mediates Chl degradation at the transcriptional level mainly by ABI3 during seed maturation, whereas, during leaf senescence, by ABI5, EEL, and ABF2/3/4 as well as ANAC016 in Arabidopsis, and by OsNAP and OsNAC2 in rice (Delmas et al., 2013; Liang et al., 2014; Sakuraba et al., 2014, 2016; Gao et al., 2016; Mao et al., 2017). Interestingly, these transcription factors have long been known to regulate drought stress/circadian clock (Sanchez et al., 2011), indicating that ABA signaling might be mainly involved in the regulation of leaf degreening-triggered by abiotic stresses. JA signaling directly regulates leaf degreening by MYC2/3/4 and ANAC019/055/072 (Zhu et al., 2015). Considering that the MYCs and ANACs are also involved in the regulation of defense responses, JA signaling likely mediates the degreening process incurred by biotic stresses. Light signal, on the other hand, inhibits leaf degreening by both maintaining the transcription of SOC1 and repressing the transcription of PIFs/reducing PIFs protein accumulation (Sakuraba et al., 2014; Song et al., 2014; Zhang et al., 2015; Chen et al., 2017). Intriguingly, major hormones share their signaling components with light, as loss-of-function mutations of major hormone signaling components (EIN2, EIN3/EIL1, ABI5, EEL, NAP, ORE1, etc.) block light signaling in regulating degreening, causing stay-green phenotypes upon light deprivation, whereas loss-of-functions of major light signaling components, PIFs, also interfere major hormone (e.g., ethylene) signaling in the promotion of degreening (Oh et al., 1997; Guo and Gan, 2006; Li et al., 2013; Kim et al., 2014; Sakuraba et al., 2014; Song et al., 2014).

Although, substantial progress has been made in exploring the molecular regulation of Chl degradation, numerous issues still await to be addressed. (1) There appears to be a "developmental window" for hormone-induced Chl degradation. Ethylene, for example, cannot readily induce leaves to degreen at their young age, and only after a certain developmental stage will leaves allow ethylene to induce their degreening (Jing et al., 2005). What is the molecular basis for the "window effect"? (2) As an inhibitor of Chl degradation, light signal is present during ethylene-, ABA-, and JA-induced or age-dependent leaf degreening (Qiu et al., 2015; Zhu et al., 2015; Gao et al., 2016), but how ethylene, ABA, or JA signaling antagonize light signaling to trigger Chl degradation? (3) There are enormous cross-talks among different hormone signaling pathways which are interweaved with light signaling in the regulation of Chl degradation. It was reported that ein3 exhibited a stay-green phenotype during MeJA treatment (Li et al., 2013), and jaz7 showed an early yellowing phenotype under dark treatment (Yu et al., 2016). More work need to be done to elucidate those cross-talks. (4) In addition to ethylene, ABA, and JA, other phytohormones are also found to be involved in the regulation of Chl degradation, with salicylic acid and brassinolide acting as promoters (Morris et al., 2000; Jeong et al., 2010), whereas cytokinin and gibberellic acid as repressors (Fletcher and Osborne, 1966; Lara et al., 2004; Kim et al., 2006). Yet, their regulatory pathways or networks are largely unexplored. (4) Thus far, studies on Chl degradation regulation mainly focus on the transcriptional level. Further investigations need to be extended to post-transcriptional levels, including the translational regulation and post-translational modification. It has been reported that PAO could be interconverted between phosphorylated and dephosphorylated status (Chung et al., 2006).

# 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 Natural Science Foundation of China (31670287, 31170218) (to BK) and National Postdoctoral Program for Innovative Talents (BX201700058) (to JC).

# REFERENCES


chlorophyll degradation in Arabidopsis. J. Integr. Plant Biol. 52, 496–504. doi: 10.1111/j.1744-7909.2010.00945.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 Zhu, Chen, Qiu and Kuai. 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.

# Interplay between Light and Plant Hormones in the Control of Arabidopsis Seedling Chlorophyll Biosynthesis

Xiaoqin Liu, Yue Li and Shangwei Zhong\*

State Key Laboratory of Protein and Plant Gene Research, School of Advanced Agricultural Sciences and School of Life Sciences, Peking University, Beijing, China

Chlorophyll biosynthesis is one of the most important cellular processes and is essential for plant photosynthesis. After germination under the soil, dark-grown seedlings are etiolated and accumulate the chlorophyll precursor protochlorophyllide (Pchlide) in cotyledons. Upon exposure to light, Pchlide is rapidly converted to chlorophyll to initiate photoautotrophic growth. In this light-regulated de-etiolation process, multiple endogenous phytohormones are also involved. Although the co-regulation of seedling greening by light and hormones has long been observed, recent studies greatly advanced our understanding of their interplay by identifying the key components connecting these pathways. The integrators, such as PHYTOCHROME-INTERACTING FACTORs, ELONGATED HYPOCOTYL 5, ETHYLENE INSENSTIVE 3 and DELLA proteins, are key transcription regulators in light or hormone signaling pathways. This review focuses on these integrators and illustrates the regulatory networks of light and hormone interactions in chlorophyll biosynthesis.

#### Edited by:

Chi-Kuang Wen, Shanghai Institutes for Biological Sciences (CAS), China

#### Reviewed by:

Rongcheng Lin, Institute of Botany (CAS), China Hao Peng, Washington State University, United States

#### \*Correspondence:

Shangwei Zhong shangwei.zhong@pku.edu.cn

#### Specialty section:

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

Received: 30 June 2017 Accepted: 03 August 2017 Published: 17 August 2017

#### Citation:

Liu X, Li Y and Zhong S (2017) Interplay between Light and Plant Hormones in the Control of Arabidopsis Seedling Chlorophyll Biosynthesis. Front. Plant Sci. 8:1433. doi: 10.3389/fpls.2017.01433 Keywords: light signaling, plant hormones, chlorophyll biosynthesis, de-etiolation, PIFs, HY5, EIN3/EIL1, DELLAs

# INTRODUCTION

Crop seeds are usually buried in soil, whereby post-germinative seedlings become etiolated and grow upward heterotrophically from seed reserves. Upon reaching the soil surface, etiolated seedlings undergo a dramatic developmental transition termed de-etiolation, which includes cotyledon opening and greening (Von Arnim and Deng, 1996; Chen et al., 2004). This transition is of particular vulnerability for plant survival, requiring rapid initiation of photoautotrophic growth without causing photooxidation (Huq et al., 2004; Zhong et al., 2014). To successfully accomplish this, chlorophyll biosynthesis must be strictly controlled.

In higher plants, chlorophyll is initially biosynthesized from glutamate, which is then converted to 5-aminolevulinic acid (ALA) and further converted to protochlorophyllide (Pchlide) (Tanaka et al., 2011). When the dark-grown seedlings are exposed to light, the rate-limiting enzymes NADPH protochlorophyllide oxidoreductases (PORs) are photoactivated and catalyze the conversion of Pchlide to chlorophyllide, which is subsequently esterified to mature chlorophyll (Fujita, 1996; Reinbothe et al., 2010). In Arabidopsis, three POR enzymes, PORA, PORB and PORC have been identified, with PORA/PORB playing the main roles in young seedlings (Buhr et al., 2008). Because Pchlide is extremely phototoxic, its amount must be stoichiometrically matched with the level of POR enzymes. Over-accumulation of the free Pchlide that cannot be converted to chlorophyll in time will result in the production of reactive oxygen species (ROS) upon light

exposure, causing photooxidative damage to the seedlings (op den Camp et al., 2003; Huq et al., 2004; Chen et al., 2013; Zhong et al., 2014). To survive, seedlings have evolved efficient ways to adjust the levels of Pchilde and POR enzymes to allow for rapid establishment of photosynthesis without causing photobleaching (op den Camp et al., 2003; Huq et al., 2004; Zhong et al., 2014). Moreover, carotenoid biosynthesis is also markedly upregulated to protect the etioplasts from photooxidative damage by quenching excess excitation energy when seedlings are exposed to light (Rodriguez-Villalon et al., 2009).

Light is the main environmental factor that regulates the pathway of chlorophyll biosynthesis, while plant hormones are also recruited to mediate the developmental switch of de-etiolation. Some key components in the light signaling pathway, such as PIFs and HY5, connect light signals to the signaling pathways of multiple phytohormones, including ethylene, gibberellin (GA) and cytokinin (CK). In this review, we concentrate on how chlorophyll biosynthesis is cooperatively regulated by light and endogenous hormone signals, focusing on the interplay between light and hormone signaling pathways during seedling de-etiolation.

### Chlorophyll Biosynthesis Regulated by Key Components in Light Signaling Pathway

Light provides plants with energy for photosynthesis and a major source of information about their environment. Both light quality and quantity are constantly monitored by plants through a group of photoreceptors (Quail, 2002; Chen et al., 2004). Among them, phytochromes (phys, including phyA-phyE in Arabidopsis) sense far-red and red light (Quail, 2002; Chen et al., 2004). The perception of light signals by phys initiates an intracellular transduction to alter the expression of nuclear genes (Quail, 2002; Chen et al., 2004; Leivar and Quail, 2011). There are two groups of transcription factors, PIFs and HY5, that mediate light-induced responses in opposite ways (Von Arnim and Deng, 1996; Chen et al., 2004; Leivar and Quail, 2011). PIFs are negative regulators and are directly targeted by photoactivated phys for degradation (Ni et al., 1998, 2014; Huq et al., 2004; Shen et al., 2008), while HY5 is stabilized by light to promote photomorphogenesis (Ang et al., 1998; Osterlund et al., 2000). During the process of de-etiolation, both PIFs and HY5 have been shown to transcriptionally regulate the chlorophyll biosynthesis pathway.

PHYTOCHROME-INTERACTING FACTORs accumulate in dark-grown seedlings and negatively regulate the tetrapyrrole metabolism of chlorophyll biosynthesis (Huq et al., 2004; Monte et al., 2004; Moon et al., 2008; Shin et al., 2009; Stephenson et al., 2009). Mutation of PIF1 and PIF3 results in an excessive amount of Pchlide in the dark and causes severe photobleaching upon light exposure (Huq et al., 2004; Monte et al., 2004; Moon et al., 2008; Shin et al., 2009; Stephenson et al., 2009). Further experiments reveal that PIF1 directly binds to the promoter of PORC, while PIF3 represses the expression of HEMA1, GUN4 and CHLH genes (Moon et al., 2008; Stephenson et al., 2009). HEMA1 is the main glutamyl-tRNA reductase that catalyzes the rate-limiting step for ALA biosynthesis, while GUN4 and CHLH promote the conversion of ALA to the chlorophyll biosynthetic branch (Stephenson and Terry, 2008; Tanaka et al., 2011). In addition, PIF5 has been shown to be involved in the negative regulation of CHLH gene expression in etiolated seedlings (Shin et al., 2009), and a large portion of nuclear-encoded chlorophyll biosynthesis genes are notably upregulated in the pifQ mutant (lacking PIF1, PIF3, PIF4 and PIF5 genes) (Leivar et al., 2009; Shin et al., 2009). PIF1 was also found to partly repress the transposase-derived transcription factor FHY3/FAR1-activated gene expression of HEMB1 that encodes the ALA dehydratase (Tang et al., 2012), and chromatinremodeling enzyme BRM interacts with PIF1 to modulate PORC expression (Zhang et al., 2017). Moreover, PIFs have been reported to directly repress the gene expression of PSY (phytoene synthase), which is the main rate-determining enzyme of carotenoid biosynthesis (Toledo-Ortiz et al., 2010). When PIFs are degraded by light, carotenoids are rapidly synthesized to coordinate with chlorophyll biosynthesis, thus facilitating the assembly of functional photosynthetic machinery (Toledo-Ortiz et al., 2010). Therefore, PIFs play important roles in the fine tuning of tetrapyrrole metabolism, directly or indirectly regulating chlorophyll biosynthesis and photosynthetic genes to optimize the seedling greening process.

ELONGATED HYPOCOTYL 5 functions downstream of the photoreceptors and central repressors in the light signaling pathway to promote seedling photomorphogenesis. In the dark, HY5 is degraded through the COP1/DET1-mediated ubiquitination degradation pathway (Ang et al., 1998; Osterlund et al., 2000). HY5 plays a vital role in the convergence of blue, red and far-red light-signal pathways for regulating the transcription levels of HEMA1 (McCormac and Terry, 2002). Several nuclearencoding photosynthetic and chlorophyll biosynthesis genes, such as CHLH, GUN4, PORC, CAO and CHL27, are the putative targets of HY5 (Lee et al., 2007). Although roots are heterotrophic organs, lots of chlorophyll accumulates in light-grown det1 and cop1 mutant roots, and HY5 mediates the process of chlorophyll synthesis in roots (Chory and Peto, 1990; Deng et al., 1992; Ang et al., 1998). In addition, a Myb-like transcription factor REVEILLE1 (RVE1) was recently found to act downstream of phyB to modulate chlorophyll biosynthesis by directly activating PORA expression (Xu et al., 2015; Jiang et al., 2016).

# Ethylene Is Crucial for Cotyledon Greening and Survival of Seedling Soil Emergence

Plant hormones are small molecules that mediate a myriad of cellular responses. Many hormones are involved in light-induced seedling greening. One prominent factor affecting chlorophyll biosynthesis is ethylene, which dramatically represses Pchlide accumulation and induces the gene expression of both PORA and PORB in etiolated seedlings (Zhong et al., 2009, 2010, 2014). Thus, ethylene plays a critical role in protecting cotyledons from photooxidative damage when the seedlings are exposed to light. The effects of ethylene are mediated by EIN3/EIL1, the master transcription factors in the ethylene signaling pathway (Chao

et al., 1997; Guo and Ecker, 2004). EIN3/EIL1 markedly repress the accumulation of Pchlide and directly bind to the promoters of PORA and PORB to activate their gene expression (Zhong et al., 2009, 2010, 2014). Genetic studies reveal that EIN3/EIL1 cooperate with PIF1 and act downstream of COP1 in promoting seedling greening (Zhong et al., 2009). The protein levels of EIN3 are enhanced by COP1 but are decreased by light (Zhong et al., 2009; Shi et al., 2016a,b). In addition, overexpressing EIN3 rescues the far-red light-triggered cotyledon greening defects (Zhong et al., 2009).

After germination in soil, the mechanical impedance of soil boosts ethylene production to adjust seedling morphogenesis to enhance the lifting capacity and protect against mechanical injuries (Zhong et al., 2014; Shen et al., 2016; Shi et al., 2016a). EIN3/EIL1 directly activate two independent pathways, an ERF1 pathway to slow down cell elongation and a PIF3 pathway to control Pchlide biosynthesis (Zhong et al., 2012, 2014). These two pathways are coupled to maintain a suitable amount of Pchilde to rapidly initiate photoautotrophic growth without causing photooxidation upon emergence (Zhong et al., 2014). When seedlings penetrate their way toward the surface, the dim light under the soil increases gradually and represses COP1 protein activity (Shi et al., 2016a). COP1 has been found to be the E3 ligase of EBF1 and EBF2, the F-box proteins of the E3 ligases for EIN3 degradation (Shi et al., 2016a). Therefore, COP1 and ethylene mediate the soil-imposed light and mechanical stress signals, respectively, to adjust EIN3 protein levels in response to soil condition changes when seedlings grow upward in the soil (Shi et al., 2016a). Interestingly, EIN3 also promotes the nuclear enrichment of COP1 protein to generate a positive feedback for EIN3 stability regulation (Yu et al., 2013, 2016). At the moment of emergence and reaching sunlight, photoactivated photoreceptor phyB directly interacts with EIN3 and rapidly degrades EIN3 by bringing it to the E3 ligases EBF1 and EBF2 (Shi et al., 2016b). As a result, the repression of photomorphogenesis by EIN3 and ethylene is rapidly lifted to initiate de-etiolation effectively.

# Gibberellin Regulates Chlorophyll Biogenesis Partially via the Light Signaling Pathway

Seedling de-etiolation is also subject to gibberellin (GA) regulation, as inhibiting gibberellin signaling can induce partial photomorphogenesis in the dark (Alabadi et al., 2004, 2008). DELLAs are a subfamily of the GRAS transcriptional regulators and negatively regulate gibberellin signaling to repress GAmediated responses (Jiang and Fu, 2007). Moreover, DELLAs inhibit the transcription activity of PIF3 and PIF4 through direct blocking of the DNA-recognition domain of these factors (de Lucas et al., 2008; Feng et al., 2008). In dark-grown seedlings, DELLAs accumulate and regulate the biosynthetic pathways of both carotenoid and chlorophyll (Cheminant et al., 2011). DELLAs upregulate the expression of genes involved in chlorophyll biosynthesis (CHLH, PORC and CAO) and photosynthesis (LHCB2.2, PSAG and PSAE-1) in a PIFdependent manner (Cheminant et al., 2011). In addition, DELLAs also positively regulate PORA and PORB gene expression independently of PIFs and repress ROS-induced photooxidative damage during de-etiolation (Cheminant et al., 2011). However, the regulation of HY5 on gibberellin-mediated chlorophyll biosynthesis seems more moderate than that of PIFs in dark conditions (Cheminant et al., 2011).

# Cytokinin Plays an Important Role in Chlorophyll Biosynthesis and Chloroplast Development

Exogenous cytokinin treatment induces cotyledon expansion and chloroplast partial differentiation (Chory et al., 1994; Vandenbussche et al., 2007). Two GATA family transcription factors, GNC and CGA1/GNL, are induced by cytokinin and regulate the expression of many chloroplast-related genes (Hudson et al., 2011; Chiang et al., 2012). Dark-grown seedlings display small etioplasts with prolamellar bodies in the absence of cytokinin, while large lens-shaped plastids contain some prothylakoid membranes in the presence of cytokinin (Chory et al., 1994). Recent reports indicate that cytokinin mediates the etioplast-to-chloroplast transition by promoting characteristic ultrastructural changes (Cortleven and Schmulling, 2015; Cortleven et al., 2016). Cytokinin signal is perceived by the receptors AHK2 and AHK3 and transduced to B-type ARR transcription factors (Argyros et al., 2008). ARRs directly regulate the expression of genes in chlorophyll biosynthesis and the light harvesting complex, such as HEMA1 and LHCB6 (Cortleven and Schmulling, 2015; Cortleven et al., 2016). As cytokinin has been reported to increase the protein levels of HY5 (Vandenbussche et al., 2007), it is possible that HY5 is a point of convergence between light and cytokinin signaling pathways.

# The Function of Other Plant Hormones in Regulating Seedling Greening

In addition to the well-documented hormones just described, other hormones are also important in regulating seedling greening. Auxin represses HY5 protein accumulation via IAA14 and its regulatory target ARFs in roots (Kobayashi et al., 2012). Moreover, chlorophyll synthesis genes are markedly activated in detached roots via cytokinin but are repressed by auxin (Kobayashi et al., 2017), suggesting that auxin signaling is also involved in the regulation of chlorophyll biosynthesis in the root greening response. However, further analyses are required to elucidate the regulatory network of auxin and light signals in regulating chlorophyll biosynthesis. Brassinosteroid (BR) is known to be involved in the process of de-etiolation. Many chlorophyll biosynthesis genes are upregulated from the microarray data of BR-insensitive bri1-116 seedlings in darkness (Sun et al., 2010). The key transcriptional factor GATA2 has been identified in mediating the crosstalk between BR and light signaling pathways (Luo et al., 2010). Recently, ABI4 was found to activate COP1 expression to repress seedling de-etiolation (Xu et al., 2016). In addition, strigolactones are reported to also be involved in light signaling via regulating the nuclear localization of COP1 (Tsuchiya et al., 2010), and jasmonate inhibits COP1 activity to promote photomorphogenesis (Zheng et al., 2017). However, the signaling pathway of ABA, strigolactones and

jasmonate in regulating chlorophyll biosynthesis remains largely unknown.

# CONCLUSION AND PERSPECTIVES

Involvement of plant hormones in light-regulated seedling greening has been known for decades. However, we have not identified the molecular links connecting light signaling to the multiple hormonal pathways until recent years. The key transcription factors of both light and hormone signaling pathways appear to be the integrators (**Figure 1**). EIN3 directly activates the gene expression of PORA/PORB and represses Pchlide accumulation to optimize the greening process. The repression of EIN3 in synthesizing Pchlide is through activating PIF3 transcription, whereas both phyB and COP1 predominantly regulate the protein levels of EIN3. PIFs play a pivotal role in integrating light and GA signals, and DELLAs directly sequester the transcription activity of PIFs. In addition, HY5 protein stability is regulated by auxin and cytokinin to coordinate these signals in mediating root greening, while COP1 could be new integrator as its nuclear localization can be regulated by ethylene, strigolactone and jasmonate hormones. Further studies, such as identifying additional integrators in light and hormonal signaling pathways and addressing how these components are integrated in regulating seedling greening, are needed. Moreover, we are only beginning to address the regulation of chloroplast development. Whether and how plant hormones regulate the etioplast-chloroplast differentiation process is critical in filling the gaps of greening. In summary, although we have not obtained a detailed network depicting how seedling greening is regulated by light and all the hormonal signals, the identification of key transcription regulators as signaling integrators has created a great starting point.

# AUTHOR CONTRIBUTIONS

SZ proposed the topic. SZ, XL, and YL collected the literature and critically assessed the information. XL and SZ wrote the manuscript.

# FUNDING

This work was supported by grants from the National Key Research and Development Program of China (2016YFA0502900) and the National Science Foundation of China (31570188) to SZ. XL was supported by a China Post-doctoral Science Foundation Grant (2016M600857) and the Outstanding Post-doctoral Fellowship of Peking-Tsinghua Center for Life Sciences.

#### REFERENCES

fpls-08-01433 August 14, 2017 Time: 11:59 # 5


# ACKNOWLEDGMENT

We apologize to our colleagues whose work could not be included because of space constraints.


release of singlet oxygen in arabidopsis. Plant Cell 15, 2320–2332. doi: 10.1105/ tpc.014662


**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 Liu, Li and Zhong. 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.

fpls-08-01433 August 14, 2017 Time: 11:59 # 6

# Cytokinins and Abscisic Acid Act Antagonistically in the Regulation of the Bud Outgrowth Pattern by Light Intensity

Adrien Corot<sup>1</sup> , Hanaé Roman<sup>1</sup> , Odile Douillet<sup>1</sup> , Hervé Autret<sup>1</sup> , Maria-Dolores Perez-Garcia<sup>1</sup> , Sylvie Citerne<sup>2</sup> , Jessica Bertheloot<sup>1</sup> , Soulaiman Sakr<sup>1</sup> , Nathalie Leduc<sup>1</sup> and Sabine Demotes-Mainard<sup>1</sup> \*

1 IRHS, Université d'Angers, INRA, Agrocampus-Ouest, SFR 4207 QUASAV, Beaucouzé, France, <sup>2</sup> Institut Jean-Pierre Bourgin Centre de Versailles-Grignon (IJPB), INRA, Agro-ParisTech, CNRS, Versailles, France

Bud outgrowth is a key process in the elaboration of yield and visual quality in rose crops. Although light intensity is well known to affect bud outgrowth, little is known on the mechanisms involved in this regulation. The objective of this work was to test if the control of bud outgrowth pattern along the stem by photosynthetic photon flux density (PPFD) is mediated by sugars, cytokinins and/or abscisic acid in intact rose plants. Rooted cuttings of Rosa hybrida 'Radrazz' were grown in growth chambers under high PPFD (530 µmol m−<sup>2</sup> s −1 ) until the floral bud visible stage. Plants were then either placed under low PPFD (90 µmol m−<sup>2</sup> s −1 ) or maintained under high PPFD. Bud outgrowth inhibition by low PPFD was associated with lower cytokinin and sugar contents and a higher abscisic acid content in the stem. Interestingly, cytokinin supply to the stem restored bud outgrowth under low PPFD. On the other hand, abscisic acid supply inhibited outgrowth under high PPFD and antagonized bud outgrowth stimulation by cytokinins under low PPFD. In contrast, application of sugars did not restore bud outgrowth under low PPFD. These results suggest that PPFD regulation of bud outgrowth in rose involves a signaling pathway in which cytokinins and abscisic acid play antagonistic roles. Sugars can act as nutritional and signaling compounds and may be involved too, but do not appear as the main regulator of the response to PPFD.

#### Keywords: bud burst, PPFD, PAR, cytokinins, ABA, sugars, Rosa hybrida L., rose bush

#### INTRODUCTION

Branching – or tillering for monocots – is a key process in the elaboration of crop yield and quality. Branching can indeed affect yield directly – through the number of fertile branches (Bredmose, 1993; Valério et al., 2009) – or indirectly by increasing the competitive ability of crops against weeds (Lemerle et al., 1996; Zhao et al., 2006) or by constraining pest infestation (Simon et al., 2012). In ornamental plants such as rose bush, plant architecture is derived from branching, and influences plant visual quality and thus consumers' preferences (Boumaza et al., 2010; Garbez et al., 2015).

**Abbreviations:** ABA, abscisic acid; CKs, cytokinins; FBV, flower bud visible; PPFD, photosynthetic photon flux density.

#### Edited by:

Yong-Ling Ruan, University of Newcastle, Australia

#### Reviewed by:

Pilar Cubas, Consejo Superior de Investigaciones Científicas (CSIC), Spain Thomas J. Bach, University of Strasbourg, France

\*Correspondence:

Sabine Demotes-Mainard sabine.demotes-mainard@inra.fr

#### Specialty section:

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

Received: 30 June 2017 Accepted: 20 September 2017 Published: 10 October 2017

#### Citation:

Corot A, Roman H, Douillet O, Autret H, Perez-Garcia M-D, Citerne S, Bertheloot J, Sakr S, Leduc N and Demotes-Mainard S (2017) Cytokinins and Abscisic Acid Act Antagonistically in the Regulation of the Bud Outgrowth Pattern by Light Intensity. Front. Plant Sci. 8:1724. doi: 10.3389/fpls.2017.01724

Bud outgrowth is a key event that determines whether a branch will develop or not. Not all axillary buds grow out along a stem because of correlative inhibitions between organs, which are processes whereby organs of a plant (apex, leaves, internodes, and other buds) can inhibit the outgrowth of a given bud (Brunel et al., 2002; Domagalska and Leyser, 2011). This produces a bud outgrowth pattern, which is a major component of plant architecture (Barthélémy and Caraglio, 2007). For example, under optimal growth conditions, the bud outgrowth pattern of a stem of the rose bush Rosa hybrida 'Radrazz' is acrotonic, meaning that apical buds grow out more frequently than median or basal buds (Demotes-Mainard et al., 2013b).

Bud outgrowth pattern can also be modified by changes in environmental factors, in particular light conditions. Photosynthetic photon flux density, red:far-red (R:FR), and blue/UV wavelengths have such effects (Leduc et al., 2014; Reddy et al., 2014; Demotes-Mainard et al., 2016; Huché-Thélier et al., 2016). The mechanisms behind light regulation of bud outgrowth are currently investigated. The effect of the R:FR ratio involves systemic auxin signaling, control of the branching repressor BRC1 or of its homolog TB1, which in turn regulates ABA signaling and synthesis in the bud (Kebrom et al., 2006; González-Grandío et al., 2013, 2017; Reddy et al., 2013; Reddy and Finlayson, 2014; Holalu and Finlayson, 2017). Far less is known on how PPFD regulates bud outgrowth pattern and this is the question we address in this paper. Whether ABA, which inhibits bud outgrowth in various species (Chatfield et al., 2000; Cline and Oh, 2006; Reddy et al., 2013; Yao and Finlayson, 2015; González-Grandío et al., 2017), is also involved in regulating the response of bud outgrowth to PPFD remains for example unknown.

Sugars promote bud outgrowth (Bonhomme et al., 2009; Henry et al., 2011; Rabot et al., 2012; Mason et al., 2014; Barbier et al., 2015), but their role in the mediation of PPFD signal is also unclear. In rose, the inhibition of bud outgrowth by low PPFD is correlated to both a low sugar content in the stem and bud and to the down-regulation of a sucrose transporter gene (Furet et al., 2014). Yet, when in a decapitated defoliated rose exposed to darkness, sugars are supplied to the bud, this cannot suppress bud outgrowth inhibition, indicating that sugars are not the main limiting factor mediating outgrowth inhibition by darkness (Roman et al., 2016).

In Arabidopsis, Su et al. (2011) showed that reduced PPFD concomitantly inhibited rosette buds and up-regulated a repressor of CK signaling in the bud. The prominent role of CKs as initial targets of the light signal controlling bud outgrowth was demonstrated recently in rose (Roman et al., 2016). Dark inhibition of bud outgrowth involves rapid repression of the transcription of CK synthesis genes in the adjacent node and up-regulation of CK degradation genes in both the node and the bud, leading to a drop of the CK content in both organs. Conversely, exogenous supply of CKs to the bud or the adjacent node fully relieves inhibition of bud outgrowth by darkness. Exogenous CKs were shown to act through the regulation of a set of genes as rapidly and intensively as exposure to white light does. In particular, CKs in the node increase the sugar sink strength of the bud by upregulating genes of sugar transport and metabolism (Roman et al., 2016).

These findings on the role of CKs were, however, obtained in response to white light or darkness in a simplified system consisting of the upper bud of a decapitated defoliated plant. In this system the bud is relieved from the apical dominance of the shoot tip and from the correlative inhibitions from other buds in an apical position, and from the influence of the leaves. The influence of the leaves is complex: mature leaves are a source of sugars for the whole plant, which promotes bud outgrowth, but young leaves can compete with buds for sugars, in addition mature leaves exert an inhibiting effect on the outgrowth of their adjacent bud (Zieslin and Halevy, 1976; Le Hir et al., 2006; Kebrom and Mullet, 2015). It remains unknown whether CKs are also key actors of the light control of bud outgrowth in whole plants. Moreover, their role was evidenced in response to extreme conditions (light vs. darkness), and whether CKs also mediate the response to PPFD variations within a range encountered under normal daylight conditions remains to be assessed. Considering the role of sugars and ABA in the control of bud outgrowth, interactions between CKs, ABA, and sugars in the control of branching pattern by PPFD need to be addressed.

We investigated these questions in rose plants subjected to high and low PPFD treatments that produce contrasted bud outgrowth patterns. We used intact plants, with buds submitted to the correlative inhibitions from other plant parts and determined the effect of PPFD on bud outgrowth pattern. Our previous studies in rose established that bud outgrowth is dependent on adjacent node resources, in particular sugars (Girault et al., 2010; Henry et al., 2011) and CKs (Roman et al., 2016) and that early light regulation of bud outgrowth takes place in the node adjacent to the bud (Roman et al., 2016). ABA which is synthesized in buds but also roots and leaves (Lacombe and Achard, 2016) circulates through the stem, and when applied to the stem ABA inhibits bud outgrowth (Arney and Mitchell, 1969; Chatfield et al., 2000; Cline and Oh, 2006). These results show the prominent role of the adjacent node on the control of the outgrowth of a bud. Therefore, to identify the early mechanisms of the control of bud outgrowth pattern by PPFD, we carried analyses on nodes. Sugar (starch, sucrose, glucose, and fructose) and hormone (CK, ABA, and IAA) contents were determined as well as the impact of exogenous supplies of sucrose, glucose, CKs, and ABA on bud outgrowth under two different light intensities. Correlations between these analyses and the establishment of a bud outgrowth pattern in response to PPFD are discussed.

#### MATERIALS AND METHODS

#### Plant Material, Growth Conditions

Single-node cuttings of R. hybrida 'Radrazz' rose bushes were obtained as in Demotes-Mainard et al. (2013a). When the second leaf of the primary axis unfolded, the young plants were transferred to two identical growth chambers (Froids et Mesures, Beaucouzé, France). All lamps were metal halide lamps (HQI, OSRAM, München, Germany). The light spectrum in

the growth chambers is presented in Supplementary Material 1, R:FR was 1.99 over the ranges (655–665 nm):(725–735 nm). Light spectrum was measured with a spectrophotometer (Avaspec-2048-6-RM, Avantes, Apeldoorn, The Netherlands). The plants were spaced 15 cm apart and surrounded by a border row. The photoperiod was 16h day/8h night, air temperature was 18◦C day/17◦C night, and humidity was 69%. The plants were sub-irrigated, and fertilized as in Demotes-Mainard et al. (2013a). The shoot apical meristem of the primary axis produced vegetative phytomers before differentiating into a terminal flower bud. The flower bud was first hidden within the unfolded young leaves until it became visible, and then referenced to as the FBV stage. It coincides in R. hybrida 'Radrazz' with the outgrowth of the first buds on the primary axis.

#### Light Treatments

All plants were exposed to a to a PPFD of 529 ± 6 µmol m−<sup>2</sup> s −1 starting when the plants were placed in the growth chamber until the FBV stage. At FBV stage the plants were randomly assigned to one of the two following light treatments. The first treatment, hereafter referred to as "high PPFD," corresponded to exposure to a continuous PPFD of 529 ± 6 µmol m−<sup>2</sup> s <sup>−</sup><sup>1</sup> until 14 days after the FBV stage. In the second treatment, hereafter referred to as "low PPFD," the plants were exposed to 89 ± 1 µmol m−<sup>2</sup> s −1 from the FBV stage to 14 days after the FBV stage. PPFD was measured with a horizontal cosine corrector quantum sensor (LI-190 Quantum Sensor, LI-COR, Lincoln, NE, United States).

### Quantification of Endogenous Sugars and Phytohormones

The fourth internode from the apex (not counting the peduncle, see Supplementary Material 2A) was collected at the FBV stage (day 0), days 4 and 8 after the FBV stage on plants grown under high or low PPFD. Samples were collected 6 h after the beginning of the light period, immediately frozen in liquid nitrogen and stored at −80◦C. Then they were lyophilized and crushed. Sucrose, glucose, fructose, and starch (enzymatic digestion step) were determined by colorimetry (Konelab), as presented in Supplementary Material 3. CKs, free ABA, and free IAA concentrations were determined as in Li-Marchetti et al. (2015). Sugar and hormone contents were determined from four replicates with five plants per light treatment each.

#### Exogenous Supply of Chemicals Methods for Enriching Plant Tissues

For all chemicals and concentrations, except sugars at high concentrations, the cotton-wick method was used to supply the compounds directly into the stem, as described in Mahieu et al. (2009). The compounds were stored in the form of a liquid solution in a reservoir and brought to the plant by a cotton wick that went through the stem through an 0.5-mm hole positioned 5 mm below the fourth node (all the organs along the primary shoot are counted downward from the apex) (Supplementary Materials 2B,C). The wick was protected from desiccation and the reservoir was re-filled with the solution if needed.

At concentrations equal or above 300 mM, the sugars crystallized along the cotton wick, so they were supplied through the leaf rachis (Lin et al., 2011). The leaflets of the fourth leaf from the top were removed, except for the two proximal leaflets, and the rachis was rapidly immersed in a liquid solution in a 1.5 ml reservoir. After 1 week the rachis was cut 0.5 cm lower. The reservoirs were re-filled if necessary. Both methods supplied the exogenous compounds continuously.

#### Sugar, CK, and ABA Supply

All chemicals were supplied from the FBV stage onward. Sugars and CKs were provided to the plants grown under the low PPFD light treatment to test if they could relieve the bud outgrowth inhibition induced by low PPFD. Sugars were provided mainly as sucrose, which is the circulating form, but also as glucose because these two forms can trigger bud outgrowth in isolated rose nodes grown in vitro (Rabot et al., 2012). Under sucrose 50, 100, 200 mM and glucose 100 mM, mannitol 100 mM was used as an osmotic control; under sucrose 300, 600, 800 mM and glucose 600 mM, mannitol 300 mM served as a control. Mannitol is not metabolized by R. hybrida and does not stimulate bud outgrowth (Henry et al., 2011; Rabot et al., 2012). These concentrations are typical of phloem sap (Nadwodnik and Lohaus, 2008; Merchant et al., 2009) and are either equal to or higher than those previously shown to be able to trigger bud outgrowth in isolated rose internodes (Henry et al., 2011; Rabot et al., 2012; Barbier et al., 2015).

Synthetic CKs were supplied as 6-benzylaminopurine (BAP) at 50 or 500 µM, solubilized in water with NaOH 40 mM. The control consisted of NaOH 40 mM alone. The concentrations of 50 and 500 µM BAP were within the range of synthetic CKs used in pea and different woody species (Dun et al., 2012; Ni et al., 2015).

To test whether ABA could inhibit bud outgrowth in favorable conditions, ABA was supplied at 1.25 and 2.5 mM with NaOH 40 mM to plants grown continuously under a high PPFD of 350 µmol m−<sup>2</sup> s <sup>−</sup><sup>1</sup> provided by white LED tubes. The control was an ABA-free solution of NaOH 40 mM. To test whether ABA could also inhibit bud outgrowth under low PPFD, ABA was supplied under low PPFD together with CKs, because buds grew out in plants supplied with CKs under the low PPFD (see "Results" section). ABA and CKs were supplied in the form of solutions containing BAP 500 µM + ABA 1.25, 2.5 or 5.0 mM + NaOH 40 mM. The control contained BAP 500 µM + NaOH 40 mM. High concentrations of ABA were applied as endogenous ABA contents were much higher than endogenous CK contents (see "Results" section).

# <sup>13</sup>C-Labeled Sugar Uptake Studies

To check that sugars supplied by both the cotton-wick method and the rachis-feeding method were indeed delivered to the stem tissues, the plants grown under low PPFD were supplied with labeled sucrose. Labeled <sup>13</sup>C-sucrose (ID 605417\_Aldrich, Sigma–Aldrich, Darmstadt, Germany) was incorporated at 3% in a solution of 200 mM sucrose supplied by a cotton wick just below node 4, or incorporated at 3.4% in a solution of 600 mM sucrose supplied through the rachis of leaf 4. Water supplied either by a

cotton wick or through the rachis served as a control. Exogenous supply started at the FBV stage onward. Days 4 and 8 after the FBV stage, the third, fourth, and fifth internodes from the apex (not counting the peduncle, Supplementary Material 2A) were sampled as described previously. Total carbon and the <sup>13</sup>C:12C ratio were determined for each sample using an elemental analyzer (NA 1500 NCS; Carlo Erba, Milan, Italy) coupled with a Delta-S isotope ratio mass spectrometer (Finnigan-Mat; Thermoquest Corp., San Jose, CA, United States). δ <sup>13</sup>C was calculated as:

$$\text{R}^{13}\text{C} \left(\% \right) \;= \left(\text{R sample}/\text{R reference} - 1\right) \times 1000 \qquad \text{(1)}$$

where R is the ratio of heavy-to-light isotope, and the R reference for <sup>13</sup>C in the PDB standard is 0.0112372.

#### Growth Measurements of the Different Buds along the Primary Axis

Three to six times a week from FBV onward, the state (dormant or growing out) of each bud along the primary axis was noted, and bud length was measured with a ruler. Buds were considered to have grown out when at least one visible leaf protruded between the two bud scales (Girault et al., 2008; Roman et al., 2016). The mean bud lengths were calculated from the sizes of outgrowing and dormant buds from a same batch. Fourteen days after the FBV stage, buds were dissected using a stereomicroscope. Bud meristem organogenic activity, which results in the formation of new organs in the bud, was assessed by counting the number of leaves in the bud, leaves being here defined as young leaves plus foliar primordia, excluding the two bud scales. Measurements were performed on 8–16 plants per experimental modality (light treatment and/or chemical exogenous supply), with three biological replicates, except otherwise stated.

#### Photosynthesis Measurements

The CO<sup>2</sup> net assimilation rate was measured as the quantity of CO<sup>2</sup> assimilated per leaf surface unit per second, and represents the difference between photosynthetic assimilation and respiration. To assess the effect of the light treatments, CO<sup>2</sup> net assimilation rates were measured both in a leaf still rapidly expanding and partially unfolded at the FBV stage (leaf 4 from the apex), and in a mature fully unfolded leaf (leaf 6 from the apex), on five plants, 3 h before and 3 h after the switch in light intensity, and then once a day for 14 days, 6 h after the beginning of the light period.

To test if the CK supply or the CK – ABA supply affected photosynthesis in plants grown under low PPFD, CO<sup>2</sup> net assimilation rates were measured in leaves of plants supplied according to the following conditions: CK 500 µM + NaOH 40 mM, CK 500 µM + ABA 1.25, 2.5, 5.0 mM + NaOH 40 mM, control NaOH 40 mM. CO<sup>2</sup> net assimilation rates were measured in the leaves positioned immediately above and below the supply point (leaves 4 and 5, respectively, Supplementary Materials 2A,B) on four plants per condition, every 2 days from FBV onward, 6 h after the beginning of the light period. Leaf positions were chosen so as to maximize the likelihood to observe an effect of the exogenous supply if it existed.

All measurements of CO<sup>2</sup> net assimilation rates were performed using an infrared gas analyser, Li-Cor-6400 (Li-Cor Inc., Lincoln, NE, United States), at a temperature of 18◦C, a relative humidity of 64%, a cuvette air flow rate of 300 ml min−<sup>1</sup> , and an ambient CO<sup>2</sup> concentration of 400 µmol mol−<sup>1</sup> . PPFDs of 540 and 90 µmol m−<sup>2</sup> s <sup>−</sup><sup>1</sup> were provided to the measured leaves for the plants grown under high and low PPFD, respectively, by a mixture of red and blue LEDs.

#### Statistical Analysis

Statistical analyses were performed using R software for Windows (R Core Team, 2014). Groups were compared using one-way ANOVA (α = 0.05) followed by Tukey's test when the ANOVA indicated significant differences at P < 0.05 and there were more than two groups.

### RESULTS

#### Decreasing PPFD Inhibits Axillary Bud Outgrowth All along the Stem

The number of outgrown buds per plant was much lower under the low PPFD treatment than under the high PPFD treatment (**Figure 1A-inset** and Supplementary Figure S1). Under high PPFD buds grew out all along the stem, displaying an acrotonic pattern with a decreasing outgrowth rate from the apical to the basal part of the stem (**Figures 1A,B**). The time course of bud outgrowth was sequential, starting from the two uppermost apical buds 2 days after the FBV stage and proceeding downward (**Figure 1C**). The uppermost two apical buds did not respond to a reduction of PPFD because they are sylleptic buds whose outgrowth is independent of environmental conditions (Huché-Thélier et al., 2011; Demotes-Mainard et al., 2013b). But, the outgrowth rate of the third topmost bud was reduced as compared to high PPFD conditions (72% vs. 100%), and outgrowth of all the buds from the fourth topmost bud and below was totally inhibited (**Figures 1A,B**). Bud 4 displayed the most contrasted outgrowth behavior between PPFD treatments. While under high PPFD all buds 4 had grown out 8 days after the FBV stage on average, none grew out in low PPFD (**Figures 1A,C**). Elongation of bud 4 was rapid and continuous under high PPFD, whereas it was slow and stopped rapidly under low PPFD (**Figure 1D**).

#### Decreasing PPFD Reduces Sugar Contents in the Internode, but Local Sugar Supply Is Not Sufficient to Restore Bud Outgrowth

To determine if sugars are involved in mediating the effect of PPFD on bud outgrowth, we analyzed the sugar content of internode 4, which is adjacent to the bud displaying the most contrasted response to PPFD treatments (**Figures 1A,B**). Internode 4 displayed a lower content in sucrose, glucose, fructose, and starch under low PPFD than under high PPFD days

4 and 8 after the FBV stage (**Figure 2**). Day 8 was the time when bud 4 grew out under high PPFD (**Figure 1C**), and day 4 was the time when we presume that conditions may have affected the inducement of bud 4 outgrowth since 4 days are required for bud outgrowth after apical dominance has been relieved under favorable light conditions in this cultivar (Girault et al., 2010). The correlation between sugar content in internode 4 and bud 4 outgrowth thus suggests that a low sugar level in the node may be involved in bud outgrowth inhibition under low PPFD.

To investigate this hypothesis, we tested if local sugar supply could relieve bud inhibition under reduced PPFD. Sugars were supplied to plants either through a cotton wick (50, 100, and 200 mM) or through the rachis for the higher concentrations (300, 600, and 800 mM). Using labeled <sup>13</sup>Csucrose, we demonstrated that sugars were indeed delivered into the stem tissues by these two methods (Supplementary Figures S2A,B). Sucrose supplied under low PPFD did not promote bud outgrowth, nor did it change the gradient of bud outgrowth along the stem as compared to the control, whatever the concentration (**Figures 3A,A-inset,D,D-inset**). Neither did it stimulate bud elongation (**Figures 3B,E**) or leaf formation into the bud (**Figures 3C,F**). Similar results were observed when glucose (100 or 800 mM) instead of sucrose was supplied (Supplementary Figures S3A–F). Altogether, these results show that although low PPFD leads to reduced sugar levels in the internode, the bud outgrowth inhibition induced by low PPFD cannot be solely relieved by local sugar supply. We therefore investigated whether CKs and ABA may limit bud outgrowth under low PPFD.

#### Decreasing PPFD Reduces CK Contents in the Internode, and CK Supply Can Relieve Low PPFD-Induced Bud Inhibition

Cytokinin (zeatin riboside-5<sup>0</sup> -monophosphate, ZRMP; isopentenyl adenosine-5<sup>0</sup> -monophosphate, iPRMP; isopentenyl adenine riboside, iPR) contents were measured in internode 4 under high and low PPFD treatments 4 and 8 days after the FBV stage. ZRMP and iPRMP are intermediate forms in the CK biosynthesis pathway, and iPR is an active form. The three CK forms followed different temporal dynamics under high PPFD:

average of four repetitions. Error bars represent standard error of the mean (SE). Different letters indicate significant differences between treatments and times

the content in ZRMP increased over the 8 days following the FBV stage, whereas the content in iPRMP was stable, and the content in iPR decreased over time (**Figures 4A–C**). Low PPFD reduced the content in all three CK forms in internode 4, 4 and 8 days after the FBV stage (**Figures 4A–C**), whereas the free IAA content in internode 4 was not affected by the PPFD level (**Figure 4D**).

(ANOVA followed by a Tukey test, P < 0.05).

In order to assess if CKs may be players in the regulation of the bud outgrowth pattern by PPFD, we tested if exogenous CKs could relieve the bud outgrowth inhibition induced by low PPFD. Synthetic CK BAP was supplied to internode 4 of plants grown under low PPFD using cotton wicks. When 50 and 500 µM CK were supplied, the number of buds that grew out per plant increased significantly as compared to the control (**Figure 5A-inset**). Increased outgrowth was observed locally close to the supply point for bud 4, but also at a distance for bud 3 (**Figure 5A**). Bud 4, whose outgrowth was totally inhibited under low PPFD in the absence of CK, displayed rates of 91 and 100% outgrowth when the plants were supplied with 50 and 500 µM of CKs, respectively, under low PPFD. Under low PPFD again, outgrowth of bud 4 was observed 7.78 ± 0.44 and 7.49 ± 0.39 days after the FBV stage in plants supplied with CKs at 50 and 500 µM, respectively, and bud 4 elongation was increased by exogenous CKs (**Figure 5B**), giving a similar date of outgrowth and dynamics of elongation as observed under high PPFD (**Figures 1C,D**). CK supply under low PPFD also increased the number of leaves formed into the bud (**Figure 5C**). Taken together, these results suggest that low PPFD inhibition of bud outgrowth is mediated at least partly through a lower stem CK content.

#### Low PPFD Increases the ABA Content, and ABA Supply Can Inhibit Bud Outgrowth under High PPFD and Antagonizes CK Stimulation of Bud Outgrowth under Low PPFD

Measurements of free ABA contents in the adjacent internode to bud 4 revealed that inhibition of bud outgrowth by low PPFD was also correlated to the maintenance of higher levels of ABA 4 and 8 days after the FBV stage (**Figure 6**).

To assess whether ABA could be involved in mediating PPFD control of bud outgrowth, we tested whether exogenous supply of ABA (0, 1.25, and 2.5 mM) to internode 4 could inhibit bud outgrowth under high PPFD (350 µmol m−<sup>2</sup> s −1 ). ABA supply at both 1.25 and 2.5 mM had a strong inhibitory effect on the outgrowth and elongation of bud 4 (**Figures 7B,C**) but not on the number of leaves formed per bud (**Figure 7D**). Interestingly, outgrowth of buds located below the point of ABA supply increased (**Figure 7A**), so that the total number of outgrown buds per plant was not affected (**Figure 7A-inset**).

To test whether a high ABA stem content could inhibit bud outgrowth under low PPFD, ABA was supplied under low PPFD together with CK, a promoter of outgrowth. ABA antagonized the stimulating effect of CK in a dose-dependent manner and decreased bud 4 outgrowth (**Figures 8A,A-inset**). With the lowest ABA concentration, only 46% of these buds grew out and their outgrowth was delayed by 2 days, occurring only 10.2 ± 0.4 days after FBV, whereas with the highest ABA concentration only 4.7% of buds 4 grew out. In the presence of CKs, ABA also drastically limited the elongation of bud

FIGURE 3 | Impact of sucrose supply on bud outgrowth pattern of plants grown under the low PPFD treatment. Sucrose was supplied through the stem below node 4 at 50, 100, and 200 mM (A–C) or through the rachis of leaf 4 at 300, 600, and 800 mM (D–F) from FBV stage onward, and mannitol (100 mM A–C or 300 mM D–F) was used as non-metabolizable control. (A,D) Effect of sucrose supply on the percentage of individuals with outgrown buds 14 days after FBV according to node position along the shoot. (A-inset,D-inset) Total number of buds per plant that grew out 14 days after FBV. (B,E) Elongation of bud 4 along time. (C,F) Number of leaves in bud 4 at FBV stage (day 0) and 14 days after FBV (day 14). Buds are numbered basipetally along the stem. Each point represents the average of three repetitions. Error bars represent the standard error of the mean (SE). Different letters indicate significant differences between sugar supplies and times, ns indicates an absence of significant difference (ANOVA, followed by a Tukey test for C, F, P < 0.05).

4 (**Figure 8B**). With the highest ABA concentration supplied together with CK, the frequencies of bud 4 outgrowth and the kinetics of bud 4 elongation were very similar to those observed under low PPFD without any exogenous supply (**Figures 1**, **8**). ABA had no effect on the number of leaves formed into the bud (**Figure 8C**). The buds below the supply point did not grow out, whatever the ABA concentrations (**Figure 8A**).

Altogether, these results indicate that low PPFD maintains high ABA levels in the stem and that stem ABA can inhibit bud outgrowth under low and high PPFD in rose. They suggest that the high ABA stem contents induced by low PPFD may contribute to the inhibition of bud outgrowth, notably by antagonizing the promoting effect of CKs on bud outgrowth.

# Neither CK Stimulation of Bud Outgrowth under Low PPFD Nor ABA Antagonism of CK Stimulation under Low PPFD Require a Change in CO<sup>2</sup> Net Assimilation

The decrease in PPFD from the FBV stage onward markedly and durably reduced photosynthetic CO<sup>2</sup> net assimilation rates as compared to high PPFD, both in mature and in still rapidly expanding leaves (Supplementary Figure S4). The literature indicates that ABA may dampen photosynthesis (Raschke and Hedrich, 1985; Zhou et al., 2006), and that CKs alone or CKs and ABA together may affect photosynthesis in a way that varies with the species, the conditions of application and the plant growth conditions (Pospíšilová, 2003; Pospíšilová and Bat'ková, 2004; Shu-Qing et al., 2004). We studied whether the effect of CKs and ABA on bud outgrowth passed through an indirect effect on photosynthetic assimilation in our conditions. Over the 2 weeks following the FBV stage, CO<sup>2</sup> net assimilation rates of leaves 4 and 5 of plants grown under low PPFD were affected neither by CK supply alone nor by combined CK and ABA supply, whatever the ABA concentration (**Figures 9A,B**). These results indicate that the effects of CKs and ABA on the bud outgrowth pattern under low PPFD are not directly mediated by changes in photosynthetic assimilation.

# DISCUSSION

## The Responses of Bud Outgrowth to PPFD Are Preceded by Changes in CK, ABA, and Sugar Stem Contents

The reduced PPFD applied from the FBV stage onward reduced the number of outgrowing buds, consistently with previous results in rose (Bredmose, 1995; Furet et al., 2014; Wubs et al., 2014). To address the question of how the CK, ABA, and sugar regulation pathways take part in the control of the branching pattern by light intensity, we focused on bud 4 that displayed the most contrasted fate with 100% or 0% outgrowth under high or low PPFD, respectively. Changes in PPFD led to variations in endogenous CK, ABA, and **s**ugar contents in internode 4, consistent with bud outgrowth response to PPFD: CKs and sugars, which promote bud outgrowth, had their levels decreased under low PPFD, while ABA, an outgrowth

inhibitor, had its content increased. The effect of PPFD on the CK stem content was consistent with the effect observed following exposure to light or darkness by Roman et al. (2016), but information on the effects of PPFD on the ABA stem content was lacking. The reduction in sugar stem content was consistent with the severe decrease in photosynthesis observed under low PPFD but may also have involved changes in partitioning. In addition, the variations in endogenous CK, ABA, and sugar contents in internode 4 preceded the outgrowth of the adjacent bud under high PPFD by at least 4 days; this supports the hypothesis that these chemicals could indeed contribute to the regulation of bud outgrowth. In rose, modifications in sugars and hormone (CK) levels and in expression of genes involved in bud outgrowth regulation were reported to be triggered before the occurrence of bud outgrowth (Girault et al., 2010; Henry et al., 2011; Rabot et al., 2012; Barbier et al., 2015; Roman et al., 2016).

### The CK Stem Content Appears to Be a Key Factor Mediating PPFD Regulation of Bud Outgrowth Pattern along the Stem in an Intact Plant

Despite the major role of CKs in promoting bud outgrowth (Dun et al., 2012; Müller et al., 2015), studies investigating their involvement in the regulation of bud outgrowth by light are scarce. Stem endogenous CKs dropped under low PPFD concomitantly to bud inhibition and conversely CK supply to

outgrown buds 14 days after FBV according to node position along the shoot. (A-inset) Total number of buds per plant that grew out 14 days after FBV. (B) Percentage of individuals whose buds 4 grew out 14 days after FBV stage. (C) Elongation of bud 4 along time. (D) Number of leaves in bud 4 at FBV stage (day 0) and 21 days after FBV (day 21). Buds are numbered basipetally along the stem. Each point represents the average of two repetitions. Error bars represent the standard error of the mean (SE). Different letters indicate significant differences between ABA supplies, ns indicates an absence of significant difference (ANOVA followed by a Tukey test, P < 0.05).

plants grown under low PPFD relieved the inhibition of bud outgrowth. CK supply stimulated bud meristem organogenic activity and bud elongation, that are the two fundamental processes of bud outgrowth. This stimulation happened in such a way that CK supply under low PPFD resulted in the same timings of bud 4 outgrowth and the same kinetics of bud 4 elongation as those observed under high PPFD. Altogether, these results are consistent with our previous findings showing that CKs are

initial targets of light in the regulation of bud outgrowth in decapitated defoliated plants (Roman et al., 2016). They further bring evidence that (i) CKs allow buds to be relieved from ecodormancy due to unfavorable environmental conditions (low PPFD) and from paradormancy (apical dominance and other correlative inhibitions, Zieslin and Halevy, 1976; Domagalska and Leyser, 2011); (ii) a reduction of CK levels not only mediates repression of bud outgrowth induced by total darkness but also by low PPFD, as can be encountered under normal daylight conditions.

The stimulating effect of CKs on bud meristem organogenic activity under low PPFD is consistent with the ability of CKs to rescue organogenic activity under darkness in shoot apical meristems of tomato and rose (Yoshida et al., 2011; Roman et al., 2016).

# The Changes in Stem CK Contents in Response to PPFD Are Not Mediated by Changes in IAA Contents

Knowing that stem IAA can downregulate stem CK contents (Tanaka et al., 2006; Shimizu-Sato et al., 2009), we wondered if the changes in stem CKs in response to PPFD may have resulted, at least partly, from a change in stem IAA. Knowledge on the effects of PPFD on stem auxin contents is poor, and the results obtained in various species and growth conditions are conflicting (Kurepin et al., 2006, 2007, 2008; Hersch et al., 2014). Here, the absence of difference in IAA contents in internode 4 between the two PPFD treatments indicates that the changes in stem CK content was not a consequence of changes in stem IAA levels. Roman et al. (2016) also observed changes in CK stem contents in response to light in the absence of variations in IAA content, but the stem IAA levels in their plant system were extremely low due to decapitation.

Both stem-synthesized CKs and root-derived CKs can contribute to bud outgrowth regulation, with conflicting results between species (Faiss et al., 1997; Tanaka et al., 2006; Müller et al., 2015). In our study we can assume that the regulation of the CK stem content by PPFD may result at least partly from local regulation of the CK metabolism in the stem, since CK synthesis genes are downregulated and CK degradation genes are upregulated in the stem of plants grown in darkness as compared to light (Roman et al., 2016). Regulation of the CK stem content may also partly result from changes in xylem-carried CKs coming from the roots, since in tobacco plants shade reduces xylem carried CKs by reducing the transpiration rate (Boonman et al., 2007).

### The High ABA Content in the Stem Induced by Reduced PPFD Can Inhibit Bud Outgrowth along the Stem and Antagonize CKs

Our results suggest that the high stem ABA content induced by low PPFD contributed to the regulation of bud outgrowth. Exogenous ABA supply to the stem inhibited bud outgrowth in rose under high PPFD, i.e., in environmental conditions favorable to outgrowth, in accordance with previous data in Arabidopsis, pea, Ipomoea nil, and tomato (Arney and Mitchell, 1969; Chatfield et al., 2000; Cline and Oh, 2006; Yao and Finlayson, 2015). To investigate further the role of ABA in mediating bud outgrowth regulation by PPFD, we showed that ABA could also inhibit bud outgrowth under low PPFD. ABA was able to antagonize the promotive effect of exogenous CKs on bud outgrowth in a dose-dependent manner under low PPFD. The antagonistic role of ABA and CKs had been evidenced in other processes such as stomatal regulation, seed germination or seedling development (Dodd, 2005; Wang et al., 2011; Guan et al., 2014), but such a dose-dependent antagonism was evidenced in this study for the first time in the regulation of bud outgrowth. Altogether, our results suggest that the effect of PPFD on bud outgrowth may be complex and that reduced PPFD inhibits bud outgrowth both by reducing the stem content in CKs (a branching inducer) and by maintaining a high content in stem ABA (a branching repressor). The high ABA content can dampen the effects of the remaining CKs. ABA was already known to mediate the response to the R:FR ratio or the photoperiod in Arabidopsis (Reddy et al., 2013; Yao and Finlayson, 2015; González-Grandío et al., 2017; Holalu and Finlayson, 2017). Our results extend its role to the mediation of environmental regulation of bud outgrowth by another light signal and species. In addition, they also point out the contribution of stem ABA in controlling bud outgrowth, and complement the previous works that focused on bud ABA.

Interestingly, ABA inhibited bud outgrowth and impaired bud elongation, but did not affect meristem organogenic activity in the bud, irrespective of the PPFD level. In contrast, CKs stimulated bud meristem organogenic activity, which indicates that the antagonistic effect of ABA and CKs was limited to certain bud outgrowth-related processes.

### Supplying Sugars Close to the Bud Cannot Alone Relieve Low PPFD-Induced Bud Inhibition

Under low PPFD, exogenous supply of both sucrose and glucose close to the bud failed to promote bud outgrowth in intact plants. These results suggest that sugars are not the main limiting factor

FIGURE 10 | Schematic representation of the relationships between CKs, ABA, IAA, and sugars in the regulation of bud outgrowth by PPFD in an intact rose plant. Low PPFD reduces stem CK content, which appears to be a key limiting factor of bud outgrowth. The level of CKs transported from the stem to the bud, where they stimulate outgrowth, is likely to be dampened under low PPFD. The CK stimulation of bud sugar sink strength, shown in decapitated rose plants (Roman et al., 2016), may also be lowered. Low PPFD maintains a high stem ABA content, ABA presumably enters the bud where it inhibits outgrowth, antagonizing the effects of CKs. Low PPFD also reduces photosynthesis and sugar stem content. However, sugar stem content is not the main limiting factor for bud outgrowth since sugar supply is not able to override bud outgrowth inhibition under low PPFD. The reduced bud sugar sink strength under low PPFD may prevent the use of sugars by the bud for outgrowth. Although IAA is known to dampen stem CK content, low PPFD effect on CK is not mediated through changes in stem IAA content.

for bud outgrowth in intact plants under low PPFD, but would act downward other factors, at least CKs and ABA. These results are consistent with those obtained in the upper bud of decapitated defoliated plants grown under darkness (Roman et al., 2016).

Cytokinins have positive effects on source-sink organ activities (Cowan et al., 2005; Peleg et al., 2011). In our case the CK supply that promoted outgrowth under low PPFD is unlikely to have acted on the global sugar content per plant as it did not change photosynthetic assimilation. CK supply to the stem is known to increase the sugar sink strength of the bud (Roman et al., 2016), in agreement with CK-mediated upregulation of sugar sink strength in other tissues (Roitsch and González, 2004). Thus, we can hypothesize that one action of CK supply in the vicinity of the bud under low PPFD was to increase bud sugar sink strength and allow buds to better compete for sugars at the expense of other plant organs. Unlike CKs, ABA affects negatively sink strength by limiting the availability of nutrients and energy (Garciarrubio et al., 1997; Parish et al., 2012). It would be of great interest to investigate in future works if ABA negatively affects the sink strength of buds and how it interacts with CKs in response to PPFD. Mason et al. (2014) demonstrated that sugar supply alone can promote bud outgrowth when plants are grown in a non-limiting light environment in which the shoot tip's high demand for sugars is the limiting factor for bud access to sugars. In our experiments the apical organs, that were in rapid expansion (Demotes-Mainard et al., 2013a), most probably intensively competed with buds for sugars too. However, sugar supply to the stem could not trigger outgrowth under low PPFD because PPFD controlled bud outgrowth through an upstream mechanism, involving CK and presumably ABA regulation. We hypothesize that at least part of this control is due to low stem CK content that restricts sugar import into the bud under low PPFD.

#### Action of Exogenous CKs and ABA on Bud Outgrowth at Different Locations along the Stem

Exogenous CK supply had an effect on the directly treated bud (bud 4) and on the first bud located above it (bud 3). The absence of effect on the buds located below the supply point is consistent with results on isolated nodes and decapitated plants (Ohkawa, 1984; Chatfield et al., 2000).

Abscisic acid supply in the stem inhibited only the local bud. We did not expect exogenous ABA supply to inhibit lower buds since ABA inhibits outgrowth in isolated nodes only when supplied from the basal side of the stem (Chatfield et al., 2000). The absence of inhibition of bud 3, located apically, may result from the dose of ABA that reached bud 3 and/or from a lower sensitivity of this bud to ABA due to its physiological state. Bud 3 was indeed much less sensitive to PPFD than the buds below, as shown by the outgrowth patterns under low and high PPFD.

When exogenous ABA totally inhibited bud 4 outgrowth under high PPFD, the outgrowth rate of the buds just below it (buds 5 and 6) increased. This increase in outgrowth may be an indirect effect of bud 4 inhibition under high PPFD: bud 4 inhibition by ABA may have both reduced the correlative inhibition exerted on lower buds (Domagalska and Leyser, 2011) and made sugars more available for these lower buds, thereby facilitating their outgrowth under high PPFD (Mason et al., 2014). In isolated internodes of Arabidopsis, ABA supplied from the apical side of the stem reduced inhibition by IAA (Chatfield et al., 2000). Such an effect cannot be ruled out but was not observed in Ipomoea nil or in sunflower (Cline and Oh, 2006).

# CONCLUSION

This study provides evidence that in intact rose plants reduced PPFD controls bud outgrowth pattern both by reducing the stem CK content and by maintaining a high stem ABA content. These two hormones regulate bud outgrowth antagonistically,

but their antagonistic effect is limited to certain bud outgrowthrelated processes since bud meristem organogenic activity is solely regulated by CKs. Although the stem sugar content was severely reduced under low PPFD, this drop did not appear to be the triggering factor of bud outgrowth inhibition under low PPFD. A model based on our results and literature is presented in **Figure 10**. Further work will contribute to investigate the relationships between CKs and ABA in the regulation cascade of the bud outgrowth pattern in response to PPFD by investigating at the molecular level regulation pathways between buds showing different sensitivity levels to PPFD variations.

### AUTHOR CONTRIBUTIONS

AC, NL, SS, SD-M conceived and designed the research. AC, OD, HA, M-DP-G, and SC carried the experimental work, collected and prepared the data. AC conducted statistical analyses. NL, SS, and SD-M directed the work. AC, HR, JB, SS, NL, and SD-M interpreted the data and wrote the paper.

### FUNDING

This research was conducted in the framework of the regional program "Objectif Végétal, Research, Education and Innovation in Pays de la Loire," supported by the French Region Pays de la Loire, Angers Loire Métropole, and the European Regional Development Fund.

# REFERENCES


#### ACKNOWLEDGMENTS

We thank the certified facility in Functional Ecology (PTEF OC 081) from UMR 1137 EEF and UR 1138 BEF in research center INRA Nancy-Lorraine for the isotopic analyzes. The PTEF facility is supported by the French National research Agency through the laboratory of Excellence ARBRE (ANR-11-LABX-0002-01). We thank Bénédicte Dubuc (UMR IRHS) and the ImHorPhen team (Phenotic platform, UMR IRHS), particularly Christian Cattanéo, Jacky Granger, Dominique Honoré, and Rémi Gardet, for rose cutting production, plant care and the maintenance of experimental facilities; Anne-Claire Grosbois (trainee UMR IRHS and Université d'Angers) for technical assistance with the ABA experiments; Rachid Boumaza (UMR IRHS) for the advice provided in statistical analyses; Jérémy Lothier, Thomas Péron (UMR IRHS), Wim van den Ende (KU Leuven), Christophe Godin (INRIA Virtual Plant), Didier Combes (INRA, UR P3F), and Elzbieta Frak (INRA, UR P3F) for fruitful discussions; PAIGE team (UMR IRHS) for the administrative work, Annie Buchwalter (a scientific translator) for reviewing the paper for English usage.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2017.01724/ full#supplementary-material

growth potential in apple trees (Malus domestica [L.] Borkh.). J. Exp. Bot. 53, 2143–2149. doi: 10.1093/jxb/erf063




dominance. Plant J. 45, 1028–1036. doi: 10.1111/j.1365-313X.2006. 02656.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 Corot, Roman, Douillet, Autret, Perez-Garcia, Citerne, Bertheloot, Sakr, Leduc and Demotes-Mainard. 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.

# Novel Protein-Protein Inhibitor Based Approach to Control Plant Ethylene Responses: Synthetic Peptides for Ripening Control

Mareike Kessenbrock<sup>1</sup> , Simone M. Klein<sup>2</sup> , Lena Müller<sup>1</sup> , Mauricio Hunsche2,3 , Georg Noga<sup>2</sup> and Georg Groth1,4 \*

1 Institute of Biochemical Plant Physiology, Heinrich Heine University Düsseldorf, Düsseldorf, Germany, <sup>2</sup> Institute of Crop Science and Resource Conservation – Horticultural Science, University of Bonn, Bonn, Germany, <sup>3</sup> COMPO EXPERT GmbH, Münster, Germany, <sup>4</sup> Bioeconomy Science Center, Forschungszentrum Jülich, Jülich, Germany

#### Edited by:

Chi-Kuang Wen, Shanghai Institutes for Biological Sciences (CAS), China

#### Reviewed by:

Bram Van De Poel, KU Leuven, Belgium Jie Le, Key Laboratory of Plant Molecular Physiology, Institute of Botany (CAS), China

\*Correspondence:

Georg Groth georg.groth@hhu.de; georg.groth@uni-duesseldorf.de

#### Specialty section:

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

Received: 19 June 2017 Accepted: 21 August 2017 Published: 05 September 2017

#### Citation:

Kessenbrock M, Klein SM, Müller L, Hunsche M, Noga G and Groth G (2017) Novel Protein-Protein Inhibitor Based Approach to Control Plant Ethylene Responses: Synthetic Peptides for Ripening Control. Front. Plant Sci. 8:1528. doi: 10.3389/fpls.2017.01528 Ethylene signaling is decisive for many plant developmental processes. Among these, control of senescence, abscission and fruit ripening are of fundamental relevance for global agriculture. Consequently, detailed knowledge of the signaling network along with the molecular processes of signal perception and transfer are expected to have high impact on future food production and agriculture. Recent advances in ethylene research have demonstrated that signaling of the plant hormone critically depends on the interaction of the ethylene receptor family with the NRAMP-like membrane protein ETHYLENE INSENSITIVE 2 (EIN2) at the ER membrane, phosphorylation-dependent proteolytic processing of ER-localized EIN2 and subsequent translocation of the cleaved EIN2 C-terminal polypeptide (EIN2-CEND) to the nucleus. EIN2 nuclear transport, but also interaction with the receptors sensing the ethylene signal, both, depend on a nuclear localization signal (NLS) located at the EIN2 C-terminus. Loss of the tight interaction between receptors and EIN2 affects ethylene signaling and impairs plant ethylene responses. Synthetic peptides derived from the NLS sequence interfere with the EIN2–receptor interaction and have utility in controlling plant ethylene responses such as ripening. Here, we report that a synthetic peptide (NOP-1) corresponding to the NLS motif of Arabidopsis EIN2 (aa 1262–1269) efficiently binds to tomato ethylene receptors LeETR4 and NR and delays ripening in the post-harvest phase when applied to the surface of sampled green fruits pre-harvest. In particular, degradation of chlorophylls was delayed by several days, as monitored by optical sensors and confirmed by analytical methods. Similarly, accumulation of β-carotene and lycopene in the fruit pulp after NOP-1 application was delayed, without having impact on the total pigment concentration in the completely ripe fruits. Likewise, the peptide had no negative effects on fruit quality. Our molecular and phenotypic studies reveal that peptide biologicals could contribute to the development of a novel family of ripening inhibitors and innovative ripening control in climacteric fruit.

Keywords: ethylene signaling, ethylene receptors, peptide, ripening control, Solanum lycopersicum (tomato), post-harvest application

# INTRODUCTION

fpls-08-01528 September 2, 2017 Time: 15:8 # 2

Worldwide, a tremendous amount of food produced for human consumption is lost or wasted until the product reaches the consumer, with about 50% of those food losses being valuable vegetables and fruits (Blanke, 2014). The tomato fruit is one of the most important climacteric fruits (Jackman et al., 1990; Vidoz et al., 2010) and has worldwide high economic and nutritional importance, mainly because of its high concentrations of carotenoids such as lycopene, β-carotene and pro-vitamin A (Canene-Adams et al., 2005; Vidoz et al., 2010) which accumulate during fruit ripening. The ripening process of climacteric fruits is characterized by a strong increase in cell respiration which is mainly regulated by the plant hormone ethylene (Alexander and Grierson, 2002; Guo and Ecker, 2004). Ripening is initiated by a burst of an auto-stimulated ethylene synthesis, with following activation of ripening related genes (Abano and Buah, 2015). This ethylene related gene expression leads to physiological, morphological and biochemical changes. In the process of fruit ripening, fruits change color, texture, firmness, flavor and aroma (Brady, 1987; Alexander and Grierson, 2002) due to degradation of pectins, cellulose and chlorophyll as well as due to a decreasing content of organic acids and increasing concentration of soluble sugars, carotenes and aroma volatiles (Brady, 1987).

Besides the traditional quality analysis of firmness and content of sugars, acids, vitamins and pigments, changes in fruit ripening and fruit quality might be evaluated by non-destructive optical methods (Abbott, 1999; Hoffmann et al., 2015). Analogous to that, consumers usually estimate fruit quality based on fruit skin color. Color development of tomatoes from green to red can be measured by monitoring chlorophyll degradation as well as lycopene and β-carotene accumulation (Alexander and Grierson, 2002). Typically, the total content of these pigments is analyzed using wet chemical procedures (Barros et al., 2007; Azeez et al., 2012; Kalogeropoulos et al., 2012), whereas nondestructive optical sensors that evaluate overall changes in fruit color based on reflection and fluorescence properties provide monitoring parameters that strongly correlate with the analytical values (McGuire, 1992; Hoffmann et al., 2015).

Control of ripening is important to ensure quality and to reduce post-harvest losses of climacteric fruits. At commercial scales, fruits are usually stored at low temperatures under controlled atmosphere to limit ethylene production and ethylene response (Watkins et al., 2000; Saltveit, 2005; Passam et al., 2007). Alternative approaches, such as genetic engineering of ethylene biosynthesis to decrease endogenous ethylene production, are under development in science and research despite of ongoing discussion in Europe about genetic engineering in general. At the production scale, fruit maturation can be delayed with aminoethoxyvinylglycine (AVG), an inhibitor of ACC-synthase, the key enzyme of ethylene biosynthesis (Saltveit, 2005). For postharvest treatment, 1-methylcyclopropene (1-MCP), a gaseous chemical with the ability to inhibit ethylene receptors and receptor-triggered ethylene response can be applied (Watkins et al., 2000; Yuan and Carbaugh, 2007).

Pathways and mechanisms for biosynthesis, perception and signal transduction of the plant hormone ethylene have been extensively studied in the model plant Arabidopsis thaliana. These studies disclosed that the ethylene signal is perceived by a family of five receptor proteins, which form homo- and heterodimers at the ER membrane and function as negative regulators of the ethylene response (Bleecker et al., 1988; Chang et al., 1993; Hua et al., 1995, 1998; Hua and Meyerowitz, 1998; Grefen et al., 2008). Although the exact output of the receptors is still obscure, genetic studies demonstrate that in the absence of ethylene, receptors activate the Raf-like protein kinase CONSTITUTIVE TRIPLE RESPONSE 1 (CTR1), another negative regulator of the pathway (Kieber et al., 1993). Downstream of the receptors and the ER associated CTR1 kinase the membrane protein ETHYLENE INSENSITIVE 2 (EIN2), which contains a highly conserved NLS (Bisson and Groth, 2011; Qiao et al., 2012) shown to mediate interaction with the up-stream receptors (Bisson and Groth, 2015; Bisson et al., 2016), implements a positive regulatory role on ethylene signaling. In the presence of ethylene, the receptors bind the hormone and become inactivated. CTR1 cannot be activated by the receptors, and the lack of CTR1 activation cannot phosphorylate EIN2. Subsequently, the C-terminal end of EIN2 (C-END) containing the NLS-motif is cleaved off by an unknown mechanism and translocated to the nucleus (Ju et al., 2012; Qiao et al., 2012; Wen et al., 2012). In the nucleus, the EIN2 C-terminus directly or indirectly stabilizes the transcription factor EIN3 (Wen et al., 2012; Li et al., 2015) and its paralogous, the EIN3-like proteins (EILs), and transcription of ethylene response genes is activated (Chao et al., 1997; Solano et al., 1998).

In analogy to the model plant Arabidopsis, tomato contains a multigene family of the ethylene receptors. In total, seven isoforms named LeETR1, LeETR2, NR, LeETR4, LeETR5, LeETR6, and LeETR7 have been identified (Wilkinson et al., 1995; Zhou et al., 1996a,b; Lashbrook et al., 1998; Tieman and Klee, 1999) which are structurally diverse sharing at the most extreme less than 50% sequence identity. Similar to their Arabidopsis relatives the tomato receptors cluster in two subfamilies. LeETR1, LeETR2 and NR forming the subfamily I are characterized by a functional histidine kinase domain and a sensor domain consisting of three transmembrane helices. An additional putative membrane-spanning domain is present in LeETR5, LeETR6, LeETR7, and possibly in LeETR4 of subfamily II which are further characterized by a degenerated histidine kinase domain. All receptors except for NR contain a C-terminal response regulator domain (Lashbrook et al., 1998; Tieman and Klee, 1999). Expression patterns vary among the different receptor isoforms. While LeETR1 is expressed constitutively in all tissues, expression of LeETR2 is bound to seed germination and leaf senescence. NR, LeETR4 and to a lower extent LeETR5 are found at high expression levels in ripening fruit (Payton et al., 1996; Lashbrook et al., 1998; Tieman and Klee, 1999), but are rapidly degraded in the presence of ethylene by a 26S proteasome dependent pathway (Kevany et al., 2007). Due to this strong post-translational regulation of their protein level by the

**Abbreviations:** ACC, 1-aminocyclopropane-1-carboxylic acid; LeETR, ethylene receptor from tomato; NLS, Nuclear localization signal; NR, NeverRipe.

plant hormone and the observed correlation of receptor content and fruit ripening (Kevany et al., 2007), these receptors are of particular interest for studying the molecular effect of ripening inhibitors targeting ethylene signaling.

Recent insights in the ethylene signaling pathway propose a novel way to interfere with fruit ripening based on a yet unknown function of the NLS in the ethylene signaling protein EIN2. Peptides such as the synthetic octapeptide LKRYKRRL (NOP-1) mimicking this NLS motif were shown to block the interaction of EIN2 and ETR1 receptors and reduce plant ethylene responses (Bisson and Groth, 2015; Bisson et al., 2016).

In this study, we demonstrate that the NOP-1 octapeptide also efficiently binds to the ripening related tomato receptors NR and LeETR4 structurally divergent from ETR1. Moreover, we provide quantitative measures of the ripening delay related to NOP-1 treatment such as pigment content, overall color analysis and fruit firmness. Our data show that surface application of NOP-1 on tomato fruits can delay ripening without impairment of fruit quality.

#### MATERIALS AND METHODS

### Cloning of Tomato Receptors LeETR4 and NR into Expression Vector pET16b

Full-length codon optimized cDNA sequences encoding tomato ethylene receptors LeETR4 and NR (UniProt ID: LeETR4 Q9XET8; NR Q41341) were ordered at GenScript United States according to published sequences (NCBI ID: LeETR4 NM\_001247276.2; NR NM\_001246965.2). Construction of expression vector pET16b (Novagen, Madison, WI, United States) carrying the target DNA sequence, an ampicillin resistance and a deca-histidine tag were performed by Gibson Assembly (Gibson et al., 2009). For amplification of linearized vector forward primer 5<sup>0</sup> -GGATCCGGCTGCTAA CAAAGC-3<sup>0</sup> and reverse primer 5<sup>0</sup> -ATGACGACCTTCGATA TGGC-3<sup>0</sup> were used. LeETR4 was amplified using forward primer 5 0 -ATCGAAGGTCGTCATATGCTGCGTACCCTGGCGAG-3<sup>0</sup> and reverse primer sequence 5<sup>0</sup> -TTAGCAGCCGCCTTACAT CAGAGCTGGATTACGGCTACCACGCA-3<sup>0</sup> . For amplification of NR forward primer 5<sup>0</sup> -ATCGAAGGTCGTCATATGGACGA TTGCATT-3<sup>0</sup> and reverse primer 5<sup>0</sup> -TTAGCAGCCGGATCC TTACAGGCTACGCTGATAACGCT-3<sup>0</sup> were used. Amplified fragments were added to Gibson Assembly Master Mix containing an exonuclease, a DNA polymerase and a ligase to assemble a circular plasmid with LeETR4 and NR coding sequence, respectively. Reaction assays were incubated at 50◦C for 10 min and at 40◦C for 60 min. Assembled plasmids were transformed into E. coli strain XL 1-blue and sequenced by Seqlab (Göttingen, Germany) to verify correctness.

# Expression of Recombinant Tomato Receptors LeETR4 and NR in E. coli

For expression of recombinant LeETR4 and NR the related pET16b expression vectors were transformed into E. coli strains C43 and BL21 (DE3), respectively. Cells were grown in 2YT medium [1.6% (w/v) peptone, 1% (w/v) yeast extract and 0.5% (w/v) NaCl] with 2% ethanol and 100 µg/mL ampicillin at 30◦C. At OD<sup>600</sup> = 0.4 temperature was reduced to 16◦C. Expression of tomato receptors was induced at OD<sup>600</sup> = 0.6 by the addition of 0.5 mM isopropyl-β-d-1-thiogalactopyranoside (IPTG). Cells were grown and harvested after 20 h (LeETR4) or 6 h (NR) by centrifugation for 15 min at 7,000 × g and 4◦C. Expression of tomato receptors was analyzed by SDS–PAGE (Laemmli, 1970) and detected by Western blotting (Towbin et al., 1979).

## Solubilization and Purification of Recombinant Tomato Receptors LeETR4 and NR

The resulting cell pellet after expression was resuspended in PBS pH 8, 10% (w/v) glycerol, 1 mM dithiothreitol and 0.002% (w/v) phenylmethylsulfonyl fluoride (PMSF). DNase I (10 µg/mL) was added before cells were broken with Constants Cell Disruption System (Constant Systems, Daventry, United Kingdom) at 2.4 kbar and 5◦C. Cell lysate was centrifuged for 30 min at 14,000 × g and 4◦C. The resulting supernatant was centrifuged again for 30 min at 40,000 × g and 4◦C. The pellet was resuspended in PBS buffer and centrifuged for 30 min at 34,000 × g and 4 ◦C. For solubilization the pellet was resuspended in 50 mM Tris/HCl pH 8, 200 mM NaCl, 1.2% (w/v) FosCholine-16, 0.002% (w/v) PMSF (buffer S) and stirred at RT and 700 rpm for 1 h. Membrane fragments were isolated by ultracentrifugation (229,600 × g, 4◦C, 30 min). The supernatant was loaded to a 5 mL HisTrap FF column operated by an ÄKTAprime plus (both GE Healthcare Life Sciences) at 4◦C equilibrated with buffer A [buffer S containing 0.015% (w/v) FosCholine-16], followed by an ATP washing step of 20 column volumes [50 mM Tris/HCl pH 8, 200 mM NaCl, 50 mM KCl, 20 mM MgCl2, 10 mM ATP and 0.002% (w/v) PMSF]. The column was washed with 50 mM imidazole and receptors eluted with 250 mM imidazole. Purified proteins were concentrated to 2.5 mL and buffer was changed to 100 mM potassium phosphate buffer pH 7.3, 300 mM NaCl, 0.015% (w/v) FosCholine-16, 0.002% (w/v) PMSF for labeling with Alexa Fluor 488-Maleimide (Thermo Fisher Scientific) on a PD-10 column (GE Healthcare Life Sciences). Alexa Fluor 488-Maleimide was applied to the protein in 2.5-fold excess and incubated for 30 min at RT. Then, buffer was changed to 50 mM Tris/HCl pH 8, 300 mM NaCl, 5% (w/v) glycerol, 0.015% (w/v) FosCholine-16, 0.002% (w/v) PMSF. Purity of LeETR4 and NR was analyzed by SDS–PAGE (Laemmli, 1970) with colloidal Coomassie staining (Dyballa and Metzger, 2009) and Western blotting (Towbin et al., 1979) using a directly conjugated Anti-His-HRP monoclonal antibody (Miltenyi Biotech, Bergisch Gladbach, Germany). Proper folding of receptors was verified by CD-spectroscopy (Classen and Groth, 2012; Kessenbrock and Groth, 2017).

### CD Spectroscopy of Recombinant Tomato Receptors

CD measurements were performed in a Jasco J715 spectropolarimeter (Jasco GmbH, Gross-Umstadt, Germany). For the far UV spectra a cylindrical quartz cuvette from Hellma

Analytics (Muellheim, Germany) with 1-mm-path-length was used. Purified tomato receptors LeETR4 and NR were dissolved to a final concentration of 0.2 mg ml−<sup>1</sup> in 10 mM potassium phosphate pH 8.0 and 0.0075% (w/v) FosCholine-16. Protein and FosCholine-16 concentrations were determined by a Direct Detect Infrared Spectrometer (Merck Chemicals GmbH, Darmstadt, Germany) (Strug et al., 2014). For detailed information on protein preparation see Kessenbrock and Groth (2017). Measurements were run at ambient temperature. Each protein sample was recorded in the range of 260–185 nm. The CD spectra were obtained by averaging ten individual spectra using a bandwidth of 1 nm at 50 nm min−<sup>1</sup> . Secondary structure content of purified proteins were calculated from the spectra by CDSSTR and CONTINLL (Provencher and Gloeckner, 1981; Johnson, 1999).

#### Binding Studies of NOP-1 at Tomato Receptors LeETR4 and NR by Microscale Thermophoresis

Binding of the NOP-1 octapeptide to purified recombinant tomato receptors LeETR4 and NR was analyzed by microscale thermophoresis (MST) (Duhr and Braun, 2006; Wienken et al., 2010; Jerabek-Willemsen et al., 2011; Seidel et al., 2013). Receptors (100 nM) labeled with Alexa Fluor 488-Maleimide (Thermo Fisher Scientific) were titrated with peptide ligand NOP-1 dissolved in 50 mM Tris–HCl pH 8.0, 300 mM NaCl at concentrations from 500 µM to 61.04 nM. Then samples were transferred into standard glass capillaries and thermophoresis was measured using a Monolith NT.115 (NanoTemper Technologies GmbH, München, Germany). MST measurements were recorded at 20% MST power for LeETR4- NOP-1 and 60% MST power for NR-NOP-1, respectively. Receptors were chemically denatured by incubation with the strong ionic detergent SDS [4% (w/v)] and the small-molecule redox reagent DTT (40 mM) for 5 min in the dark at RT and served as control to confirm specific and selective binding of the ligand. All measurements were run in triplicate.

### Fruit Material, Treatments and Storage Conditions

Tomato fruits (Solanum lycopersicum L.) of the cultivar 'Lyterno' (Rijk Zwaan, De Lier, Netherlands) were harvested at the maturity stage "green" (USDA; 2005) from tomato plants which where cultivated in a commercial-like greenhouse at the research station Campus Klein-Altendorf (University of Bonn, Germany). At the early development stages, the trusses were manually thinned to six fruits per truss, according to the common practice aiming standardized fruit size and quality. For the experiment, the last two fruits of the fourth truss, counted from the bottom, were chosen for evaluations. At the beginning of the experiment, 150 fruits showing similar color and size were divided into four treatments (n = 25 fruits per treatment). On each fruit, a transparent polyethylene film was placed at the equatorial zone to demark the four evaluation points of 2.8 cm diameter each.

The four treatments were as follows: (1) control; (2) NOP-1, 400 µM; (3) NOP-1, 1000 µM; (4) NOP-1, 2000 µM. The NOP-1 peptide (GenScript, Piscataway, NJ, United States) was dissolved in 5 ml deionized water. On each fruit, a total of 200 microdroplets (0.5 µL each) of the peptide solution were gently deposited (50 micro droplets on each marked area of the tomato fruit) with a Hamilton microdispenser (Hamilton Bonaduz AG, Bonaduz, Switzerland). Fruits of the control treatment received an equal number of microdroplets of deionized water. After application of the droplets fruits of each treatment were allocated in storage boxes thereby avoiding fruit-to-fruit contact, and stored at room temperature (19 ± 2 ◦C).

### Non-destructive Measurements of Fruit Color at Ripening

Starting at the beginning of the experiment until 28 days after treatments (DAT), fruit ripening was evaluated and monitored twice a week with two non-destructive sensors using the principles of light reflectance and fluorescence emission. Evaluations were done on the marked fruit zones (n = 4 areas/fruit, n = 25 fruits per treatment).

Changes of the surface color over time were determined with a portable spectrophotometer (CM-503d, Konica Minolta Inc., Tokyo, Japan), which has a sensing area of 7 mm<sup>2</sup> . Based on the CIELAB model (McGuire, 1992), the recorded parameters are converted into the hue◦ index. Hue◦ was calculated according to the following formula:

$$\text{hue}^{\circ} = \tan^{-1} \left( \frac{b^\*}{a^\*} \right),$$

where 'a ∗ ' and 'b ∗ ' are defined as color coordinates, provided that 'a ∗ ' is the point of the green-red axis and 'b ∗ ' is defined as the point of the yellow-blue axis. On the basis of the above, the resulting angle is converted into the corresponding color value.

#### Fluorescence Based Analysis of Fruit Maturity

Pigment fluorescence was applied as second non-destructive technique to address fruit maturation. To this end, a handheld device (Multiplex <sup>R</sup> 3, Force-A, Orsay, France) equipped with light-emitting diodes (LEDs) with UV (375 nm), blue (475 nm), green (510 nm) and red (635 nm) excitation was used. Fluorescence was detected in the blue-green (BGF, 425–475 nm), red (RF, 680–690 nm) and near-infrared (FRF, 720–755 nm) spectral regions (Ben Ghozlen et al., 2010). Based on the absolute fluorescence signals recorded in a detection diameter of approximately 2 cm, simple and complex fluorescence ratios were calculated. The parameter Simple Fluorescence Ratio excited with red light (SFR\_R) as a measure of the chlorophyll content is as follows:

SFR\_R=(FRF\_R/RF\_R) (according to Ben Ghozlen et al., 2010).

#### Determination of Chlorophyll, β-Carotene and Lycopene

Determination of the pigment content in the fruits was done weekly on five fruits each treatment. The concentrations

of β-carotene, lycopene and chlorophyll were analyzed from freeze-dried and ground fruit samples, as described below. Concentrations of chlorophyll, β-carotene and lycopene were analyzed according to the method of Nagata and Yamashita (1992) as described by Barros et al. (2007), Azeez et al. (2012) and Kalogeropoulos et al. (2012) with the following modifications. Briefly, 1.5 ml of the solvent Aceton:Hexane (4:6) was added to 0.1 g of the freeze-dried and ground material, homogenized and centrifuged for 10 min at 16,100 × g (CENTRIFUGE 5415 R, Eppendorf AG, Hamburg, Germany). Next, 2 ml of the solvent was added to the supernatant and the absorption of the solutions was determined at 435, 505, 645, and 663 nm in a LAMBDA 35 spectrophotometer (PerkinElmer <sup>R</sup> , Waltham, MA, United States). Concentrations of chlorophyll, β-carotene and lycopene were calculated according to Nagata and Yamashita (1992) by the following equations:

chlorophyll a - mg/100 ml = 0.999A<sup>663</sup> − 0.0989A<sup>645</sup>

chlorophyll b - mg/100 ml = −0.328A<sup>663</sup> + 1.77A<sup>645</sup>

β − carotene - mg/100 ml = 0.216A<sup>663</sup> − 1.22A<sup>645</sup> − 0.304A<sup>505</sup> − 0.452A<sup>453</sup>

lycopene - mg/100 ml = −0.0458A<sup>663</sup> + 0.204A<sup>645</sup> + 0.372A<sup>505</sup> − 0.0806A<sup>453</sup>

Total chlorophyll was calculated by adding chlorophyll a and chlorophyll b content.

#### Statistical Analysis

All results are expressed as mean ± SE. Analyses of variance were determined with one-way ANOVA (α ≤ 0.05). In case of statistical significance, the Tukey's HSD (α ≤ 0.05) was applied to establish the differences among means. Statistical analyses were carried out using SPSS 22.0.

#### RESULTS

#### Expression and Purification of Tomato Receptors LeETR4 and NR

Codon optimized synthetic genes encoding full-length ethylene receptors LeETR4 and NR were each cloned into expression vector pET16b by Gibson Assembly Cloning (Gibson et al., 2009). Expression vectors encoding the tomato receptors were transformed into cells of E. coli strains BL21 (DE3) and C43 (DE3) which have been successfully applied for the expression of different members of the ethylene receptor family from A. thaliana, Lycopersicon esculentum and Physcomitrella patens in previous studies (Voet-van-Vormizeele and Groth, 2008; Classen and Groth, 2012). Protein expression was induced by the addition of 0.5 mM IPTG. Optimum expression was obtained for LeETR4 after 20 h in C43 (DE3) at 16◦C, while systematic analysis of expression parameter for NR showed best expression after 6 h in BL21 (DE3) at 16◦C (**Figure 1**).

Receptors were localized in the membrane fractions of the host and solubilized from these membranes by the mild detergent Fos-Choline-16. After solubilization, receptors were purified in a single chromatography step on Ni–NTA agarose (GE Healthcare Life Sciences, Munich, Germany). Purification of the recombinant tomato receptors was analyzed by SDS–PAGE. The related protein gels (**Figure 2**) show prominent bands at 90 and 70 kDa corresponding to the molecular weight of LeETR4 (88 kDa) and NR (74 kDa). Besides, two minor contaminations were detected in the lower MW range at 55 and 32 kDa, respectively. Hence, both receptors have been successfully purified from their heterologous host. Identity of the receptors was confirmed by antibodies directed against the deca-histidine tag in both proteins.

#### Secondary Protein Structure and Functional Folding of Purified LeETR4 and NR

Folding and protein secondary structure of purified recombinant tomato receptors were probed by CD spectroscopy. The corresponding spectra of LeETR4 and NR shown in **Figure 3** are typical of partially helical proteins, displaying two minima at approximately 209 and 222 nm with an isosbestic point at 202 nm. Overall, the spectra of both receptor proteins are highly similar and correspond to previous CD data on receptor orthologs from Arabidopsis and Physcomitrella (Classen and Groth, 2012). Secondary structure calculations by CDSSTR and CONTINLL suggest an α-helix content of 34% and a β-sheet percentage of 18–20% for LeETR4. Similar numbers of 41– 42% α-helix and 14% β-sheet structure were obtained for NR. Consequently, CD spectroscopic measurements verify that the purified receptors adopt a well-folded structure and are indicative for a native conformation of the recombinant tomato proteins.

#### Binding of NOP-1 to Purified Recombinant Tomato Receptors LeETR4 and NR

Analysis of protein–ligand interactions by MST was used to monitor and to quantify the interaction of synthetic NOP-1 octapeptide with purified recombinant LeETR4 and NR, respectively. Ligand binding and the related dissociation constant with the isolated receptors were deduced from changes in thermophoresis upon addition of the NOP-1 octapeptide (**Figure 4**). In analogy to previous studies on receptors from Arabidopsis and constitutively expressed tomato receptor LeETR1 (Bisson et al., 2016) clear changes of the thermophoretic signal were observed upon addition of the synthetic peptide with purified recombinant LeETR4 and NR receptors that are expressed at high levels at fruit ripening. Selectivity of the peptide–receptor interaction was probed in MST studies with chemically denatured receptor proteins. In these experiments, no change in thermophoresis was detected upon addition of NOP-1 (**Figure 4**). The apparent dissociation constant (Kd) calculated from the changes in thermophoresis induced by different

amounts of NOP-1 added to fluorescently labeled tomato receptors was 4.15 ± 0.85 µM for LeETR4 and 23.52 ± 1.99 µM for NR, respectively (**Figure 4**). Both numbers are in the lower micromolar range and together with the negative controls on denatured receptor proteins are indicative of efficient and specific binding of NOP-1 to tomato receptors LeETR4 and NR.

### Impact of NOP-1 on Fruit Ripening and Fruit Quality

For many fruits and vegetables color development is the most important external characteristic to assess ripeness and postharvest life. Color change from green to red was slowed down in tomato fruits treated with NOP-1 (1000 µM), as indicated by the hue◦ index which was significantly higher

on DAT 4, 7, and 9 in this treatment group as compared to control (untreated) fruits (**Figure 5A**). For treatment with 400 or 2000 µM NOP-1 slower color change as compared to control fruits were observed only at DAT 4, and to a smaller extent at DAT 7. Thereafter, color change of treated fruits was similar to controls. From DAT 14 onwards, there were no significant differences in hue◦ among all evaluated treatments. Similar to the observed effects on the hue index, application of 1000 µM NOP-1 also resulted in a significantly higher Simple Fluorescence Index (SFR\_R, estimating chlorophyll concentration) on DAT 4, 7, and 9 (**Figure 5B**). A slight increase in SFR\_R compared to non-treated fruits was further observed

for fruits treated with 400 and 2000 µM NOP-1 at DAT 7. Finally, all treatments reached similar values on DAT 14 and no further changes were observed until the end of the experiment (DAT 23).

Color development is another sensitive but invasive measure to monitor ripening. The chlorophyll content of tomato fruits over time is shown in **Figure 6A**. Fruits treated with 1000 µM NOP-1 showed higher concentrations of total chlorophyll compared to non-treated controls throughout the experiment. While total chlorophyll decreased rapidly to about 30% after DAT 9 in controls, only 50% of the pigment originally present was degraded in fruits treated with 1000 µM at this time. However, over time chlorophyll breakdown in fruits treated with 1000 µM NOP-1 converged to chlorophyll degradation in controls and had essentially ceased after DAT

16. Surprisingly, fruits treated with 400 and 2000 µM NOP-1 showed a more pronounced chlorophyll breakdown on DAT 9 than non-treated controls. However, with further progression of the experiment chlorophyll degradation in these fruits ceased and chlorophyll levels adapted to controls on DAT 16–24.

In contrast to the observed degradation in chlorophyll concentration of β-carotene increased in all treatments throughout the experiment (**Figure 6B**). However, fruits treated with 1000 µM NOP-1 showed a slower increase in β-carotene concentration and revealed significantly lower levels of this pigment at DAT 9 when compared to all other treatments which essentially showed the same pattern for the increase of this carotenoid during ripening. Over time β-carotene in the fruits treated with 1000 µM NOP-1 increased to control levels. All treatments showed comparable levels of this carotene on DAT 16 and concentration of this pigment remained constant throughout

FIGURE 6 | (A) Total chlorophyll content of control tomato fruits ( ) and tomato fruits treated with NOP-1 400 <sup>µ</sup>M (#), 1000 <sup>µ</sup>M (H) or 2000 <sup>µ</sup>M (1). Concentration of <sup>β</sup>-carotene (B) and lycopene (C) of control tomato fruits ( ) and tomato fruits treated with NOP-1 400 <sup>µ</sup>M (#), 1000 <sup>µ</sup>M (H) or 2000 <sup>µ</sup><sup>M</sup> (1). Data are means ± SE, n = 5, different letters indicate significant differences between treatments on each measurement day (Tukey's HSD, α ≤ 0.05).

the further experiment. Slightly higher concentrations of β-carotene were observed for the 400 µM treatment at DAT 23–24.

The concentration of lycopene principally responsible for the characteristic deep-red color of ripe tomato fruits, increased significantly in all treatment groups (**Figure 6C**). Similar to the pattern observed for β-carotene, treatment with 1000 µM NOP-1 showed the largest delay in pigment accumulation at DAT 9 where only 33% of the lycopene level of non-treated controls was measured. Treatment with 400 µM of the synthetic peptide still resulted in a delay of lycopene accumulation of about 69% compared to non-treated controls. According to the pattern observed with β-carotene, lycopene levels in all treatments adjusted to similar concentrations after DAT 16 and stayed constant for the rest of the experiment. Highest concentrations of lycopene were measured for all treatments in completely ripe tomatoes at DAT 24. Total color development of fruits treated at different concentrations of NOP-1 is illustrated by visual images of whole fruits (**Figure 7**).

Fruit soften during ripening due to biochemical processes resulting in the breakdown of cell-wall polymers. Hence, firmness is an indirect measurement of ripeness and represents one of the most important variables for fruit quality. Consequently, we determined fruit firmness in tomato fruits treated with NOP-1 and non-treated controls using a non-destructive sensor and a shore scale ranging from 0 to 100 units. Firmness decreased from 80–89 shore to 55–50 shore in the course of the experiment (**Figure 8**). Firmest fruits were observed at the beginning of the experiment, softest fruits were measured at the end. Control, treatment with NOP-1 at concentrations of 400 and 2000 µM showed a continuous decrease in fruit firmness over time, whereas firmness was unaffected at early stages (DAT 4–9) in fruits treated with 1000 µM NOP-1. However, after the initial lag phase firmness decreased to similar levels in these fruits as observed for the other treatments on DAT 16 and later stages. All treatments showed similar numbers for fruit firmness and thereby comparable fruit quality at the end of the experiment on DAT 25.

#### DISCUSSION

Previous studies on Arabidopsis demonstrated that the small basic peptide NOP-1 derived from the natural NLS-sequence of the ethylene regulator protein EIN2 is able to disrupt ethylene signaling and inhibit plant ethylene responses. Protein–protein interaction studies on recombinant purified proteins EIN2 and ETR1 and related FRET studies in planta suggest that the inhibitory peptide competes for binding of EIN2 at the receptors offering a novel way to interfere with ethylene signal transduction and ethylene responses in planta (Bisson and Groth, 2015; Bisson et al., 2016). The high conservation of the NLS-motif among the plant kingdom (see Supplementary Figure S1 in Bisson et al., 2016) and the level of homology in ethylene receptors open up new avenues for ripening control of fruits and vegetables by biological peptides in modern agriculture and horticulture. Initial studies on tomato, a climacteric fruit of high economic and nutritional impact serving as model to study fruit ripening, support these ideas and confirm the results obtained with the Arabidopsis genetic model.

FIGURE 7 | Visual images of fruits treated at different concentrations of NOP-1. Representative photos of whole fruits treated with 400, 1000 and 2000 µM NOP-1 on DAT 1, 7, 10, 14, and 16. Control fruits are depicted in the upper row.

In order to further evaluate the potential of the inhibitory peptide identified in previous studies, we have further analyzed molecular and physiological effects of the basic NLS-derived peptide NOP-1 on tomato. Our studies with purified recombinant receptors LeETR4 and NR, which are both highly expressed in ripening fruit, reveal efficient binding of the peptide to both receptors and thereby confirm that the NOP-1 octapeptide may interact with receptors from both receptor subfamilies. Both receptors interact with the peptide at affinities in the lower µM-range, but interaction with LeETR4 seems to be stronger.

Noteworthy, previous studies demonstrate an even stronger binding affinity of NOP-1 at tomato receptor LeETR1 – a receptor of subfamily I continuously expressed throughout the plant. However, LeETR4 and NR are quite different from LeETR1 with respect to kinase activity as well as in the number and their type of phosphorylation sites (Kamiyoshihara et al., 2012). Keeping further in mind the actual sequence identities of Arabidopsis ETR1 and tomato LeETR1/NR/LeETR4 of 81, 69, and 41%, the observed range in affinities of different receptors and receptor subfamilies for NOP-1 is not surprising at all. Bearing in mind that NOP-1 was derived from the NLS motif in EIN2 the different binding affinities observed with the peptide may also suggest that receptors have different affinities for the EIN2 central hub.

In associated post-harvest studies, we have evaluated changes in color and texture development of tomato fruits treated with different concentrations of NOP-1 and the impact of the

NLS-derived peptide on fruit ripening and fruit quality. Our studies show clear effects on fruit ripening at concentrations of 1000 µM, whereas effects on color development at 400 µM are substantially less pronounced and manifest only at early ripening stages (DAT 9 – see **Figure 6C**). In total, a concentration of 400 µM NOP-1 applied to the fruit surface as microdroplets seems to be too low for significant inhibition of the ethylene transduction cascade, whereas 1000 µM allows a ripening delay, as expressed by the chlorophyll degradation and lycopene and β-carotene accumulation (**Figure 6**). When applied at 2000 µM concentration no effect of NOP-1 was observed which might be related to concentration-dependent aggregation or changes in secondary structure (Garg et al., 2013), which may impair uptake of the peptide by the fruit surface. Alternatively, the fact that no significant positive effect on ripening delay was observed at 2000 µM concentration in contrast to 1000 µM NOP-1 may be explained by differences in droplet–surface–uptake interactions at the different concentrations. In this case, the solution concentration of 1000 µM showed higher uptake of NOP-1 at the same contact surface area (i.e., tomato fruit surfaces), possibly due to the optimum dose/concentration/interfacial area which apparently causes the maximum penetration. For 2000 µM droplet the NOP-1 loading was twice as high at the same water volume. Consequently, this may have reduced the droplet viscosity on the cuticle, also apparent by its modified macroscopic appearance and altered biomechanical properties (Domínguez et al., 2011; Burkhardt and Hunsche, 2013). In consequence, penetration and uptake of NOP1 may have been reduced, resulting in a less effective 2000 µM treatment.

Improving shelf life and nutritional quality of tomato fruits is difficult to achieve with the methods currently in use. Maintaining adequate storage conditions is expensive and might cause chilling injury if used improperly (Passam et al., 2007). Even though genetic modifications reducing gene expression of proteins involved in ethylene synthesis are possible in principle (Abano and Buah, 2015) these procedures are banned by law throughout Europe and have low acceptance among European

#### REFERENCES


consumers. Chemical methods such as application of AVG to inhibit ethylene biosynthesis or treatment with 1-MCP to inhibit ethylene response significantly delay ripening and slow down lycopene synthesis and chlorophyll breakdown (Saltveit, 2005; Passam et al., 2007). The drawbacks of these methods correlating with the restricted use of both chemicals on tomatoes in Europe also relate to quality losses in taste development or to complete arrest in maturation, as observed in some cases (Passam et al., 2007).

In summary, our study shows that the NOP-1 octapeptide derived from the NLS-motif at the EIN2 C-terminus is a potent inhibitor of the maturation process in tomato. The peptide efficiently binds to different receptor isoforms and, when applied to the surface of immature fruit, successfully delays the ripening process without impairment of final overall fruit quality at the fully mature stage. This novel approach to delay fruit ripening is making use of a synthetic peptide that corresponds to the highly conserved NLS-motif in all known EIN2 sequences and holds great promise to control processes such as ripening or senescence in horticultural and agricultural applications.

#### AUTHOR CONTRIBUTIONS

GG conceived the project. GG, GN, and MH planned, designed and supervised the research. MK, SK, and LM performed the experiments and contributed equally to this work. All authors contributed to data analysis and the writing of the manuscript.

#### ACKNOWLEDGMENTS

This work was funded in part by a grant from the Ministry of Innovation, Science and Research within the framework of the NRW Strategieprojekt BioSC (No. 313/323-400-002 13) and by the Deutsche Forschungsgemeinschaft, CRC 1208, project B06 to GG.


Canene-Adams, K., Campbell, J. K., Zaripheh, S., Jeffery, E. H., and Erdmam, J. W. Jr. (2005). The tomato as a functional food. J. Nutr. 135, 1226–1230.


ethylene signaling. Methods Mol. Biol. 1573, 141–159. doi: 10.1007/978-1-4939- 6854-1-12



Zhou, D., Mattoo, A., and Tucker, M. (1996b). Molecular cloning of a tomato cDNA encoding an ethylene receptor. Plant Physiol. 110, 1435–1436.

**Conflict of Interest Statement:** MH is associate professor at the University of Bonn and Global Head of R&D of the company COMPO EXPERT GmbH.

The other 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 Kessenbrock, Klein, Müller, Hunsche, Noga and Groth. 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.

# Diverse Roles of Ethylene in Regulating Agronomic Traits in Rice

Cui-Cui Yin<sup>1</sup> , He Zhao1,2, Biao Ma<sup>1</sup> , Shou-Yi Chen<sup>1</sup> and Jin-Song Zhang1,2 \*

<sup>1</sup> State Key Lab of Plant Genomics, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing, China, <sup>2</sup> University of Chinese Academy of Sciences, Beijing, China

Gaseous hormone ethylene has diverse effects in various plant processes. These processes include seed germination, plant growth, senescence, fruit ripening, biotic and abiotic stresses responses, and many other aspects. The biosynthesis and signaling of ethylene have been extensively studied in model Arabidopsis in the past two decades. However, knowledge about the ethylene signaling mechanism in crops and roles of ethylene in regulation of crop agronomic traits are still limited. Our recent findings demonstrate that rice possesses both conserved and diverged mechanism for ethylene signaling compared with Arabidopsis. Here, we mainly focused on the recent advances in ethylene regulation of important agronomic traits. Of special emphasis is its impact on rice growth, flowering, grain filling, and grain size control. Similarly, the influence of ethylene on other relevant crops will be compared. Additionally, interactions of ethylene with other hormones will also be discussed in terms of crop growth and development. Increasing insights into the roles and mechanisms of ethylene in regulating agronomic traits will contribute to improvement of crop production through precise manipulation of

#### Edited by:

Chi-Kuang Wen, Shanghai Institutes for Biological Sciences (CAS), China

#### Reviewed by:

Caren Chang, University of Maryland, College Park, United States Rongfeng Huang, Institute of Biotechnology (CAAS), China

#### \*Correspondence:

Jin-Song Zhang jszhang@genetics.ac.cn

#### Specialty section:

This article was submitted to Crop Science and Horticulture, a section of the journal Frontiers in Plant Science

Received: 30 June 2017 Accepted: 12 September 2017 Published: 26 September 2017

#### Citation:

Yin C-C, Zhao H, Ma B, Chen S-Y and Zhang J-S (2017) Diverse Roles of Ethylene in Regulating Agronomic Traits in Rice. Front. Plant Sci. 8:1676. doi: 10.3389/fpls.2017.01676 ethylene actions in crops.

Keywords: ethylene, agronomic traits, rice, crops, stress

# INTRODUCTION

Ethylene is a simple gaseous phytohormone present in plants and regulates plant growth and developmental processes ranging from germination to senescence (Bleecker and Kende, 2000). Dark-grown Arabidopsis seedlings treated with saturated ethylene exhibit triple response that consists of inhibiton of hypocotyls and roots and exaggeration of the curvature apical hooks (Guzmán and Ecker, 1990). Significant progress has been made in the ethylene signal pathway after the discovery of ethylene-insensitive and constitutive ethylene response mutants from the dicotyledonous model Arabidopsis. In Arabidopsis, ethylene is perceived by its five receptors (Hua and Meyerowitz, 1998). Then receptors and another negative regulator, CONSTITUTIVE TRIPLE-RESPONSE1 (CTR1) are inactivated, and the central signal transducer ETHYLENE-INSENSITIVE 2 (EIN2) C-terminal end (CEND) (Lin et al., 2009a) is dephosphorylated and cleaved. The cleaved CEND is translocated into the nucleus (Qiao et al., 2012; Wen et al., 2012) and the P-body. EIN2 CEND mediates translation repression of ETHYLENE-INSENSITIVE (EIN3) Binding F-Box1/2 (EBF1/2) at P-body and consequently activates ethylene response (Li et al., 2015). In the nucleus, the master transcription factors EIN3 and ETHYLENE INSENSITIVE-LIKE 1 (EIL1) lead to the ethylene-induced transcription activation and ethylene response (Chao et al., 1997). Interestingly, the EIN2 CEND also contributes to the downstream signaling through elevating the acetylation at H3K4 and H3K23 (Zhang et al., 2016a). Although these studies shed better light on how ethylene regulates plant growth and development and adaption to environment, there are still limitations due to the difference between dicotyledonous and monocotyledonous plants.

Semi-aquatic rice has a "double response" (inhibition of root growth but promotion of coleoptile elongation) of darkgrown seedling upon ethylene treatment (Ma et al., 2013). The ethylene growth response kinetics in monocots, including millet, barley and rice, is distinct from that in the dicots such as Arabidopsis and tomato (Kim et al., 2012). There are five ethylene receptors in rice genome. OsERS1 and OsERS2 belong to subfamily I with a conserved histidine kinase domain, whereas OsETR2, OsETR3, and OsETR4 belong to subfamily II with a diverged kinase domain (Yang et al., 2015a). But rice and many other monocotyledonous plants do not have ETR1-type ethylene receptor and OsETR2 have Ser/Thr kinase activity (Wuriyanghan et al., 2009). In rice, OsRTH1, OsRTH2, and OsRTH3 are identified and they are homologous to AtRTH1 in Arabidopsis. OsRTH1 and AtRTE1 have the highest sequence identity, and ectopic transgenic analysis indicates that only OsRTH1 is able to mimic the function of AtRTE1 (Zhang et al., 2012). AtCTR1 is encoded by a single gene in Arabidopsis. Its loss-of-function mutation leads to a constitutive ethylene response (Huang et al., 2003). In rice, three homologs OsCTR1, OsCTR2 and OsCTR3 were identified, and OsCTR2 is most closely related to AtCTR1 (Wang et al., 2013; Yang et al., 2015a). Similar to that in Arabidopsis, OsEIN2 is a positive regulator in ethylene signal pathway of rice (Ma et al., 2013). EILs are the master transcription factors in ethylene signaling pathway, with six members found in rice (Yang et al., 2015a). However, only OsEIL1 and OsEIL2 regulate root and coleoptile ethylene response respectively, with the left members having no role in ethylene response (Yang et al., 2015b). These results indicate that OsEIL1 and OsEIL2 have divergent function in growth and development in rice. Moreover, both OsEIL1 and OsEIL2 negatively affect rice salt tolerance by promoting transcription of HIGH-AFFINITY K <sup>+</sup> TRANSPORTER 2;1 (HKT2;1) and Na<sup>+</sup> absorption in roots (Yang et al., 2015b). Carotenoid isomerase (CRTISO) catalyzes the conversion of prolycopene to all-trans-lycopene (Fang et al., 2008). ABA4 drives the conversion of zeaxanthin to neoxanthin but no enzyme activity is detected in Arabidopsis (North et al., 2007). Its homologous gene MHZ4/ABA4 mutation reduced abscisic acid (ABA) production in rice (Ma et al., 2014). Both MHZ5/CRTISO and MHZ4/ABA4 are involved in ABA biosynthesis. The study of the two ethylene response mutants mhz4/aba4 and mhz5/crtiso indicates that ethylene inhibits root growth through ABA pathway in rice (Ma et al., 2014; Yin et al., 2015). In contrast, in Arabidopsis, ethylene inhibition of root elongation is ABA-independent (Beaudoin et al., 2000). Besides, MHZ5/CRTISO and MHZ4/ABA4 mediated ABA pathway inhibits rice coleoptile elongation likely through downregulating ethylene signaling pathway (Ma et al., 2014; Yin et al., 2015). These studies indicate that the ethylene signal transduction has conserved and divergent aspects between Arabidopsis and rice (**Figure 1**). Furthermore, rice ethylene response is also different from other monocots such as maize, wheat, sorghum, and Brachypodium distachyon (Yang et al., 2015a). Therefore, identification of rice ethylene signal transduction components and analysis of the interactions between ethylene and other phytohormones will shed better

FIGURE 1 | A diagram of ethylene signal transduction pathway in rice. In the dark, rice etiolated seedlings have "double response" with ethylene treatment. Ethylene promotes coleoptile elongation but inhibits root growth in rice. A linear ethylene signaling pathway has been found in rice etiolated seedlings which is similar to that in Arabidopsis. Rice has homologs of Arabidopsis ethylene signaling pathway, such as ethylene receptors, OsCTRs, MHZ7/OsEIN2, and MHZ6/OsEIL1 and OsEIL2. In contrast to that in Arabidopsis, the AtEIN3/EIL1 homologs OsEIL1 and OsEIL2 have divergent functions in rice coleoptile and root growth. (A) Ethylene promotes coleoptile/mesocotyl elongation through OsEIL2 which inhibits GY1/EG1-mediated jasmonate (JA) biosynthesis. JA pathway acts downstream of ethylene signaling pathway to inhibit cell elongation. On the other hand, MHZ5/CRTISO and MHZ4/ABA4-mediated ABA pathway acts upstream of ethylene signaling pathway to inhibit transcription of MHZ7/OsEIN2 and inhibit coleoptile elongation. (B) Ethylene regulates root growth through MHZ6/OsEIL1 function. In Arabidopsis, AtEIN3 and AtEIL1 have functional redundancy in root inhibition. Moreover, ethylene inhibits root growth partially through MHZ5/CRTISO and MHZ4/ABA4-mediated ABA pathway. Arrows and T-bars indicate direct or indirect activation and inhibition, respectively. Dotted lines indicate several steps involved that are not shown in the diagram.

light on how ethylene is specifically involved in rice growth and adaptation.

Rice, as a major staple crop, feeds more than half of the world's population (Yu et al., 2016). To satisfy the increasing global demand of the growing population, a 50% increase in rice production will be required (Alexandratos and Bruinsma, 2012). Rice shares close synteny and collinearity with other important cereal crops (Gale and Devos, 1998). Rice has the smallest genome of the major cereals and rich genetic diversity (Sasaki et al., 2005). In addition, the sequence of rice entire genome (Sasaki et al., 2005; Du et al., 2017) provides basis for identification of homologous genes for other crops. Rice is an annual grass, and can finish a life cycle (**Figure 2**) in 5 months or less. Ethylene regulates several stages of life cycle of rice.

Previous research indicates that three phytohormones including gibberellins (GA), cytokinins, and brassinosteroids regulate several agronomically important traits of rice, such as plant height (Sasaki et al., 2002), grain number (Ashikari et al., 2005) and leaf erectness. Ethylene plays a subtle role in plant growth, particularly in adaption to stressful environmental conditions (Peng et al., 2014; Tao et al., 2015). Abiotic and biotic stresses affect growth of crops at various stages and ultimately result in loss of yield. Reproductive processes including flowering, grain filling, and maturation are highly affected by abiotic stresses. Ethylene-insensitive mutants offer a great opportunity for understanding ethylene signal transduction in rice (Ma et al., 2013; Yang et al., 2015b; Yin et al., 2015). Then, how does the ethylene signaling affect agronomic traits? Ethylene impacts on fruit ripening have been studied in a series of plant species, including tomato and apple, etc. (Alexander and Grierson, 2002; Lee et al., 2010; Li et al., 2017). Jasmonate (JA) promotes ethylene biosynthesis to promote apple fruit ripening (Li et al., 2017). This review will highlight the impact of ethylene on crop (especially for rice) agronomic traits, emphasizing that the topics should be further investigated.

# ETHYLENE AND SEEDLING GROWTH

Ethylene has a biphasic role, stimulating and inhibiting growth depending on the species, organs/tissue, developmental stages, and environmental conditions (Hattori et al., 2009; Zhong et al., 2012; Yu et al., 2013). In Arabidopsis, ethylene inhibits hypocotyl elongation through activating the transcription factors WAVED-DAMPENED 5 (WDL5) (Sun et al., 2015) and ETHYLENE RESPONSE FACTOR 1 (ERF1) (Zhong et al., 2012; Shi et al., 2016a,b) in the dark or low light intensities. Transcription factor HYPOCOTYL 5 (HY5) also participates in this process which is degraded by the E3 ligase CONSTITUTIVE PHOTOMORPHOGENIC 1 (COP1) (Osterlund et al., 2000). In the light, AtEIN3 acts upstream of COP1 to influence COP1 localization and HY5 stabilization, and balances the equilibrium between ethylene and light signaling and ultimately promotes

hypocotyl growth (Yu et al., 2013; Shi et al., 2016a,b). These analyses indicate that ethylene signaling pathway is indispensable for plant emergence from the soil.

In rice, coleoptile and mesocotyl are essential protection structures for seedling emergence from the soil. Ethylene promotes coleoptile elongation not only in the dark but also in the light (Ma et al., 2013; Wang et al., 2013; Yin et al., 2015). This fact is different from that in Arabidopsis where ethylene plays an opposite role in hypocotyl elongation in dark and in light. In addition, ethylene promotes the curvature of rice coleoptile in the light (Wang et al., 2013). Obviously, identification and characterization of the component of ethylene signaling pathway in rice will benefit us in unraveling its mechanisms in rice. Progress has been made in the function of ethylene on rice growth and development and agronomic traits using the forward and reverse genetic approaches.

The rice ethylene receptor loss-of-function mutants Osers2, Osetr2, and Osetr3 in Zhonghua background all exhibited mild enhanced ethylene response in coleoptile elongation (Wuriyanghan et al., 2009). However, loss-offunction Osers1, Osers2, and Osetr2 mutants in the strain Dongjin (DJ) background showed no apparent phenotype in coleoptile but an enhanced ethylene response in root growth inhibition (Ma et al., 2014; Yin et al., 2015). These results suggest that the rice ethylene receptors may have functional redundancy and background/subspecies-specific features. In Arabidopsis, the ethylene insensitivity phenotype of REVERSION-TO-ETHYLENE SENSITIVITY 1 (RTE1) overexpression lines depends on ETR1 (Zhou et al., 2007). The OsRTH1 overexpression lines exhibited ethylene insensitivity in coleoptile elongation in dark-grown seedlings and coleoptile curvature in light-grown seedlings (Zhang et al., 2012). As the key negative component, CTR's loss of function mutant Osctr2 had a stronger exaggerated coleoptile curvature on ethylene treatment and this phenotype is probably not due to different ethylene production but resulted from enhanced ethylene sensitivity (Wang et al., 2013). Loss-of-function mutant Osein2/mhz7 exhibited complete coleoptile and root insensitivity to ethylene. The OsEIN2-overexpression lines showed constitutive ethylene response without ethylene treatment and enhanced ethylene response in the presence of ethylene (Ma et al., 2013). OsEIL2- RNAi transgenic lines exhibited coleoptile insensitivity to ethylene, and loss of function in Oseil1/mhz6 led to ethylene insensitivity in roots (Yang et al., 2015b). These features indicate that OsEIL1/MHZ6 and OsEIL2 have divergent functions in seedling growth in rice, which is different from those in Arabidopsis where EIN3/EIL1 showed no apparent organ specificity in ethylene responses.

Upon ethylene perception, OsEIL2 directly bound to the promoter of GAOYAO/EXTRA GLUME (GY1/EG1) (Zhang et al., 2016b) and inhibited GY1/EG1-mediated jasmonic acid biosynthesis and ultimately promoted mesocotyl and coleoptile elongation in etiolated rice seedlings (Xiong et al., 2017). In addition, the promotion of ethylene responses in the coleoptile is correlated with a decrease in the levels of OsEIL2 and an increase in rice F-box protein OsEBF2 transcripts upon treatment with 10 ppm ethylene for 3 h, whereas no significant change was observed in the transcript levels of OsEIL1 (Wang X. et al., 2009; Kim et al., 2012). Whether OsEIL2 exerts its effects on mesocotyl/coleoptile growth through regulation at levels of transcriptional repression activity, its own gene transcription and/or protein degradation need further investigation in rice. Besides, the MHZ4/ABA4 and MHZ5/CRTISO-mediated ABA pathway likely inhibits coleoptile elongation through decreasing the transcription of OsEIN2 and acts upstream of ethylene signaling pathway (Ma et al., 2014; Yin et al., 2015).

Taken together, the crosstalk between ethylene and light signaling pathway is extensive and complex in plant emergence from soil (Shi et al., 2016a,b; Xiong et al., 2017; Yu and Huang, 2017). The growth of rice coleoptile plays a vital role in earlystage rice seedling growth and development, and ethylene has a major role in this process. What is the molecular mechanism of ethylene in promoting rice coleoptile elongation through OsEIL2? Are there any specific downstream factors, e.g., OsERFs, involved in this process? How does ethylene collaborate with light in regulating these processes? Are the mechanisms different from that in Arabidopsis? Revealing the molecular mechanism of ethylene control on the elongation of rice coleoptile/mesocotyl will provide new genetic resources and theoretical basis for breeding new cultivars suitable for growth in dry land.

In addition to regulating etiolated seedlings, ethylene influences rice seedling growth in light. OsRTH1 overexpression had an inhibitory effect on ethylene-induced leaf elongation and adventitious root growth in rice (Zhang et al., 2012). Osctr2 produced more adventitious root regardless of ethylene than control cultivar DJ (Wang et al., 2013). When grown in water, compared to the curved roots in control seedlings, the root of Osein2/mhz7 mutant was longer and straight, while the shoots of Osein2/mhz7 were significantly shorter than control cultivar (Ma et al., 2013). This short shoot phenotype is consistent with that observed in OsEIN2-RNAi plants by Jun et al. (2004). The phenotype of Oseil1/mhz6 seedling was similar to that of Osein2/mhz7 mutants above. However, it should be noted that, at mature stage, the plant height of Oseil1/mhz6 mutant, OsEIL1-overexpressing plants and OsEIL2-RNAi lines showed no significant difference compared to control plants, while the OsEIL2-overexpressing transgenic plants were apparently shorter than control plants (Yang et al., 2015b). The mhz4/aba4 mutant plants were taller than WT and MHZ4-overexpression transgenic lines were shorter than WT. Double mutant analysis revealed that MHZ4/ABA4 negatively regulates plant height in an OsEIN2 dependent manner (Ma et al., 2014). In contrast, the plant height of mhz5/crtiso is shorter than WT due to the shorter length of all internodes (Yin et al., 2015). Double mutant analysis revealed that MHZ5/CRTISO regulate plant height in an OsEIN2-independent manner.

#### ETHYLENE AND FLOWERING

The control of plant flowering time is essential to produce sufficient seed for propagation and can be influenced by environmental factors (e.g., day length, temperature, etc.) and endogenous developmental cues of plants. The major genes

controlling flowering time are conserved and divergent between Arabidopsis and rice (Higgins et al., 2010). The flowering time regulatory networks of the two model plants were compared in the review provided by Shrestha et al. (2014). The two evolutionarily distant plant species had different photoperiod requirement. The B-type response regulator EARLY HEADING DATE 1 (EHD1) in rice has no ortholog genes in Arabidopsis and it acts independent of HEADING DATE1 (Hd1). This is a unique flowering pathway in rice (Doi et al., 2004).

Ethylene regulates plant flowering (Abeles et al., 1992; Iqbal et al., 2017). In Arabidopsis, the ethylene-overproducer mutant eto1 exhibited early flowering, whereas the dominant gain-offunction etr1 mutants, and loss-of-function ein2-1 and ein3-1 mutants showed delayed flowering (Ogawara et al., 2003). However, the loss-of-function mutation of the key negative regulator Ser/Thr kinase gene AtCTR1 led to delayed flowering (Achard et al., 2007). The discrepancy of the role of ethylene signaling components in flowering requires further investigation. In rice, the OsETR2-RNAi lines and loss-of- function Osetr2 mutant exhibited early flowering and heading time, and the latter showed enhanced ethylene sensitivity compared to that of control. Overexpression of OsETR2 transgenic lines exhibited reduced ethylene sensitivity and showed late flowering time compared with that of WT. The delayed flowering is associated with the higher transcript level of OsGI (OsGIGANTEA) and RCN1 (TERMINAL FLOWER1/CENTRORADIALIS-like). Recessive loss-of-function Osetr3 mutant also showed early flowering (Wuriyanghan et al., 2009). At mature stage, the OsETR2-overexpression lines matured later than controls and RNAi-lines (Wuriyanghan et al., 2009). Both loss-of-function Osctr2 mutants and the 35S:OsCTR21−<sup>513</sup> lines exhibited late flowering compared with the control plants (Dongjin and ZH11) (Wang et al., 2013). The panicles of Osein2/mhz7 mutant seemed to have more green grains and matured later than control plants (Ma et al., 2013). These studies indicate that ethylene signaling likely leads to early flowering in Arabidopsis and in rice. Alterations of ABA-biosynthesis pathway in mhz4/aba4, MHZ4 overexpressing plants, and mhz5/crtiso resulted in delayed heading time in comparison with control plants (Ma et al., 2014; Yin et al., 2015).

Ethylene affects plant flowering time by interaction with other classical phytohormones. It has been reported that the inhibitory effect of auxin in Pharbitis nil flowering results from the stimulation of ethylene production (Kesy et al., 2008). It is likely that ethylene inhibits flowering probably depending on the ABA level which is influenced by ethylene in Pharbitis nil (Wilmowicz et al., 2008). In Arabidopsis, gibberellin plays an important role in initiating flowering through promoting specific transcription factors under short days and long days, respectively (Wilson et al., 1992; Porri et al., 2012). DELLA-domain proteins are transcription factors and function to repress gibberellin (GA) responses in plants (Fleet and Sun, 2005). In Arabidopsis, ethylene promotes EIN3 accumulation and results in delayed flowering in a DELLA-dependent manner under short day conditions (Achard et al., 2007). These findings suggest that mechanism of ethylene regulation of plant flowering time is complex and needs further investigation.

# ETHYLENE AND GRAIN FILLING

Seed maturation process usually involves grain filling and fruit ripening. Previous studies mainly focus on the mechanism of ethylene on climacteric plant fruit ripening (Barry and Giovannoni, 2007). Ethylene has an auto-inhibitory effect during vegetative growth and an auto-stimulatory effect during fruit ripening (Lelièvre et al., 1998). However, ethylene effect on grain filling of crops is largely obscure and needs elucidation. Rice grain mainly consists of starch, protein, and other metabolites. The two major reserves of starch and protein largely determine the quality and yield of rice (Duan and Sun, 2005). Increasing the number of spikelets per panicle would likely expand yield sink capacity but might result in poor grain filling and low yield. Effect of ethylene on filling of compact-panicle and laxpanicle in rice has been studied. The differential grain filling of the superior and inferior spikelets probably resulted from the higher content of ethylene in compact-panicle rice cultivar (Yang et al., 2006b; Panda et al., 2015, 2016). The basal spikelets produce more ethylene at anthesis, and the expression of ethylene receptors and ethylene signaling transducers retained longer at post-anthesis compared to apical spikelets (Sekhar et al., 2015). Higher ethylene concentration negatively correlated with cell division rate, grain filling and starch concentration, but positively correlated with soluble sugar concentration of the growing endosperm (Panda et al., 2009; Kuanar et al., 2010; Sekhar et al., 2015). Ethylene affects the activity of the enzymes for starch synthesis (Naik and Mohapatra, 2000). The downregulation of essential proteins for rice cell progression and division, such as importin-α, elongation factor 1-β and cell division control protein 48 (CDC48), might limit the sink capacity and lead to poor grain filling in the inferior spikelets compared to that in the superior spikelets (Das et al., 2016). Ethylene negatively regulated grain filling by decreasing the activities of sucrose synthase (SuSase) and starch synthase (StSase) (Zhao et al., 2007). These analyses suggest a negative impact of ethylene on grain filling. Inhibitors including 1-MCP, AVG (aminoethoxyvinylglycine), and STS of ethylene biosynthesis or signaling have been useful in investigating the function of ethylene in rice grain filling. 1-MCP and AVG treatment increased cell number and size, and starch synthesizing enzyme gene expression, and ultimately improved grain filling (Mohapatra and Mohapatra, 2006; Panda et al., 2016). Furthermore, 1-MCP treatment enhanced expression of cell cycle regulators such as cyclin dependent kinase (CDK), cyclin (CYC) and cyclin dependent kinase inhibitor (CKI) and ultimately increased grain filling (Panda et al., 2016).

Using rice lines altered in ethylene signaling, regulation of grain filling by ethylene can be characterized. Ethylene receptor OsETR2 delays the flowering time and influences starch accumulation in the grains through decreasing transcription level of a monosaccharide transporter gene, ultimately leading to a lower seed-setting rate. In addition, both loss-of-function mutants Osetr2 and Osetr3 reduced starch accumulation in stems. The seeds of OsETR2-RNAi lines ripe earlier than that of seeds from control and overexpression lines (Wuriyanghan et al., 2009). The higher expression level of OsETR4 at 6 days after anthesis during spikelet development probably correlated with

grain-filling (Sekhar et al., 2015). The weight of grains of Osctr2 is slightly lower than that of DJ plants (Wang et al., 2013). Compared to WT, the total grain weight per plant decreased in Osein2/mhz7 mutant and OsEIN2-overexpression lines (Ma et al., 2013). OsEILs, acting downstream of OsEIN2, induce ethylene responsive factors (ERF) and/or other transcription activation/inhibition events (Hattori et al., 2009; Ma et al., 2013; Yang et al., 2015b). Sekhar et al. (2015) found that OsEIL1 is negatively associated with grain filling in the inferior spikelets of lax-panicles. Our recent research found that ethylene induced the expression of MHZ5/CRTISO and MHZ4/ABA4 (Ma et al., 2014; Yin et al., 2015). Moreover, grain filling is hindered in mhz4/aba4 mutant (Ma et al., 2014). Carotenoid metabolism is tightly related to tomato fruit ripening. The ethylene-induced carotenoid biosynthesis may also be beneficial for rice grain filling. These results indicate that alteration of ethylene signaling pathway affects rice grain filling.

Ethylene may work in coordination with other hormones to influence rice /grain filling. Numerous data suggest that ethylene negatively regulates rice inferior grain filling through interaction with ABA and other hormones. Under moderate water-stressed condition, higher ratio of ABA to ethylene/ACC concentrations was beneficial for the grain-filling rate in wheat and rice (Yang et al., 2002b, 2004, 2006a). Our results indicate that mutants of mhz4/aba4 and mhz5/crtiso have decreased ABA contents but higher ethylene production (Ma et al., 2014; Yin et al., 2015). The high ratio of ethylene to ABA inhibits the expression of starch synthesis genes and their enzyme activities and leads to poor grain filling (Zhu et al., 2011). Similar results have also been obtained in wheat. When treated with spermidine and spermine, the endogenous ethylene content decreased and led to poor the grain filling and lower grain weight (Liu et al., 2013). Furthermore, under drought conditions, spermidine and spermine significantly increased zeatin, zeatin riboside and ABA content, decreased ethylene content in grains, and improved grain-filling rate in wheat (Liu et al., 2016). Furthermore, the balance of ethylene and ABA plays a pivotal role in grain-filling rate (Yang et al., 2004). Besides, an increase in ABA content but a decrease in GA content, and the altered balance of hormones in rice grains controlled by moderate water deficit, could enhance the weight of the grains (Yang et al., 2001). Cytokinins accelerate cell division during the early phase of rice grain development and consequently influence grain filling (Yang et al., 2000, 2002a). In addition, ethylene inhibited the filling of grains in the basal spikelets while ABA and auxin stimulated this process (Zhang et al., 2009; Kuanar et al., 2010). IAA content is also associated with cell division at the early grain filling stage (Yang et al., 2001).

Taken together, the grain filling is a complicated process and may be regulated by ethylene and other hormones through interactions (Zhao et al., 2007; Liu et al., 2013, 2016). Roles of the hormones from exogenous application studies and from the endogenous determination studies have their own limitations and advantages. Genetic approaches should be more reliable among these, and mutants and transgenic materials should be more valuable in elucidating roles of hormones in grain filling. However, the molecular mechanisms by which ethylene regulates grain filling are largely obscure and need further investigation.

# ETHYLENE AND GRAIN SIZE

As one of the factors determining grain weight, grain size is specified by grain length, grain width, length-to-width ratio, and grain thickness. On the other hand, 1000-grain weight (KGW) is the most reliable trait in assessing grain weight (Huang et al., 2013; Zuo and Li, 2014). Previous studies highlight the pivotal roles of cytokinin (Ashikari et al., 2005), brassinosteroid (BR) (Hong et al., 2003; Tanabe et al., 2005; Sakamoto et al., 2006; Che et al., 2015), and auxin (Ishimaru et al., 2013) on rice grain production. Next the genetic and molecular regulation of ethylene on grain size is discussed.

Rice ethylene response mutants are ideal tools to study the effects of ethylene on agronomic traits. The 1000-grain weights of OsETR2-RNAi lines were substantially higher than that of control; however, this parameter was reduced or similar in ETR2-overexpression plants (Wuriyanghan et al., 2009). As a central component of ethylene signal transduction, OsEIN2/MHZ7 regulates several agronomic traits and plays a pivotal role in rice production. Compared to WT, the total number of panicles was significantly reduced and the KGW of total grains decreased in OsEIN2/MHZ7-overexpression lines. More precisely, OsEIN2/MHZ7 also affects rice grain size. In four allelic mutants of mhz7, the length and width of wellfilled grains decreased compared to WT grains except the allelic mutant mhz7-1. In MHZ7-overexpressing plants, grain length was increased compared to that in WT. Ratio of grain length/grain width in mhz7 mutants was higher or similar with that of WT (Ma et al., 2013). Similarly, loss-of-function Oseil1/mhz6 mutant exhibits a significant reduction in grain length and width compared to WT grains. OsEIL1/MHZ6 overexpression increases grain size and 1000-grain weight (Yang et al., 2015b). The grain size and KGW of OsEIL2-OX lines were smaller than those of WT plants whereas the two parameters of the OsEIL2-RNAi plants exhibited some fluctuation compared with that of WT (Yang et al., 2015b). From these analyses, we propose that ethylene regulates rice agronomic traits under a subtle control. Substantially, ethylene insensitivity may lead to reduced grain size and grain weight, and enhanced ethylene response may be related to larger grains and/or more grain weight, depending on the individual signaling genes. The seedsetting rate and KGW of mhz4/aba4 and mhz5/crtiso were all significantly reduced. Besides, the two mutants showed an increase in grain length and the mhz4 mutant exhibited a decrease in grain thickness (Ma et al., 2014; Yin et al., 2015). The mhz5-3 Osein2 double mutant showed significant decreases in both seed-setting rate and 1000-grain weight, which is similar to that of mhz5/crtiso mutant plants (Yin et al., 2015). These observations suggest that MHZ5/CRTISO regulates seed-setting rate and 1000-grain weight in rice in an OsEIN2-independent manner.

Compared to other cereals, our knowledge about the molecular mechanism that regulates grain size in rice is more than that of wheat, sorghum and maize (Li et al., 2010; Zuo and Li, 2014). ZmIPT2, an isopentenyl transferase involved in cytokinin biosynthesis, is associated with kernel weight and final grain yield (Weng et al., 2013). Several key genes associated with seed size

have been cloned and studied in Arabidopsis. AtAP2 is involved in the increase of total protein and seed oil contents and ultimately regulating Arabidopsis seed size although it negatively affects plant fertility and growth (Jofuku et al., 2005). The mutation of NIMA-related kinase NEK6 leads to increased ethylene production and rounder seeds, while AtNEK6 overexpression suppressed expression of several ethylene biosynthesis- and signaling-related genes and resulted in smaller seeds and reduced 1000-grain weight. These results indicate that NEK6 negatively regulates Arabidopsis seed size (Zhang et al., 2011). Endosperm is the main component of a rice grain, whereas an Arabidopsis seed contains mostly the embryo. This difference suggests that grain size control may be different in various plant species.

### ETHYLENE AND LEAF SENESCENCE

Senescence of vegetative organs is essential for reserve remobilization to developing grain. The main function of leaf senescence is nutrient recycling and this confers an adaptive advantage. To some extent, leaf senescence is beneficial for flowering and seed production. Ethylene is a positive regulator of leaf and flower senescence (Bleecker and Kende, 2000). Studies show that ethylene is one of the most important hormones that affects plant leaf senescence (reviewed by Iqbal et al., 2017). In Arabidopsis, the two mutant alleles of EIN2 oresara2 (ore2) and (ore3) showed delayed senescence, indicating that ethylene affects developmental leaf senescence (Oh et al., 1997). The AtEIN2 positively accelerates leaf senescence by repressing microRNA164 (miRNA164) and regulating the transcription of ORE1/NAC2. The transcription factor ORE1/NAC2 is one of the targets of miRNA164 (Kim et al., 2009). AtEIN3 acts downstream of AtEIN2 to directly repress the expression of miRNA164 and increase ORE1/NAC2 gene transcription to promote leaf senescence (Li et al., 2013). In addition, AtEIN2 affects leaf senescence by regulating the senescence-associated NAC transcription factors (TFs), including ANAC019, AtNAP, ANAC047, ANAC055, ORS1, and ORE1/NAC2. AtEIN3 acts downstream of AtEIN2 and activates ORE1/NAC2 and AtNAP. Furthermore, ORE1/NAC2 and AtNAP affect leaf senescence by activating transcript of common and/or distinct downstream NAC TF factors (Kim et al., 2014). Recent research indicates that EIN3, ORE1/NAC2 and chlorophyll catabolic gene (CCGs) formed a coherent feed-forward loop in the process of ethyleneregulated chlorophyll degradation (Qiu et al., 2015). These results suggest that ethylene regulation of leaf senescence is a highly intricate process and further studies are required to reveal the role of ethylene in leaf senescence.

It is reported that an F-box protein (containing a Kelch repeat motif) OsFBK12 interacted with Oryza sativa S-PHASE KINASE-ASSOCIATED PROTEIN-LIKE PROTEIN (OSK1) to form an SCF complex and degrade its substrate, such as S-ADENOSYL-L-METHIONINE SYNTHETASE1 (SAMS1). The degradation of OsSAMS1 results in a lower content of ethylene, ultimately suppressing leaf senescence and increasing seed size and grain number (Chen et al., 2013). Overexpression of OsRTH1 prevented ethylene-induced leaf senescence through repressing the expression of ethylene-inducible gene, such as submergence 1C (Sub1c), alcohol dehydrogenase 2 (ADH2), and glutathione S-transferase (SC129) (Zhang et al., 2012). Loss-of-function Osctr2 mutant leaves exhibited a much stronger senescence phenotype than Dongjin with or without ethylene treatment. 1-MCP treatment could delay the ethylene-induced senescence in Dongjin but not in Osctr2 (Wang et al., 2013). Osein2/mhz7 mutant leaf remained green whereas the overexpression lines turned yellow under the treatment of dark-induced and natural senescence. Some senescence-associated genes (SAGs) were upregulated, including OsL36, OsL43, OsL85, OsL55, OsNAC1/2 (Ma et al., 2013). Overexpression of the submergence tolerance gene SUB1A delayed dark-induced flag-leaf senescence by limiting ethylene production and responsiveness to the other two positive senescence regulatory hormones JA and salicylic acid (Fukao et al., 2012). Ethylene regulates plant leaf senescence through interactions with other hormones. Ethylene interacts with JA to regulate rice leaf senescence by a direct regulatory cascade of OsCOI1b-OsJAZ-OsEIN3-OsORE1 (Lee et al., 2015). In rice, the AtNAP homologous gene OsNAP promotes leaf senescence and OsNAP is induced by ABA but not 1-aminocyclopropane-1-carboxylic acid (Liang et al., 2014). These studies suggest that the mechanism of NAC TFs involved in leaf senescence may be different in rice and Arabidopsis and need to be further explored.

#### ETHYLENE AND ARCHITECTURE ESTABLISHMENT

Leaf angle is an important morphological trait of plant architecture and influences rice cultivation and grain yield (Sinclair and Sheehy, 1999). Several plant hormones, e.g., ethylene, GA, and auxin are involved in leaf angle formation, and most of these phytohormones regulate the leaf angle through interaction with BRs (Cao and Chen, 1995; Wang L. et al., 2009). In addition, Cao and Chen (1995) suggested that ethylene may be involved in BR-induced rice leaf epinasty. However, Takeno and Pharis (1982) suggest that ethylene is not effective in rice lamina inclination in the intact leaves. Further study is required to reveal the role of ethylene in the BR-induced rice leaf epinasty. Rice ethylene response mutant may be useful for studying the relationship between leaf angle and ethylene. At mature stage, the OsETR2-overexpression lines had erect panicles compared with controls and RNAi-lines (Wuriyanghan et al., 2009). Osein2/mhz7 mutants were more erect than WT plants (Ma et al., 2013).

Great progress has been achieved in identifying important genes associated with rice tiller and panicle branches (for review Liang et al., 2014). Effective tillers produce panicles and play an important role in determining rice yield. The number of panicles per plant is one of the major factors determining rice yield. Previous studies showed that cytokinin (Ashikari et al., 2005), auxin (Xia et al., 2012), and SL (Lin et al., 2009b), and BR (Tong et al., 2009) regulate rice tiller and affect yield. The OsETR2 overexpression lines had reduced effective panicles and seed setting rates whereas the RNAi lines had no significant difference

FIGURE 3 | Diverse roles of ethylene in regulating agronomic traits in rice. In rice, ethylene regulates a wide variety of major agronomic traits ranging from emerging from the soil to grain filling and senescence. Ethylene promotes growth of coleoptiles/mesocotyls by partially inhibiting JA biosynthesis. In contrast, JA promotes rice leaf senescence through a cascade of OsCOI1b-OsJAZ-OsEIN3-OsORE1. JA and ethylene synergistically accelerate rice leaf senescence by activating common and differential SAGs. Ethylene positively regulates rice grain size/weight, flowering, tillering and leaf angle but negatively affects rice grain filling and salt tolerance. The lower ratio of ethylene to ABA is beneficial for rice grain filling. Upon submergence, SUB1A negatively regulates GA response by inhibiting the degradation of SLR1/SLRL1 to restrict plant growth in submergence-tolerance species. On the contrary, ethylene promotes internode elongation through increasing transcription of SK1/2 and GA production to escape flooding in deepwater rice. OsGI1, OsGIGANTEA1; RCN1, TERMINAL FLOWER1/CENTRORADIALIS-like; GY1/EG1, GAOYAO/EXTRA GLUME; HKT2;1, HIGH-AFFINITY K<sup>+</sup> TRANSPORTER2;1; SLR1, Slender Rice-1; SLRL1 SLR-like 1; ABA, abscisic acid; GA, gibberellin; JA, jasmonate; COI1, coronatine insensitive 1; JAZ, jasmonate ZIM-domain protein. Blue lines indicate the leaf senescence regulatory pathway. Purple lines indicate the submergence tolerance regulatory pathway. Arrows and T-bars indicate direct or indirect activation and suppression, respectively. Dotted lines indicate several or unknown steps involved in the pathway.

compared with the controls (Wuriyanghan et al., 2009). Both Osctr2 and 35S:OsCTR21−<sup>513</sup> transgenic lines produced more effective tillers than the wild type (Dongjin and ZH11) (Wang et al., 2013). The ABA-deficient mutant mhz4/aba4 plants had the same effective tiller number compared to WT, whereas the MHZ4-overexpressing lines produced more tillers than WT (Ma et al., 2014). In contrast, the other ethylene abnormal mutant mhz5/crtiso had excessive tillers, smaller panicles, and fewer primary and secondary branches in panicles compared with WT plants. The tiller number of double mutant mhz5-3 Osein2 was the same as mhz5-3, indicating that MHZ5/CRTISO regulates plant tiller in an OsEIN2-independent manner (Yin et al., 2015). In addition, the strigolactone biosynthesis is also impaired in mhz5/crtiso, and the SL mainly regulates plant tiller/branch (Lin et al., 2009b; Yin et al., 2015). These results suggest that ethylene directly or indirectly regulates rice tiller number.

# ETHYLENE AND SUBMERGENCE

Ethylene is the primary signal for rice and other semi-aquatic plants to adapt to flooding. Concentration changes of abscisic acid, gibberellin, and auxin are also required to gain fast growth under water (Raskin and Kende, 1984; Cox et al., 2004; Saika et al., 2007; Jackson, 2008). Rice has an abundant genetic diversity and evolved two mechanisms to escape flooding. First, increased ethylene upregulates the expression of snorkel1/2 (SK1/2) via OsEIL1b/OsEIL2 binding to the SK1 and SK2 promoters and GA content to promote internode elongation in deepwater rice during flooding (Hattori et al., 2009). Second, SUB1A, an ERF transcription factor existed in limited rice accessions, negatively regulates GA response through restriction of Slender Rice-1 (SLR1) and SLR-like 1 (SLRL1) degradation to inhibit plant elongation during flash floods at the seedling stage in water-tolerance rice (Fukao and Bailey-Serres, 2008).

SUB1A promotes plant recovery from drought at the vegetative stage through decrease of leaf water loss and lipid peroxidation, and increase of gene transcript level associated with acclimation to dehydration. In addition, overexpression of SUB1A could augment ABA responsiveness to enhance plant tolerance to drought (Fukao et al., 2011). Fukao and Xiong (2013) have reviewed the intricate regulatory mechanisms of rice plants acclimating the submergence and drought stress, and ethylene plays a key role in this process. Furthermore, the SUB1A is beneficial for plant physiological recovery upon desubmergence due to the greater capability for non-photochemical quenchingmediated photoprotection (Alpuerto et al., 2016). These results suggest that plant had orchestrating regulatory mechanisms to enhance plant adaptation.

#### CONCLUSION AND PERSPECTIVES

In this review, we summarized roles of ethylene-related factors and genes in regulation of seedling growth, flowering, grain filling, grain size, leaf senescence, leaf angle, tiller and submergence in rice (summarized in **Figure 3**). At present, several environmental cues, including global warming, heat stress, drought, chilling and salinity, affect crop productivity drastically and threaten global food security. Ethylene signaling is indispensable for plant response and adaptation. It has been proven that EIN3/EIL1 is the integration node of several hormones in plant growth and development and adaption to biotic and abiotic stresses (Kazan, 2015; Wawrzynska and Sirko, ´ 2016; Quan et al., 2017). Ethylene promotes salt tolerance in Arabidopsis (Lei et al., 2011). However, the research of Oseil1/mhz6 mutants and OsEIL2-RNAi lines indicates that ethylene signaling negatively regulates salt tolerance in rice (Yang et al., 2015b; **Figure 3**). The roles of ethylene in plant response to salinity has been reviewed by Tao and the divergence between Arabidopsis and rice in the regulation of salinity response by ethylene probably due to the evolutionary divergence under different growing conditions or different plant species (Tao et al., 2015).

In conclusion, higher concentrations of ethylene impair rice grain filling. Similar results were found in wheat. In mature wheat plants, increased ethylene production is associated with decreased 1000-grain weight and hastened maturity (Beltrano et al., 1999). Down-regulating the ethylene biosynthesis pathway

#### REFERENCES


can significantly improve the maize grain yield under abiotic stress (Habben et al., 2014). It should be noted that while higher production of ethylene negatively affects several agronomic traits in wheat and rice, overexpression of OsEIL1/MHZ6 increased grain size and 1000-grain weight (Yang et al., 2015b). OsEIN2/MHZ7 overexpression increased grain size but not grain weight (Ma et al., 2013), and OsETR2-RNAi lines and loss-offunction mutant Osetr2 produced higher 1000-grain weight than control (Wuriyanghan et al., 2009). This discrepancy suggests that the ethylene biosynthesis and signaling pathways may have different functions in grain size control or the ethylene effect is governed not only by the ethylene production, but also by the tissue sensitivity to ethylene. These discrepancies may also be due to the fact that different plants/stages/methods were used. Although each study has its own contribution toward a special area, genetic analysis should be regarded as an effective approach. While progress has been made in recent years in understanding the molecular basis of ethylene signaling in rice, many fundamental questions remain unanswered. Further investigation is needed to confirm the effects of ethylene on crop agronomic traits, especially grain/yield-related traits, using ethylene biosynthesis and signaling mutant or transgenic plants with an alteration of hormone action. How does ethylene signaling connect to grain filling, storage accumulation, and grain size control? The integration and crosstalk points should be identified. Elucidation of how ethylene influences rice grain development may contribute to the advancement of comparative biological studies in other cereals, and assist in breeding of novel high-yield/quality cultivars.

#### AUTHOR CONTRIBUTIONS

C-CY, BM, S-YC, and J-SZ conceived the topic. C-CY wrote the manuscript. J-SZ revised the final version. HZ drew the picture of rice life cycle.

### ACKNOWLEDGMENTS

This work is supported by the National Natural Science Foundation of China (31530004, 31600980, and 31670274), the 973 project (2015CB755702) and the State Key Lab of Plant Genomics.




**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 Yin, Zhao, Ma, Chen and Zhang. 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.

fpls-08-01676 September 23, 2017 Time: 15:18 # 12

# Comparative Physiological, Biochemical, and Genetic Responses to Prolonged Waterlogging Stress in Okra and Maize Given Exogenous Ethylene Priming

Emuejevoke Vwioko<sup>1</sup> , Onyekachukwu Adinkwu<sup>1</sup> and Mohamed A. El-Esawi 2, 3 \*

*<sup>1</sup> Department of Plant Biotechnology, Faculty of Life Sciences, University of Benin, Benin, Nigeria, <sup>2</sup> Botany Department, Faculty of Science, Tanta University, Tanta, Egypt, <sup>3</sup> The Sainsbury Laboratory, University of Cambridge, Cambridge, United Kingdom*

#### Edited by:

*Chi-Kuang Wen, Shanghai Institutes for Biological Sciences (CAS), China*

#### Reviewed by:

*Rakesh Kumar Shukla, Central Institute of Medicinal and Aromatic Plants (CIMAP), India Islam Abd El-Daim, Agricultural Research Center, Egypt*

\*Correspondence:

*Mohamed A. El-Esawi mohamed.elesawi@ science.tanta.edu.eg*

#### Specialty section:

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

Received: *04 May 2017* Accepted: *14 August 2017* Published: *25 September 2017*

#### Citation:

*Vwioko E, Adinkwu O and El-Esawi MA (2017) Comparative Physiological, Biochemical, and Genetic Responses to Prolonged Waterlogging Stress in Okra and Maize Given Exogenous Ethylene Priming. Front. Physiol. 8:632. doi: 10.3389/fphys.2017.00632* Waterlogging is an environmental challenge affecting crops worldwide. Ethylene induces the expression of genes linked to important agronomic traits under waterlogged conditions. The ability of okra (*Abelmoschus esculentus* L. Moench.) and maize (*Zea mays* L.) given exogenous ethylene priming to tolerate prolonged waterlogged conditions was investigated in this study. The investigation was carried out as field experiments using 3 week-old plants grouped into four treatments; control, waterlogged plants, ethylene priming of plants before waterlogging, and ethylene priming of plants after waterlogging. Different growth parameters were recorded. Soil chemical and bacterial analyses were performed. The activity and gene expression of antioxidant enzymes were studied. The ethylene biosynthetic genes expression analysis and root anatomy of surviving okra plants were also carried out. Results revealed that okra and maize plants showed increase in their height under waterlogged conditions. Ethylene priming and waterlogged conditions induced early production of adventitious roots in okra and maize. Maize survival lasted between 5 and 9 weeks under waterlogging without reaching the flowering stage. However, okra survived up to 15 weeks under waterlogging producing flower buds and fruits in all treatments. Variable changes were also recorded for total soluble phenolics of soil. Cross sections of waterlogged okra roots showed the formation of a dark peripheral layer and numerous large aerenchyma cells which may have assisted in trapping oxygen required for survival. The activity and gene expression levels of antioxidant enzymes were studied and showed higher increases in the root and leaf tissues of okra and maize subjected to both waterlogging and ethylene priming, as compared to control or waterlogged condition. Quantitative RT-PCR analysis also showed that the ethylene biosynthetic gene expression levels in all okra and maize tissues were up-regulated and showed much higher levels under ethylene-treated waterlogged conditions than those expressed under control or waterlogged conditions at all time points. These results indicate that okra and maize tissues respond to the conditions of waterlogging and exogenous ethylene priming by inducing their ethylene biosynthetic genes expression in order to enhance ethylene production and tolerate the prolonged waterlogging stress. In conclusion, this study revealed that exogenously generated ethylene gas as a priming treatment before or after waterlogging could enhance waterlogging tolerance in maize and okra crops.

Keywords: waterlogging, ethylene priming, gene expression, Abelmoschus esculentus, Zea mays

### INTRODUCTION

Waterlogging of upland communities in Southern Nigeria is a highly relevant environmental issue which is not welladdressed by agricultural experts. The flooding and waterlogging incidence of 2012 was economically devastating and many indigenous fishing and farming communities were sacked in Delta State (Nigeria). FAO (2002) also reported that over 10% of the global arable lands are affected by waterlogging and flooding annually. Hence, waterlogging has become a challenge to many agricultural zones of the world known for growing crops, including wheat, barley, maize, millet, sorghum, cassava, and leafy vegetables (Sayre et al., 1994). Waterlogged soil environments adversely affect the normal functioning of terrestrial plants and ecology (Setter and Waters, 2003). Widespread incidences of waterlogging are reported often in countries, including Pakistan, Nepal, India, Indonesia, Malaysia, Bangladesh and China, where the adverse effects are very pronounced in rice-wheat rotation farming systems (Samad et al., 2001).

Plants under waterlogged conditions are affected by gas exchange limitations, essential nutrient deficiencies and micronutrient toxicities such as Fe, Mn, and Cu (Setter et al., 2006). The shoot system of plants exhibit certain features, including wilting, premature yellowing of leaves (senescence), stunted plant height, epinasty, stem deformation, shoot length alteration, and leaf area reduction. Furthermore, crops including wheat show sterile floret and reduced grain yield and kernel weight (Hossain and Uddin, 2011). Root extension growth in terrestrial plants is arrested because the root tip is intolerant to hypoxic (low oxygen) or anoxic (absence of oxygen) conditions. Prolonged waterlogging will lead to alteration of chemical and physical properties of soil. These soil factors include pH, EC, hydraulic conductance, soil structure, porosity, and organics (Syversten et al., 1983; Setter et al., 2009). The oxygen-deficient soil environment leads to changes in composition and decomposition activities of microbes. The nutrient recycling process is hampered and nutrient deficiency conditions become eminent. The potential energy needed for nutrient uptake by plant roots comes primarily from aerobic respiration (Ferreira et al., 2008). In waterlogged soil, hypoxic, or anoxic conditions affect aerobic respiration and initiate anaerobiosis. In addition, synthesis and translocation of growth regulators, photosynthesis and carbohydrate partitioning are also negatively affected (Ferrer et al., 2005). These physiological impedances culminate in the reduced yield of crops under waterlogged conditions. In general, plant responses to waterlogged and flooding conditions are reported to include anatomical, physiological and molecular changes (Voesenek et al., 2006).

Ethylene is generated by plant tissues under abiotic stress such as drought and waterlogging (El-Esawi, 2016a,b,c). It is synthesized from methionine which forms S-adenosylmethionine. ACC synthase (ACS) then catalyzes the conversion of S-adenosylmethionine to 1-aminocyclopropane-1-carboxylate (ACC). ACC oxidase (ACO) then oxidizes ACC to generate ethylene (Geisler-Lee et al., 2010). Multigene families encode both ACS and ACO in several plant species such as maize (Zea mays L.) and okra (Abelmoschus esculentus L. Moench.). Among the well enunciated roles played by ethylene in waterlogged condition, induction of gene expression linked to leaf senescence, aerenchyma formation, adventitious roots, and epinasty are paramount (Jackson, 2008; Vidoz et al., 2010; Sasidharan and Voesenek, 2015) as morphological responses. These responses were observed with concomitant increase in endogenous ethylene synthesis in crops, including maize, barley, wheat, and soybean (He et al., 1994; Watkin et al., 1998; Drew et al., 2000). Exogenous ethylene treatment resulted in enhanced aerenchyma formation in rice (Takahashi et al., 2014). Under well-drained soil, aerenchyma formation is not observed in the root tissue of maize, whereas waterlogged condition induces aerenchyma formation in maize. Aerenchyma formation is attributed to the activity of ethylene in programmed cell death (PCD, Yamauchi et al., 2014).

Since the farming communities in Delta State (Nigeria) were substantially affected by the flood incidence of 2012, many farmlands were waterlogged for a period of at least 4 weeks and many crops generally did not tolerate this stress. In some of the riverine areas, farmlands were submerged for more than 2 weeks. Waterlogged conditions resulting from episodes of flooding in this region will reoccur as evidence of climate change. Therefore, the need to identify crops from the commonly grown crop species that can tolerate long periods of waterlogging has become foremost. Okra and maize are economically important food crops worldwide. The main objective of the current study was to assess okra and maize plants given exogenous ethylene priming for tolerating long periods of waterlogging. To achieve this objective, recording different plant growth parameters as well as root anatomy and soil chemical and bacterial analyses were conducted. Additionally, the activity and expression levels of antioxidant enzymes as well as the ethylene biosynthetic gene expression in okra and maize tissues were studied.

#### MATERIALS AND METHODS

#### Plant Material

The seeds of okra variety Clemson spineless (produced by Technism and packed in France) and maize variety Oba-98, were used in this study.

# Preparation of Soil Samples and Experimental Pots

Composite soil sample was obtained from the Faculty of Agriculture Demonstration Farm, University of Benin, Nigeria. Five kilograms of soil was weighed into each experimental pot. Forty pots were prepared for the study. These pots were not perforated underneath so that water may be retained during the waterlogging experiment.

#### Seed Viability Test, Raising Plants in Nursery and Transplanting

Test for seed viability was carried out following the floatation method. A large number of seeds were put in a bowl of water and allowed to stand for 10 min. Only seeds that sank down were taken as viable. The viable seeds were sown in a nursery to raise 2 week old plants that were later transplanted into prepared experimental pots for the waterlogged experiment. The transplanted crops were allowed to acclimatize to the pot environment for 1 week.

# Exposure of Plants to Ethylene Gas

Based on preliminary experiments, ethylene priming of 350–400 ppm showed promising results. Okra and maize plants were, therefore, exposed to ethylene gas (375 ppm) for 1 h in the chamber. This was taken as ethylene priming of the young plants. For the purpose of ethylene priming, plants were divided into two groups; ethylene priming before waterlogging (EBW) or after waterlogging treatments, respectively.

### Experimental Design and Waterlogging Conditions

The experimental plants were categorized into four groups; control (C), waterlogged (W), ethylene primed before waterlogging (EBW) and ethylene primed after waterlogging (EAW) for maize and okra used in this study. The plants grouped as control (C) were neither subjected to waterlogging condition nor ethylene priming. The group of waterlogged condition (W) consists of plants subjected to waterlogged condition without ethylene priming. The group of EBW was made up of plants given ethylene priming before subjecting to waterlogged condition. The group of EAW was made up of plants subjected to waterlogged condition, and after 2 days in this waterlogged condition, ethylene priming was carried out. The experiment pots were arranged as a completely randomized block design with five replicates. For waterlogging condition, each experimental pot was filled with water to cover the soil up to 1–2 cm above the soil surface, and this water level was maintained by adding water when necessary. Waterlogging condition was sustained for 4 weeks. The experimental data captured in the study were recorded in number of weeks counted from the first day plants were subjected to waterlogged condition.

#### Growth Traits Measured

Growth traits measured included plant height, number of surviving leaves per plant, stem girth, number of adventitious roots formed, survival percentage, distance between first formed adventitious root and the soil level, extension of stem-root junction from the soil level, number of flower buds formed, and number of fruits produced.

# Soil Chemical Analysis

Analysis for soil factors, including pH, sulfate, phosphorus, organic carbon, nitrogen, manganese, iron and total soluble phenolics, were carried out following appropriate standard methods. The soil chemical analyses were done for the composite soil sample used for the study and for soil samples after plant growth. pH was determined in a soil-water slurry (ratio 1:3) (Ademoroti, 1996). Sulfate determination was carried out by a modification of the methods of Appiah and Ahenkorah (1989) and Ben Mussa et al. (2009). Phosphorus determination was performed using the method of Bray and Kurtz (1945). Organic carbon was conducted using Walkley-Black chromic acid wet oxidation method (Bremner and Jenkinson, 1960). Total nitrogen was performed using the Kjeldahl method (Bremner, 1960). Iron content was done using hydroxylamine and 1,10- phenanthroline procedure (Islam et al., 2009). Manganese was conducted using permanganate oxidation procedure (Islam et al., 2009). Total soluble phenolics analysis was performed using modification of citrate extraction procedure followed by Folin–Ciocalteau colorimetric method (Blum, 1997).

# Soil Bacterial Analysis

A weight of 10 g soil sample was measured into a beaker and mixed with 90 ml sterilized water. By serial dilution from the stock sample, 10−<sup>4</sup> dilute solution was prepared. The pour plate technique was used for inoculation on a sterilized nutrient agar (NA), impregnated with antifungal agent, for the growth of bacterial isolates only. The plates were incubated at 37◦C for 24–48 h. After incubation, total viable colonies were recorded for respective microbial isolates and expressed colony forming units per gram (cfu/g). The isolation, characterization and identification of bacterial isolates were performed following the reported methods (Buchanan and Gibbons, 1974).

#### Root Anatomy

The roots of harvested plants were cut and washed to prepare microscope slides to observe internal tissues. The root segments were embedded in paraffin wax and allowed to solidify. By clamping in the microtome, sections were cut and dewaxed. Eosin and hematoxylin stains were applied to the sections, respectively, to allow the cytoplasm and nucleus of cells to appear distinct under the microscope. Excess stains were washed off using increasing concentrations of ethanol sequentially at intervals before oven drying. After oven drying, the slides were ready for viewing and photographing.

# Antioxidant Enzymes Assay

Activities of Ascorbate peroxidase (APX) and catalase (CAT) enzymes were estimated in the leaf or root tissues of okra and maize following the protocol of Zhang and Kirkham (1996). Briefly, 0.25 g of leaf or root tissues were homogenized in 3 ml of a mixture (EDTA (0.2 mM), PBS (50 mM) and 1% PVP), and then centrifuged. The supernatant was used for measuring the absorbance at 240 nm (for CAT) or 290 nm (APX). Superoxide dismutase (SOD) activity was also estimated in the leaf or root tissues of okra and maize using the method of Bradford (1976). In brief, leaf or root plant tissues were homogenized with a phosphate buffer (0.2 M) then centrifuged. The supernatant was used for measuring the absorbance at 560 nm.

#### RNA Isolation, cDNA Synthesis, and Quantitative RT-PCR

Quantitative real-time PCR (qRT-PCR) analysis was carried out to assess the expression levels of antioxidant enzymes (APX, CAT, and SOD) in root or leaf tissues collected at several time points (3 days, 1 week, 2 weeks, 4 weeks) from the beginning of treatments (C, W, EBW, EAW). qRT-PCR analysis was also done to evaluate the expression levels of ACS, ACO, and ethylene receptor (ETR2) genes in the tissues of roots, hypocotyls and epicotyls of okra and maize collected at several time points from the beginning of treatments. Total RNA was isolated from these tissues using RNeasy Plant Mini kit (Qiagen), and RNase-Free DNase Set (Qiagen) was utilized to get rid of DNA. cDNA synthesis was performed using Reverse Transcription kit (Qiagen). Quantitative RT-PCR was conducted in triplicates (3 biological replicates and three technical repeats) using QuantiTect SYBR Green PCR kit (Qiagen) following the manufacturer's protocol. PCR amplification was accomplished under the following conditions: 95◦ C for 15 min; 50 cycles of 95 ◦ C for 30 s, 62◦ C for 30 s, 72◦ C for 2 min; and 72◦C for 5 min. The gene specific-primer sequences used for amplification (Geisler-Lee et al., 2010; Hemavathi et al., 2011; Habib et al., 2016; Neta et al., 2016) are shown in Supplementary Table 1. Amplification specificity was then tested using melting curve analysis. UBIQUITIN (UBQ1, Chen et al., 2013) was used as a housekeeping gene, and the genes expression levels were measured using 2−11Ct method.

#### Statistical Analysis

Mean and standard deviation were calculated from the data collected. One-way analysis of variance was carried out. Significance of mean values was done using the Duncan multiple range (DMR) test.

# RESULTS

#### Plant Height

The values obtained for plant height indicated that the plants showed increase extension growth under waterlogged conditions. Okra plants showed high tolerance to waterlogging. Ten weeks under waterlogged conditions, the highest and lowest values for plant height were 32.81 and 29.80 cm for control and EAW treatments, respectively (**Figure 1A**; Supplementary Table 2). By 5 weeks under waterlogged conditions, all maize plants died under waterlogged (W) condition (**Figure 1B**; Supplementary Table 3). By 9 WAF, maize plants were observed only in EBW treatments.

#### Number of Surviving Leaves

In okra plants, the number of leaves began to reduce immediately after 4WAF. Ten weeks under waterlogged condition, okra plants had a maximum of 3–4 healthy looking leaves per surviving plant (**Figure 1C**; Supplementary Table 4). The older leaves were lost first. The reduction of leaves in maize plants began 1 week after waterlogging. By 7WAF, all maize plants in two treatments (W,

EAW) had lost all their leaves, and wilting of plants from the apical portions was apparent. Surviving maize plants in EBW treatments had average of two leaves (**Figure 1D**; Supplementary Table 5).

#### Stem Girth

Stem girth measurements for okra plants showed increase from 2WAF to 10WAF. The highest and lowest average values for stem girth of plants 10 weeks after flooding were 1.85 and 1.55 cm for EBW and EAW treated plants, respectively (**Figure 2A**; Supplementary Table 6). Maize plants (EBW) survived up to 9WAF and had an average value of 1.30 cm for stem girth (**Figure 2B**; Supplementary Table 7).

#### Survival Percentage

As the duration of waterlogged condition progressed, the number of surviving okra and maize plants reduced for the treatments. Ten weeks after flooding, the highest and lowest survival rates of okra plants were 43 and 20% for control and EAW, respectively (**Figure 2C**). Seven weeks after flooding, the average survival of maize plants was 20%. Only waterlogged maize plants survived under EBW treatment by 9WAF (**Figure 2D**).

#### Number of Adventitious Roots

Okra plants grown under EBW and EAW treatments exhibited the highest average number of adventitious root formed per plant (5 per plant). The lowest number, three per plant, was observed in the control plants. Adventitious roots were not recorded for control plants until 7WAF (**Figure 3A**; Supplementary Table 8). Maize plants under EBW and EAW treatments initiated adventitious roots earlier than others by 3WAF (**Figure 3B**; Supplementary Table 9). The average number of adventitious roots formed per maize plant grown under EBW and EAW treatments was higher than that of control plants until 5WAF.

### Extension of Stem-Root Junction from the Soil Level

The extension of root-stem junction provides information on how the roots can extend above soil level to expose some portions of root to oxygen in order to ameliorate the anaerobic condition. Seven weeks after flooding of okra plants, the highest and lowest values for distance of root-stem junction above the soil level were recorded in EBW and control conditions, respectively (**Figure 3C**; Supplementary Table 10). Five weeks after flooding of maize plants, the lowest and highest values of root-stem junction above soil surface were 1.50 and 3.40 cm for EAW and EBW, respectively (**Figure 3D**; Supplementary Table 11). The maize plants under waterlogged (W) treatment died.

#### Number of Flower Buds and Fruits

All maize plants in this study failed to exhibit reproductive capacity or flower formation under the different treatments. Records of flower buds formed (**Figure 3E**; Supplementary Table 12) and fruits (**Figure 3F**; Supplementary Table 13) were taken for okra plants that tolerated waterlogged conditions. Okra plants under control and EBW treatments formed an average of four flower buds per plant by 14WAF. The average number of fruits per plant was 3 or 4 by 15WAF.

maize. Within the same WAF time point, values with similar alphabet do not differ significantly. Error bars represent SD, *n* = 5.

FIGURE 3 | Number of adventitious roots, distance (cm) of root-stem junction and number of buds and fruits of *Abelmoschus esculentus* and *Zea mays* subjected to waterlogged condition. (A,C,E,F) for Okra, and (B,D) for maize. Within the same WAF time point, values with similar alphabet do not differ significantly. Error bars represent SD, *n* = 5.

#### Soil Chemical and Bacterial Analyses

The changes in values of some selected soil parameters investigated were minimal (**Table 1**). The average phosphorus content of okra and maize plant grown under waterlogged conditions showed a reduction, when compared with the composite sample. Total soluble phenolics generally increased after plant growth. Values recorded for soil samples of C (maize), W (maize), EBW (maize), and C (okra) were >100 mg/Kg. Other soil samples gave lower values. Soil pH values generally ranged between 6 and 7. The bacterial isolates recorded for soil samples following maize and okra growth under waterlogged conditions, are also shown in Supplementary Table 14. The TABLE 1 | Values obtained for soil factors in composite soil sample before study and in the different experimental soil samples after plant growth under waterlogged condition.


*Values* = *mean (SD); C* = *control; W* = *under waterlogged; EBW* = *ethylene priming before waterlogged; EAW* = *ethylene priming after waterlogged.*

lowest and highest values for average bacterial counts were 6.4 × 10<sup>4</sup> and 9.7 × 10<sup>4</sup> . Micrococcus leutus and Micrococcus varians were detected in 7 out of 8 soil samples. Other bacterial species include Arthrobacter sp, Serratia sp, Pseudomonas sp, and Bacillus spp.

#### Anatomy of Okra Roots

The cross section of okra roots showed a dark colored peripheral layer on the roots of waterlogged plants. This dark layer incompletely surrounded control root. All root sections showed the presence of air channels (lacunae) in the cortex. In **Figure 4A** (control), the cells around the air channels appear to possess thick walls. This control root was not submerged in water and as such the hypoxic condition would be milder than the others under water. Therefore, less change was exerted on the cellulose cell wall. **Figure 4B** (EBW) shows the air channels surrounded by cells whose walls indicate changes. The walls are thin but wall strengthening materials appear to be intact. More aerenchyma cells may be observe. **Figure 4C** (EAW) shows air channels surrounded by cells with wavy outlines of thin walls. This suggests an indication of collapsed wall strengthening materials. **Figure 4D** (W) shows air channels surrounded by cells with thin walls. Fewer aerenchyma cells may be seen.

### Activity and Expression Analyses of Antioxidant Enzymes

The effect of waterlogged conditions and exogenous ethylene priming on the activity and expression level of antioxidant enzymes (APX, CAT, and SOD) was studied in okra and maize root and leaf tissues collected at several time points (3 days, 1 week, 2 weeks, 4 weeks) from the onset of flooding. The activity of APX enzyme greatly increased in the root and leaf tissues of okra and maize subjected to waterlogging and ethylene priming (EBW, EAW), as compared to control (C) or waterlogged condition (W) (**Figure 5**). Additionally, the APX activity of okra exhibited a continuous increase under waterlogged condition (W) at all-time points, as compared to control (C).

Similarly, the activity of CAT enzyme greatly increased in the root and leaf tissues of okra and maize subjected to waterlogging and ethylene priming, as compared to control or waterlogged conditions (**Figure 5**). However, its activity in okra leaf was much higher than that in the root. On the other hand, the activity of SOD enzyme slightly increased in maize root and leaf tissues subjected to waterlogging and ethylene priming, whereas it highly increased in okra root and leaf under these conditions (**Figure 5**).

Quantitative real-time PCR analysis was also conducted to evaluate the expression levels of these antioxidant enzymes in the root and leaf tissues of okra and maize subjected to waterlogging and ethylene priming, as well as to determine the correlation with the enzyme proteins. The expression analysis revealed that APX, CAT, SOD genes showed higher expression levels in the root and leaf tissues of okra and maize subjected to waterlogging and ethylene priming (EBW, EAW), as compared to control (C) or waterlogged condition (W) (**Figure 6**). mRNA expression results of these enzymes were in a complete concordance with their specific activities.

### Expression Analysis of Ethylene ACS and ACO Genes

Quantitative real-time PCR analysis was carried out to study the effect of waterlogging conditions and exogenous ethylene treatment on the regulation of ethylene biosynthetic pathway genes in the roots, hypocotyls and epicotyls of okra and maize collected at several time points from the onset of flooding.

The expression levels of ACS1, ACS4, ACS6, ACO1, ACO3, and ETR2 genes were assessed in okra roots, hypocotyls and epicotyls collected at several time points (12 h, 24 h, 2 days, 4 days, 1 week, 3 weeks, 6 weeks, 9 weeks) from the onset of flooding (**Figures 7**–**9**). Interestingly, waterlogging and exogenous ethylene priming (EBW, EAW) significantly induced the expression levels of all ACS, ACO, and ETR2 genes in all okra tissues until 9 weeks of waterlogging, as compared to the control (C) or waterlogging (W) conditions (**Figures 7**–**9**).

The expression levels of ZmACS2, ZmACS6, ZmACS7, ZmACO20, ZmACO31, and ZmETR2 genes were also studied in maize roots, hypocotyls and epicotyls collected at several time points (12 h, 24 h, 2 days, 3 days, 5 days, 1 week, 3 weeks, 5 weeks) from the onset of flooding (**Figures 10**–**12**). Interestingly, waterlogging and exogenous ethylene priming (EBW, EAW)

FIGURE 4 | Cross sections of roots of *Abelmoschus esculentus* plants grown under waterlogged condition showing presence of aerenchyma cells and air cavities. (A) okra plant (control) without waterlogging, (B) okra plant given ethylene priming before waterlogging, (C) okra plant given ethylene priming after waterlogging, (D) okra plant under waterlogging only.

induced the expression levels of ACS, ACO, and ETR2 genes in all maize tissues until 3 or 5 weeks of waterlogging, as compared to the control (C) or waterlogged conditions (W) (**Figures 10**–**12**). The results indicate that the high expression levels of ACS, ACO, and ETR2 genes lasted for longer periods after waterlogging and ethylene priming in okra, as compared to maize.

# DISCUSSION

Water is indispensable to the survival of plants as it supports plant growth and functions. However, flooding or waterlogging threatens plants. The negative impact of flooding leaves the society and environment devastated. Plant biodiversity, distribution of natural species and worldwide food production decline because terrestrial plant species, including cultivated crops, are sensitive to flood conditions (Normile, 2008). The harmful effect of flooding is attributed to the fact that gas diffusion in water, as a medium of transport within living systems, is too low to allow terrestrial plants survival for a long time. Two vital plant processes, respiration and photosynthesis, are affected (Sasidharan and Voesenek, 2015). Sasidharan and Voesenek (2015) reported that the energy crisis resulting from hampered aerobic metabolism leads to an imbalance between consumption and production, resulting in plant mortality. Waterlogging describes only the situation where the soil and plant roots are covered by water. In this study, the values for plant height as growth parameter evaluated were higher in control plants than in plants subjected to waterlogged condition. The waterlogged conditions stimulated early formation of adventitious roots of okra plants in this study. Generally, the plants suffered loss of leaves in the process of surviving the condition.

Waterlogged conditions reduce available oxygen to plant root cells simply by displacement of air pockets in the soil. Therefore, there is a steep oxygen gradient between root and shoot portions. Roots in waterlogged soils suffer rapid oxygen depletion due to respiration both of the roots and the rootassociated microbiome (Vashisht et al., 2011). These waterlogged roots switch to the inefficient anaerobic fermentation, consuming available carbohydrate reserves for the generation of needed ATP to remain alive and functioning. As the hypoxic or anoxic

situation continues, impaired membrane integrity, starvation and diffusion of phytotoxic compounds into the root cells combine to hinder root growth and function (Sauter, 2013). Since the roots cannot transport water and nutrients efficiently under hypoxic or anoxic conditions, shoot functions are affected and visible symptoms such as wilting, senescence, and death may be observed (Sasidharan and Voesenek, 2015).

In sensitive genotypes, waterlogged conditions may cause induction and initiation of crop tolerance features or adaptive traits that may improve aeration and ameliorate root

week after flooding). (A,C,E) for Okra, and (B,D,F) for maize. Error bars represent SD, *n* = 6.

hypoxia or anoxia in order to preserve root function and plant survival. Those adaptive features observed included development of a lignin/suberin barrier in the root to reduce oxygen loss and commit its transport to the tip of the root (Shiono et al., 2011), enhanced development of aerenchyma cells to promote root aeration (Takahashi et al., 2014) and development of adventitious roots rich in aerenchyma (Sauter, 2013). Maize and okra plants are sensitive to waterlogged conditions. Both species developed adventitious roots. The root extension growth also enabled the root-stem junction to rise above the soil surface, exposing portions of the root to molecular oxygen aerially. This should be considered as another feature to survive the waterlogged conditions.

Phytohormones are important to plants in signals and responses to stress. Many plant responses are based on the sensitivity of the tissues involved. Many cells of higher plants can carry out synthesis of ethylene in low amount (Abeles, 1992). Root hypoxia (low oxygen) or anoxia (oxygen deficient) induced by waterlogging limits the conversion of ACC to ethylene because the ACC oxidase needs molecular oxygen to complete the synthesis of ethylene (Jackson, 2008). Hypoxia or anoxia favors the accumulation of ACC in flooded roots, which is transported to the shoot, where it is usually converted to ethylene. Bradford and Yang (1980) classified ACC as the primary signal transferred from roots to shoots during the early period of hypoxia. The conversion of ACC to ethylene in the presence of molecular oxygen triggers the development of adaptive systemic responses by shoots of waterlogged plants and these include aerenchyma cells and adventitious root formation (Fukao and Bailey-Serres, 2008; Rajhi et al., 2011; Yamauchi et al., 2014). Molecular oxygen, generated from photosynthetic activities of

the aerial shoot, diffuses downward to the aerenchyma cells and aerenchyma-rich adventitious roots under hypoxic or anoxic condition, as an amelioration mechanism in mesophytes (Irfan et al., 2010). Rajhi et al. (2011) stated that application of ethylene induced the development of aerenchyma in maize plants. Under hypoxia, Takahashi et al. (2014) observed further augmentation of aerenchyma development in rice, which is dependent on treatment with ethylene or ACC. Degree of aerenchyma induction and formation by endogenous ethylene treatment varies between species. Aerenchyma formation is completed as a genetically programmed death process in the cortex of roots, which high levels of reactive oxygen species (ROS) are capable of triggering (Steffens et al., 2011; Yamauchi et al., 2014). ROS is one component of the ethylene-mediated signaling

network (Sasidharan and Voesenek, 2015). Separately, ROS or ethylene induces ectopic cell death only, whereas ethylenemediated ROS is important in the formation of aerenchyma. Aerenchyma, formed in the root cortex, provides aeration, and enhances survival. The plants under waterlogged conditions developed more adventitious roots than the control. Also, the plants given ethylene priming produced higher number of adventitious roots per plant. This supports the hypothesis that plants develop morphological adaptative features under waterlogged conditions.

The waterlogging-induced production of crop adventitious roots, which functionally replace the damaged soil-borne roots, improves the shoot-root diffusion of gases (Sasidharan and Voesenek, 2015). Ethylene has been reported to be important in adventitious root formation. McNamara and Mitchell (1989) concluded that auxin interacts with ethylene to induce the

development of adventitious roots, but the role of ethylene may differ relying on the plant species (Vidoz et al., 2010). Ethylene induces an increase in auxin-sensitivity of root forming tissue during adventitious root formation in waterlogged Rumex palustris (Visser et al., 1996), whereas the induction of adventitious root production during waterlogging requires ethylene perception by the Never Ripe receptor in tomato (Vidoz et al., 2010). The transport of ACC from hypoxia or anoxia roots to the shoot stimulates ACC synthase genes to make ACC synthase enzyme available to complete the biosynthesis of ethylene in the molecular oxygen-mediated process. The elevated level of ethylene in the stem reprogrammes auxin transport in the shoot, directing the flow of auxin toward the submerged stem to initiate the growth of adventitious roots. Adventitious

root production is hindered if auxin transport is inhibited (Vidoz et al., 2010).

The present study has shown that okra survived better than maize under waterlogged condition. The okra plants were able to flower and fruit, whereas maize plants did not survive beyond 9WAF and were unable to flower and form fruits during this period. Some maize plants given EBW were the ones that survived up to 9WAF. The highest population of soil bacterial flora in a soil sample after plant growth was detected in EBW (maize), along with an absence of Micrococcus spp. The survival of maize plants under EBW treatment indicates that EBW was beneficial to plants. The changes in soil factors following plant growth of maize and okra did not reflect a major difference between the two crops. The results of cross-sections of okra

roots showed presence of aerenchyma cells. The root tissues of okra plants possess large air channels which may have assisted in trapping oxygen useful for survival. Formation of aerenchyma improves the porosity of roots (Videmsek et al., 2006). Differences in porosity of roots of a genotype have been reported (Thomson et al., 1990) and this can be linked to the size and number of aerenchyma present. Porosity allows more gas to be present within the internal tissues of root. Justin and Armstrong (1987) reported that since porosity promotes internal movement of gases, plant roots adapted to anaerobic conditions should possess higher porosity as a characteristic. The formation of the dark peripheral layer (suspected to be impervious) and numerous large aerenchyma by waterlogged roots of okra suggest promising support for tolerance of anaerobic conditions. In addition, exogenous ethylene priming provides a beneficial influence on growth of okra under waterlogged conditions.

In the current study, the specific activities and transcript levels of three antioxidant enzymes were also studied in okra and maize root and leaf tissues subjected to waterlogging and ethylene priming. APX, CAT, and SOD enzyme activities increased in the root and leaf tissues of okra and maize subjected to both waterlogging and ethylene priming, as compared to control or waterlogged condition, indicating the efficient role of antioxidant enzymes in tolerating waterlogging stress. mRNA expression levels of these enzymes revealed a positive correlation with their specific activities.

In the present study, qRT-PCR analysis also showed that the expression levels of ACS, ACO, and ETR2 genes in all okra and maize tissues were up-regulated and showed much higher levels under EBW and EAW treatments than those expressed under control or waterlogged conditions at all-time points. This indicates that okra and maize tissues respond to conditions of waterlogging and exogenous ethylene priming by inducing their ethylene biosynthetic genes expression in order to enhance ethylene production and tolerate the prolonged waterlogging stress. In conclusion, this study revealed that exogenously

#### REFERENCES


generated ethylene gas as a priming treatment before or after waterlogging could enhance waterlogging tolerance in maize and okra crops.

#### AUTHOR CONTRIBUTIONS

ME and EV designed the study. ME, EV, and OA carried out the experiments, analyzed the data, and wrote the manuscript. All authors revised and approved the final manuscript.

#### ACKNOWLEDGMENTS

This research was supported by University of Benin in Nigeria and Tanta University in Egypt. We would also like to thank Dr. Thomas Jonesman (University of Cambridge, UK) for his appreciated help.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fphys. 2017.00632/full#supplementary-material

and Crop Improvement, eds M. Anis and N. Ahmad (Singapore: Springer), 523–545.


**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 Vwioko, Adinkwu and El-Esawi. 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.

# Endogenous Ethylene Concentration Is Not a Major Determinant of Fruit Abscission in Heat-Stressed Cotton (Gossypium hirsutum L.)

Ullah Najeeb<sup>1</sup> \*, Muhammad Sarwar <sup>2</sup> , Brian J. Atwell 1, 3, Michael P. Bange1, 4 and Daniel K. Y. Tan<sup>1</sup>

*<sup>1</sup> Faculty of Science, Plant Breeding Institute, Sydney Institute of Agriculture, School of Life and Environmental Sciences, University of Sydney, Sydney, NSW, Australia, <sup>2</sup> Agronomic Research Institute, Ayub Agricultural Research Institute, Faisalabad, Pakistan, <sup>3</sup> Department of Biological Sciences, Faculty of Science and Engineering, Macquarie University, Sydney NSW, Australia, <sup>4</sup> Commonwealth Scientific and Industrial Research Organisation Agriculture and Food, Australian Cotton Research Institute, Narrabri, NSW, Australia*

#### Edited by:

*Chi-Kuang Wen, Shanghai Institutes for Biological Sciences (CAS), China*

#### Reviewed by:

*Basharat Ali, University of Bonn, Germany Roberta Paradiso, University of Naples Federico II, Italy*

> \*Correspondence: *Ullah Najeeb najeb\_ullah@yahoo.com*

#### Specialty section:

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

Received: *11 June 2017* Accepted: *04 September 2017* Published: *21 September 2017*

#### Citation:

*Najeeb U, Sarwar M, Atwell BJ, Bange MP and Tan DKY (2017) Endogenous Ethylene Concentration Is Not a Major Determinant of Fruit Abscission in Heat-Stressed Cotton (Gossypium hirsutum L.). Front. Plant Sci. 8:1615. doi: 10.3389/fpls.2017.01615* We investigated the role of ethylene in the response of cotton to high temperature using cotton genotypes with genetically interrupted ethylene metabolism. In the first experiment, Sicot 71BRF and 5B (a lintless variant with compromised ethylene metabolism) were exposed to 45◦C, either by instantaneous heat shock or by ramping temperatures by 3◦C daily for 1 week. One day prior to the start of heat treatment, half the plants were sprayed with 0.8 mM of the ethylene synthesis inhibitor, aminoethoxyvinylglycine (AVG). In a subsequent experiment, Sicot 71BRF and a putatively heat-tolerant line, CIM 448, were exposed to 36 or 45◦C for 1 week, and half the plants were sprayed with 20 µM of the ethylene precursor, 1-aminocyclopropane-1-carboxylic acid, (ACC). High temperature exposure of plants in both experiments was performed at the peak reproductive phase (65–68 days after sowing). Elevated temperature (heat shock or ramping to 45◦C) significantly reduced production and retention of fruits in all cotton lines used in this study. At the termination of heat treatment, cotton plants exposed to 45◦C had at least 50% fewer fruits than plants under optimum temperature in all three genotypes, while plants at 36◦C remained unaffected. Heat-stressed plants continued producing new squares (fruiting buds) after termination of heat stress but these squares did not turn into cotton bolls due to pollen infertility. *In vitro* inhibition of pollen germination by high temperatures supported this observation. Leaf photosynthesis (*P*n) of heat-stressed plants (45◦C) measured at the end of heat treatments remained significantly inhibited, despite an increased leaf stomatal conductance (*g*s), suggesting that high temperature impairs *P*<sup>n</sup> independently of stomatal behavior. Metabolic injury was supported by high relative cellular injury and low photosystem II yield of the heat-stressed plants, indicating that high temperature impaired photosynthetic electron transport. Both heat shock and ramping of heat significantly reduced ethylene release from cotton leaf tissues measured at the end of heat treatment but modulating ethylene production via AVG or ACC application had no significant effect on fruit production or retention in heat-stressed cotton plants. Instead, high temperature accelerated fruit abortion by impairing pollen development and/or restricting leaf photosynthesis.

Keywords: elevated temperature, ethylene manipulation, heat shock, photosynthesis, pollen germination

#### INTRODUCTION

With rising atmospheric temperatures, there has been an increased concern about protecting crops from extreme weather events such as, heat shock (short bursts of very high temperatures), which are expected to become more common in near future (Zheng et al., 2012). For example, Australia has been experiencing an increased frequency of very hot (>40◦C) daytime temperatures since the 1990s (CSIRO and Bureau of Meteorology, 2012). Cotton is widely cultivated under hot, semiarid climatic conditions hence crops often experience yield losses due to episodes of high air temperatures during reproductive phase. Optimum temperature for cotton growth has been determined as 20–30◦C (Reddy et al., 1991a), with temperatures above this optimum negatively influencing productivity. For example, a lint yield reduction of 110 kg ha−<sup>1</sup> is expected for each 1 ◦C rise in maximum day temperature (Singh et al., 2007) due to accelerated fruit shedding during the post-stress period (Ehlig and Lemert, 1973).

Elevated temperature influences various processes during reproductive growth such as, pollen formation, pollen germination, and fertilization, with failure leading to abscission of reproductive structures (Kakani et al., 2005). For example, day temperatures above 40◦C can induce abscission of cotton flowers and young squares developed during heat stress (Reddy et al., 1991a). In addition to pollen abortion, high temperature induces lint yield losses by influencing key physiological processes such as, photosynthesis, stomatal behavior and cell membrane thermo-stability (Bibi et al., 2008; Cottee et al., 2010). In brief, yield losses in heat-stressed cotton crops could be result of impaired development of reproductive parts (pollen or pistil) and/or competition for photo-assimilates (Zhao et al., 2005).

Fruit abscission is mediated by enzymes that hydrolyse pectinrich middle lamella in the abscission zone (Sawicki et al., 2015); ethylene is invoked as a key factor that triggers the expression of these hydrolytic genes (Zhu et al., 2011). Furthermore, ethylene release is accelerated by stress factors such as, temperature, excessive light, and waterlogging (Jackson, 1985; Hyodo, 1991). Thus, we hypothesized that ethylene plays a role in fruit abscission in cotton stressed at high temperatures. Previously, we observed a strong positive correlation between waterlogginginduced fruit abscission and ethylene concentrations in cotton tissues (Najeeb et al., 2015). Such stress-induced damage can be modified by regulation of the synthesis (Grichko and Glick, 2001) or perception (Wilkinson et al., 1997) of ethylene through genetic manipulation. Furthermore, ethylene metabolism can be suppressed by blockers of ethylene synthesis or perception such as, aminoethoxyvinylglycine (AVG) and 1-methylcyclopropene (1-MCP), effectively increasing tolerance to abiotic stresses in cotton (Bange et al., 2010; Kawakami et al., 2013). Earlier studies suggested positive effects of modulating ethylene production through 1-MCP on different crops including soybean and wheat (Hays et al., 2007; Djanaguiraman et al., 2011). Variable effects 1-MCP application have been observed on cotton crops e.g., Kawakami et al. (2013) observed increased lint yield in cotton in response to 1-MCP application but Scheiner et al. (2007) suggested no effect of 1-MCP on cotton lint yield.

The present study explores the relationship between ethylene metabolism and fruit abscission in heat-stressed cotton, testing the hypothesis that elevated ethylene levels are involved in heat-induced fruit loss. The first experiment attempted to minimize heat-induced fruit abscission by manipulating ethylene production using: (1) AVG to block ethylene synthesis and (2) the lintless cotton mutant (5B), which produces very few fibers in the boll. Since fiber development and elongation in cotton is regulated by ethylene (Shi et al., 2006), it is proposed that ethylene metabolism is impaired in this cotton mutant. Earlier studies suggested a lesion in the pathway of ethylene synthesis in developing ovules of lintless mutants (Wang et al., 2011; Gilbert et al., 2013). While we cannot confirm the location of the genetic block in mutant 5B, it has a definitive phenotype and thus constituted an ideal genetic tool to test the role of ethylene under heat. In a separate glasshouse experiment, the effects of an ethylene precursor 1-aminocyclopropane-1-carboxylic acid (ACC) were investigated in two cotton cultivars with naturally contrasting heat tolerance to see whether induced ethylene release exacerbated sensitivity to heat. The main objectives of these experiments were (1) to study how cotton plants respond to elevated temperatures (heat shock and ramping heat) and (2) to explore the impact of ethylene on heat tolerance in cotton.

#### MATERIALS AND METHODS

#### Plant Material and Growth Conditions Experiment 1

Two cotton genotypes, Sicot 71 BRF and 5B line, were used in this glasshouse experiment. Sicot 71BRF is a commercial cotton cultivar [Gossypium hirsutum L. (Bollgard II <sup>R</sup> Roundup Ready Flex <sup>R</sup> ), CSIRO Australia; (Stiller, 2008)] that is widely cultivated in Australian cotton growing regions. The 5B line was originally separated from a fully linted cotton cultivar B1278 as a spontaneous mutant, having little or no lint on the seeds (Dr. Alistair Low unpublished, CSIRO Irrigation Research, Griffith, NSW). Consistent with a disruption to ethylene metabolism, the 5B mutant line exhibited a tendency to retain fruits in response to soil waterlogging (Najeeb et al., unpublished data), while fruit abscission is normally strongly associated with increased ethylene production from vegetative tissues of waterlogged cotton plants (Najeeb et al., 2015).

Seeds were sown in plastic pots (25 × 20 cm; height × diameter) containing finely mixed red silt loam Ferrosol soil from Robertson, NSW, Australia. Plants were grown under similar climatic and cultural conditions as reported in an earlier study (Najeeb et al., 2015). In brief, plants were grown at 28/20◦C day/night temperature, 50–70% relative humidity, 14/10 light/dark photoperiod under natural light. The light intensity during the day cycle was maintained to a minimum of 400 µmol m−<sup>2</sup> s <sup>−</sup><sup>1</sup> using supplemented light (Philips Contempa High Pressure Sodium lamps). At peak reproductive growth phase (65 days after sowing, DAS), the plants were either transferred to growth cabinets at 45/30◦C, day/night temperature directly (heat shock) or to growth cabinets where day temperature was increased to 45◦C by an incremental rise of 3◦C per day (e.g., 33, 36, 39, 42, and 45◦C-ramping heat). Plants were transferred to the high temperature growth cabinets at 9:00–9:30 a.m., and were exposed to 45◦C day temperature for 1 week. Notably, plants under ramping heat experienced a cumulative 11 days of temperatures higher than optimum (cf. 7 days for the heat-shock treatment). Control plants were grown under optimal temperature (28/20◦C, day/night). Soil was maintained at field capacity while humidity levels in growth chambers were kept at >50% by placing water-filled plastic trays on the cabinet floor.

One day prior to the start of heat treatments, the plants in each set were sprayed with either 0.8 or 0 mM (water) AVG formulated as ReTain <sup>R</sup> (Sumitomo Chemicals Australia) + one drop of detergent (Supplementary Table 1). AVG application rates have been optimized in earlier glasshouse and field experiments (Bange et al., 2010; Najeeb et al., 2016) and the rate applied here was effective in suppressing ethylene production and fruit loss in waterlogged cotton crops (Najeeb et al., 2015).

#### Experiment 2

In this experiment, a relatively heat-tolerant cotton cultivar CIM 448 (Rahman, 2005) was used along with Sicot 71BRF. The plants were exposed to either 45/30◦C or 36/30◦C day/night temperature (each applied as a heat shock without ramping) for 7 days at the peak reproductive phase (68 DAS). Control plants were grown under 28/20◦C day/night temperature. As the plants were exposed to heat shock, an immediate effect of heat on ethylene synthesis was expected. To modulate ethylene synthesis, 20 µM ACC was sprayed on cotton plants (Supplementary Table 1). Plants readily convert ACC into ethylene so ACC was applied on 2 consecutive days, namely 10 µM ACC 1 day prior to heat treatment and 10 µM at the start of heat treatment. After heat-treatment, all plants were returned to optimum conditions (28/20◦C day/night temperature) to study the recovery in reproductive physiology.

#### Fruit Production and Retention

Data on fruit production and retention were collected the day before (−1 day) then 8 and 15 days after heat treatments (DHT). With no significant variation in the number fruits at −1 and 8 DHT, data are presented only for −1 and 15 DHT. All plants were mapped for fruit number and type (squares and green bolls) on individual fruiting position and branches. Pollinated flowers and bolls were assigned as green bolls and young non-pollinated flowers as squares. Fruit retention (FR) was calculated from the ratio of retained fruits to total fruiting sites on each plant.

# Leaf Gas Exchange and Chlorophyll Fluorescence

Leaf CO<sup>2</sup> exchange parameters such as, photosynthesis (Pn) and stomatal conductance (gs) were measured at the termination of heat treatments. Data were collected from the youngest fully expanded leaves between 9:00 and 13:00 h using a Li-6400 portable photosynthesis system (Li-Cor Ltd, Lincoln, NE, USA) with a pulse-amplitude modulated (PAM) leaf chamber head. The gas exchange measurements were recorded at 1,800 µmol photon m−<sup>2</sup> s <sup>−</sup><sup>1</sup> photosynthetically active radiation (PAR) and a CO<sup>2</sup> concentration of 400 µmol mol−<sup>1</sup> . Temperature inside the leaf chamber was adjusted according to the corresponding treatment temperature e.g., 28, 36, or 45◦C.

Chlorophyll fluorescence was measured using a photosynthesis yield analyser (MINI- PAM, Walz, Effeltrich, Germany).

### Relative Cell Membrane Injury

Leaf discs (14-mm diameter) were excised from the interveinal section of fully expanded leaves between 13:00 and 14:30 h at the end of heat treatments. The discs were triple rinsed with distilled water to remove any exogenous electrolytes, and then placed into glass vials containing 2 mL of MilliQ water. Vials were incubated in water bath at 25 and 55◦C. Electrical conductivity (EC) was measured using a waterproof ECT calibrated conductivity meter before incubation (IEC) and after incubation (FEC).

Relative cellular injury (RCI%) was calculated according to Cottee et al. (2007) using the following equation:

$$RCI = \text{ 1} - \text{ [1} - \text{ (IECt/FECt)}/1 - \text{ (IECc/FECc)} \times 100$$

Where IECt and FECt are the initial and final EC values of heattreated vials, respectively, while IECc and FECc are the initial and final EC values of control vials, respectively.

# In Vitro Pollen Germination

Flowers used for pollen viability assay were collected from the first fruiting position during peak reproductive growth phases (65–75 DAS) on a daily basis during both experiments. Pollen was collected between 8:00 and 9:00 a.m. and directly sprinkled on solidified nutrient media (Taylor, 1972), consisting of 1.5 g agar, 30 g sucrose, and (mg) KNO<sup>3</sup> (5.3), MnSO<sup>4</sup> (51.7), H3BO<sup>3</sup> (10.3), MgSO4·7H2O (10.3) in 100 mL of deionised water (Kakani et al., 2005). Each replicate comprised three flowers from the same plant, with four plants per treatment for each genotype. Pollen germination percentage was calculated by counting the proportion of pollen grains and pollen tube length was measured using ImageJ software.

#### Ethylene Measurements

Youngest fully expanded leaves were collected from heated and control plants (three leaves per plant) 24, 48 h, and 7 d after heat treatment. In addition, ethylene was measured from leaf tissues 12 h after spraying ACC. Significantly higher ethylene concentrations were recorded from ACC-treated plants after 12 and 24 h of treatment under optimum temperatures but no change in ethylene concentration was observed in the later measurements. With no significant variation in the ethylene levels at different measurement times (except ACC-treated plants at 12 and 24 h under optimum temperatures), data were only presented for plants 7 d after heat treatment. This ethylene measurement time was selected from our earlier observations, where high temperature caused maximum damage to cotton plants. Ethylene concentrations from leaf tissues were measured using a protocol adopted by Najeeb et al. (2015). Leaf tissues were placed in 50-mL glass vials containing rubber septa. One milliliter gas samples were withdrawn using gas-tight syringes after 20–30 min (Jackson and Campbell, 1976) and injected into PYE series 104 gas chromatograph fitted with a flame ionization detector (FID) and equipped with activated aluminum coated glass column. Fresh biomass of the leaf tissues was determined after ethylene detection, and ethylene synthesis rates were calculated as nmol g−<sup>1</sup> FW h−<sup>1</sup> .

#### Data Analysis

Data for different growth parameters was statistically analyzed by JMP v. 9 (SAS Institute, Cary, NC, USA) statistical program. A linear mixed model REML (Residual Maximum Likelihood) was applied to assess the individual and interactive effects of temperature, genotype, AVG, and ACC, while the respective means were compared using Tukey's HSD (honestly significant difference) test. Data for each experiment and measurement time were separately analyzed.

To identify the parameters that best describe high temperature-induced fruit loss in cotton, a principal component analysis (PCA) was performed. Values of Pn, leaf ethylene concentrations, RCI, and number of green bolls of the three cotton cultivars under various treatment conditions were included in the PCA. This analysis estimates and then ranks principal components (PC) for contribution to the variation in data by consolidating the relationships among measured physiological variables.

#### RESULTS

#### Fruit Production

The numbers of green bolls (pollinated flowers + bolls) were significantly reduced with exposure to heat shock (45◦C) or ramping high temperature (Supplementary Table 2). At the end of heat treatment, Sicot 71BRF plants exposed to heat (shock or ramping heat) had 80% fewer green bolls (averaged across AVG and H2O treatments) compared with plants under optimum temperature (**Figure 1A**). Similarly, 5B plants showed a 52– 66% reduction in green bolls in response to heat shock and ramping heat, respectively, compared with the plants under control temperature (**Figure 1A**). In the 15 d after heat-shock, plants produced new green bolls, albeit fewer than the controls. On the other hand, after a period of ramping heat up to 45◦C, plants did not produce any additional green bolls over the subsequent 15 d (**Figure 1B**). AVG had no significant effect on numbers of green bolls in 5B or Sicot 71BRF under any treatment condition (**Figure 1A**).

In the 2nd experiment, Sicot 71BRF displayed a similar response to 45◦C heat shock as observed in Experiment 1 i.e., stressed plants experienced 75 and 69% reduction (averaged across ACC and H2O treatments) in green bolls at −1 and 15 DHT, respectively (**Figures 1C,D**). In addition, plants exposed to 36◦C produced significantly fewer green bolls than plants at optimal temperature at 15 DHT, although the reduction was substantially less than that was seen at 45◦C (**Figures 1C,D**). In CIM 448, a significant reduction in green bolls was also recorded in response to 45◦C both at 1 and 15 DHT. In contrast, CIM 448 plants exposed to 36◦C showed no significant reduction in green bolls at −1 or 15 DHT (**Figures 1C,D**). ACC significantly reduced green bolls of Sicot 71 BRF at −1 DHT under all temperatures but it induced more green bolls in both cultivars at 15 DHT under optimum temperature (**Figures 1C,D**). This increased green boll production was observed mainly at the top of the plant canopy.

The number of squares in both genotypes (Sicot 71BRF and 5B) was significantly reduced by heat shock or ramping heat at −1 DHT. Significantly greater loss of squares resulted from ramping heat (62% loss, averaged across both genotypes) than heat shock (15% loss, averaged across both genotypes; **Figure 2A**). In Sicot 71BRF, the number of squares in control and ramping-heat plants were significantly reduced at 15 DHT, although, heat-shock plants produced significantly more squares than both control and ramping heat plants over this period. In contrast, all heat-stressed 5B plants (heat shock and ramping heat without AVG spraying) produced significantly more squares than the control at 15 DHT (**Figure 2B**). AVG-treated heatstressed (heat shock or ramping heat) Sicot 71BRF plants bore significantly more squares than their respective non-AVG treated plants at −1 DHT but the effect at 15 DHT was significant only on heat-shock plants. AVG had no significant effect on number of squares in 5B plants.

By the end of heat treatment (−1 DHT) in the 2nd experiment, heat shock at 45◦C reduced number of squares in both genotypes (Sicot 71BRF and CIM 448) while plants exposed to 36◦C produced marginally more squares than control plants (**Figure 2C**). In stark contrast, 45◦C-treated plants produced significantly more squares than plants at 36 and 28◦C 15 DHT (**Figure 2D**). ACC-treated plants of both cotton genotypes exposed to 28 and 36◦C treatments produced slightly more squares at −1 DHT than non-ACC treated plants (P = 0.047) but the effect was non-significant at 15 DHT, as the plants supported only a few squares at this stage (**Figures 2C,D**).

#### Fruit Retention

High temperature (heat shock or ramping) significantly reduced fruit retention (FR) in all the studied genotypes. The greatest reduction in FR was observed under ramping heat, which caused 29 and 42% reduction in FR of Sicot 71BRF and 5B, respectively, compared with their respective controls (optimum temperature) at −1 DHT. Similarly, plants exposed to heat shock (45◦C) had 20% (averaged across all the three studied

genotypes) lower FR compared with their respective control at −1 DHT (**Figures 3A,C**). Further reduction in FR of heatstressed plants of all the studied genotypes was observed at 15 DHT (**Figures 3B,D**). Although, exposure to 36◦C caused no significant effect on FR of Sicot 71BRF and CIM 448 at −1 DHT, it significantly reduced FR of Sicot 71BRF at 15 DHT. AVG slightly increased (Genotype × Temp × AVG; P < 0.05) FR of Sicot 71BRF plants under ambient conditions only but no significant effect of ACC was observed on FR of any genotype under any heat treatment (Supplementary Table 3).

#### Leaf Gas Exchange and Chlorophyll Fluorescence

Elevated temperature significantly reduced photosynthetic capacity of both cotton genotypes (Sicot 71BRF and 5B), although the response to heat stress varied in these two genotypes (Genotype × Temp, P < 0.05; Supplementary Table 4). Heat shock caused relatively greater reduction in P<sup>n</sup> of Sicot 71 BRF (20% lower than control) than in 5B (13% lower than control; **Table 1**). By contrast, 5B plants exposed to ramping heat experienced a greater reduction in P<sup>n</sup> (49% lower than control) than Sicot 71 BRF (18% lower than control; **Table 1**). Along the same lines, 36◦C had no significant effect on P<sup>n</sup> of Sicot 71BRF leaves, but it significantly increased P<sup>n</sup> in CIM 448 leaves. ACC had no significant effect on P<sup>n</sup> of cotton genotypes under any treatment condition (**Table 2**).

As for Pn, stomatal conductance (gs) of cotton genotypes was variably affected by heat shock and ramping heat (Supplementary Table 4). Sicot 71BRF plants had 25 and 22% higher g<sup>s</sup> under heat shock and ramping heat stress, respectively, compared with control plants (**Table 1**) while g<sup>s</sup> of 5B plants significantly increased (49% higher than control) in response to heat-shock and but decreased (68% lower than control) under ramping heat (**Table 1**). Sicot 71BRF plants exposed to 45 and 36◦C had 32 and 98% higher g<sup>s</sup> , respectively, than control. In a similar manner to Sicot 71BRF, CIM 448 plants exhibited 48 and 92% increase in g<sup>s</sup> under 45 and 36◦C, respectively, compared with control (**Table 2**). AVG and ACC application had no significant effect on g<sup>s</sup> of cotton genotypes under optimum conditions but AVG slightly increased g<sup>s</sup> of Sicot 71BRF under ramping heat and decreased under heat shock (Supplementary Table 4).

Heat shock (45◦C) equally inhibited the PSII yield of cotton leaves in both glasshouse experiments, causing 25% (averaged across three genotypes) reduction in PSII yield compared with their respective controls (**Tables 1**, **2**). Ramping heat also significantly reduced PSII yield of cotton leaves and the reduction was relatively greater in 5B (58% reduction over control) than Sicot 71BRF (40% reduction over control). In contrast, no

significant effect of 36◦C was observed on PSII yield of Sicot 71BRF or CIM 448 leaves. ACC or AVG application had no significant effect on PSII yield of cotton leaves under any treatment conditions (Supplementary Table 4).

#### Relative Cell Membrane Injury

High temperature treatment (heat shock and ramping heat) significantly increased RCI in cotton leaf tissues (**Table 1**). For example, heat-stressed Sicot 71BRF and 5B plants had 25 and 20% (averaged across heat shock and ramping heat) higher RCI at 40 and 55◦C, respectively, compared with control plants. AVG had no significant effect on leaf RCI of any cotton genotype under any temperature, except RCI of heat-shock plants at 40◦C, where AVG treated leaves had significantly lower RCI than non-AVG treated leaves at the same temperature.

In the 2nd experiment, RCI of Sicot 71BRF and CIM 448 was significantly increased by 36 or 45◦C treatments but the effect of 45◦C was relatively >36◦C (**Table 2**). ACC further increased the RCI of heat-stressed plants (Supplementary Table 5).

#### In Vitro Pollen Development

Cotton plants exposed to heat shock or ramping heat for 1 week did not produce any viable pollen up to 2 weeks after the termination of stress. Therefore, pollen germination was tested only from the plants growing under control conditions. Similarly, no significant effect of prior treatment of plants with AVG or ACC was observed on subsequent pollen germination (data not shown).

No significant change in pollen germination percentage of the two genotypes (Sicot 71BRF and 5B) was observed by increasing incubation temperature from 28 to 30◦C, however, further increase in incubation temperature significantly reduced the pollen germination rate, and no pollen germination was observed at ≥39◦C (**Figures 4A,B**). Pollen tube growth was also inhibited by increasing temperature over the same range.

In the 2nd experiment, pollen germination was tested only at three temperatures (28, 30, and 36◦C). Pollen germination and pollen tube length of Sicot 71BRF exhibited a similar trend to that of the first experiment, showing no significant change by increasing incubation temperature from 28 to 30◦C and a three-fold reduction at 36◦C. CIM 448 plants had significantly greater pollen germination (%) than Sicot 71BRF under different temperatures; e.g., germination (%) of CIM 448 pollen was almost twice that of Sicot 71BRF at 36◦C, although no significant difference in pollen tube length of both cotton genotypes was observed at this temperature (**Figures 4C,D**).

#### Ethylene Production

Elevated temperature (45◦C) significantly reduced ethylene release from leaf tissues of cotton cultivars Sicot 71BRF and CIM 448, although its effect on 5B was non-significant (**Figures 5A,B**). Similarly, ramping high temperature significantly reduced ethylene release from leaf tissues of Sicot 71BRF and 5B genotypes and this reduction was relatively greater than that of heat-shock treatment. No significant change in ethylene release was observed in any cotton cultivar in response to 36◦C. AVG also significantly reduced ethylene release from leaf tissues of cotton cultivars under all treatment conditions. Although, there was a significant increase in ethylene production from ACCtreated cotton leaves when measured 24 h after ACC application (data not shown), no significant change in ethylene production was observed at the termination of heat treatment (7 days after ACC application; **Figure 5B**).

### Relationship between Fruit Production and Leaf Physiology

The mechanism of heat-induced fruit loss was explored by studying the relationships between number of green bolls and the major yield-affecting variables e.g., leaf Pn, ethylene and RCI. Irrespective of treatment conditions, number of green bolls in the studied genotypes showed a strong positive relationship with photosynthesis (**Figures 6A,B**) and a negative relationship with RCI (**Figures 6C,D**). In contrast, no significant correlation was observed between fruit number and leaf ethylene release in this study (**Figures 6E,F**).

Relationships among various components was further explored using multivariate analysis. PCA was conducted based on cotton genotypes (**Figure 7A**) and ethylene regulator application (**Figure 7B**). The loading matrix of PCA also indicated a strong positive correlation between P<sup>n</sup> and green bolls, which were negatively correlated with leaf RCI (**Figures 7A,B**). Variation caused by different treatments and genotypes were mainly explained by first principal components (71.7%, PC1) followed by second principal components (16%, PC2; **Figures 7A,B**). The eigenvectors for PC1 and PC2 were

PC1 = 0.536X1, 0.522X2, −0.523X3, 0.409X4 and PC2 = −0.386X1, −0.169X2, 0.138X3, 0.896X4

where X1 is number of green bolls; X2 is leaf Pn; X3 is leaf RCI and X4 is leaf ethylene production.


TABLE 1 | Changes in various physiological components of different cotton genotypes in response to heat shock (45◦C), ramping heat (45◦C), and aminoethoxyvinylglycine (AVG) application in Experiment 1.

*Data were collected from the topmost fully expanded leaves at the termination of 7 d heat treatment. gs, stomatal conductance (mol H*2*O m*−*<sup>2</sup> s* −*1 ), Pn, rate of photosynthesis (*µ*mol CO<sup>2</sup> m*−*<sup>2</sup> s* −*1 ), PSII yield, photosystem II yield, RCI, % relative cell injury. Values are the mean of four independent replications with standard error in parenthesis (brackets) and sample size* = *4. Means sharing same letters within each column are not significantly different at* α = *0.05.*

TABLE 2 | Changes in various physiological componets of different cotton genotypes in response to heat shock (36 and 45◦C) and 1-aminocylopropane-1-carboxlyic acid (ACC) application in experiment 2.


*Data were collected from the topmost fully expanded leaves at the termination of 7 d heat treatment. gs, stomatal conductance (mol H*2*O m*−*<sup>2</sup> s* −*1 ), Pn, rate of photosynthesis (*µ*mol CO<sup>2</sup> m*−*<sup>2</sup> s* −*1 ), PSII yield, photosystem II yield, RCI, % relative cell injury. Values are the mean of four independent replications with standard error in parenthesis (brackets) and sample size* = *4. Means sharing same letters within the column are not significantly different at* α = *0.05.*

Significantly higher values of eigenvectors for leaf P<sup>n</sup> and RCI indicated that PC1 is an index of greater number of fruits in cotton genotypes with high P<sup>n</sup> and lower RCI, and it mainly separated the plants because of temperature treatments (**Figures 7A,B**). A distinctive clustering pattern of the variables was observed in PC1 e.g., the genotypes under control 28 and 36◦C were clustered mainly on the right hand side of the biplot, showing higher number of green bolls and photosynthesis. On the other hand, the plants exposed to 45◦C either through heat shock or ramping high temperature were clustered on the lefthand side, indicating higher level of RCI. PC2 was an index of the subtle differences in leaf ethylene and number of green bolls and thus could explain the separation of variables in terms of ethylene production.

#### DISCUSSION

Significant losses in cotton lint yield has been reported under high temperature in numerous glasshouse and field studies (Ehlig and Lemert, 1973; Kakani et al., 2005; Zhao et al., 2005). Reddy et al. (1991b) suggest that a rise in maximum day temperature above 30◦C will induce abscission of squares and flowers, while cotton plants shed most of their fruits when temperatures rise above 40◦C. The first objective of the experiments reported here was to identify the pattern of fruit loss in selected cotton genotypes in response to sudden and graduated exposure to high temperatures. Surprisingly, sudden exposure to high temperatures had less impact on fruit production than heat imposed after ramping, even though the duration of heating in

ramping treatments was 11 d compared with 7 d in heat-shocked plants. In spite of this, the period that plants were held at 45◦C was the same in both treatments and there was no evidence of acclimation when plants were raised to this temperature incrementally.

Plants were also surprisingly resilient after the heat-shock treatment. Plants exposed to 45◦C supported significantly fewer squares at −1 DHT but they produced significantly more squares than non-stressed plants by 15 DHT since most of the squares in non-stressed plants had turned into green bolls at this stage. However, these young squares may not contribute to final lint yield due to delayed maturity, especially in mechanized cotton production systems, where cotton is harvested as a determinate crop (Da Costa and Cothren, 2011; Da Costa et al., 2011). On the other hand, plants in the ramped heat treatment had no more squares than control plants at 15 DHT, indicating that this long-term heat treatment irreversibly damaged reproductive development. This assertion was also supported by significantly greater reduction in P<sup>n</sup> after ramping heat than after heat shock. Despite continuing square production, heat-shocked (45◦C) plants were unable to produce a viable reproductive flower up to 15 DHT, indicating that high temperature specifically impaired

the early phases of pollen development, leading to production of abnormal (sterile) pollen. This was supported by in vitro pollen germination assays, where heat-stressed plants did not produce any viable pollen at 15 DHT. Similar data have been proposed by Fisher (1975) who found that high temperature induced pollen sterility and subsequent fruit abscission in cotton during the late reproductive phase. Song et al. (2015) indicated that sporogenous cell to tetrad formation phases (16– 24 days prior to anthesis) are the most sensitive stages of pollen development to high temperature. Thus, yield losses in heatstressed cotton can be minimized by protecting early phases of pollen development. HSP101 protein plays a key role in thermotolerance acquisition in plants, however, the gene coding for this protein is not normally expressed in pollen tissues. Burke and Chen (2015) has achieved a degree of success in increasing thermotolerance in cotton and tobacco pollen through introduction of a gene (AtHSP101) encoding a heat shock protein. The resultant transgenic plants also exhibited better growth and yield stability compared with wild type plants at high temperature.

Reduced photosynthesis in response to high temperature (heat shock and ramping), as observed in this study, has been documented (Lu et al., 1997; Hejnák et al., 2015). However, in contrast to Perry et al. (1983) who suggested 33◦C as a threshold temperature for diminished photosynthesis in cotton, leaf P<sup>n</sup> was unchanged up to 36◦C in Sicot 71BRF and CIM 448, with PSII yield unaffected and g<sup>s</sup> significantly increased. This reflects the development of modern and more heat-tolerant cotton cultivars. Zhao et al. (2005) also observed undiminished P<sup>n</sup> up to 36◦C. Increasing g<sup>s</sup> with temperatures up to 45◦C show that stomatal resistance did not constrain P<sup>n</sup> even at high temperatures, indicating that lower P<sup>n</sup> was caused by

other factors such as, temperature-induced injury to electron transport (Wise et al., 2004), inhibited activation of rubisco (Scafaro et al., 2016) or lower mesophyll conductance (Scafaro et al., 2011). However, negative effects of heat on P<sup>n</sup> in 5B was seen after plants were ramped to 45◦C, at least in part because of reduced stomatal closure. The PSII complex is very sensitive to abiotic stresses and elevated temperature can damage components of PSII complex of the thylakoid membranes by increasing fluidity of thylakoid membranes (Schrader et al., 2004). The current study in cotton also showed major decreases in PSII yield accelerated abscission of fruits. This finding was also supported by increased RCI in cotton leaves under elevated temperature (45◦C) in the present study. Thus heat-induced impaired electron flow and energy transfer pathway is a likely reason for photosynthetic inhibition in cotton leaves (Cottee et al., 2014).

Heat-induced damage to cotton growth has been linked with peroxidation of lipid membranes (Bibi et al., 2008)—a key site for sensing high and low temperature conditions in plants (Örvar et al., 2000). As a result, cell membrane thermostability has often been used as an indicator of heat stress tolerance in cotton (Cottee et al., 2007). A strong negative correlation between green bolls and RCI in the present study supports the case for heat-induced cellular damage across all genotypes. For example, less fruit loss in Sicot 71BRF than 5B under ramping heat could be associated with more resilient membranes in the commercial genotype and better photosynthetic performance. Nonetheless, at 36◦C fruits of Sicot 71BRF began to abort without any significant change in photosynthetic performance, indicating that photoassimilate partitioning or more specific effects in the reproductive machinery are also important in fruit set. In particular, genotypic variation in reproductive potential in response to higher temperature was closely associated with pollen viability, as suggested by Kakani et al. (2005), as seen in relatively better pollen germination in the more heat-tolerant genotype (CIM 448) under 36◦C.

The second objective of these experiments was to establish a link between ethylene and fruit abscission in heat-stressed cotton. Despite numerous reports on the effect of anti-ethylene agents (e.g., 1-MCP) on heat-stressed cotton (Oosterhuis et al. (2009), limited work has been done to practically measure ethylene concentrations from cotton tissues (Kawakami et al., 2013). In contrast to wheat (Hays et al., 2007) and soybean (Djanaguiraman et al., 2011), where high temperature induces ethylene production, significantly lower ethylene concentration in heat-stressed cotton tissues was recorded in this study. Similarly, Kawakami et al. (2013) observed that heat stress can significantly reduce ethylene production from reproductive tissues of cotton (38/20◦C) 2 days after anthesis and proposed that fruit development could be increased by regulating ethylene metabolism. However, significant loss of fruits in heat-stressed plants despite lower ethylene concentration in the present study suggested a limited role of ethylene in regulating heatinduced damage to cotton fruits. This was supported by the observation that there was no significant effect of AVG or ACC (ethylene regulators) on fruit production in cotton under high temperatures.

PCA indicated that clustering of variables on PC1 (72% of total variations) was mainly controlled by temperature rather than chemical application. A negative relationship (eigenvectors) between leaf ethylene and green bolls on PC2 (16% of total variations) indicates that ethylene might play a minor role reducing fruit number of cotton. However, this separation on PC1 was mainly observed in Sicot 71BRF under 28◦C (**Figures 7A,B**), and can be explained by the positive effects of AVG on fruit production rather than ethylene-induced fruit loss under control conditions. AVG was effective in increasing fruit production in non-stressed cotton via improved carbon assimilation (Najeeb et al., 2015). Although positive effects of an ethylene action inhibitor (1-MCP) has been reported on the growth and yield of heat stressed cotton, no data are available on the effect of ethylene synthesis inhibitor (AVG) on heat-stressed cotton. For example, Kawakami et al. (2010) suggested that 1-MCP can increase lint yield of heat-stressed cotton by increasing boll weight but not fruit abscission, while Oosterhuis et al. (2009) observed significantly more fruits in heat-stressed cotton in response to 1-MCP application. Similarly, Chen et al. (2015) and Kawakami et al. (2013) observed positive effects of 1-MCP on lipid membrane and leaf chlorophyll of heatstressed cotton but Scheiner et al. (2007) observed no significant effect of 1-MCP on these parameters. Thus, it is hard to conclude that inhibiting ethylene production or action in cotton tissues can protect cotton crops from heat-induced injury. Further, most of these 1-MCP application experiments were conducted under field conditions (where potentially multiple stresses including drought are present), which complicates the role of anti-ethylene agent in protecting cotton crops from high temperatures.

Based on the inhibitory role of ethylene at high temperatures, we further tested whether elevated ethylene may protect cotton from heat-induced injury. Earlier studies suggested positive effects of ethylene on heat-stressed plants e.g., Arabidopsis (Larkindale and Knight, 2002) and tomato (Firon et al., 2012). In the present study, ACC slightly increased g<sup>s</sup> and P<sup>n</sup> of heatstressed plants, indicating that ACC might regulate stomatal behavior of cotton by modulating ABA synthesis (Ahmed et al., 2006). In addition, it increased green bolls of cotton at 15 DHT under 28 and 36◦C. However, it is difficult to conclude that this increase was induced by ACC treatment, since no significant increase in leaf ethylene was detected at the end of heat treatment. Instead, accelerated fruit production in ACC treated plants at 15 DHT could be result of tendency of cotton plants to compensate ACC-induced early loss of flowers (Stewart et al., 2001). This was also supported by production of significantly more squares at −1 DHT (7 days after ACC application) but the effect became non-significant at 15 DHT.

#### CONCLUSIONS

Elevated temperature (45◦C) significantly impaired various physiological processes of cotton, leading to pollen sterility and flower abscission. Cotton plants increased leaf stomatal conductance but could not protect photosynthetic machinery from heat-induced (45◦C) injury. High temperature and AVG reduced ethylene production, while ACC caused a transitory increase in ethylene production from cotton tissues but these treatments (ACC or AVG) could not affect abscission of fruits. Similarly, the ethylene-defective genotype 5B had an impaired tolerance to heat when compared with the commercial genotypes tested. Further heat-stressed (45◦C) cotton plants (irrespective of ACC or AVG treatment) did not produce any viable pollen up to 2–3 weeks after stress, indicating that ethylene might not play any direct role in heat-induced yield damage to cotton. In contrast, high temperature induces fruit abscission either directly by impairing pollen development and/or indirectly by inhibiting photoassimilate supply to developing fruits.

### AUTHOR CONTRIBUTIONS

Concept development, experimental planning, revision of initial draft: DT, MB, and BA. Experimental execution, data collection, analysis, and initial writing: UN and MS.

#### ACKNOWLEDGMENTS

This study was financially supported by the Cruiser R and D fund co-funded by Syngenta Crop Protection Australia and Cotton Seed Distributors. The authors are thankful to the University of Sydney, Australia for a USYDIS research grant and Macquarie University, Australia for technical

#### REFERENCES


support and glasshouse/laboratory facilities. We are also thankful to CSIRO Plant Industry for providing cotton seeds and Phil Glover, Sumitomo Chemicals, Australia for the supply of AVG. The authors are also highly thankful to Higher Education Commission of Pakistan for supporting MS through International Research Support Initiative Program (IRSIP).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2017. 01615/full#supplementary-material


in cotton by inhibiting ethylene synthesis and sustaining photosynthetic capacity. Plant Growth Regul. 76, 83–98. doi: 10.1007/s10725-015- 0037-y


**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 Najeeb, Sarwar, Atwell, Bange and Tan. 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.

**196**

# Enhancing Brassinosteroid Signaling via Overexpression of Tomato (Solanum lycopersicum) SlBRI1 Improves Major Agronomic Traits

Shuming Nie1,2† , Shuhua Huang<sup>1</sup>† , Shufen Wang<sup>1</sup> , Dandan Cheng<sup>1</sup> , Jianwei Liu<sup>1</sup> , Siqi Lv<sup>1</sup> , Qi Li<sup>1</sup> and Xiaofeng Wang<sup>1</sup> \*

<sup>1</sup> State Key Laboratory of Crop Stress Biology for Arid Areas, College of Horticulture, Northwest A&F University, Yangling, China, <sup>2</sup> Qinghai Key Laboratory of Vegetable Genetics and Physiology, Xining, China

Brassinosteroids (BRs) play important roles in plant growth, development, and stress responses through the receptor, Brassinosteroid-insensitive 1 (BRI1), which perceives BRs and initiates BR signaling. There is considerable potential agricultural value in regulating BR signaling in crops. In this study, we investigated the effects of overexpressing the tomato (Solanum lycopersicum) BRI1 gene, SlBRI1, on major agronomic traits, such as seed germination, vegetative growth, fruit ethylene production, carotenoid accumulation, yield, and quality attributes. SlBRI1 overexpression enhanced the endogenous BR signaling intensity thereby increasing the seed germination rate, lateral root number, hypocotyl length, CO<sup>2</sup> assimilation, plant height, and flower size. The transgenic plants also showed an increase in fruit yield and fruit number per plant, although the mean weight of individual fruit was reduced, compared with wild type. SlBRI1 overexpression also promoted fruit ripening and ethylene production, and caused an increase in levels of carotenoids, ascorbic acid, soluble solids, and soluble sugars during fruit ripening. An increased BR signaling intensity mediated by SlBRI1 overexpression was therefore positively correlated with carotenoid accumulation and fruit nutritional quality. Our results indicate that enhancing BR signaling by overexpression of SlBRI1 in tomato has the potential to improve multiple major agronomic traits.

Keywords: agronomic traits, Brassinosteroids, nutritional quality, signaling regulation, SlBRI1, tomato

# INTRODUCTION

Brassinosteroids (BRs) are a group of steroid phytohormones that play important roles in plant growth, development, and stress responses (Zhu et al., 2013; Xia et al., 2015). Studies to date, mostly using the model plant Arabidopsis thaliana, have shown that BRs bind to the receptor Brassinosteroid Insensitive 1 (BRI1), activating it and initiating a signal transduction cascade. BR-deficient or BR-insensitive mutants generally have serious developmental deficiencies, such as dwarfed stature, dark green leaves, short petioles and hypocotyls, delayed flowering and senescence, reduced male fertility, decreased seed setting rates and photosynthetic capacity, as well as perturbed phytohormone balances. Furthermore, BR-insensitive mutants have been reported to have increased transcript levels of BR biosynthetic genes, as well as a higher BR content (Chory et al., 1991; Li and Chory, 1997; Noguchi et al., 1999; Wang et al., 2008; Kim and Wang, 2010).

#### Edited by:

Yunde Zhao, University of California, San Diego, United States

#### Reviewed by:

Guang Wu, Shaanxi Normal University, China Rosario Muleo, Università degli Studi della Tuscia, Italy

> \*Correspondence: Xiaofeng Wang wangxff99@nwsuaf.edu.cn †These authors have contributed equally to this work.

#### Specialty section:

This article was submitted to Crop Science and Horticulture, a section of the journal Frontiers in Plant Science

Received: 09 June 2017 Accepted: 25 July 2017 Published: 10 August 2017

#### Citation:

Nie S, Huang S, Wang S, Cheng D, Liu J, Lv S, Li Q and Wang X (2017) Enhancing Brassinosteroid Signaling via Overexpression of Tomato (Solanum lycopersicum) SlBRI1 Improves Major Agronomic Traits. Front. Plant Sci. 8:1386. doi: 10.3389/fpls.2017.01386

**197**

Brassinosteroids influence many important agronomic traits associated with growth, photosynthesis, architecture, and yield and crop quality. The application of BR analogs has been shown to promote plant growth, photosynthesis, fruit carotenoid accumulation, and quality attributes of fruit, while inhibiting BR biosynthesis with the compound Brassinazole has the opposite effect (Yu et al., 2004; Wu et al., 2008; Xia et al., 2009a). BRs increase photosynthetic capacity by activating RuBisCO activase (Xia et al., 2009a), and BR treatment induces ethylene production in fruit by inducing the expression of ethylene biosynthetic genes, thereby regulating ripening (Zhu et al., 2015). In addition, exogenous BRs have been shown to enhance resistance of plants to abiotic and biotic stresses, such as drought, low and high temperatures, high salinity, and pathogen or nematode attack (Collins, 2007; Xia et al., 2009b; Ahammed et al., 2012; Xi et al., 2013).

However, in an agronomic context, BR application is very expensive and its effects are unstable, which have limited its potential value in crop production. An alternative strategy is the genetic manipulation of BR biosynthesis or signaling is to alter BR content or its signaling intensity in crop plants (Wang et al., 2016). Overexpression of DWARF, a BR biosynthetic gene, in A. thaliana, tomato (Solanum lycopersicum), rice (Oryza sativa), and maize (Zea mays), has been reported to result in higher endogenous BR levels, increased growth, nutrition quality and yield (Wang et al., 2014; Li et al., 2015; Eremina et al., 2016). Furthermore, overexpression in tomato of AtBZR1-1D, a transcription factor from A. thaliana involved in BR mediated signaling, increased carotenoid accumulation and other quality attributes in fruit (Liu et al., 2014), while constitutive activation of BR signaling enhanced freezing resistance by regulating cold responsive gene expression in A. thaliana (Eremina et al., 2016). Accordingly, there is considerable interest in enhancing endogenous BR levels or associated signaling as a means to improve crop yield, nutritional quality, and stress tolerance (Divi and Krishna, 2009).

The perception of BRs by BRI1 and consequent BR signal transduction in A. thaliana has been investigated using biochemical, genetic, and proteomic approaches (Ryu et al., 2010). Endogenous overexpression of A. thaliana AtBRI1 was reported to cause in an increase in petiole length and sensitivity to BRs (Fujimura and Samet, 2001), while overexpression of TaBRI1 (Triticum aestivum) in A. thaliana resulted in early flowering, greater silique size and seed yield, and an increase in root sensitivity to BR (Singh et al., 2016). The phenotype of the tomato bri1 (cu-3) mutant is similar to that of the A. thaliana bri1 mutant, and constitutive overexpression of SlBRI1 in tomato was reported to rescue the cu-3 dwarf phenotype (Koka et al., 2000; Holton et al., 2007). Furthermore, the transgenic lines had longer internodes and hypocotyls, and their leaves were more ovoid and showed reduced serration compared to the wild type (WT) (Koka et al., 2000; Holton et al., 2007). However, when SlBRI1 was overexpressed in the A. thaliana AtBRI1 mutant (bri1- 5), it did not fully complement the mutation (Holton et al., 2007), suggesting that the functions of different BRI1 orthologs may vary somewhat; an idea that needs to be further investigated.

To date, the potential for using SlBRI1 and BR signaling in tomato to enhance major traits related to architecture, yield, and fruit nutritional quality has not been determined. In this current study, we generated BRI1-overexpressing tomato plants exhibiting increased BR signaling intensity and evaluated seed germination, growth, developmental characteristics, fruit nutritional quality and yield. Based on these results, we discuss the feasibility of improving tomato growth and fruit nutritional quality by genetic manipulation of SlBRI1.

#### MATERIALS AND METHODS

#### Plant Material and Growth Conditions

For the seed germination experiment, seeds from S. lycopersicum (cv. Micro-Tom; used here as the WT), and from the T2 generation transgenic lines (see below for details of generating the transgenic plants), were imbibed on Petri dishes with 0.74% agar and maintained at 28◦C for 3 days in darkness, and then maintained at 25◦C for 4 days under a photosynthetic photon flux density (PPFD) of 200 µmol m−<sup>2</sup> s −1 . For the phenotype evaluation, seeds were germinated at 28◦C in Petri dishes lined with two layers of filter paper moistened with deionized water. The germinated seeds were then sown and plants were also grown in plastic pots (7 cm × 7 cm × 8 cm), filled with 105 g of a mixture of peat and vermiculite (v/v = 7:3) in trays. Seedlings and plants were grown in growth chambers, at a temperature 25/20◦C, a PPFD of 500 µmol m−<sup>2</sup> s −1 , and a 12-h photoperiod. Flowers were tagged at the full-bloom stage to synchronize developmental comparisons. Fruits were identified as being at the breaker (B) stage when they displayed the first sign of external color change. Fruit from the WT and transgenic lines took 54 and 45 days, respectively, to develop from anthesis to the B stage. Fruit were considered to be at the mature green (MG) stage 3 days before B, while fruit at 3 days after B were considered to be at the P (pink) stage and those at 6 days after B were at the MR (mature red) stage.

#### Constructs and Plant Transformation

The tomato SlBRI1 gene was PCR-amplified from tomato (S. lycopersicum cv. Micro-Tom) cDNA based on a UniGene sequence (Solyc04g051510), and a Flag peptide encoding sequence was added to the 3<sup>0</sup> end. The SlBRI1 cDNA sequence was cloned into the KpnI and XbaI sites of the binary pBI121vector, which carries the kanamycin resistance gene for bacterial and plant selection, and drives transgene expression using the constitutive cauliflower mosaic virus 35S promoter. The construct was transformed into the Agrobacterium tumefaciens strain GV3101 by electroporation, which was then used to transformed tomato cotyledon explants (Park et al., 2003). Three homozygous transgenic tomato lines (BRI1: OX4, BRI1: OX6, and BRI1: OX37), exhibiting high SlBRI1 expression levels, were chosen for subsequent phenotypic and molecular characterization.

# CO<sup>2</sup> Assimilation Rate Detection

fpls-08-01386 August 8, 2017 Time: 15:42 # 3

The CO<sup>2</sup> assimilation rate (Pn) was determined using a portable photosynthesis system (LI-6400-40, Li-Cor, Lincoln, NE, United States). All measurements were carried out at 400 µmol (CO2) mol−<sup>1</sup> , at 22◦C, and 800 µmol m−<sup>2</sup> s −1 light intensity.

# Phenotypic Characterization

Plant height, expansion diameter of plant, leaf area, length and width of the third leaf were determined at 30 days after sowing and of the sixth leaf at 60 days after sowing. The fresh weight (FW) of the shoot, leaf mass per area, leaf thickness, stigma length, and petal length were measured at 30 days after sowing. The third and sixth leaves were photographed with a digital camera (Canon G15, Japan) at 30 and 60 days after sowing. Leaf areas were calculated using ImageJ (Collins, 2007) software. Leaf thicknesses of WT and transgenic plants were measured using light microscopic observation of paraffin sections, generated as in Yang et al. (2011), with an Olympus microscope (BX53, Japan). The third leaf was selected and at least six sections were analyzed. The growth and fruit characteristics of 15 plants per transgenic line and of the WT genotype were observed.

#### Ethylene Detection

To measure ethylene production, B, P, and R stage fruit of approximately the same size were harvested and kept in sealed plastic bags for 3 h at room temperature (each biological replicate corresponded to five fruits). One milliliter of gas was removed from each plastic bag with a syringe, and ethylene production was determined using a gas chromatograph, which main specifications were FID detector, DB-130 m, 0.32 µm, 0.25 µm chromatographic column (Agilent 7890B, United States). All samples involved three biological replicates.

#### Gene Expression Analysis

RNA from various organs of WT and transgenic plants was extracted using the RNAiso Plus (TaKaRa, Japan) kit, according to the manufacturer's instructions. Residual DNA was removed with DnaseI (Thermo, United States). One microgram total RNA was reverse-transcribed using a cDNA synthesis Kit (Roche, Switzerland), following the manufacturer's instructions. The cDNA concentrations were normalized to the expression levels of the tomato ubiquitin gene (X58253.1) for subsequent RT-PCR analysis. The qPCR experiments were performed using a Real-Time PCR system (Bio-Rad) with SYBR Green Master Mix (Vazyme, Nanjing, China). Relative gene expressions were calculated using the 2−11CT method (Livak and Schmittgen, 2001) and the tomato UBI3 gene was used as the internal control. The experiment had three independent biological samples per line and each sample had four technical replications. Genespecific qPCR primers are listed in Supplementary Table S2. The WT and transgenic plants were represented by four replicates for each sample.

#### Determination of Carotenoid Content

Carotenoids were extracted and analyzed as previously described (Liu et al., 2014) with modifications. Tomato pericarp (0.5 g) was ground to a fine powder in liquid nitrogen and extracted with 15 mL of hexane/acetone/ethanol (1:1:1) solution containing 0.01% butylated hydroxytoluene (BHT) on a magnetic stirrer for 30 min. The extracts were centrifuged at 4000 × g at 4 ◦C for 10 min, and the supernatant transferred to 50 mL tubes. Fifteen milliliter of hexane/acetone/ethanol was added to the sediment and the extraction process was repeated. The supernatants from the two extractions were merged, before hexane was added up to a total volume of 25 mL. The absorbance of 5 mL of the mixture was measured at 450 nm (UV-1750; Shimadzu Corporation) and this value used to represent total carotenoid levels. The remaining mixture was frozen and evaporated using a vacuum evaporator (RE-52AA). The residue was dissolved in 1 mL ethyl acetate and 20 µL were injected onto a high pressure liquid chromatograph (HPLC, Shimadzu [LC 2010A, Japan]) using an autosampler. All the procedures were performed under dim light. A Nova-Pak C18 column (4.6 mm × 150 mm) and ultraviolet detector were used with a mobile phase of methanol/acetonitrile (45/55 by volume) at a flow rate of 1 mL/min. The total retention time was 25 min. Absorbance was detected at 450 nm. Standards (lutein, lycopene, and β-carotene; Sigma, United States) were prepared and used to identify and quantify the corresponding carotenoids. The carotenoid content was expressed as µg/g FW of tomato fruit.

#### Western Blot Analysis

Proteins was extracted from young leaves (0.2 g) of 25-day-old transgenic and WT plants with 2 × SDS gel loading buffer, as in Wang et al. (2016).

#### Fruit Firmness

To measure fruit firmness, R stage fruits of the approximately same size and location were harvested (each biological replicate consisted of five fruits) and firmness was determined with a hand-held penetrometer (FHM-1, Japan). Data values are the means ± SD of six independent biological samples.

#### Soluble Solid Contents

Soluble solid contents (oBRIX) of total fruit juice from freshly harvested R stage fruits were determined with a hand-held refractometer (Chengdu Optical Instrument Factory, Chengdu, China). Data values are the means ± SD of six independent biological samples.

# Soluble Sugar Contents

The soluble sugar contents were measured as previously described (Liu et al., 2014). The experiment had three independent biological samples per line and each sample had four technical replications.

#### Ascorbic Acid Contents

Ascorbic acid contents were determined as previously described with minor modifications (Hu et al., 2016). Frozen pericarp (1 g) of R stage fruits was ground to a fine powder in liquid nitrogen and extracted with 5 mL of 6% (w/v)

trichloroacetic acid (TCA). The homogenates were centrifuged at 4000 × g for 15 min at 4◦C. The supernatant (0.4 mL) was transferred to a 10 mL tube containing 0.4 mL of 5 mM DL-dithiothreitol (DTT) in 0.4 M potassium phosphate buffer, pH 7.4. The 10 mL tube was incubated for 30 min at 37◦C, and 0.2 ml 0.5% (w/v) N-ethylmaleimide (NEM) added, followed by incubation for 1 min at room temperature. Color reagent (1.6 mL) was then added to the mixture, which was incubated for 1 h at 37◦C. The absorbance was recorded at 550 nm (Shimadzu UV-1750). The color reagent was prepared as follows: solution A, 31% orthophosphoric acid, 4.6% (w/v) TCA, and 0.6% (w/v) iron chloride (FeCl3); solution B, 4% 2, 2-dipyridyl (w/v in 70% ethanol). Solutions A and B were mixed in a ratio of 2.75:1 before use. The experiment had three independent biological samples per line.

#### Statistical Analysis

The data were analyzed using SPSS version 17.0 and a Student's t-test. Mean and standard error values were calculated for the variables comparison. Values of P < 0.05 and P < 0.01 were considered statistically significant.

# RESULTS

### SlBRI1 Is Ubiquitously Expressed

SlBRI1 expression in different organs was determined using RT-PCR, and we observed that expression was very low in roots, higher in flowers and red fruits, and highest in stems and leaves (**Figure 1A**), suggesting that SlBRI1 is involved in regulating both vegetative and reproductive growth.

### Overexpression of SlBRI1 Increases the Endogenous BR Signal Intensity

We generated transgenic tomato plants expressing SLBRI1 under the control of the constitutive CaMV 35S promoter. Transcript levels of SlBRI1 in the three transgenic lines BRI1:OX4, BRI1:OX6, and BRI1:OX37 were 6.0, 11.6, and 23.4 times higher, respectively, than in WT plants (**Figure 1C**). The expression levels of the BR biosynthetic gene, DWARF, in these transgenic lines were 3.7, 2.9, and 3.5 times lower, respectively, than in WT plants for DWARF (**Figure 1D**), while those of another BR biosynthetic gene, CPD, were 1.52, 1.23, and 1.51 times lower, respectively (**Figure 1E**). Similarly, Western blot analysis indicated that SLBRI1 protein levels in the BRI1:OX4 and BRI1:OX6 lines were lower than in the BRI1:OX37 line (**Figure 1B**), while the SLBRI1-Flag fusion protein was, as expected, not detected in the WT plants. Taken together, these results indicate that overexpression of SLBRI1 resulted in increased transcription and expression of SLBRI1 and caused a feedback-inhibition of BR biosynthetic genes, which consequently showed decreased expression. This is consistent with an increase in the intensity of endogenous BR signaling in the SLBRI1 overexpressing transgenic lines (He et al., 2005).

# Overexpression of SlBRI1 Causes Precocious Seed Germination and Accelerated Seedling Growth

To test whether the SLBRI1 overexpression mediated increase in BR signaling affected seed germination seeds of WT, BRI1:OX4, BRI1:OX6, and BRI1:OX37 were germinated on agar and were monitored for 7 dyas. BRI1:OX4, BRI1:OX6, and BRI1:OX37 seeds germinated earlier than WT seeds (**Figure 2A**). Additionally, radicle lengths increased in proportion to the increase in SLBRI1 expression, such that the radicle lengths in the BRI1:OX4, BRI1:OX6, and BRI1:OX37 lines were 1.9-, 2.1-, and 2.7-fold greater, respectively, than in WT, 3 days after germination (**Figure 2C**). Compared with WT, the transgenic lines showed increased root and hypocotyl length, number of lateral roots and seedling weight after 7 days of germination (**Figures 2B–G**), highlighting the critical role of BR signaling in seed germination and early growth vigor.

# Plant Growth and Photosynthesis Are Enhanced in SLBRI1 Overexpression Lines

To elucidate the effects of the SLBRI1 overexpression mediated increase in BR signaling on the growth of tomato plants, we observed the phenotypes of WT, BRI1:OX4, BRI1:OX6, and BRI1:OX37 plants grown in a growth chamber. The transgenic plants exhibited increased plant height, expansion diameter, internode length, and hypocotyl length 30 days after sowing (**Figures 3A,C,E,G,H**), and also 60 days after sowing in the case of the BRI1:OX37 line (**Figures 3B,D,F**). The transgenic plants had a larger leaf area than WT at the early growth stage (**Figures 4A,C**), and also longer leaves throughout development. The transgenic plants also exhibited a reduced leaf thickness and leaf mass per area compared with WT (**Figure 4**), as well as larger flowers with longer petals and style (**Figure 5**). Finally, gas exchange analyses indicated that the SLBRI1-overexpressing plants had an increased CO<sup>2</sup> assimilation rate and FW 30 days after sowing (**Figures 3I,J**). Taken together, these results indicated that the increase in BR signaling due to SLBRI1 overexpression promoted various aspects of growth and development, as well as photosynthetic capacity.

#### Overexpression of SlBRI1 Causes an Increase in Ethylene Production and in the Expression of Ethylene Biosynthetic Genes

To test whether SLBRI1 expression level is important for the regulation of ethylene production, we measured ethylene accumulation in the fruits of the transgenic lines and WT by gas chromatography. Ethylene production in BRI1:OX4, BRI1:OX6, and BRI1:OX37 were 65.1, 40.5, and 62.7% higher, respectively, than in WT at the B stage (**Figure 6A**). Ethylene production in the transgenic lines was still higher than in WT up to MR stage. We also measured the transcript abundance of the ethylene biosynthetic genes ACO1, ACS2, and ACS4, which are important for ethylene production in tomato fruit (Barry et al., 1996; Barry et al., 2000; Klee and Giovannoni, 2011). Compared with

7 days. (C) Radicle length at 3 days. (D) Lateral root number at 7 days. (E) Root length at 7 days. (F) Hypocotyl length at 7 days. (G) Weight of seedlings at 7 days. Data values are the means ± SD of 15 independent biological samples. Different letters indicate significant differences according to Student's t-test (P < 0.05).

WT fruit, ACO1 and ACS2 transcript levels in the fruits of the three transgenic lines were markedly higher than in WT at the B and P stages (**Figures 6B,C**), with ACO1 expression being 69-, 50-, and 62-fold higher at the B stage. Similarly, ACS4 transcript levels were higher in the transgenic lines than in WT at the B and P stages (**Figure 6D**). These results showed that SLBRI1 overexpression mediated increases in BR signaling promote ethylene production during tomato fruit ripening by upregulating the expression of ethylene biosynthetic genes.

### Overexpression of SlBRI1 Leads to Higher Fruit Carotenoid Accumulation and Expression of Lycopene Biosynthetic Genes

To test whether the SLBRI1 overexpression mediated increase in BR signaling affects carotenoid accumulation, total carotenoid levels were determined by spectrophotometry. We observed that total carotenoid accumulation was significantly greater in the SLBRI1-overexpressing lines (**Figure 7A**). Furthermore, when lycopene, β-carotene, and lutein accumulation in the pericarp of fruit at the MR stage from the transgenic lines and WT was measured, using HPLC, the SLBRI1 transgenic lines showed increased levels of both lycopene and lutein. The lycopene abundance was 120, 84, and 79% greater in the pericarp of BRI1:OX4, BRI1:OX6, and BRI1:OX37, respectively, than in WT (**Figure 7B**), while the lutein levels were 22, 140, and 182% greater, respectively (**Figure 7C**). We observed no significant difference in β-carotene levels between the transgenic lines and WT (**Figure 7D**). Taken together, these results indicated that carotenoid accumulation in tomato fruits correlated with the BR signaling strength in the SLBRI1 overexpression lines.

To investigate whether changes in the expression of genes in the carotenoid biosynthetic pathway correlated with changes in carotenoid content, we used qRT-PCR to determine the transcript levels of SlBRI1 (**Figure 8A**), SlDXS (**Figure 8B**), SlGGPS (**Figure 8C**), SlPSY1 (**Figure 8D**), and SlCYC-B (**Figure 8E**) in the pericarp of fruit at four stages. SlBRI1 expression in the three transgenic lines was significantly greater at all four stages, with SlBRI1 transcript abundance in BRI1:OX37 being 18-, 109-, 119-, and 159-fold higher than in WT at the MG, B, P, and MR stages, respectively (**Figure 8A**). Lycopene biosynthesis related genes (SlDXS, SlGGPS, and SlPSY1) were also all markedly upregulated in the three transgenic lines at the B, P, and MR stages, with the exception of SlGGPS in the BRI1:OX4 and BRI1:OX6 lines at the R stage (**Figures 8B–D**). Additionally, SlDXS expression was also significantly up-regulated at the MG stage (**Figure 8B**). SlCYC-B encodes an enzyme that catalyzes the conversion of lycopene to β-carotene (Hirschberg, 2001), and its expression in the transgenic lines was higher in the B and P stages compared to WT, but was significantly lower at the MR stage (**Figure 8E**). These results are consistent with previous reports (Römer and Fraser, 2005; Dellapenna and Pogson, 2006), that the expression of genes encoding enzymes involved in carotenoid biosynthesis determining the total levels of carotenoids, and that SlBRI1 overexpression cause an increase in BR signaling, which is important for carotenoid accumulation.

# Overexpression of SlBRI1 Accelerates Fruit Ripening and Improves Fruit Quality

We also investigated whether the SlBRI1 expression level affected fruit quality and yield. The transgenic lines exhibited a slight increase in fruit yield and number per plant compared with WT, but a decrease in individual fruit weight. SlBRI1 overexpression resulted in early flowering and a significant reduction in the ripening time (Supplementary Table S1). Levels of ascorbic acid in the fruit of the BRI1:OX4, BRI1:OX6, and BRI1:OX37 lines were 17, 17, and 59% higher, respectively, than in WT (**Figure 9A**), and the increase in total soluble solids was 27, 22, and 29%, respectively (**Figure 9B**). Furthermore, the contents of soluble sugars in fruit from the BRI1:OX4, BRI1:OX6, and BRI1:OX37 lines were 21, 30, and 22% higher than in WT (**Figure 9C**). The firmness of fruit from the BRI1:OX37 line was greater than WT, while no difference was observed between the BRI1:OX4 and BRI1:OX6 lines compared to WT (**Figure 9D**).

#### DISCUSSION

The role of BRs and BR signaling in plant growth and development is well established, based mostly on studies of A. thaliana (Zhu et al., 2013). Their role as regulators of growth, development, fruit yield, and quality have also been investigated in tomato (Liu et al., 2014; Li et al., 2015; Zhu et al., 2015). However, little is known about the potential role of SlBRI1 and BR signaling in regulating tomato architecture or fruit nutrient content. We generated tomato SlBRI1-overexpressing plants, which we concluded showed increased BR signaling intensity based on the following evidence: SlBRI1 mRNA and protein levels in the overexpressing plants were significantly higher than in WT (**Figures 1B,C**); the expression levels of the DWARF and CPD genes were markedly inhibited in the transgenic plants (**Figures 1D,E**) (He et al., 2005); and the hypocotyls and roots of the transgenic plants were significantly longer than WT (**Figures 2E,F**), which is known to closely correlate with the BR signaling intensity (Koka et al., 2000). Taken together, these results indicated that the transgenic plants had increased BR signaling, which resulted in expression of feedback-inhibited BR biosynthetic genes and an increase in hypocotyl and root length.

Known BR mutants include mutants in BR biosynthetic genes causing BR deficiency and in BR receptor genes, causing BR insensitivity, and these mutant types both have a dwarfed phenotype, dark green leaves, and malformed stems and flowers in A. thaliana and tomato (Chory et al., 1991; Szekeres et al., 1996; Li and Chory, 1997; Montoya et al., 2002). In contrast, SlBRI1 overexpression in tomato, as shown here, caused increased plant height, internode length, and hypocotyl length (**Figure 3**). DWARF overexpression has been shown to cause an increase in endogenous BR levels and also resulted in increased plant height, but resulted in a more slender and compact plant architecture at

later developmental stages (Li et al., 2015). In contrast, we found that SlBRI1 overexpression enhanced the expansion diameter of the BRI1:OX37 lines throughout development (**Figure 3**). Earlier studies showed that a mutation in a receptor AtBRI1 gene in A. thaliana resulted in a reduction in petiole and leaf length, whereas overexpression of AtBRI1 resulted in longer petioles, slender leaves, and increased sensitivity to exogenous BRs (Fujimura and Samet, 2001). We also observed longer petioles and a slender leaf phenotype, as well as decreased leaf thickness (**Figure 4**), which has been seen in DWARF overexpressing tomato plants (Li et al., 2015). Furthermore, our SlBRI1-overexpressing plants exhibited larger flowers, with longer petals and stigmas than the WT (**Figure 5**). Therefore, the regulatory role of the endogenous BR level and BR signaling in plant growth and development may differ in different organs and developmental stages of tomato. We also found that the net photosynthetic rate and FW increased with an increase in SlBRI1 expression (**Figure 3**), indicating that the increase in BR signaling caused by SlBRI1 overexpression is associated with a regulation of photosynthetic capacity in tomato.

Ethylene plays a critical role in the regulation of tomato fruit ripening, and previous studies have shown that BR treatment induces ethylene production by regulating the expression of genes involved in ethylene biosynthesis (Zhu et al., 2015). In addition, overexpression of the DWARF gene was also shown to promote ethylene production and ripening in tomato fruit (Li et al., 2015). In this current study, our results indicated that the SlBRI1 overexpression mediated increase in BR signaling promoted ethylene production and fruit ripening (**Figure 6** and Supplementary Table S1). However, the molecular mechanism underlying the increased endogenous BR levels or BR signaling to promote ethylene production has not yet been investigated. The ethylene biosynthetic genes ACO1, ACS2, and ACS4 play critical roles in regulating ethylene production during fruit ripening (Barry et al., 1996, 2000; Klee and Giovannoni, 2011), and we observed that the mRNA levels of ACO1 and ACS2

were significantly higher in the transgenic plants than in WT (**Figure 6**). BRs have also been shown to induce the transcription of ACS genes and increase ACS protein stability (Zimmermann et al., 2004; Hansen et al., 2009). We conclude that increased BR levels or BR signaling may up-regulate the expression of ethylene biosynthetic genes and in turn enhance the ethylene production rate, thereby promoting fruit ripening and the transcription of ripening-related genes.

Carotenoid accumulation has been shown to be increased by BRs and decreased by application of the drug Brassinazole (Zhu et al., 2015). In addition, overexpression of the DWARF gene was reported to cause an increase in BR levels, thereby

enhancing carotenoid accumulation in tomato fruit (Li et al., 2015). Finally, overexpression of AtBZR1-1D in tomato, is known to promote carotenoid accumulation by regulating the expression of carotenoid biosynthetic genes (Liu et al., 2014). In this study, we found that SlBRI1 overexpression markedly increased fruit carotenoid accumulation (except β-carotene) (**Figure 7**). Previous studies have shown that the expression levels of genes encoding the biosynthetic enzymes involved in carotenoid biosynthesis directly regulate carotenoid accumulation (Römer and Fraser, 2005; Dellapenna and Pogson, 2006; Sandmann et al., 2006). Here, we used qRT-PCR to measure the transcript levels of such genes, and found that SlDXS, SLGGPS, and SlPSY1 expression were significantly enhanced in SlBRI1-overexpressing plants compared to WT (**Figure 8**). Our results suggested that SlBRI1 overexpression mediated increases in BR signaling may directly or indirectly up-regulate the expression of carotenoid biosynthetic genes, thereby elevating carotenoid accumulation.

Yield and nutrient quality are important factors in tomato production. We showed that the SlBRI1-overexpressing plants had slightly increased fruit yield and fruit number, but lower individual fruit weight (Supplementary Table S1). In contrast, DWARF overexpression generated larger fruits, and decreased fruit numbers and fruit yield per plant (Li et al., 2015). These results indicate that enhanced endogenous BR levels or increased endogenous BR signaling had different regulatory effects on fruit size and yield. SlBRI1 overexpression resulted in early flowering and accelerated ethylene production and ripening in the fruits (Supplementary Table S1), and DWARF overexpressing plants showed similar phenotypes (Li et al., 2015). These results indicate that increased endogenous BR levels or BR signaling might both accelerate fruit ripening by promoting ethylene production. Finally, previous studies have shown that application of BRs or AtBZR1-1D overexpression in tomato caused improved fruit quality attributes (Liu et al., 2014; Zhu et al., 2015). We observed that SlBRI1 overexpression resulted in an increase in soluble solids, soluble sugars, and ascorbic acid levels, as well as the fruit firmness, during tomato fruit ripening (**Figure 9**), indicating that SlBRI1 overexpression mediated increases in BR signaling can promote fruit ripening and quality.

#### REFERENCES


#### CONCLUSION

SlBRI1 overexpression increased the endogenous BR signaling intensity and improved fruit carotenoid accumulation and quality attributes in transgenic tomato. The increased BR signaling either directly up-regulated the expression of key ethylene biosynthetic genes to promote fruit ripening, or directly up-regulated the expression of carotenoid biosynthetic genes and ripening-related genes. Taken together, our results indicate that endogenous BR signaling plays an important role in regulating growth and fruit nutrient quality in tomato. Therefore, these results underline the potential for improving growth and fruit nutritional quality by genetically manipulating SlBRI1 expression levels which was feasible in tomato. In addition, the constitutive activation of BR signaling enhanced freezing resistance by regulating cold responsive gene expression in Arabidopsis (Eremina et al., 2016). Further study will confirm increased BR signaling intensity whether can also improve freezing resistance and other stress response in tomato.

#### AUTHOR CONTRIBUTIONS

SN wrote the main manuscript text. SN and XW contributed to the conception of the study. SN, SH, and SW performed the experiments. SN, SH, DC, JL, SL, and QL performed the data analyses and helped perform the analysis with constructive discussions. All the authors gave the final approval for submission of the manuscript.

## FUNDING

This research was financially supported by the National Natural Science Foundation of China (Nos. 31501771 and 31672142).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2017.01386/ 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 © 2017 Nie, Huang, Wang, Cheng, Liu, Lv, Li and Wang. 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.

fpls-08-01386 August 8, 2017 Time: 15:42 # 12

\*

# Flowering of Woody Bamboo in Tissue Culture Systems

#### Jin-Ling Yuan<sup>1</sup> , Jin-Jun Yue<sup>1</sup> , Xiao-Ping Gu<sup>1</sup> and Choun-Sea Lin<sup>2</sup>

*<sup>1</sup> Research Institute of Subtropical Forestry, Chinese Academy of Forestry, Hangzhou, China, <sup>2</sup> Agricultural Biotechnology Research Center, Academia Sinica, Taipei, Taiwan*

Flowering and subsequent seed set are not only normal activities in the life of most plants, but constitute the very reason for their existence. Woody bamboos can take a long time to flower, even over 100 years. This makes it difficult to breed bamboo, since flowering time cannot be predicted and passing through each generation takes too long. Another unique characteristic of woody bamboo is that a bamboo stand will often flower synchronously, both disrupting the supply chain within the bamboo industry and affecting local ecology. Therefore, an understanding of the mechanism that initiates bamboo flowering is important not only for biology research, but also for the bamboo industry. Induction of flowering *in vitro* is an effective way to both shorten the flowering period and control the flowering time, and has been shown for several species of bamboo. The use of controlled tissue culture systems allows investigation into the mechanism of bamboo flowering and facilitates selective breeding. Here, after a brief introduction of flowering in bamboo, we review the research on *in vitro* flowering of bamboo, including our current understanding of the effects of plant growth regulators and medium components on flower induction and how *in vitro* bamboo flowers can be used in research.

#### Edited by:

*Chi-Kuang Wen, Shanghai Institutes for Biological Sciences (CAS), China*

#### Reviewed by:

*Yulong Ding, Nanjing Forestry University, China Hong-Hwa Chen, National Cheng Kung University, Taiwan*

#### \*Correspondence:

*Choun-Sea Lin cslin99@gate.sinica.edu.tw*

#### Specialty section:

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

Received: *23 June 2017* Accepted: *30 August 2017* Published: *14 September 2017*

#### Citation:

*Yuan J-L, Yue J-J, Gu X-P and Lin C-S (2017) Flowering of Woody Bamboo in Tissue Culture Systems. Front. Plant Sci. 8:1589. doi: 10.3389/fpls.2017.01589* Keywords: flowering induction, in vitro seed set, in vitro hybridization, bamboo reproduction, plant growth regulators

#### INTRODUCTION

Flowering, fruiting, and seed development are the most fundamental processes of sexual propagation in plants. Most flowering plants pass from seed germination to a brief period as a seedling, to a vegetative or juvenile phase that is predominated by growth, and then onto a reproductive phase, during which plants have the capacity to produce the components required for flowering and seed production (Huijser and Schmid, 2011). The length of the plant juvenile phase varies widely. Usually, herbaceous plants have a short juvenile phase (within 1–2 seasons), complete their life cycle within a few years, and die after seed production (Feng et al., 2016). However, woody plants have a long juvenile phase (many years), remain alive after flowering, and can flower every year after reaching maturity (Wendling et al., 2014a,b).

Compared with these two types of plants, woody bamboos which were identified as monopodial with leptomorph rhizome (**Figures 1A,C**) and sympodial with pachymorph rhizome (**Figures 1B,D**; McClure, 1966), have a unique flowering behavior. Woody bamboos have a very long juvenile phase (decades), similar to woody plants. However, woody bamboo only flowers once and dies after seed production (monocarpy) (McClure, 1966; Janzen, 1976).

There are advantages and disadvantages to this unique flowering behavior, particularly for the bamboo industry. During the long vegetative phase, bamboo stalks (building materials) and young shoots (edible vegetable) can be continually harvested for many years. Propagation costs can be

**209**

reduced in bamboo species that grow by rhizomes, such as monopodial bamboos, yielding an entire plantation composed of plants of the same genetic background. These monoculture plantations consist of plants initiated at the same time and often flower en masse, disrupting the supply chain and causing huge economic losses. This gregarious flowering not only takes that plantation out of service (Sarma et al., 2010) but also causes ecological and enviromental challenges. For example, a bamboo grove that has recently flowered does not provide food for the giant panda (Li and Denich, 2004). Furthermore, once fruit has set, this new food source can lead to overpopulation of rats, which in the past have over-consumed the seeds, leaving the bamboo forest unrecovered (Nag, 1999).

Since it is difficult to predict flowering time and to time the flowering of two bamboo accessions for hybridization, it is difficult for breeders to use select varieties for bamboo breeding (John and Nadgauda, 1999; Singh et al., 2013). Without genetic recombination through cross-pollination, genome diversity is limited and genetic studies are nearly impossible. Furthermore, bamboo classification is debated and confusing since plant taxonomy often relies on the morphology and anatomy of flowers and fruits and speciation depends on sexual incompatibility (Bhattacharya et al., 2006, 2009).

Over the years, many researchers have tried to manipulate bamboo flowering not only for research and industrial purposes but also to manage the environmental impact. Because of the size of woody bamboo, it is difficult to establish controlled environments for scientific research. The first case of in vitro bamboo flowering and seed production was reported less than 30 years ago (Nadgauda et al., 1990). Compared with in vivo flowering, there are many advantages to in vitro flowering. Firstly, the plantlets can be incubated in a sterile, controlled environment, which can reduce interference from biotic and abiotic stresses and uncontrolled pollination. Secondly, the size of the plantlet is relatively small within the incubation container, allowing addition of plant growth regulators to the whole plantlet. Thirdly, flowering can be induced when desired. Here, we review recent reports on in vitro bamboo flowering.

#### FLOWER INDUCTION IN VITRO: SPECIES AND EXPLANT TYPES

To date, 13 bamboo species have been induced to flower in vitro (**Table 1**). Five of the species belong to the genus Bambusa: B. arundinacea (Nadgauda et al., 1990, 1997; Ansari et al., 1996; Joshi and Nadgauda, 1997), B. edulis (Lin and Chang, 1998, **Figure 1E**), B. multiplex (Prutpongse and Gavinlertvatana, 1992), B. oldhamii (Ho and Chang, 1998; Zhang and Wang, 2001), and B. vulgaris (Rout and Das, 1994). Six of the species belong to the genus Dendrocalamus: D. brandisii (Nadgauda et al., 1990), D. giganteus (Rout and Das, 1994; Ramanayake et al., 2001), D. hamiltonii (Chambers et al., 1991; Kaur et al., 2015), D. latiflorus (Zhang and Wang, 2001; Lin et al., 2006, 2007b), D. membranaceus (Prutpongse and Gavinlertvatana, 1992), and D. strictus (Rout and Das, 1994). The last two accessions are Cephalostachyum pergracile (Prutpongse and Gavinlertvatana, 1992) and an intergenus hybrid, B. pervariabilis × D. latiflorus (Zhang and Wang, 2001). Based on these reports, all the bamboo species that have been shown to flower in vitro are sympodial bamboos. There has not been a report on in vitro flowering for a monopodial bamboo. Actually, there are only few reports on monopodial bamboo tissue culture (Gielis, 1999; Wang et al., 2005; Pei et al., 2011; Mudoi et al., 2013; Yuan et al., 2013), regardless of its better cold-tolerance and other merits over sympodials. Although Hassan and Debergh (1987) originally reported tissue culture protocols for P. viridis, they retracted the article because of a taxonomy issue. Therefore, work remains to be done to develop in vitro flowering tissue culture protocols for important monopodial species.

Different species showed different responses in the same medium. In medium supplemented with 6-benzylaminopurine (BA) and coconut milk, B. arundinacea showed a 70% flowering rate, D. brandisii only 40%, and D. strictus did not flower (Nadgauda et al., 1990). D. brandisii and Dendrocalamopsis oldhamii (=B. oldhamii) did not flower in a medium that could induce flowering in B. pervariabilis × D. latiflorus (Zhang and Wang, 2001). B. edulis flowered in a medium supplemented with 0.1 mg/L thidiazuron (TDZ), but B. oldhamii only proliferated multiple shoots (Lin and Chang, 1998; Lin et al., 2007a). Those reports indicated that different species will not induce flowering in a uniform medium, and the medium components for certain bamboo must be screened through purposely designed experiments.

The time to in vitro flowering also varies between different species, across a range including 45 days (B. arundinacea, Ansari et al., 1996), just under 12 months (B. edulis, Lin and Chang, 1998), 29 months (D. giganteus, Ramanayake et al., 2001), and three years (D. latiflorus, Zhang and Wang, 2001). Although it still takes years for some of the bamboos to flower, in vitro culture nevertheless dramatically reduces bamboo flowering times compared to those in the field.

Another key factor in in vitro flowering is the choice of explant used to establish the in vitro culture and the explant selected for subsequent micropropagation. Current protocols use shoot meristems (Lin and Chang, 1998; Ramanayake et al., 2001; Lin et al., 2010) and seedlings (Nadgauda et al., 1990, 1997; Chambers et al., 1991; Ansari et al., 1996; Joshi and Nadgauda, 1997; Singh et al., 2000; Zhang and Wang, 2001). However, sourcing of seeds is not predictable, and bamboo seeds often do not have unique or widely diverse genetic backgrounds. The use of meristems from superior bamboo lines is a better strategy that will support bamboo breeding.

# FLOWER INDUCTION IN VITRO: CONTROL BY PLANT GROWTH REGULATORS

Plant growth regulators are critical to in vitro bamboo flowering. In vitro flowering of bamboo can be induced by cytokinins, as has been shown in D. brandisii (Nadgauda et al., 1990), D. giganteus (Rout and Das, 1994; Ramanayake et al., 2001), D. hamiltonii (Chambers et al., 1991), D. latiflorus (Zhang and Wang, 2001; Lin

et al., 2007b), D. strictus (Rout and Das, 1994; Singh et al., 2000), B. arundinacea (Nadgauda et al., 1990, 1997; Joshi and Nadgauda, 1997), B. edulis (Lin and Chang, 1998; Lin C. C. et al., 2003), B. multiplex (Prutpongse and Gavinlertvatana, 1992), and B. vulgaris (Rout and Das, 1994, **Table 1**). The effects of cytokinins on in vitro bamboo flowering are species dependent. For example, kinetin (Kin) could not induce flowering in B. arundinacea (Joshi and Nadgauda, 1997) or D. latiflorus (Zhang and Wang, 2001), but could for B. edulis plantlets with multiple shoots (Lin C. C. et al., 2003). Similar positive results were observed with zeatin (ZT) treatment of B. arundinacea and B. edulis (Joshi and Nadgauda, 1997; Lin C. S. et al., 2003). In B. arundinacea, flowering only occurred in medium containing BA combined with either ZT, adenine sulfate (Ads), Kin, or isopentyl adenine (2iP), but not those containing only one of the listed cytokinins without BA (Joshi and Nadgauda, 1997). D. strictus could not flower with 5 mg/L BA alone (Nadgauda et al., 1990), but a combination of cytokinin (Ads), auxin [Indole-3-butyric acid (IBA)] and Gibberellic acid (GA3) could induce flowering and seed formation (Rout and Das, 1994). In B. edulis, cytokinins are important not only for flower induction but also for inflorescence proliferation (Lin C. S. et al., 2003). The inflorescences could multiply when treated with different kinds of cytokinins, such as BA (Lin et al., 2004b). According to these results, cytokinins play positive roles in bamboo flowering.

Interestingly, auxins play an opposite role in bamboo flowering. In medium containing 0.1 mg/L TDZ, flowering of B. edulis plantlets with multiple shoots was inhibited by naphthaleneacetic acid (NAA) (Lin C. S. et al., 2003). When using in vitro inflorescences as explants, auxin-only medium increased the floret size, and also induced adventitious roots and caused 35% more vegetative shoots to emerge. These rooted vegetative plantlets could be transplanted to the greenhouse and survive (Lin et al., 2005). These results indicated that auxin plays a negative role in bamboo flowering and inflorescence proliferation in vitro.

Other plant growth regulators and medium components have also been investigated, such as the ethylene precursor 1 amino-cycliopropane-1-carboxylic acid (ACC), acetic salicylic acid, gibberrellin, the gibberrellin synthesis inhibitor ancymidol (Lin, 1998), coconut water (Zhang and Wang, 2001), sucrose, nitrogen at various concentrations (Lin C. C. et al., 2003), and the pH of the medium (Joshi and Nadgauda, 1997). These treatments led to only slight effects on flower induction.

#### TABLE 1 | *In vitro* flowering of bamboo species: explant types, medium components, and results.


*(Continued)*

TABLE 1 | Continued


*MS, Murashige and Skoog medium; 2,4-D, 2,4-dichlorophenoxyacetic acid; BPA, N-Benzyl-9-(2-tetrahydropyranyl) adenine; CW, coconut water.*

### FERTILITY OF IN VITRO-INDUCED FLOWERS

Seeds could be obtained from in vitro flowers of B. arundinacea, D. brandisii, B. vulgaris, D. giganteus, and D. strictus (Nadgauda et al., 1990, 1997; Rout and Das, 1994). While D. strictus could produce fertile pollen grains (Singh et al., 2000), in vitro anthers of B. edulis could not (Lin and Chang, 1998; Lin C. S. et al., 2003; Lin et al., 2004a). In B. edulis, the effects of different plant growth regulators on fertility were analyzed. Although auxin treatments promoted anther emergeance outside of glumes, no fertile pollen or seeds were obtained (Lin et al., 2004b). During normal in vivo flowering, D. strictus and B. multiplex have good fertility and easily produce seeds (Nadgauda et al., 1993; Yuan et al., 2011), but there is no report of seed set in Bambusa edulis, reflective of the in vitro results. Therefore, we speculate that the differential fertility in vitro may be related to genetic characteristics of the bamboo species. There is evidence that B. edulis is an intergenus hybrid between Bambusa and Dendrocalamus (Ye, 2010; Zheng, 2014), meaning that B. edulis cannot produce gametes with the correct chromosome number for seed set. Due to its long juvenility, it is difficult to conduct cytogentics in bamboo using reproductive organs, such as anthers. Therefore, most karyotyping has been conducted using root tips (Chen R. Y., 2003), although these experiments may have resulted in unreliable chromosome counts in bamboo.

### APPLICATIONS OF BAMBOO IN VITRO FLOWERING—CLONING OF FLOWER-RELATED GENES

Bamboo flowers produced in vitro provide an important material for flower-related molecular and cell biology studies. D. latiflorus spikelets have been used to identify numerous full-length cDNAs of the flowering-related MADS genes (Chen Y. Y., 2003). From a B. oldhamii cDNA library, 4,470 (floral tissue) and 3,878 (vegetative tissue) ESTs were published (Lin et al., 2010). Using proteomic analysis of bamboo flowers, 128 differentially expressed proteins in floral meristems were identified (Kaur et al., 2015). To do such studies on gene and protein expression in floral organs, flowers must be readily available in sufficient quantity.

With next generation sequencing, it has become easier to investigate non-model plant transcriptomes. One such transcriptome that has been explored is that of the in vitroproduced flowers of B. edulis. Using this transcriptome and sequences from a bacterial artificial chromosome (BAC) library, 16 full-length Type II MADS (BeMADS) genes were identified. The gene structures and amino acid sequences were highly similar to rice MADS homologs (77–92%). Most importantly, all of the predicted proteins contain M, I, K, and C domains, definitive of type II MADS (Shih et al., 2014). When the whole genome of moso bamboo was published (Peng et al., 2013), 34 MADS genes were identified (Peng et al., 2013; Cheng et al., 2017). However, the protein lengths and exon numbers were unlike the other Poaceae MADS. Five genes did not have the M domain (PheMADS56-4, PheMADS21, PheMADS14, PheMADS29, and PheMADS90; Cheng et al., 2017), while others were very short and contained only the M domain (PheMADS1, PheMADS5, PheMADS64, PheMADS65; Cheng et al., 2017). Thus far it is unknown whether these differences are due to the starting materials (DNA from in vivo flowers in Cheng et al., 2017 vs. RNA from in vitro flowers in Shih et al., 2014).

Because MADS proteins are transcription factors, they will form complexes that go to the nucleus. However, most fluorescently tagged BeMADS proteins cannot enter the nucleus when expressed in either Arabidopsis protoplasts or in bamboo leaves, but can when expressed in lemmas (Shih et al., 2014). This indicated that correct results can only be shown in the correct materials. Therefore, in vitro bamboo flowers are very important for investigations into bamboo reproduction.

#### MOVING FORWARD

Due to the flowering characteristics unique to bamboo (long juvenile phase, mass flowering, and death after flowering), establishment of controllable in vitro bamboo flowering is absolutely required to facilitate timely and effective bamboo breeding. While only self-crosses have thus far successfully produced seeds in vitro (Nadgauda et al., 1990), advances in technology, new induction protocols, or alternative hybridization strategies can further the realization of this goal. For example, D. latiflorus and B. edulis plantlets induced to flower in vitro were successfully transferred to the greenhouse, where they continued flowering (Zhang and Wang, 2001; Lin C. C. et al., 2003; Lin et al., 2005). Perhaps parental bamboo accessions could be induced to flower in vitro and transplanted to the greenhouse for further hybridization with other bamboos that are flowering, whether they were induced in vitro or in vivo. This transplantion strategy avoids the limitations of in vitro hybridization, such as high humidity or low wind- or insect-mediated pollination rates. In vitro flowers can also be maintained in tissue culture to preserve those flowering bamboos that cannot survive in the field. Compared with bamboo vegetative tissues, it is easier to establish bamboo reproductive tissues in a tissue culture system (Lin and Chang, 1998).

Furthermore, our study of bamboo flowering indicated that the use of standard model plant material (ex. Arabidopsis) gives

#### REFERENCES


misleading results for bamboo (Shih et al., 2014). While many bamboo flower-related genes have been identified via genomics, the mechanisms of flowering, the expression of floral genes and proteins, and other functional analyses must be done in bamboo reproductive tissues. Stable and readily available sources of in vitro reproductive tissues offer many advantages for further experiments such as genetic transformation.

As the situation stands today, in vitro flowering in bamboo is limited to sympodial bamboos, and only B. edulis has thus far been investigated systematically (Lin and Chang, 1998; Lin C. C. et al., 2003). This is a challenge, but our research community hopes to apply the knowledge and techniques reviewed above to further develop tissue culture and in vitro flowering protocols for monopodial bamboos, especially for moso bamboo, which has a longer juvenile phase and is the most important monopodial bamboo species for the bamboo industry in subtropical and temperate regions. Furthermore, the work outlined above represents the current state from which researchers can refine floral induction protocols to predictably induce fertile in vitro bamboo flowers.

#### AUTHOR CONTRIBUTIONS

CSL organized and prepared this manuscript. JLY, JJY, CSL, and XPG contributed to the writing.

### FUNDING

This work was supported by the Natural Science Foundation of China (31500551); the Fundamental Research Funds for the Central Non-profit Research Institute of Chinese Academy of Forestry (CAFYBB2016QB008); Zhejiang Science and Technology Major Program on Agricultural New Variety Breeding (2016C02056-8); and the Fundamental Research Funds for the Central Non-profit Research Institute of Subtropical Forestry-Chinese Academy of Forestry (RISF2014001) to JLY. This work was supported by Ministry of Science and Technology, Taiwan (105-2313-B-001-007-MY3), and Academia Sinica, Taiwan, to CSL.

#### ACKNOWLEDGMENTS

We thank Anita K. Snyder and Miranda Loney for English editing.

subsp. spathiflorus at population level. Plant Syst. Evol. 282, 13–20. doi: 10.1007/s00606-008-0092-1


Dissertation/doctor's, Kunming Institute of Botany; Chinese Academy of Science, Kunming.


Zheng, C., H. (2014). Speciation of Bambusa-Dendrocalamus-Gigantochloa (BDG) complex—Hybridization revealed by Molecular Evidence. Dissertation/doctor's, Chinese Academy of Science, Beijing.

**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 Yuan, Yue, Gu and Lin. 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.

# Two FERONIA-Like Receptor Kinases Regulate Apple Fruit Ripening by Modulating Ethylene Production

Meiru Jia† , Ping Du† , Ning Ding, Qing Zhang, Sinian Xing, Lingzhi Wei, Yaoyao Zhao, Wenwen Mao, Jizheng Li, Bingbing Li\* and Wensuo Jia\*

*College of Horticulture, China Agricultural University, Beijing, China*

Ethylene has long been known to be a critical signal controlling the ripening of climacteric fruits; however, the signaling mechanism underlying ethylene production during fruit development is unknown. Here, we report that two FERONIA-like receptor kinases (FERLs) regulate fruit ripening by modulating ethylene production in the climacteric fruit, apple (*Malus*×*domestica*). Bioinformatic analysis indicated that the apple genome contains 14 members of the FER family (*MdFERL1–17*), of these 17 FERLs, *MdFERL6* was expressed at the highest level in fruit. Heterologous expression of *MdFERL6* or *MdFERL1*, the apple homolog of Arabidopsis *FER*, in another climacteric fruit, tomato (*Solanum lycopersicum*) fruit delayed ripening and suppressed ethylene production. Overexpression and antisense expression of *MdFERL6* in apple fruit calli inhibited and promoted ethylene production, respectively. Additionally, virus-induced gene silencing (VIGS) of *SlFERL1,* the tomato homolog of *FER,* promoted tomato fruit ripening and ethylene production. Both MdFERL6 and MdFERL1 physically interacted with MdSAMS (S-adenosylmethionine synthase), a key enzyme in the ethylene biosynthesis pathway. *MdFERL6* was expressed at high levels during early fruit development, but dramatically declined when fruit ripening commenced, implying that MdFERL6 might limit ethylene production prior to fruit development and the ethylene production burst during fruit ripening. These results indicate that FERLs regulate apple and tomato fruit ripening, shedding light on the molecular mechanisms underlying ripening in climacteric fruit.

#### Keywords: apple, ethylene production, fruit ripening, FERONIA-like receptor kinase, tomato, VIGS

#### INTRODUCTION

Fleshy fruits are physiologically classified as climacteric or non-climacteric. Climacteric fruits exhibit an increase in respiration at the onset of ripening (Nitsch, 1953; Coombe, 1976; Brady, 1987). Studies of the mechanisms regulating fruit ripening began in the 1920s (Brady, 1987) and a major focus has been identifying the critical internal factors or signals governing this process. Ethylene has long been known to be a critical signal controlling the ripening of climacteric fruit (Biale, 1964; Burg and Burg, 1965; Alexander and Grierson, 2002), which exhibit a large increase in ethylene production at the onset of ripening. Exposure to exogenous ethylene can initiate the ripening of climacteric fruits (Seymour et al., 2013), and its effect is so great that limiting ethylene production in fruits or ethylene exposure for harvested fruit has become a major concern in the commercial cultivation industry (Brady, 1987).

#### Edited by:

*Yong-Ling Ruan, University of Newcastle, Australia*

#### Reviewed by:

*Hiroshi Ezura, University of Tsukuba, Japan Benedetto Ruperti, University of Padua, Italy Shaohua Zeng, South China Institute of Botany (CAS), China*

#### \*Correspondence:

*Bingbing Li libingbing@cau.edu.cn Wensuo Jia jiaws@cau.edu.cn † These authors have contributed equally to this work.*

#### Specialty section:

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

Received: *13 March 2017* Accepted: *28 July 2017* Published: *10 August 2017*

#### Citation:

*Jia M, Du P, Ding N, Zhang Q, Xing S, Wei L, Zhao Y, Mao W, Li J, Li B and Jia W (2017) Two FERONIA-Like Receptor Kinases Regulate Apple Fruit Ripening by Modulating Ethylene Production. Front. Plant Sci. 8:1406. doi: 10.3389/fpls.2017.01406*

**217**

The ripening of fleshy fruits is a complex process that involves dramatic changes in physiological and biochemical metabolism (Giovannoni, 2001; Seymour et al., 2013). While early studies focused on these changing patterns, more recent research has aimed to establish the molecular basis for the regulation of fruit development and ripening (Seymour et al., 2013). Several genes associated with the regulation of fruit development and ripening have been identified, but as most encode enzymes or transcription factors, little is known about the signaling mechanisms that underlie these processes to control the changes in physiological and biochemical metabolism (Fischer and Bennett, 1991; Giovannoni, 2001; Klee and Giovannoni, 2011; Seymour et al., 2013).

Most studies of fruit development-associated signaling have focused on the molecular perception of and response to ethylene in the model plant, tomato (Solanum lycopersicum; Felix et al., 1991; Wang et al., 2002; Bram et al., 2015). The mechanisms controlling the increase in ethylene production are essential for its function as a signal regulating fruit development and ripening. Ethylene production is primarily controlled by the activities of the key enzymes in its biosynthesis pathway. In the context of cellular signaling, enzyme activity may be regulated at the transcriptional or post-transcriptional level by upstream signaling cascades. While ethylene biosynthesis and its downstream signaling pathways have been studied extensively, little is known about the signaling cascades upstream of ethylene biosynthesis. However, it was shown that ethylene biosynthesis is induced by both cell wall fragments and wall-lysing enzymes. Wall fragments were implicated as possible regulators of ethylene biosynthesis in ripening citrus fruit (Citrus sp.; Baldwin and Biggs, 1988), tomato fruit (Baldwin and Pressey, 1988; Brecht and Huber, 1988), and cultured pear (Pyrus sp.) cells (Tong et al., 1986), and the wall-lysing enzymes polygalacturonase and endoxylanase were reported to induce ethylene biosynthesis in tomato fruit (Baldwin and Pressey, 1988) and tobacco (Nicotiana tabacum) leaves (Fuchs et al., 1989), respectively. Although the precise mechanism by which cell wall degradation induces ethylene biosynthesis is not known, these related studies suggest that receptors or sensors that initiate the signaling cascade upstream of ethylene biosynthesis exist in the membrane–wall complex.

Plant receptor-like kinases (RLKs) are transmembrane proteins with putative amino-terminal extracellular domains and carboxyl-terminal intracellular kinase domains (Shiu and Bleecker, 2001a; Humphrey et al., 2007; Boisson-Dernier et al., 2011; Cheung and Wu, 2011; Lindner et al., 2012). The Arabidopsis thaliana genome contains more than 600 RLK homologs with diverse functions (Shiu and Bleecker, 2001a,b, 2003). We hypothesize that RLKs function as membrane–wall complex-anchored receptors that regulate ethylene biosynthesis during fruit development and ripening. The RLK superfamily contains a subfamily with extracellular malectin or malectinlike domains that putatively bind oligosaccharides (Verica and He, 2002; Annabelle et al., 2006; Alexandre et al., 2010; Schallus et al., 2010). Oligosaccharides have increasingly been demonstrated to function as signals that activate either defensive or developmental processes in plants (Clarence and Farmer, 1991; Schallus et al., 2008), and therefore elucidating whether malectin or maletin domain-containing RLKs are implicated in the regulation of fruit development and ripening is a key research target. In a previous study, we found that the strawberry genome (Fragaria vesca) contains 62 RLK members (designated FvMRLKs) that contain malectin or malectin-like domains, and a preliminary analysis suggested that more than half of the FvMRLKs could be implicated in strawberry fruit development and ripening (Zhang et al., 2016).

Recently, several malectin domain-containing RLKs have been demonstrated to play important roles in Arabidopsis growth and development (Nissen et al., 2016); among these genes, FERONIA (FER) is of particular interest because of its diverse roles in the regulation of a series of crucial biological processes. FER was first identified for its role in fertilization (Huck et al., 2003). In Arabidopsis, FER directly interacts with Rho of plants guanine nucleotide exchange factors (RopGEFs), activating the ROP GTPase and mediating the production of reactive oxygen species (ROS) at the entrance of the female gametophyte, thereby inducing pollen tube rupture and sperm release (Kessler et al., 2010; Duan et al., 2014; Ngo et al., 2014). FER is believed to function in a variety of other important processes, such as root hair elongation (Duan et al., 2010; Huang et al., 2013), starch accumulation (Yang et al., 2015), sugar utilization (Pu et al., 2017), seed development (Yu et al., 2014), pathogen resistance (Keinath et al., 2010; Kessler et al., 2010), and vegetative growth (Guo et al., 2009; Deslauriers and Larsen, 2010). Notably, FER has been demonstrated to play a pivotal role in the cross-talk signaling among the phytohormones that play crucial roles in Arabidopsis, including abscisic acid (ABA; Yu et al., 2012; Chen et al., 2016), auxin (Duan et al., 2010), brassinosteroids (BR), and ethylene (Guo et al., 2009; Deslauriers and Larsen, 2010). Importantly, a recent study by Mao et al. (2015) reported that FER was able to physically interact with S-adenosylmethionine synthetase (SAMS) and thereby suppress ethylene biosynthesis in Arabidopsis.

Given the involvement of FER in a variety of important processes, especially in the cross-talk between phytohormones and in ethylene biosynthesis in Arabidopsis, we were particularly interested to determining whether some FER-like RLKs regulate fruit development and ripening. Apple (Malus × domestica) is one of the most popular and important fruit crops worldwide; however, due to difficulties in molecular research using apple fruit, basic research into fruit development and ripening has commonly been conducted using tomato as a model plant. In the present study, we searched the apple genome database and identified 14 FER-like RLKs. We demonstrated that two of these proteins are important regulators of fruit development and ripening. Interestingly, two of the FER-like RLKs were found to be regulators of ethylene production in fruits. This work sheds light on the mechanisms underlying climacteric fruit development and ripening, by suggesting that Feronia-like RLKs represent upstream signaling components that regulate ethylene production during ripening.

# MATERIALS AND METHODS

# Plant Materials and Growth Conditions

Tomatoes (Solanum lycopersicum L. cv. Micro Tom) were planted in pots (diameter, 20 cm; depth, 20 cm) containing a mixture of nutrient soil, vermiculite, and organic fertilizer (4:2:1, v/v/v). Tomato seedlings were grown under standard conditions: 60% relative humidity and 25◦C/18◦C (day/night) under a 12 h/12 h light/dark cycle. Plants were watered daily to the drip point. Apple (Malus × domestica) callus tissue derived from the "Golden Delicious" cultivar was grown on Murashige and Skoog (MS) medium at 27◦C ± 1 ◦C in darkness, and subcultured at 10-day intervals before being used for gene transformation.

# Gene Isolation and Sequence Analysis

The cDNA sequences of full-length Arabidopsis CrRLK1Ls were obtained from TAIR (http://www.arabidopsis.org). To identify FER-like (FERL) genes in apple or rice, the coding sequence of FER (At3g51550) was used as a query to BLAST the apple genome database (http://genomics.research.iasma.it/) or NCBI (https:// www.ncbi.nlm.nih.gov/). Phylogenetic trees were constructed using the Neighbor-Joining (NJ) method in MEGA 4.0.2 software (Tamura et al., 2007), with 1000 bootstrap replicates performed to evaluate the reliability of the different phylogenetic groups. The deduced amino acid sequences of the MdFERLs were aligned using ClustalX 2.0.12 (Larkin et al., 2007) with default settings. The alignments were edited and marked using GeneDoc (Nicholas et al., 1997).

To isolate MdFERL1 and MdFERL6, total RNA was extracted from the fruit using an EZNATotal RNA Kit (Omega Biotek), according to the manufacturer's instructions. First-strand cDNA synthesis was conducted from 1 µg of total RNA. Full-length MdFERL1 and MdFERL6 were cloned from cDNA by PCR using Q5 High-Fidelity DNA Polymerase (New England Biolabs) and the following conditions: 98◦C for 30 s (one cycle); 98◦C for 30 s, 59◦C for 25 s, and 72◦C for 5 min (35 cycles); with a final extension of 72◦C for 2 min. The PCR products were subcloned into the pCambia1304 vector and used to produce positive colonies. The primer sequences and GenBank accession numbers are shown in Supplemental Table S1.

# Quantitative Reverse Transcription PCR (RT-qPCR)

RT-qPCR was performed using SYBR Premix Ex TaqTM (Takara Bio) in a 7500 Real-Time PCR System (Applied Biosystems). The expression level of each gene was analyzed in three pools of five fruits, with three measurement replicates per pool. ACTIN was used as an internal control. The 2−11CT method was used to determine transcription levels, where 1CT represents the difference between the cycle threshold values of the target and reference genes (Schmittgen and Livak, 2008).

# Spatio-Temporal Expression of MdFERLs and Ripening Marker Genes

The whole process from fruit set to ripening was classified into five stages based on days post-anthesis (DPA), namely 60DPA, 85DPA, 105DPA, 130DPA, and 155DPA. The fruit was frozen in liquid nitrogen, and kept at –80◦C until required for gene expression analysis. The expression of MdFERL and ripening marker genes was assessed by RT-qPCR analysis, using primers designed in Primer3 Plus (http://www.primer3plus.com/ cgi-bin/dev/primer3plus.cgi). Primer sequences are shown in Supplemental Tables S2, S3, S5. The data from five fruits were combined as an individual sample and each sample was analyzed in triplicate.

# Analysis of MdFERL Expression Changes in Response to 1-MCP and ACC Treatment

For the apple fruit treatment, 5 g of fruit disks (10 mm in diameter and 1 mm in thickness) was prepared and combined from six fruits at the 105 DPA stage. The disk samples were vacuum infiltrated for 30 min in equilibration buffer (10 mM MgCl2, 50 mM MES-Tris (pH 5.5), 5 mM CaCl2, 10 mM EDTA, 200 mM mannitol, and 5 mM vitamin C) (Archbold, 1988). The samples were then shaken for 6 h at 25◦C in equilibration buffer containing 1 mM ACC or 3 µM 1-MCP. After incubation in the dark for 6 h, the samples were washed and frozen immediately in liquid nitrogen, and maintained at –80◦C until required. Each individual analysis was conducted with three sample replicates. Primers used for the RT-qPCR analysis of ripening-related genes are provided in Supplemental Table S2.

To treat apple calli with 1-MCP, the calli were transfected with MdFERL6-AS and empty pCambia1304 vector, and then cultivated on MS medium at 27◦C in darkness. Three days after the transfection, calli were treated with 3 µM 1-MCP for 6 h and expression of the selected genes was analyzed as described above.

# Transfection of Tomato Fruit and Apple Callus by Agroinfiltration

To construct vectors for the overexpression (OE) of MdFERL1 (MDP0000445374) and MdFERL6 (MDP0000465341) (abbreviated hereafter as MdFERL1-OE and MdFERL6-OE), full-length MdFERL1 and MdFERL6 were respectively cloned into the plant expression vector pCambia1304 using the KpnI and EcoRI restriction sites. The empty pCambia1304 vector and the overexpression vectors MdFERL1-OE and MdFERL6- OE were transformed individually into the Agrobacterium tumefaciens strain EHA105 (Lazo et al., 1991). For the MdFERL6-antisense (as) vector, a 621-bp fragment near the 5′ end of MdFERL6 cDNA was amplified by PCR using the primer pair 5′ -GAATTCGTCCACGAACGTGTCTC -3′ (with an EcoRI restriction site; forward) and 5′ - GGTACCGGATCGTTGATCTCAGAG -3′ (with a KpnI restriction site; reverse). The obtained fragment was sequenced and forward-cloned into the EcoRI and KpnI restriction sites of pCambia1304. To generate the TRV-VIGS (virus-induced gene silencing) vectors, a 595-bp fragment of SlFERL1 (GenBank accession number KY435602) was PCR-amplified from tomato cDNA. The resulting product was cloned into the vector pTRV2 to generate pTRV2-SlFERL1. The A. tumefaciens strain GV3101 containing vectors pTRV1 or pTRV2 and their derivatives was used for the VIGS experiments.

The transformed strains were grown at 28◦C in Luria-Bertani liquid medium containing 10 mM MES, 20 µM acetosyringone, and the appropriate antibiotics. When the culture reached an optical density at 600 nm of approximately 0.8, the A. tumefaciens cells were harvested, resuspended in infection buffer (10 mM MgCl2, 10 mM MES (pH 5.6), and 200 mM acetosyringone), and shaken for 2 h at room temperature before being used for infiltration. For VIGS infiltration, each A. tumefaciens strain containing pTRV1 and pTRV2 or their derivative vectors was mixed at a ratio of 1:1 and infiltrated into the carpopodium of tomato fruits attached to the plant about 18 days after pollination, using a 1-mL needle-less syringe. Pairs of fruit at the same developmental stage and with similar phenotypes were selected; for each pair, one fruit was transfected with MdFERL1-OE, MdFERL6-OE, or SlFERL1-VIGS, while the other was transfected with an empty vector as the control. For each gene, 25 pairs of fruit were injected. To examine the effect of MdFERL1- OE, MdFERL6-OE, and SlFERL1-VIGS on the expression of ripening-related genes, the fruits were collected 10 dayd after transfection for MdFERL1-OE and MdFERL6-OE, and 12 days after transfection for SlFERL1-VIGS.

Callus tissues were subcultured three times at 15-day intervals before being used for gene transformation. The fresh calli were soaked for 20 min in an A. tumefaciens solution containing the MdFERL6-OE or MdFERL6-AS constructs, or the empty vector. The calluses were co-cultured on MS medium at 25◦C.

### Determination of Fruit Ripening-Associated Physiological Parameters

Flesh firmness was measured after skin had been removed on opposite sides of the fruit using a GY-4 fruit hardness tester (Zhejiang Top Instrument). The contents of fruit pigment, flavonoid, and total phenol were determined as described (Fuleki and Francis, 1968; Lees and Francis, 1971). The soluble sugar content was determined as described by Jia et al. (2011). Titratable acidity was calculated as malic acid. The total titratable acidity was determined using the acid-base titration method (Kafkas et al., 2007). Fruit aroma production was characterized by performing a headspace solid-phase microextraction and gas chromatography-mass spectrometry, as described by Dong et al. (2013).

#### Measurements of Ethylene Production

To measure ethylene production, intact fruits were enclosed in a gas-tight container (0.86 L, 24◦C) equipped with septa, and 1 mL of headspace gas was sampled using a gas syringe. The ethylene concentration was measured using a gas chromatograph (GC-17A, Shimadzu). Six fruits were assessed as one sample, and three replicates were performed.

#### Bimolecular Fluorescence Complementation (BiFC) and Subcellular Localization Assays

For BiFC assays, the full-length coding sequences of MdFERL1, MdFERL6, MdSAMS, MdETR2, MdETR5, MdACS3, and MdACS5 were amplified and individually cloned into pCambia1300-YFP(n/c) vectors. To determine the subcellular localization of MdFERL6, the full-length cDNA sequence was amplified and fused to green fluorescent protein (GFP) in the pMDC83 plant expression vector. A coexpression analysis was conducted in tobacco (Nicotiana tabacum) leaves as described by Schütze et al. (2009). Fluorescence was observed 3 days after transformation using a confocal laser-scanning microscope (Fluoview FV1000, Olympus). Maize (Zea mays) protoplast isolation and transformation were carried out according to a protocol described by Sheen et al. (1995), with minor modifications (Li et al., 2012). The primers used are listed in Supplemental Table S4.

### Co-Immunoprecipitation (CO-IP) Assay

For CoIP assays, the vector pMDC83:His-MdFERL6-GFP was used to generate the fusion protein His-MdFERL6-GFP, and the vector pCambia1300:Myc-MdSAMS-YFPn was used to generate the fusion protein Myc-MdSAMS-YFPn. The primers used are listed in Supplemental Table S4. His-MdFERL6-GFP and Myc-MdSAMS-YFPn were co-transformed into apple calli, using cotransformation of His-MdFERL6-GFP and pCambia1300-Myc-YFPn or Myc-MdSAMS-YFPn and pMDC83-GFP as control. The total proteins were extracted 3 days after transformation using extraction buffer (25 mM Tris–HCl, 1 mM EDTA, 10% glycerin, 0.5% Triton X-100, and 1 mM DTT). Protein A+G-Sepharose beads (100 µL, ComWin Biotech) were incubated with 10 µL polyclonal anti-Myc or anti-His antibody (ComWin Biotech) for 2 h at 4◦C. An equal amount of anti-Myc antibody-coupled Protein A+G-Sepharose beads was added to total protein samples expressing Myc-MdSAMS-YFPn/His-MdFERL6-GFP and Myc-MdSAMS-YFPn/pMDC83-GFP, and detected with anti-His antibody. An equal amount of anti-His antibody coupled Protein A+G-Sepharose beads was added to total protein samples expressing Myc-MdSAMS-YFPn/His-MdFERL6-GFP and His-MdFERL6-GFP/ pCambia1300-Myc-YFPn and detected with anti-Myc antibody.

#### Ethionine Sensitivity Assay

Apple calli were transfected with MdFERL6-OE, MdFERL6-AS, and empty vector, and then cultivated on MS medium containing 0, 100, and 200 µM L-ethionine. The growth status of calli was used to reflect the activity of SAM synthase. After a 7 days incubation at 25◦C, the weights of the calli were measured and MdFERL6 and MdSAMS expression were determined by qRT-PCR.

#### Statistical Analysis

Samples were analyzed in triplicate, and the data were noted as the mean + SD. Data were analyzed by Student's t-test using SAS software (version 8.1, USA), and the least significant difference at a 0.05 level of probability was used to explore the effect of P input on parameters. P < 0.05 was considered as significantly different, P < 0.01 was considered as highly significantly different. The statistical significance of differences between samples in **Figure 6C** was tested with ANOVA using the SAS software package 8.02 and the least significant difference (LSD) was used to compare the means.

# RESULTS

# Genome-Wide Identification of Apple FER-Like Genes

The FER-like receptor kinases (FERLs) belong to the CrRLK1L family. To identify FER-like genes in apple, we first conducted a genome-wide screen for members of the MdCrRLK1L family. The MdCrRLK1L family consists of 34 members, which can be divided into six clades (**Figure 1**). A comparison of the MdCrRLK1L family with the AtCrRLK1Ls of Arabidopsis thaliana and the OsCrRLK1Ls of rice (Oryza sativa) indicated that the functionally characterized representative members of AtCrRLK1L, i.e., FERONIA, NAXUR, HERK, and THESEUS, were distributed in different clades, with FER being located in clade II. Interestingly, all members of clade I were apple CrRLK1Ls, suggesting that this group of CrRLK1Ls could be specific to the apple genome. Compared to other genes, the proteins encoded by members of clades I and clade II showed a relatively higher amino acid sequence identity with FER, and hence were designated as FERLs (FERONIA-Like receptor kinase). There were 17 MdFERL members, designated

as MdFERL1–17. Clades I and II contained 12 and 5 MdFERLs, respectively. When the MdFERLs were aligned, all members were found to have an extracellular domain containing a malectin domain, which is predicted to bind carbohydrates.

# Expression of MdFERLs and Ripening-Related Marker Genes during Apple Fruit Development and Ripening

The expression patterns of FERLs during fruit development and ripening are commonly correlated with their corresponding functions. To investigate the potential roles of MdFERLs in apple, we first examined their expression patterns in relation to fruit development and ripening. Apple fruit development and ripening occur between anthesis and the time when fruits are ready to be harvested (i.e., during a period of about 160 days) (**Figure 2A**). A preliminary examination by RT-PCR detected only six MdFERL members with a relatively high transcript level in the fruits (Supplemental Figure S1A); therefore, these six genes were further analyzed by RT-qPCR. Two of these genes, MdFERL1 and MdFERL6, were expressed more than 10- to 1,000-fold higher than the others. Strikingly, unlike the expression patterns for the other MdFERL members examined, the transcript level of MdFERL6 was high at the early stages of fruit development, but quickly declined after about 105 DPA (**Figure 2B**). It has been reported that ethylene production in "golden" apple fruit peaks at around 105 DAF, then remains high during the fruit development and ripening stages (Li et al., 2015). Accordingly, we found that most ethylene biosynthesis and signaling transduction genes, such as MdACO1, MdACO2, MdACS1, MdACS3, MdERF2, and MdERF3, were significantly upregulated at around 105 DPA and remained high from then onwards (**Figure 2C**). Comprehensive analysis of the unique expression pattern of MdFERL6 and of ethylene evolution in apple fruit, suggests that MdFERL6 regulates ethylene production in apple fruit.

To further characterize the onset of ripening in apple fruit, we evaluated the expression of ripening-related genes such as key transcription factor genes and structural genes involved in pigment accumulation, starch degradation, sugar metabolism, acid metabolism, and fruit softening (**Figure 2C**). Most ripeningrelated genes examined, including the softening genes MdBG, MdEXP1, and MdXYL1, pigment metabolism genes MdMYB10, MdCHS, MdDFR, and MdUFGT, and the sugar metabolism gene MdSPS1, were strongly up-regulated after 105 DPA. Starch biosynthesis genes did not follow this pattern, and were downregulated at this time point. Collectively, these results suggest that MdFERL6 is a regulator of ethylene-mediated fruit ripening.

#### Expression of MdFERLs in Response to 1-MCP and ACC Treatment

To further explore the relationship between FERLs and ethylene in apple fruit, we examined the effects of 1-MCP, an inhibitor of ethylene perception, and ACC (1-aminocyclopropanecarboxylic acid), a precursor of ethylene, on the expression of the six MdFERLs evaluated earlier. As shown in **Figure 3**, MdFERL1, MdFERL6, MdFERL8, and MdFERL12 were upregulated by 1-MCP and suppressed by ACC treatment; however, MdFERL7 could only be induced by 1-MCP and MdFERL14 was only suppressed by ACC (**Figure 3**).

### Effect of MdFERL1 and MdFERL6 Manipulation on Physiological Parameters and Molecular Events Associated with Fruit Ripening and Quality

As mentioned above, MdFERL1 and MdFERL6 have the highest transcript levels in apple fruit among the FERLs examined. MdFERL1 has the highest level of amino acid identity with Arabidopsis FER, while MdFERL6 belongs to the group of CrRLK1Ls that appears to be specific to the apple genome. We thus examined the potential roles of MdFERL1 and MdFERL6 in fruit ripening and development. Our attempt to manipulate the expression of these genes in a variety of apple species under different conditions failed. However, methods of transgenic manipulation are well established in tomato; therefore, we heterologously expressed MdFERL1 and MdFERL6 in tomato fruit. Expressing either MdFERL1 or MdFERL6 in tomato fruit delayed ripening in comparison with the control fruit transformed with empty vector (**Figure 4A**). Since a dramatic increase in ethylene biosynthesis is a critical signal controlling apple and tomato fruit development and ripening, we examined the effects of MdFERL1 and MdFERL6 on ethylene production. The heterologous expression of MdFERL1 or MdFERL6 both resulted in a significant reduction in ethylene production in comparison with the control (**Figure 4B**). Notably, the regulatory effect of MdFERL6 on both fruit ripening (**Figure 4A**) and ethylene production (**Figure 4B**) was much stronger than that of MdFERL1. These results, in combination with the finding that MdFERL6 was potentially specific to the apple genome and that the expression pattern of MdFERL6 was most tightly associated with fruit ripening, prompted us to select this gene for further studies using overexpression and antisense manipulation in apple fruit calli. We found that MdFERL6 overexpression greatly suppressed ethylene production, and conversely, that antisense expression greatly promoted ethylene production (**Figure 4C**).

Since the heterologous expression of MdFERL1/6 in tomato fruits in combination with the overexpression and antisense expression of MdFERL6 in apple fruit calli strongly implicated FERLs in fruit ripening and ethylene production, we investigated the effect of FERL downregulation on fruit ripening. Using VIGS, we investigated the effect of downregulating tomato FERL expression on fruit ripening. A search of the tomato genome identified a total of six FERLs, designated SlFERL1–6, which contained many relatively conserved domains (Supplemental Figure S2). SlFERL1 shared 55.3% amino acid sequence identity with FER, and could be considered a FER homolog; therefore, we selected SlFERL1 for VIGS manipulation. Unexpectedly, VIGS of SlFERL1 caused a great decrease in the transcript levels of not only SlFERL1, but also of the other five SlFERLs (**Figure 5A**). Furthermore, VIGS of SlFERL1 greatly promoted fruit ripening (**Figure 5B**) and ethylene production (**Figure 5C**).

Fruit ripening involves dramatic changes in a series of physiological parameters, such as pigment content, sugar/acid

content, aroma, flavor, and texture. The effect of FERL manipulation on fruit ripening described above is reflected by a change in pigment accumulation. To decipher the roles of FERLs in fruit development and ripening, we examined their effects on some related physiological parameters. Given that VIGS-mediated silencing of SlFERL1 downregulated the expression of several other FERL members, and importantly, that it affected fruit ripening more strongly than OE, we analyzed

various physiological parameters of the VIGS line. We found that SlFERL1-VIGS affected several major physiological parameters associated with fruit ripening and quality (**Table 1**). Strikingly, while SlFERL1-VIGS had only a small effect on fructose content, it had a major impact on both sucrose and glucose contents.

To further probe the mechanisms by which MdFERLs influence fruit development and ripening, we examined the altered expression of important fruit development genes following the manipulation of the MdFERLs. Given that ethylene production was greatly affected by the changes in the expression of the MdFERLs, we focused on genes associated with ethylene biosynthesis and signaling responses, such as ACO, ACS, ERF, E4, and E8. We also examined the expression of several marker genes of fruit ripening and quality, such as CHS, F3H, ANS, and PSY, which function in pigment metabolism, SS and SPS, which are involved in fruit sugar metabolism (fruit quality), PG, PME, XYL, and EXPs, which function in cell wall metabolism (fruit texture), and RIN, CNR, and HB1, which encode ripening-related transcription factors in tomato. Heterologous expression of MdFERL1 and MdFERL6 in tomato fruits significantly suppressed the expression of most of the genes examined, except for SPS1 and SS, which were upregulated rather than suppressed, and EXP1, which was not altered (**Figure 6A**). In accordance with these observations, VIGS of SlFERL1 resulted in an increase in the transcript levels of most of the genes examined, particularly ACS, E4, and E8, but the expression of SS and SPS1 decreased, while no change was observed for ACO1 or ACO2 (**Figure 6B**). To further reveal how MdFERL6 regulates fruit ripening and to identify its precise role in ethylene-mediated fruit ripening, we examined the expression of various genes involved in the regulation of ethylene biosynthesis, ethylene signaling transduction, and fruit ripening in apple fruit calli in which the levels of MdFERL6 had been manipulated. Overexpression of MdFERL6 caused a dramatic decrease in the expression of the ethylene biosynthesis genes MdACO1, MdACS1, MdACS5; ethylene signal transduction genes MdCTR1, MdETR2, and MdETR5; and fruit ripening-related genes MdCEL4, MdPG1, and MdXYL1 (**Figure 6C**). By contrast, antisense expression of MdFERL6 caused a great increase in the expression of these genes (**Figure 6D**). Furthermore, the expression of most of the analyzed genes was suppressed by 1-MCP treatment. Interestingly, MdACO2, MdACS5, MdETR5, MdERF1, MdCEL4, MdPG1, MdEXP1, and MdPME1 expression was less sensitive to 1-MCP treatment in apple calli harboring the MdFERL6 antisense construct (MdFERL6-AS) than in nontransgenic control calli treated with 1-MCP, while the expression of MdSS1 and MdSPS1 was more sensitive to 1-MCP treatment, suggesting that all of these genes are regulated by ethylene through MdFERL6-mediated pathways. However, compared to non-treated MdFERL6-AS calli, the expression of MdACS1, MdACS3, MdETR2, MdCTR1, MdERF2, MdXYL1, and MdEXP1 was not significantly altered by 1-MCP, suggesting that MdFERL6 also regulates fruit ripening via an ethylene-independent mechanism or that MdFERL6 functions downstream of the examined signaling proteins (e.g., MdCTR1 and MdERF2; **Figure 6D**). The intricate roles of MdFERL6 in the regulation of fruit ripening and ethylene signal transduction merit further investigation.

#### MdFERl1 and MdFERL6 Physically Interact with MdSAMS

Given that MdFERL1 and MdFERL6 were shown to regulate ethylene production as well as ethylene responses, we further tested the possibility that these enzymes physically interact with key enzymes in the ethylene biosynthesis pathway and the signaling proteins involved in the ethylene response. The ethylene receptor ETRs are central proteins in ethylene signal transduction, and are localized to the endoplasmic reticulum. Interestingly, in addition to its localization to the membrane, MdFERL6 was also found to localize to the ER (**Figure 7A**). We thus tested the possibility that MdFERL6 interacts with ethylene receptors, using a BiFC assay. We also tested whether three well-characterized key enzymes in the ethylene biosynthesis pathway, SAMS, ACS, and ACO, could interact with MdFERL1 and MdFERL6. While fluorescence was not observed in the control (combination of empty vectors) or in the combinations of MdFERL6 and MdACS and of MdFERL6 and MdACO, strong fluorescence appeared when MdSAMS-YFP<sup>n</sup> was combined with MdFERL1 and MdFERL6 (**Figures 7B,C**). Furthermore, MdFERL6 did not appear to interact with the ethylene receptor MdETRs. The interaction between MdFERL1/6and SAMS further indicate that MdFERL1 and MdFERL6 modulate ethylene production during fruit development and ripening. To test whether MdFERL6 could interact with MdSAMS in apple calli, we conducted a co-immunoprecipitation (Co-IP) assay. As shown in **Figure 7D**, the Myc-MdSAMS protein complex (43 kD) was co-immunoprecipitated by His-MdFERL6 with anti-His antibody, and the His-MdFERL6-GFP complex (98 kD) was coimmunoprecipitated by Myc-MdSAMS with anti-Myc antibody. These results suggest that MdFERL1 and MdFERL6 may interact in vivo.

#### Ethionine Activity Assay in MdFERL6-OE and MdFERL6-AS Apple Calli

Having demonstrated that MdFER6 physically interacts with MdSAMS in apple, we further examined the roles of the MdFERL-SAMS protein complex in ethylene production by performing an ethionine activity assay. Ethionine is a toxic analog of methionine (Met). When cultivated with ethionine, plants with higher SAMS activity absorb more ethionine, resulting in toxic symptoms. Thus, plant growth status can be used as an indicator of SAMS activity (Mao et al., 2015). As shown in **Figure 8**, when cultivated in the presence of ethionine, the growth status of MdFERL6-OE and MdFERL6-AS transgenic calli was altered. MdFERL6-AS transgenic calli were more

sensitive than control calli to ethionine treatment. After 7 days of cultivation, MdFERL6-AS transgenic calli had become pale and gained less weight than control calli, suggesting that SMAS has higher activity in MdFERL6-AS transgenic calli than in the control. Furthermore, after 7 days of cultivation, MdFERL6-OE transgenic calli had become dark yellow and, for the 200 µM treatment, had gained more weight than the control, indicating that overexpression of MdFERL6 suppresses SAMS activity (**Figures 8A,B**). Furthermore, the expression of MdSAMS was increased in MdFERL6-AS and repressed in MdFERL6-OE, suggesting that MdFERL6 affects SAM synthesis through both activity level and transcription level (**Figure 8C**).

# DISCUSSION

The first receptor-like protein kinase with a carbohydratebinding domain, malectin, was identified from Catharanthus roseus, and thus named CrRLK1 (Schulze-Muth et al., 1996). Since then, CrRLK1-like receptor kinases (CrRLK1Ls) have increasingly attracted the interest of plant biologists. The genome size of rice is nearly four times that of Arabidopsis thaliana (Hong et al., 1997; Arabidopsis Genome Initiative, 2000), yet the number of CrRLK1Ls in rice (20 genes) is only slightly larger than that in Arabidopsis (17 genes; Kessler et al., 2010; Haruta et al., 2014). By contrast, the genome size of apple is only slightly larger than that of rice (Hong et al., 1997; Velasco et al., 2010), yet in the present study, we found that the CrRLK1L family in apple contains 34 members, and is therefore much larger than its counterpart in Arabidopsis and rice. In the context of plant morphology and anatomy, one of the major differences between apple and Arabidopsis or rice is the production of fleshy fruit, and accordingly, we were interested in determining whether the large difference in CrRLK1L family size between these species is correlated with fruit development. Our phylogenetic tree of CrRLK1L members from three species contained seven clades, five of which contained members from all three species. Clade I was not only specific to apple, but also contained more MdRLKs (12 members) than the other clades, while clade II contained the AtFER homolog, MdFERL1. We thus focused on MdFERLs in clades I and II in the present study.

Apple is one of the most popular and economically important fruit crops worldwide, and knowledge of the molecular mechanisms underlying apple fruit development and ripening could have beneficial applications in the apple industry. Difficulties in apple transgenesis, such as the long time required to obtain transgenic fruit and the lack of an efficient



*The SlFERL1-VIGS construct was transfected into tomato at 18 days post-anthesis. At 12 days after transfection, fruits were detached and analyzed. Values are means* ± *SD of two samples (each sample includes five fruits). Asterisks denote significant differences compared with the control sample (i.e., VIGS-C) at P* < *0.01, using Student's t-test. C, control; FW, fresh weight.*

transformation system, have meant that molecular studies of apple fruit development and ripening have been limited, with most studies being performed in a tomato model. Despite extensive efforts, we failed to develop a transient transgenic system that could be successfully used for gene manipulation in apple fruits. In the present study, we examined apple fruit development and ripening using three approaches: (1) the heterologous expression of MdFERL1 and MdFERL6 in tomato fruit, (2) VIGS-mediated manipulation of SlFERL1 in tomato fruit, and (3) overexpression and antisense transformation of the apple genes in apple fruit calli. Heterologous expression of MdFERL1 and MdFERL6 in tomato fruits delayed fruit ripening, while downregulation of SlFERL1 expression dramatically promoted tomato fruit ripening. The two systems corroborate each other, collectively demonstrating the pivotal roles of MdFERL1 and MdFERL6 in fruit development and ripening. Strikingly, when compared with the heterologous expression experiment, the effect of SlFERL1 VIGS on fruit ripening was much stronger. To elucidate this response, we examined the effect of SlFERL1 VIGS on the transcript levels of the other SlFERLs we identified in the tomato genome, and found it to cause a significant decrease in the expression of all five related genes. This implies that multiple SlFERL genes play important roles in fruit development and ripening. In this study, we identified 17 MdFERLs in the apple genome, and demonstrated that two of these members, MdFERL1 and MdFERL6, regulate fruit development and ripening. To fully understand the roles of MdFERLs in these processes, the contribution of the remaining MdFERLs should be examined.

Since its identification as an essential regulator of female fertility (Huck et al., 2003; Rotman et al., 2003), FER has been found to control a series of different developmental and biological processes. Besides its critical roles in fertilization (Escobar-Restrepo et al., 2007; Duan et al., 2014; Ngo et al., 2014; Liu et al., 2016), it has also been implicated in the regulation of cell elongation (Guo et al., 2009), root hair growth (Duan et al., 2010), seed size (Yu et al., 2014), plant defense (Keinath et al., 2010; Kessler et al., 2010), and

#### FIGURE 6 | Continued

analyzed 10 days after transfection. Control samples were transfected with the empty vector (pCambia1304). RT-qPCR was conducted using *SlACTIN* as an internal control. Values are means + SD of three biological replicates. Double asterisks denote a significant difference at \*\**P* < 0.01 and \**P* < 0.05 using Student's *t*-test. (B) Effect of *SlFERL1*-VIGS (virus-induced gene silencing) on the expression of tomato ripening-related genes. The VIGS constructs were injected into fruits at 18 DPA, and gene expression was examined 12 days after transfection, using RT-qPCR. *SlACTIN* was used as the internal control. Values are means + SD of three biological replicates. Double asterisks denote a significant difference at \*\**P* < 0.01 and \**P* < 0.05 using Student's *t*-test. (C) Effect of *MdFERL6*-OE on the expression of ripening-related genes in apple calli. RT-qPCR was conducted using *MdACTIN* as an internal control. Values are means + SD of three biological replicates; different lowercase letters represent significant differences based on ANOVA (*P* < 0.05). (D) Effect of *MdFERL6*-AS and *MdFERL6*-AS combined with 1-MCP on the expression of ripening-related genes in apple calli. RT-qPCR was conducted using *MdACTIN* as an internal control. Values are means + SD of three biological replicates; different lowercase letters represent significant differences based on ANOVA (*P* < 0.05).

mechanosensing and calcium signaling (Shih et al., 2014). In the context of metabolism, FER was found to be a multifunctional regulator of starch accumulation (Yang et al., 2015) and sugar metabolism (Yeats et al., 2016; Pu et al., 2017). Recently, FER has been implicated in the cross-talk signaling of various phytohormones, including in ABA (Yu et al., 2012; Chen et al., 2016), auxin (Duan et al., 2010), BR, and ethylene (Guo et al., 2009; Deslauriers and Larsen, 2010) signaling. FER may also interact with SAM/SAMS, thereby suppressing ethylene production in Arabidopsis (Mao et al., 2015). We found that MdFERL6 could physically interact with MdSAMS, leading to changes in MdSAMS expression and MdSAMS activity in apple (**Figures 7**, **8**). Phosphorylation site analysis identified five sites in MdSAMS, 134T, 153T, 250T, 271S, and 284Y, that can be potentially be phosphorylated by kinases (http://kinasephos. mbc.nctu.edu.tw/index.php). However, the mechanism by which MdFERL6 and MdSAMS interact, including the involvement of phosphorylation, remains to be elucidated. A recent study showed that a FERONIA-like kinase (MDP0000493959), which was identified as MdFERL12 in our study, was transcriptionally downregulated by ethylene during postharvest ripening and senescence of apple fruit (Zermiani et al., 2015). We found that MdFERL12 was expressed at lower levels than MdFERL1 and MdFERL6 in apple fruit (**Figure 3**, Supplemental Figure S1B); however, MdFERL12 could be induced by 1-MCP and other ripening-related signals such as mannitol, PEG, and high temperature (Supplemental Figure S3), suggesting that FERLs have diverse functions in regulating ethylene production and signal transduction, and that their mechanisms need to be further revealed.

Although FER has been extensively studied as described above, a potential role of FERLs in fruit development has not been reported. The present study demonstrates that FERLs are critical regulators of fruit development and ripening, further indicating the importance of FERLs as universal regulators of plant growth and development. In contrast to Arabidopsis, which has just one FER gene, the apple genome contains 14 FERLs. If each individual MdFERL member controls multiple developmental or biological processes, the mechanisms by which these proteins collectively or synergistically regulate apple plant growth and development may be of great significance, and should be thoroughly investigated.

Ethylene has long been known to be the critical signal controlling climacteric fruit ripening (Biale, 1964; Burg and Burg, 1965; Alexander and Grierson, 2002; Bram et al., 2015). One of the important features of ethylene signaling is its large increase in production during fruit ripening; therefore, any cellular internal factor controlling ethylene production should be investigated as a critical regulator of climacteric fruit development and ripening. It has been well established that there are three key enzymes in the ethylene biosynthesis pathway; SAM synthase, ACC Synthase (ACS), and ACC oxidase (ACO). SAM synthase catalyzes the first step in the ethylene biosynthesis pathway, namely the formation of S-adenosylmethionine from methionine and ATP. ACS catalyzes the formation of ACC from S-adenosylmethionine. In the present study, we revealed that both MdFERL1 and MdFERL6 physically interact with SAMS, but not with ACS or ACO. The present study suggests that multiple FER-like protein kinases physically interact with SAMS; such a variety of interactions is not possible in Arabidopsis, as it contains just one member of this family. The finding that different MdFERLs jointly regulate the same biological process reflects the significance of the existence of so many FERLs in the apple genome. As ACC is the immediate precursor of ethylene, the rate-limiting step of ethylene biosynthesis is, under most conditions, considered to be the conversion of S-adenosylmethionine to ACC by ACS (Wang et al., 2002). Thus, the regulation of ethylene synthesis via SAMS rather than ACS or ACO suggests that other rate-limiting steps exist in this pathway. In a previous study, we found that the amount of ABA that accumulated under drought stress was controlled by enzymes that function in the early stages of ABA biosynthesis, rather than by 9-cis-epoxycarotenoid dioxygenase (NCED), an enzyme that catalyzes the production of xanthin, the immediate precursor of ABA (Ren et al., 2007). Once ABA starts to accumulate, its immediate precursors are rapidly exhausted, to the extent that the early stages in the pathway become the ratelimiting steps that determine how much ABA accumulates, rather than the NCED-catalyzed step. A similar mechanism operates in ethylene production during climacteric fruit ripening; the regulation of ethylene production by FER-like protein kinases occurs at an early stage of ethylene biosynthesis (at the step catalyzed by SAMS), rather than at the downstream enzymes ACS and ACO.

In climacteric fruits, ethylene production occurs via two systems, the origins of which are key to our understanding of the mechanisms regulating fruit development and ripening. The first system is basal low rate of ethylene production, which occurs until fruit ripening commences, while the second system

C-, and N-terminus, respectively, of yellow fluorescent protein (YFP; designated as YFPc or YFPn, respectively). Different combinations of the fused constructs were co-transformed into tobacco (*Nicotiana tabacum*) cells, and visualized using confocal microscopy. YFP and bright-field were excited at 488 nm and 543 nm, respectively. Bars = 20 µm. (C) BiFC analysis of the physical interaction between MdFERL6 and MdACS3, MdACS5, MdETR2, and MdETR5. MdFERL6 was fused to the C-terminus of YFP, while the other proteins were fused to the N-terminus of YFP. (D) Co-immunoprecipitation (Co-IP) assay of the interaction between MdFERL6 and MdSAMS in apple calli. MdFERL6 was fused with GFP/His using the pMDC83 vector, and MdSAMS was fused with Myc using the pCambia1300 vector. (The molecular weight of Myc-MdSAMS is 43 kD, Myc is 1.2 kD, His-MdFERL6 is 98 kD, and 6 × His is 0.6 kD).

generates the large increase in ethylene production during fruit ripening (McMurchie et al., 1972; Hoffman and Yang, 1980; Brady, 1987). Given that both MdFERL1 and MdFERL6 suppress ethylene production, it is reasonable to propose that MdFERL1 and MdFERL6 are implicated in the first system, acting to inhibit ethylene biosynthesis before fruit ripening occurs. Nevertheless, the transcript levels of both MdFERL1 and MdFERL6 declined as apple fruit growth and ripening progressed, suggesting that MdFERL1 and MdFERL6 are likely associated with the regulation of the ethylene production spike during ripening. Notably, in contrast to the continual decline in MdFERL1 expression throughout the process of fruit set to ripening, the transcript level of MdFERL6 remained high until about 100 DPA, after which it decreased dramatically. Compared with MdFERL1, MdFERL6 had a stronger effect on fruit ripening in transgenic tomatoes, suggesting that it is the decline of the MdFERL transcripts at the later development stages that is most important for the regulation of fruit ripening. On the other hand, the different patterns of MdFERL1 and MdFERL6 expression implies that ethylene production is tightly modulated throughout the process, from fruit set to ripening; individual MdFERLs may play different roles in the two phases of ethylene production, and they collectively or synergistically determine the spatio-temporal changes in ethylene content.

In summary, we found that FERLs in apple can be categorized into two groups; some share a relatively high level of amino acid sequence identity with the FERLs in both Arabidopsis and rice, while others are less similar and appear to be specific to the apple genome. MdFERL1 and MdFERL6 belong to the first and second of these groups, respectively, and were the most highly

expressed of all the MdFERLs in developing fruit. The following four findings support the notion that MdFERL1 and MdFERL6 regulate fruit ripening: (1) heterologous expression of MdFERL1 and MdFERL6 delayed fruit ripening and suppressed ethylene production in tomato; (2) overexpression and antisense silencing of MdFERL6 suppressed and promoted ethylene production, respectively; (3) VIGS ofSlFERL1 promoted fruit ripening and ethylene production; and (4) MdFERL1 and MdFERL6 were capable of physically interacting with the key ethylene biosynthesis enzyme, SAMS. As MdFERL6 and MdFERL1 suppress ethylene production during fruit development and ripening, their high transcription levels in the early stages of fruit growth followed by their dramatic decreases (especially for MdFERL6) during fruit ripening imply that they contribute to the regulation of ethylene production throughout development; i.e., they maintain ethylene at relatively low levels during the early developmental stages, and boost ethylene levels when fruit ripening commences. MdFERL1 and MdFERL6 therefore function as negative regulators of climacteric fruit ripening. This work has, for the first time, demonstrated that FER-like protein kinases are implicated in fleshy fruit development and ripening, providing new insights into the molecular basis of these processes. Nevertheless, this is a preliminary study; more research is required to elucidate the molecular recognition and signaling cascades of MdFERLs during fruit development and ripening.

### AUTHOR CONTRIBUTIONS

MJ and PD, performed most of the experiments; ND, QZ, and SX, contributed to some of the experiments; LW, YZ, WM, and JL, provided technical assistance; BL, designed the experiments and contributed to the date analysis; WJ, conceived the project, supervised the experiments, and complemented the writing.

#### FUNDING

National Natural Science Foundation of China.

#### REFERENCES


#### ACKNOWLEDGMENTS

This work was supported by the National Natural Science Foundation of China (Grant No. 31471851, 31672133, 31572104), Fok Ying-Tong Education Foundation, China (Grant No. 151027), the 111 Project (Grant No. B17043), and Beijing Natural Science Foundation (Grant No. 6171001).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2017. 01406/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 © 2017 Jia, Du, Ding, Zhang, Xing, Wei, Zhao, Mao, Li, Li and Jia. 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.

# Gibberellins Promote Brassinosteroids Action and Both Increase Heterosis for Plant Height in Maize (Zea mays L.)

Songlin Hu<sup>1</sup> \*, Cuiling Wang<sup>2</sup> , Darlene L. Sanchez<sup>1</sup> , Alexander E. Lipka<sup>3</sup> , Peng Liu<sup>4</sup> , Yanhai Yin<sup>5</sup> , Michael Blanco<sup>6</sup> and Thomas Lübberstedt<sup>1</sup>

<sup>1</sup> Department of Agronomy, Iowa State University, Ames, IA, United States, <sup>2</sup> Department of Agronomy, Henan University of Science and Technology, Luoyang, China, <sup>3</sup> Department of Crop Sciences, University of Illinois at Urbana–Champaign, Champaign, IL, United States, <sup>4</sup> Department of Statistics, Iowa State University, Ames, IA, United States, <sup>5</sup> Department of Genetics, Development and Cell biology, Iowa State University, Ames, IA, United States, <sup>6</sup> Plant Introduction Research Unit, Department of Agronomy, United States Department of Agriculture – Agricultural Research Service, Iowa State University, Ames, IA, United States

#### Edited by:

Chi-Kuang Wen, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, China

#### Reviewed by:

Hao Peng, Washington State University, United States Xuehui Huang, Shanghai Normal University, China

> \*Correspondence: Songlin Hu hsonglin1@gmail.com

#### Specialty section:

This article was submitted to Crop Science and Horticulture, a section of the journal Frontiers in Plant Science

> Received: 03 April 2017 Accepted: 30 May 2017 Published: 20 June 2017

#### Citation:

Hu S, Wang C, Sanchez DL, Lipka AE, Liu P, Yin Y, Blanco M and Lübberstedt T (2017) Gibberellins Promote Brassinosteroids Action and Both Increase Heterosis for Plant Height in Maize (Zea mays L.). Front. Plant Sci. 8:1039. doi: 10.3389/fpls.2017.01039 Brassinosteroids (BRs) and Gibberellins (GAs) are two classes of plant hormones affecting plant height (PHT). Thus, manipulation of BR and GA levels or signaling enables optimization of crop grain and biomass yields. We established backcross (BC) families, selected for increased PHT, in two elite maize inbred backgrounds. Various exotic accessions used in the germplasm enhancement in maize project served as donors. BC1-derived doubled haploid lines in the same two elite maize inbred backgrounds established without selection for plant height were included for comparison. We conducted genome-wide association studies to explore the genetic control of PHT by BR and GA. In addition, we used BR and GA inhibitors to compare the relationship between PHT, BR, and GA in inbred lines and heterozygotes from a physiological and biological perspective. A total of 73 genomic loci were discovered to be associated with PHT, with seven co-localized with GA, and two co-localized with BR candidate genes. PHT determined in field trials was significantly correlated with seedling stage BR and GA inhibitor responses. However, this observation was only true for maize heterozygotes, not for inbred lines. Path analysis results suggest that heterozygosity increases GA levels, which in turn promote BR levels. Thus, at least part of heterosis for PHT in maize can be explained by increased GA and BR levels, and seedling stage hormone inhibitor response is promising to predict heterosis for PHT.

Keywords: brassinosteroid, gibberellin, plant height, genome-wide association study, heterosis

# INTRODUCTION

Increasing demand for biomass production led to a paradigm shift regarding plant height (PHT) from dwarfs to giants (Fernandez et al., 2009). Tall maize varieties are more desirable, if maize is used as dual purpose or biomass crop. For dual-purpose maize, increased stover biomass adds value: grain is harvested for food or feed, stover for bioenergy conversion. While PHT increasing alleles contribute to increasing biofuel production, they also increase the risk of lodging.

To overcome this negative side effect, breeders can either produce maize varieties with increased lignin level or a stronger root system to stabilize plants (Fernandez et al., 2009). Increased lignin levels are not desirable for biochemical, but favorable for thermochemical conversion of biomass (Mendu et al., 2011).

Plant height is an important agronomic trait in modern maize and more generally cereal breeding programs, and has been manipulated during maize domestication as it shows significant correlations with different agronomic traits such as grain yield (Teng et al., 2013; Abdel-Ghani et al., 2016). For grain maize production, breeders prefer a short stature, as high yielding maize varieties need to be lodging-tolerant under high nitrogen levels and high density planting conditions. Breeders use semi-dwarf genes to moderately decrease PHT in cereals, such as the green revolution genes sd-1 in rice (Sasaki et al., 2002) and Rht in wheat (Peng et al., 1999). In maize, several maize dwarf genes have been well-characterized, but are not intentionally used in breeding programs due to their adverse impact on grain yield, such as dwarf3 (Winkler and Helentjaris, 1995), dwarf8 and dwarf9 (Thornsberry et al., 2001; Lawit et al., 2010), nana plant1 (Hartwig et al., 2011), and brd1 (Makarevitch et al., 2012). All these genes cause defects in either the brassinosteroid (BR) or the gibberellin (GA) pathway, stressing the importance of these two plant hormones in the control of PHT. It was reported that BR and GA have the most direct effects on PHT without major negative pleiotropic effects as compared to other plant hormones (Fernandez et al., 2009).

Brassinosteroids are steroid hormones found throughout the plant kingdom, similar to animal steroid hormones. They promote cell growth, even if at low concentrations, by regulating cell division and elongation (Clouse, 1996). The biosynthesis pathway for brassinolide (BL), the most active BR, is wellestablished by characterization of BR-deficient mutants in model species such as Arabidopsis, pea (Pisum sativum), and tomato (Solanum lycopersicum). Enzymes involved in BL biosynthesis include DWARF7 (DWF7), DIMINUTO (DIM1), FAD dependent oxidase, DEETIOLATED2 (DET2) steroid 5a-reductase and several cytochrome p450 monooxygenases, such as DWF4, CONSTITUTIVE PHOTOMORPHOGENESIS AND DWARFISM (CPD), and DWF (Fernandez et al., 2009). Signal transduction (Clouse, 2011) initiates with binding of BR to BRASSINOSTEROID INSENSITIVE 1 (BRI1), a plasma membrane-localized leucine rich repeat (LRR) receptor kinase. In the absence of BR, the negative regulator (BRI1 KINASE INHIBITOR 1) BKI1 binds and inhibits BRI1 function (Jaillais and Chory, 2010). Transcription factors (bri1- EMS-suppressor 1) BES1/(bri1 ETHLYMETHANESULFONATE SUPPRESSOR1 and BRASSINAZOLE RESISTANT1) BZR1 are phosphorylated by (BRASSINOSTEROID INSENSITIVE2) BIN2 and are inhibited by several mechanisms (Choe et al., 2002). Binding of BR to BRI1 leads to release of BKI1, which interacts with 14-3-3 and thus promotes nuclear accumulation of BES1 (Wang et al., 2011). Association of BRI1 with co-receptor (BRASSINOSTEROID INSENSITIVE 1) BAK1 activates BRI1 kinase activity (Gou et al., 2012). Activated BRI1 signals through (BR-Signaling Kinases) BSK1 and (Constitutive Differential Growth1) CDG1 kinases as well as (BRI1-Supressor1) BSU1 phosphatase to inhibit BIN2 kinase activity (Kim et al., 2011). The inhibition of BIN2 and action of (protein phosphatase 2A) PP2A phosphatase allow accumulation of BES1/BZR1 in the nucleus, which regulates gene expression in combination with other transcription regulators (Tang et al., 2010).

Gibberellins are cyclic diterpene compounds that promote stem elongation, and mutants in GA synthesis or signaling show dwarf phenotypes. GA synthesis begins with transgeranylgeranyl diphosphate and requires six enzymatic steps for the formation of bioactive GA1 or GA4 (Yamaguchi, 2008). A deficiency in any step leading to the production of bioactive GAs causes dwarfism. GA homeostasis is maintained through a balance of anabolic and catabolic activities. Four rice genes are primarily responsible for this regulation through a feedback transcriptional control mechanism: (GA 20-oxidase) GA20ox, GA3ox, GA2ox, and (epioxidation catalyzed by ELONGATEDUPPER INTERNODES1) EUI1 (Appleford et al., 2007). All of them are proven targets for manipulating GA levels. For example, the recessive tall rice mutant elongated uppermost internode (eui) (Rutger and Carnahan, 1981) exhibits a rapid and enhanced elongation of internodes, particularly the uppermost internode during the heading stage. It has been shown that EUI controls bioactive GA levels to modulate internode elongation in a tissue- and developmental stage–specific manner (Zhu et al., 2006). Central to GA signaling are DELLA proteins, which are negative regulators that repress GA-induced gene transcription in the absence of GA signaling (Zentella et al., 2007). GA signaling induces gene expression by targeting the DELLA proteins for degradation (GIBBERELLIN INSENSITIVE DWARF1) GID1 and GID2 are positive regulators of GA signaling and (SLENDER RICE1) SLR1 is a DELLA protein and is a negative regulator (Sasaki et al., 2003) of GA signaling. As with metabolism, both positive and negative regulators of GA signaling have the demonstrated potential for dramatic effects on PHT.

Plant hormone inhibitors are powerful tools for elucidating plant hormone functions during plant development. For example, BR inhibitor Propiconazole (Pcz) has successfully been employed for studying BR control of sex determination and PHT in maize (Hartwig et al., 2012). GA inhibitor Uniconazole (Ucz) was used to explore root growth and nitrogen transfer in soybean (Yan et al., 2013). Instead of using hormone pathway mutants, plant hormone inhibitors can phenocopy hormone-deficient mutants in crops, with the advantage that deficiency levels of hormones can be controlled. Moreover, plant hormone inhibitors can help with identification and characterization of hormone deficient mutants without prior knowledge of the mutant phenotype (Hartwig et al., 2012). In maize, Pcz and Ucz are two popular plant hormone inhibitors of BR and GA, respectively, due to their easy accessibility and low costs (Rademacher, 2000; Hartwig et al., 2012). Application of Pcz and Ucz reduce mesocotyl elongation, and genotypes with elevated BR or GA level are more tolerant to Pcz or Ucz, resulting in alleviated reduction of mesocotyl length (Hartwig et al., 2012). Therefore, Pcz and Ucz can be employed to explore the relationship between morphological traits, BR and GA activities in maize.

In this study, two sets of backcross (BC) libraries derived by introgression of a diverse set of tropical maize accessions into inbred lines, representing two major maize heterotic groups (Iowa Stiff Stalk and Non-Stiff Stalk), were evaluated for PHT. Moreover, we applied BR and GA inhibitors Pcz and Ucz at seedling stage to compare BR and GA inhibitor responses between tall and short maize BC families within each library, and between the tallest and shortest individuals within BC families. In addition, two sets of doubled haploid (DH) libraries derived from the same parents producing the two BC libraries were tested for PHT and BR/GA activities as a comparison. Our objectives were (1) to evaluate the PHT performance of the two BC libraries and the two DH libraries; (2) to conduct a genome-wide association study (GWAS) to investigate the genomic regions associated with BR, GA, and PHT in these BC families; and (3) to apply Pcz and Ucz treatments to these two BC libraries and two DH libraries to investigate the relationship between BR, GA, and PHT in both inbred lines (the two DH libraries) and heterozygotes (the two BC libraries).

# MATERIALS AND METHODS

#### Plant Materials

Two libraries (refer to an enriched genetic diversity) of phenotypic-selected introgression families (PIFs): PIFB47 (PHB47 as recurrent parent) and PIFZ51 (PHZ51 as recurrent parent) were used in this study. PHB47 and PHZ51 are two elite expired PVP (Plant Variety Protection) inbred lines which belong to Iowa Stiff Stalk and Non-Stiff Stalk heterotic groups, respectively. Donor parents were tropical or subtropical accessions from the germplasm enhancement of maize (GEM) project (Salhuana and Pollak, 2006), of which 42 different accessions were used for PIFB47 and 46 for PIFZ51, with five accessions being used for both PIFB47 and PIFZ51 (**Supplementary Table S1**). The process of constructing PIF libraries was described in a previous study (Abdel-Ghani et al., 2016). Briefly, in each backcross (BC) generation, phenotypic selection for PHT (selection for tallness) and flowering time (synchrony with recurrent parent) was carried out to accumulate PHT increasing alleles from donor parents into elite maize background (PHB47 and PHZ51) and to minimize confounding effects between flowering time and PHT. As a result, 75 and 71 PIFs were produced for PIFB47 and PIFZ51, respectively (**Supplementary Table S1**). In addition, two doubled haploid libraries (DHB47 and DHZ51) were used as unselected groups (without any phenotype selection) for comparison, including 103 and 66 BGEM (DH lines from GEM project) lines, respectively. The method used to create the BGEM lines was described in a previous study (Brenner et al., 2012). Briefly, donor parents from GEM project were used for producing BGEM lines as for PIFs, and same recurrent parents (PHB47 and PHZ51) were used for backcrossing. Different from PIF development, BGEM lines were produced (induced and doubled) from BC<sup>1</sup> individuals, and there was no phenotypic selection during the backcross process. The hybrid of PHB47 × PHZ51 was included in field experiments for comparison.

# Field Experiments

#### Experiment 1: PHT and Flowering Time Characterization of PIFs

PIFs together with their recurrent parents were evaluated for PHT across 3 years (2013, 2014, 2015), with one location in 2013 (Plant Introduction Station: PSI, Ames, IA, United States), three locations in 2014 [Agronomy farm (AG), Ames, IA, United States PSI, and Neely-Kinyon Memorial Research and Demonstration Farm (NK) – Greenfield, IA, United States] and one location in 2015 (AG) with a randomized complete block design (RCBD) within each location (two blocks per location). All locations were under standard farm management practices (tillage, nutrient management, etc.), with 15 plants per plot and 0.75 m between plots. PIFs were compared with their recurrent parents and hybrid of PHB47 × PHZ51 for PHT and flowering time. Two weeks after tasseling, a representative plant with median PHT (MPHT) was selected per plot, and PHT was recorded from ground to tassel tip. Across years and locations, multiple data points for each PIF were collected. In 2014, the shortest (SPHT) and the tallest individual (TPHT) within each PIF were also measured. In 2014 and 2015, flowering time traits – days to tasseling (DTT), days to silking (DTS), and anthesis-silking interval (ASI) were recorded for each plot, when at least 50% of the plants shed pollen or showed silks.

#### Experiment 2: PHT and Flowering Time Characterization of BGEM Lines

DHB47, DHZ51 together with their recurrent parent were characterized for PHT and flowering time in 2014 and 2015 with a RCBD design with two replications at AG. In each plot, one representative individual grown in the middle of the plot was used to measure PHT, and flowering time was measured in the same way as for PIFs.

#### Experiment 3: Phenotypic and Genotypic Characterization for the Two Tallest Individuals within Each PIF

In 2014 at PSI, leaf samples of the two tallest plants from each PIF together with six PHB47 and six PHZ51 individuals (304 individuals in total) were collected after tasseling for genotyping. PHT, ear height (EHT), node number (NNode), leaf angle of the first leaf below the flag leaf (LA) and tassel length (TL) were measured for each genotyped individual. EHT was measured from the ground to the primary ear node. NNode was scored as the number of nodes between the top brace root node and the flag leaf excluding the variation in brace root nodes and subterranean nodes. LA was measured as the angle between horizontal position and the midrib of the first leaf below the flag leaf. TL was measured from the non-branching node present below the lowermost primary branch to the tip of central spike. In 2015, the tallest and shortest individual from each PIF was selected to produce backcross seed for hormone inhibitor response comparison.

#### Hormone Inhibitor Test Experiments

All PIFs and BGEM lines were evaluated for responses to BR and GA inhibitors in three independent experiments

(BR, GA, and water). All entries were treated with 80 µM BR inhibitor Pcz, 80 µM GA inhibitor Ucz and water (mock treatment), with three replications per experiment in a completely randomized design (CRD). Each treatment was applied to 12 kernels from each PIF, and 6 kernels from each BGEM line for each replication within each experiment. Seed was soaked for 24 h, then transferred to paper rolls (Kumar et al., 2012) containing the same soaking solution (i.e., water, Pcz, or Ucz). Rolls were placed in buckets with a cover in a growth chamber without light at 25◦C. After 8 days, seedlings were removed from the growth chamber and mesocotyl length was measured from the root-shoot transition zone to the first node for each seedling (Hartwig et al., 2012) as an indicator for hormone inhibitor response. The average mesocotyl length for the 12 (or 6) seedlings under Pcz, Ucz, and water treatment was recorded as MP (mesocotyl length with Pcz treatment), MU (mesocotyl length with Ucz treatment), and MW (mesocotyl length with water treatment), respectively. BR inhibitor response was defined as MP/MW, and GA inhibitor response as MU/MW according to a previous protocol (Hartwig et al., 2012; Huang et al., 2013), as shown in Supplementary Figure S1. Seed harvested from the tallest and the shortest individual within each PIF in 2015 was tested using the same protocol described above in three independent experiments.

#### DNA Extraction and Genotyping

DNA was extracted using Qiagen BioSprint96 at the Genomic Technology Facility at Iowa State University. Samples were genotyped with the MaizeSNP3K chip (3072 SNPs across the whole genome), which is a subset of the Illumina MaizeSNP50 BeadChip (Ganal et al., 2011). Genotyping was executed using the Illumina GoldenGate SNP genotyping platform (Fan et al., 2006) at the National Maize Improvement Center of China (China Agricultural University). DNA from all BGEM lines was extracted using 1–10 kernels per line, and genotyped with 8523 (chip-based) SNPs across the genome, in which 7739 polymorphic markers were used in the analysis, with 920 markers overlapping with the MaizeSNP3K subset across the genome. Both DNA extraction and SNP chip genotyping of the BGEM lines were done at KWS SAAT SE (company in Germany).

## Marker Quality Control

The 12 genotyped recurrent parent individuals (six PHB47 and six PHZ51) were analyzed for marker quality control: SNP markers missing across all RP individuals or showing heterozygous state across at least three (out of six) recurrent parent individuals were excluded, as residual heterozygosity is random and unlikely to frequently occur at the same genomic position within a fixed inbred line. Subsequently, all PIF genotypes were jointly analyzed. SNP markers with >5% missing data were discarded. For missing SNP markers, the LD k-nearest neighbor algorithm (LD KNNi imputation) was used for imputation using TASSEL5.2.15 (Money et al., 2015).

# Phenotype and Genotype Statistical Analysis

Phenotypic data was analyzed using SAS 9.3 software (SAS Institute). For each evaluated trait, variance component estimates were obtained from a mixed linear model (MLM) fitted across all environments in SAS PROC MIXDED. Variance components (σ 2 g , σ 2 g×s , σ 2 s ) were estimated where σ 2 g , σ 2 g×s , σ 2 s corresponds to genotypic variance, genotype by environmental interaction variance, and error variance respectively. Entry mean-based heritability was calculated from variance components estimates as h 2 = σ 2 g σ 2 <sup>g</sup>+σ 2 g×s /n+σ 2 s /rn where r is the number of replications within each location, and n is the number of locations (Holland et al., 2003). Comparisons between PIFs and recurrent parents or between DH lines and recurrent parents were done using Fisher's least significant difference (LSD) with a significance level of 0.05. Pearson's correlation coefficients were used to assess the relationship between different traits with the SAS PROC CORR function. Linear regression analysis was performed using the SAS PROC GLM function. Graphs were obtained using ggplot2 in R (Wickham, 2009).

#### Genome-Wide Association Study

We conducted GWAS for all measured field traits and BR/GA inhibitor responses of the PIFs (292 genotypes) with 2930 polymorphic genome-wide SNPs (minor allele frequency > 5%). Population structure was estimated using Structure 2.3.4 software (Pritchard et al., 2000). The parameter settings included a burnin length of 50,000 followed by 50,000 iterations of setting K (clusters – number of subpopulations) from 1 to 10, with five replications for each K (Pace et al., 2014). The most probable K value was picked by plotting each K (x-axis) with its estimated Ln probability of data (y-axis). The K value was selected when the estimated Ln probability of data reached a plateau (Pritchard et al., 2000).

To balance false positives and false negatives, we used three models for GWAS analysis: (1) a generalized linear model (GLM) + Q, with the covariates Q from STRUCTURE included in the model as fixed effects to account for population structure; (2) a MLM (Yu et al., 2006) with population structure and kinship as covariates; and (3) FarmCPU (Fixed and random model Circulating Probability Unification) with kinship and population structure as covariates, but with additional algorithms solving the confounding problems between testing markers and covariates (Liu et al., 2016). GLM was conducted with the software program TASSEL 5.0 (Bradbury et al., 2007). MLM was used as implemented in GAPIT (Genome Association and Prediction Integrated Tool-R package) (Lipka et al., 2012). FarmCPU was applied with R package FarmCPU (Liu et al., 2016). Statistical program simpleM implemented in R was used to account for multiple testing (Johnson et al., 2010). The threshold level was based on an effective number of independent tests m, and m was used in a similar way as the Bonferroni correction method (Pace et al., 2014). In this study, for a family-wise error rate of 0.05 the threshold for significant trait-marker associations was set as 4.44 × 10−<sup>5</sup> (multiple testing threshold level).

# Comparison with Published PHT QTL Regions

SNP markers associated with PHT were compared with previously published PHT quantitative trait loci (QTL). The dataset for comparison is based on a maize PHT QTL metaanalysis aimed at identifying the most likely position of PHT QTL that are consistently found across studies. This dataset summarized published PHT QTL into 40 hot spots across the maize genome based on the maize B73 RefGen\_v2 physical map (Wang et al., 2006). PHT associated SNP markers from this study were compared with these hot spots to study co-localizations.

#### BR and GA Candidate Genes

The information about the genes involved in BR and GA biosynthesis and signaling pathway was collected from Arabidopsis and rice (model species) and listed in **Supplementary Table S2**. The protein sequence of these genes was obtained from the National Center for Biotechnology Information (NCBI) databases, and was used to identify orthologous genes in maize through BLASTP in Gramene (Altschul et al., 1997). Based on BLASTP score, % Identity, and E-value, hits were ranked. From the most likely hit, a reverse BLAST search was conducted: maize genes identified using the approach described above were blasted back to model species to identify orthologous genes, the goal was to confirm that the gene with the highest score was the original BR/GA pathway gene from model species. If the gene identified was not the original gene (used to find maize candidate genes), it was discarded. Finally, each BR/GA gene from the model species was aligned with at least one candidate gene in maize, with a few genes aligned with 2–4 maize candidate genes (which are likely due to gene duplications in maize). Finally, we compared the GA candidate genes found in this study with GA candidate genes published with another method (Song et al., 2011), which used all previously reported genes encoding GA metabolism enzymes in maize and other species as BLAST queries.

# Co-localization of BR and GA Candidate Genes and PHT Associated SNPs

We used a bin size of 1 Mb according to a previous protocol (Yang et al., 2015) around each PHT associated SNP to capture BR/GA candidate genes. Moreover, we used a re-sampling based non-parametric method to test the hypothesis, that the observed number of BR or GA candidate genes co-localized with PHT associated SNP markers using a 1 Mb bin size is not different from randomly occurrence. First, we defined the number of PHT associated SNP markers capturing BR/GA candidate genes using a 1 Mb bin size as "observed value." Second, the same number of PHT associated SNP markers from GWAS results was randomly sampled (sampling with replacement) from the whole set of SNP markers. In each random sample, all sampled SNP markers used a bin size of 1 Mb to capture BR/GA candidate genes. The number of SNP markers capturing a BR or GA candidate gene was recorded as test statistic. This procedure was repeated for 10,000 times, and the number of test statistics exceeding the observed value was divided by the total number of simulations to compute a P-value (Johnson et al., 2011; Yang et al., 2015).

# Path Analysis between BR, GA, PHT, and Heterozygosity

For each PIF, we recorded the BR inhibitor response, GA inhibitor response, PHT performance and the level of heterozygosity (the percentage of heterozygous markers). We used path analysis implemented in IBM SPSS Amos 20 (Arbuckle and Amos, 2006; Jansen et al., 2013) to investigate the direct or indirect correlations between these four variables by fitting proposed path models according to the covariance structure of the underlying data (Jansen et al., 2013). The goal is to explore the hypothesis that heterozygosity is impacting PHT through the regulation of BR and GA level in maize heterozygotes. The correlations between each pair of these four variables are used to calculate total, direct, and indirect effects between the observed variables. We started with a proposed model with the hypothesis that the level of heterozygosity indirectly impacts PHT through the regulation of BR and GA instead of a direct correlation. In other words, there are no other factors connecting heterozygosity and PHT except for BR and GA. Based on the data structure, path analysis suggests to add new links or to delete old links between variables. The final best fit was defined by a stringent criterion with Chi-squared P-value > 0.05, root mean square error of approximation (RMSEA) < 0.05, comparative fit index (CFI) > 0.95, goodness of fit index (GFI) > 0.8, adjusted goodness of fit index (AGFI) > 0.9, normed fit index (NFI) > 0.9, relative fit index (RFI) > 0.9, incremental fit index (IFI) > 0.9. If more than one model fit all criteria, then selection was based on a minimum Akaike's information criterion (AIC) value (Thompson, 1988; Jansen et al., 2013).

# RESULTS

# Phenotypic Characterization

All statements about statistical inferences are based on a type I error rate (or family-wise error rate) of 5% unless P-values are specifically given in the Section "Results." As described in the Section "Materials and Methods," two libraries of phenotypic-selected introgression families (PIFs): PIFB47 (PHB47 as recurrent parent) and PIFZ51 (PHZ51 as recurrent parent), and two doubled haploid libraries (DHB47 and DHZ51, without any phenotype selection) were compared with their recurrent parent. Most PIFs were significantly taller than their recurrent parent (**Figure 1**) and showed consistent PHT performance (heritability: h<sup>2</sup> = 0.81) across 3 years. The percentages of PIFs significantly taller than their recurrent parent in each year were: 91.6% (2013), 92.5% (2014), 92.0% (2015) for PIFB47; and 85.7% (2013), 85.5% (2014), and 90.0% (2015) for PIFZ51. Even for the shortest individual within each PIF (SPHT), 91.2 and 59.4% of these individuals were taller than recurrent parent within PIFB47 or PIFZ51, respectively. In contrast, the proportion of taller BGEM lines (DH lines from DHB47 and DHZ51: see Materials and Methods) compared with their recurrent parent was only 23% for DHZ51 and 18%

for DHB47. PIFB47 was on average 23 cm taller than PIFZ51, and PHB47 was on average 8 cm taller than PHZ51. For the BGEM lines, DHB47 was on average 2 cm taller than DHZ51. None of the PIFs were significantly taller than the F<sup>1</sup> hybrid of PHB47 × PHZ51.

All PIFs flowered within 3 days compared to the respective recurrent parent for both days to tasseling (DDT) and DTS. The average ASI was 0.3 and 0.6 days for PIFB47 and PIFZ51, respectively, and thus similar to PHB47 (0.2 days) and PHZ51 (0.4 days). Differences of DDT and DTS between BGEM lines and recurrent parent ranged from 0–13 days in DHB47 and 0– 15 days in DHZ51. The F<sup>1</sup> hybrid of PHB47 × PHZ51 flowered 4 or 6 days earlier than PHB47 and PHZ51, respectively. Selection for PHT affected other agronomic traits including ear height (EHT), number of nodes (NNode), and tassel length (TL) but not leaf angle (LA). EHT, NNode, and TL increased along with an increased PHT (**Table 1**) and LA is not associated with other traits. Summary statistics for other traits are listed in **Supplementary Table S3**.

#### Genotypic Characterization

We defined donor genome proportion (DGP) of PIFs as 50% of heterozygosity (percentage of heterozygous markers) in this study, as for each heterozygous locus, only one allele is from the donor parent. On average, PIFs had a higher DGP than expected by chance (**Table 2**). For example, the expected DGP



<sup>∗</sup>α = 0.05 and ∗∗α = 0.01, respectively. PHT, EHT, TL, NNode and LA represents for plant height, ear height, tassle length, number of nodes, and leaf angle, respectively.

without selection is on average 0.78% for BC<sup>6</sup> individuals, but the observed DGP for BC<sup>6</sup> PIFs was 17.2 and 6.2% in PIFB47 and PIFZ51, respectively. On average, PIFB47 contained 12.9% higher DGP compared to PIFZ51. In contrast, BGEM lines showed a slightly reduced average DGP of 22.3% in DHB47 and 23.8% in DHZ51, compared with the expected 25% of BC1-DH. DGP and PHT are significantly correlated for both PIFB47 (r = 0.72) and PIFZ51 (r = 0.70). However, there was no significant linear relationship between DGP and PHT for BGEM lines.

#### Hormone Inhibitor Responses of PIFs and BGEM Lines

We applied BR and GA inhibitors to PIFB47, PIFZ51, DHB47, and DHZ51 to calculate the BR and GA inhibitor responses. The BR and GA inhibitor Pcz and Ucz greatly reduced mesocotyl length compared to water treatment (**Figure 2**). Heritabilities of BR inhibitor response (defined as MP/MW; see Materials and Methods) and GA inhibitor response (MU/MW) were 0.74 and 0.80, respectively. Ucz showed stronger effects than Pcz for reducing the mesocotyl length for both PIFs and BGEM lines (**Figure 3**). PHZ51 showed stronger BR and GA inhibitor response compared to PHB47. MP/MW was 0.19 and 0.35 for PHB47 and PHZ51, and MU/MW was 0.07 and 0.20 for PHB47 and PHZ51, respectively. 69% (PIFB47) and 45% (PIFZ51) of PIFs had increased BR inhibitor response (more tolerant to the BR inhibitor, in other words with an elevated BR level

TABLE 2 | Genetic characterization of PIFB47 and PIFZ51.


DGP represents for donor genome proportion.

or signaling) compared to their respective recurrent parent. GA inbititor response was 68% (PIFB47) and 38% (PIFZ51) increased for PIFs, compared to their respective recurrent parent. In contrast, only 6–13% of BGEM lines showed significant increased BR/GA inhibitor response compared to their recurrent parent, and the BR and GA inhibitor responses of the recurrent parents were always close to the median of DHB47 and DHZ51 (without phenotype selection) (**Figure 3**). The F<sup>1</sup> hybrid of PHB47 × PHZ51 showed significantly higher hormone inhibitor response for both BR and GA than most of the PIFs. Only one PIF from PIFZ51 showed higher BR inhibitor response than the hybrid of PHB47 × PHZ51 (**Figure 3**).

#### BR and GA Control of PHT

Within both PIFB47 and PIFZ51: BR and GA inhibitor responses, and PHT were significantly correlated (**Table 3**). PIFs with stronger BR or GA inhibitor response were taller in the field. In other words, the PIFs which were more tolerant to BR and GA inhibitors were also taller. In contrast, no significant correlation was found between BR inhibitor response and PHT or between GA inhibitor response and PHT for BGEM lines. This indicates that the relationship between BR, GA, and PHT was different in heterozygotes verses inbred lines. When assessed between the tallest individual and the shortest individual within PIFs, seed harvested from the tallest individual showed stronger hormone inhibitor response (either BR or GA or both) compared with the seed harvested from the shortest individual (**Supplementary Table S4**). For example, seed from the tallest individual was more tolerant to GA inhibitor compared to the seed from the shortest one of PIF113 (P = 0.0003) (**Figure 4**). These PIFs contained a high level of heterozygosity (%) with on average 40.4% (PIFB47) and 17.8% (PIFZ51).

Path analysis was used to connect the level of heterozygosity, BR inhibitor response, GA inhibitor response, and PHT separately within both PIFB47 and PIFZ51. Starting from the proposed model (see Materials and Methods), new links were added and old links were deleted based on the correlation structure. Path analysis for PIFB47 and PIFZ51 arrived at the same final best model, for which the direct correlation between heterozygosity and BR was deleted, and new links between BR and GA, and between heterozygosity and PHT was added (**Figure 5**) compared to the original proposed model. This was the only model fitting all criteria and it was associated with the lowest AIC value. According to the final model a high level of heterozygosity has a direct positive effect on PHT and two indirect paths to impact PHT: (i) a higher level of heterozygosity increases the GA level (synthesis or signaling), and increased GA promotes PHT; (ii) increased GA causes an increased BR level/signaling, which then promotes PHT.

#### Genome-Wide Association Analysis

Three sub-populations were obtained for joint analysis of PIFB47 and PIFZ51, with two subpopulations containing PHB47 and PHZ51, respectively (with limited donor introgression), and a mixed subpopulation containing a large (∼30%) DGP. With three GWAS models MLM, GLM+Q, FarmCPU for

PHB47 × PHZ51 were treated with gibberellin inhibitor Ucz (green), brassinosteroid inhibitor Pcz (orange), and Water (blue), respectively.

balancing false positives and false negatives (see Materials and Methods), we found that none of the SNP markers was found to be significantly associated with BR/GA inhibitor responses, PHT and other agronomic traits with a MLM (Q+K) model, which corresponds to previous studies that MLM is a very stringent model. With the FarmCPU method [Q+K model, but with controlled confounding effect between covariates (Q, K) and testing markers], we found one SNP SYN38535 on Chromosome 5 (282,014,45 bp; P-value 1.39 × 10−15) significantly associated with PHT (**Figure 6**) and one SNP PZE-108005623 on Chromosome 8 (567,899,5 bp; P-value 5.45 × 10−<sup>8</sup> ) significantly associated with GA inhibitor response (**Figure 6**). QQ-plot showed that population structure and kinship controlled false positives effectively (**Figure 6**). The number of associated loci was limited either due to the model stringency or limited number of markers, but detected loci has a high probability to be true positive. With a GLM+Q model, 73 SNPs were significantly associated with PHT (**Supplementary Table S5**) including SYN38535 from FarmCPU, and 22 of them overlapping with the 40 published PHT QTL hot spots. This corresponds to previous studies that GLM+Q model is less stringent compared to MLM and FarmCPU. Of these 73 SNPs, 41 were located within genes and these gene functions were summarized in **Supplementary** **Table S6** according to the function of their orthologes in rice and Arabidopsis. Among these genes, the four genes GRMZM2G348452, GRMZM2G082191, GRMZM2G100423, and GRMZM2G025171 are promising for further investigation, as the orthologes of these four genes – cytokinin oxidase, receptorlike protein kinase, cytochrome P450 and ATP synthase in Arabidopsis and rice are functioning in PHT control (Li et al., 2002; Komorisono et al., 2005; Zhu et al., 2006; Gao et al., 2014).

For GA and BR inhibitor response, we found five, and two significant SNPs with GLM+Q method, respectively (**Supplementary Table S5**), with one overlapping SNP marker SYN19928 on Chromosome 8 (128,133,446 bp) associated with both BR (P-value 3.62 × 10−<sup>5</sup> ) and GA (P-value 1.66 × 10−<sup>5</sup> ) inhibitor response. This SNP marker was found to be the closest linked marker in our dataset to a BR signaling pathway candidate gene ZmBSU1 with a 200 kb physical distance. For the other four SNPs associated with GA inhibitor response, three were close to GA candidate genes: PZE-108005623 (overlapped with FarmCPU method) and PZA00058.6 (Chromosome 8; 602,357,7 bp) are clustered and 2 Mb away from GA signaling candidate gene ZmRBX1A. Marker PZE-108070783 (Chromosome 8; 123,786,918 bp) is 3 Mb away from GA signaling candidate gene ZmCUL1. These three candidate genes ZmBSU1, ZmRBX1A,

and ZmCUL1 are expressed at seedling stage with the same absolute expression value of 3475.53 (Sekhon et al., 2011). For the other traits, four markers were associated with EHT, one with NNode, and one with LA (**Supplementary Table S5**). None of these markers were close to any BR or GA candidate genes.

### Co-localization of BR and GA Candidate Genes with PHT Associated SNP Markers

We systematically used ortholog information from Arabidopsis and rice to find all GA and BR candidate genes in maize. Compared with a different method, which used all previously reported genes encoding GA metabolism enzymes in maize and other species as BLAST queries to find GA candidate genes, all the GA candidate genes found in this study matched with their results (Song et al., 2011). With 1 Mb bin size, seven PHT associated SNP markers co-localized with GA candidate genes (**Table 4**). Except for marker PZA02388.01 and PZE-108073190, which are the third and second closest SNP markers to the candidate gene, all the other SNP markers are the closest SNP markers included in our marker dataset to the candidate gene. The probability of finding seven PHT associated SNP markers co-localized with GA candidate genes by chance with a 1 Mb bin size within our marker dataset is P = 0.041 based on a nonparametric resampling method. We found 2 BR candidate genes co-localized with PHT associated SNP markers with 1 Mb bin size (**Table 4**), and both of them belongs to a BR catabolic gene family BAS1.



∗∗α = 0.01, ∗∗∗α = 0.001, respectively; BR and GA represents for brassinosteroid and gibberellin inhibitor response, respectively; PHT represents for plant height.

fpls-08-01039 June 20, 2017 Time: 12:35 # 9

FIGURE 4 | Hormone inhibitor response of seed from tallest and shortest individuals of PIF113. Green, orange, and blue represents for gibberellin inhibitor Ucz, brassinosteroid inhibitor Pcz, and water treatment, respectively.

# DISCUSSION

#### Impact of Heterozygosity on Increased PHT in PIFs

Phenotypic selection for PHT and flowering time was successful in this study. Most PIFs (∼90%) were significantly taller than their respective recurrent parent (**Figure 1**), while flowering undistinguishable from their recurrent parents. Selection for PHT resulted in high levels of heterozygosity (**Table 2**). The percentage of heterozygous markers for PIFB47 and PIFZ51 after four to six generations of backcrossing was on average 34 and 9%, respectively. Percentage of heterozygosity was correlated with PHT in both PIFB47 and PIFZ51 (r ∼ 0.7). In contrast, only about 20% of BGEM lines were significantly taller than their recurrent parent, and there was no significant correlation between DGP and PHT. Without selection, the observed DGP is around 23%, close to the expected 25% for BC<sup>1</sup> derived doubled haploid lines.

Stronger selection was observed in PIFB47. PIFB47 was on average 23 cm taller than PIFZ51, and this difference was larger than the difference between PHB47 and PHZ51 (8 cm). Moreover, PIFB47 contained on average 25% more heterozygous markers than PIFZ51. In comparison, DHB47 was on average 2 cm taller than DHZ51, and both of them had an observed DGP around 23%. In addition, we compared a subset of PIFs from PIFB47 and PIFZ51, which were derived from the same set of donor parents backcrossed with both PHB47 and PHZ51. On average, these PIFs were 30 cm taller for PIFB47 compared to PIFZ51. The stiff stalk group (including PHB47) is more distantly related to tropical germplasm than the non-stiff stalk group (including PHZ51) (Liu et al., 2003). Thus, the chance of heterotic effects increases for crosses between PHB47 and tropical germplasm. Previous studies for U.S. maize germplasm showed that panmictic midparent heterosis (PMPH) linearly increased with increasing genetic distance (Moll et al., 1962), and PMPH also increased with increasing genetic distances among tropical and U.S. germplasm, unless affected by maladaptation problems (Moll et al., 1965). An increased divergence between two genotypes of a heterotic pattern increases the probability to select for complementary favorable alleles at different loci (Reif et al., 2005), which is consistent with our findings of accumulating more exotic regions at heterozygous state in PHB47 background.

chromosomes in maize and Y-axis represents for negative log10-transformed P-values. (B) X and Y axis represents for the expected and observed negative log10-transformed P-values for FarmCPU model, respectively.

# BR and GA Inhibitor Responses Are Correlated with PHT for PIFs

Brassinosteroid and GA are two classes of plant hormones that are regarded as major pathways controlling PHT (Wang and Li, 2008). The effect of BR and GA on PHT is manifested primarily through enhanced internode elongation resulting from both increased cell elongation and cell division (Zhang et al., 2007; Fernandez et al., 2009). Previous studies focused on two aspects to connect BR, GA, and PHT in maize from a genetic perspective: (1) identification of genes with profound effect on PHT leading to dwarf mutants, caused by defects in BR or GA synthesis or signaling pathway genes (Peng et al., 1999; Wang and Li, 2008; Fernandez et al., 2009; Hartwig et al., 2011; Makarevitch et al., 2012); (2) identification of PHT QTL, co-localized with BR and GA candidate genes (Fernandez et al., 2009; Weng et al., 2011; Teng et al., 2013; Peiffer et al., 2014). In this study, we found seven co-localized GA candidate genes with PHT associated SNP markers using a 1 Mb bin size, and this observed frequency of colocalization is significantly different from random co-segregation. In addition, two BR candidate genes were found to be colocalized with two PHT associated SNPs. With a diverse panel of 7000 accessions, Yang et al. (2015) successfully discovered colocalizations with 1 Mb bin size. Considering the limited number of backcross generations for generating PIFs, 1 Mb is a reasonable cutoff to serve as a linkage block in this study. In maize, there are around 32,000 genes predicted, with a 2700 Mb genome size of maize (Schnable et al., 2009). On average, there are around 32000/2700 = 12 genes within a 1 Mb region. Our assumption is, that BR and GA candidate genes are promising candidates for PHT control. In sorghum, two recent studies have used the Sequenom (SQNM) MassARRAY iPLEX platform (Gabriel et al., 2009) to develop molecular markers from GA and BR candidate genes (Perez et al., 2014; Zhao et al., 2016) for association studies, this method is promising to be applied in maize populations with low linkage disequilibrium but do not add more information for populations with extended linkage blocks such as in PIFs.

Previous studies used Pcz and Ucz as BR and GA inhibitors to phenocopy the effect of BR and GA biosynthetic gene mutations (Hartwig et al., 2012). With Pcz and Ucz treatment, mesocotyl length of maize seedlings is significantly reduced, and genotypes with higher BR and GA biosynthesis or signaling show increased tolerance to corresponding inhibitors. In this study, we applied Pcz and Ucz to compare the BR and GA inhibitor responses among PIFs and BGEM lines. We found that the pattern of hormone inhibitor responses was very similar to the pattern of PHT performance (**Figures 1**, **3**), as PIFs are on average more tolerant to both BR and GA inhibitors compared to their respective recurrent parents and BGEM lines, and PIFs are also much taller compared to respective recurrent parents and BGEM lines. In addition, when compared between different PIFs – taller PIFs are associated with stronger BR and GA inhibitor response within both PIFB47 and PIFZ51 (**Table 3**). Even within the same PIF, there were significant differences for hormone inhibitor response between offspring from the tallest and the shortest plants. In this study, the only selection applied was for PHT, not for hormone inhibitor response. This indicates that (1) "stronger" hormone pathway genes were directly selected for by selecting for increased PHT, or (2) there was a common pathway contributing to both taller PHT and increased hormone inhibitor response for PIFs.

### Crosstalk between BR and GA

We found significant correlations between BR and GA inhibitor responses for both the PIFs and BGEM lines (**Table 3**). PIFs and BGEM lines that are more tolerant to one of these two hormone inhibitors also tend to show tolerance to the other hormone inhibitor. Pathway analysis suggested that there was a direct crosstalk between the BR and GA pathways, supported by recent studies in Arabidopsis and rice (Bai et al., 2012; Gallego-Bartolomé et al., 2012; Li et al., 2012; Tong et al., 2014; Hofmann, 2015; Unterholzner et al., 2015). In addition, we found the same SNP, co-localized with a BR candidate gene ZmBSU1, to be associated with both BR and GA. This indicates that BSU1 may have a function in the interaction between BR and GA. A previous study showed that BZR1, a transcription factor activated by BR signaling and a DELLA protein, which inhibits the GA signaling pathway, are the key genes mediating crosstalk between BR and GA signaling pathways (Li et al., 2002; Bai et al., 2012). Since BSU1 is activating BZR1 activity (Ye et al., 2011), it explains detection of BSU1 as a modulator.

# Prediction of Heterosis in PHT by Early Monitoring of BR and GA Levels

Heterosis and inbreeding depression are considered two sides of the same coin (Mather and Jinks, 2013) and both of them are defined and quantified in relation to a reference


population. Any effect of inbreeding in a population will increase heterosis by the same amount (Miranda and Brasil, 1997). Heterosis is clearly related to heterozygosity, but it has long been debated how heterozygosity results in heterosis (Li et al., 2001). Inbreeding depression is caused by increased homozygosity of individuals (Charlesworth and Willis, 2009), as increased levels of homozygosity accumulate detrimental recessive mutations and reduce heterozygote advantages. Positive correlations between trait expression and level of heterozygosity are recognized as suggestive evidence for heterosis and inbreeding depression, and inbreeding coefficients estimated using homozygous SNPs was found to correlate well with pedigree inbreeding coefficient to infer inbreeding depression (Kim et al., 2015). In this study, we found significant positive correlations between PHT and the level of marker heterozygosity (r ∼ 0.7). Path analysis indicates that the level of heterozygosity is directly and positively correlated with GA levels (r ∼ 0.5), and seedling stage BR/GA inhibitor responses were positively correlated with field PHT for PIFs (r ∼ 0.5), but not BGEM lines (r ∼ 0.1). We were not able to directly correlate BR/GA with heterosis in PHT, as we only have recurrent parent and backcross progeny information (exotic donor parents are segregating accessions without available seed source and characterization). However, we were able to calculate the level of homozygosity, and correlate the level of homozygosity with PHT. Increased inbreeding significantly reduced PHT, with correlation (between the level of homozygosity and PHT) around 0.7. As the level of homozygosity is closely correlated with inbreeding depression (Kim et al., 2015), and inbreeding depression is the flip side of heterosis, we conclude that BR and GA promote heterosis for PHT (α = 0.05) in this study.

Previous studies have shown that GA levels are correlated with seedling heterosis: (1) hybrids were associated with higher GA levels than parental inbred lines in maize (Rood et al., 1988, 1992), sorghum (Rood et al., 1992), poplar (Bate et al., 1988) and rice (Ma et al., 2011), and when heterosis was not displayed due to unfavorable environmental conditions, the hybrid contained equal levels of endogenous GA-like substances (Rood et al., 1985) as the inbred parents; (2) maize inbred lines were more responsive to external application of GAs compared to hybrid progeny, suggesting a deficiency of endogenous GAs in inbred lines (Rood et al., 1988); (3) GA biosynthesis and positive signaling components were up-regulated in hybrids, whereas genes deactivating bioactive GAs, and negative GA signaling components were down-regulated, which together increase seedling heterosis (Ma et al., 2011). In addition to seedling heterosis, GA content was also found to be correlated with heterosis in PHT: it was reported that increased elongation of the uppermost internode contributed most to heterosis for PHT in wheat hybrids (Zhang et al., 2007). Examination of GA levels and activities in the uppermost internode tissue of wheat hybrids revealed, that (1) genes promoting GA biosynthesis were upregulated and GA deactivating genes were down-regulated, which resulted in higher level of GA in hybrids; (2) upregulation of GA receptors GID1 (for GA INSENSITIVE DWARF1) and positive regulator GAMYB, and down-regulation of negative component GAI resulted in enhanced GA sensitivity; (3) GA promoted the expression of expansion genes such as gibberellins induced proteins (GIPs) and endoxyloglucan transferase (XET), which promoted cell division and cell elongation, finally contributed to increased internode elongation and heterosis in PHT. After 4 days of germinating rice seed, the content of GA<sup>4</sup> started to be significantly higher in hybrids compared to their inbred parents (Ma et al., 2011). At both seedling and adult stages, GA levels were increased in hybrids. This explains why the seedling stage GA level was correlated with heterosis in PHT for PIFs. There are only few correlation studies between BR and heterosis available, although BRs function in both cell division and elongation in model species rice and Arabidopsis (Szekeres et al., 1996; Hu et al., 2000; Zhiponova et al., 2013; Tong et al., 2014). Despite BRs' broad effects on the physiological and developmental processes of plants, they were not widely recognized as plant hormones until the mid-1990s (Clouse, 1996; Divi and Krishna, 2009). We observed that the BR level was increased for PIFs with higher levels of heterozygosity, and path analysis indicated that this was due to an indirect effect from the crosstalk with the GA pathway. Nonetheless, BR was shown to directly increase PHT. BR and GA coordinately promote plant growth and development by jointly regulating the expression of specific groups of genes (Yang et al., 2004). Heterosis for PHT is also affected by other factors. For example, we did not measure the auxin response due to the complexity of the auxin biosynthesis pathway (Zhao, 2010), but polar auxin transport has been shown to affect PHT (Wang and Li, 2008).

Both DNA markers, transcripts, and metabolites have been evaluated for prediction of heterosis with various approaches (Andorf et al., 2010; Maenhout et al., 2010; Steinfath et al., 2010). Molecular markers associated with genomic regions contributing to heterosis can be identified with linkage or association mapping methods, and used in a linear regression approach to predict heterosis (Vuylsteke et al., 2000; Basunanda et al., 2010; Meyer et al., 2010; Tang et al., 2010). The advantage of transcriptome based prediction of heterosis over DNA markers is that transcript abundancies are resulted from the integration of the whole genome information, including DNA methylation level and histone modification status, thus the prediction accuracy is higher (Scholten and Thiemann, 2013). For metabolism based prediction of heterosis, metabolic markers were used for the prediction of biomass heterosis in Arabidopsis (Andorf et al., 2010), as metabolite levels are the result of more genes than those represented by genetic markers. Here, we used plant hormone inhibitors to assess the BR and GA level in different maize genotypes, and this information is significantly correlated with heterosis in PHT. With both seedling stage BR and GA inhibitor responses incorporated into a linear model, the prediction accuracies (r) are 0.62 in PIFB47 and 0.63 in PIFZ51. Our results indicate the possibility to use seedling stage hormone inhibitor response to predict heterosis for PHT in maize breeding projects, especially for biomass maize production. However, it needs to be noted that our prediction is based on the heterozygotes (PIFs) per se, instead of using their parental information. In other words, previous studies used molecular data from inbred parents to predict heterosis, whereas our prediction was from the same genotypes. To further investigate the usage of hormone based prediction, we measured 201 hybrids for

PHT across four replications, which were derived from BGEM lines (DHB47) × PHZ51, BGEM lines (DHZ51) × PHB47, and PHB47 × PHZ51. We calculated the Mid-parent heterosis as we measured PHT for F1, BGEM lines, PHB47 and PHZ51. We used parental BR and GA inhibitor response to predict Mid-parent heterosis, with data available in **Supplementary Table S7**. Additional phenotypic and genotypic data supporting this research is available in **Supplementary Tables S8–S10**. Neither BR nor GA inhibitor responses of the inbred parents were associated with heterosis in PHT, with correlations less than 0.1. Thus, our hormone based prediction method can only be used to predict adult performance based on seedling information of the same genotype, instead of using inbred parental information to predict heterosis in hybrids.

#### CONCLUSION

In this study, we used phenotypic selection to improve PHT and broadened the genetic variation of two maize heterotic groups (Stiff Stalk and Non-Stiff Stalk) by adapting multiple tropical and subtropical accessions into these two genetic pools. We found that phenotypic selection of PHT was genotype dependent, and stronger selection was observed when crosses were made between Stiff Stalk and tropical germplasm. Our main founding is that, for heterozygotes, GA activities were elevated with an increased level of heterozygosity. Increased GA promotes BR and they together lead to increased heterosis in maize PHT. If this can be generalized for other populations of hybrids (in addition to the crosses between temperate lines and tropical accessions), early stage monitoring of plant hormones is promising for predicting PHT for maize heterozygotes without growing all seed in the field.

#### AUTHOR CONTRIBUTIONS

SH and TL conceived the study, designed the experiments, discussed the results and finalized the manuscript; CW identified the candidate genes in BR and GA pathway; DS collected genotype and phenotype data for doubled haploids as a comparison; AL, PL helped with statistical analysis. YY

#### REFERENCES


established protocol for measuring BR/GA inhibitor response. TL, AL, PL, YY, MB edited the manuscript; all authors read and approved the final manuscript.

#### ACKNOWLEDGMENTS

We are grateful to China Scholarship Council for SH and CW funding. In addition, authors would also like to thank USDA's National Institute of Food and Agriculture (Project Numbers: IOW04314, IOW01018), as well as the Plant Sciences Institute, RF Baker Center for Plant Breeding and K. J. Frey Chair in Agronomy at Iowa State University for funding this work. Finally, we are grateful for Dr. Mingliang Xu and Dr. Yanling Guo for their help with genotyping, and Dr. Candice A. C. Gardner for providing doubled haploid seed.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2017.01039/ full#supplementary-material

TABLE S1 | Donor parent information and backcross generations for PIFs.

TABLE S2 | Maize candidate genes in Brassinosteroid (BR) and Gibberellin (GA) biosynthesis and signaling pathway.

TABLE S3 | Summary statistics for other agronomic traits (besides plant height) in PIFB47 and PIFZ51.

TABLE S4 | Comparison between the seed from the tallest/shortest individual within one PIF for hormone inhibitor response.

TABLE S5 | Markers associated with PHT, EHT, NNode, and LA from association study.

TABLE S6 | Plant height (PHT) associated SNP marker annotations and candidate gene functions.

TABLE S7 | Raw data for the prediction of Mid-parent heterosis for plant height.

TABLE S8 | Raw field phenotype data.

TABLE S9 | Raw data for Table 3.

TABLE S10 | Raw genotypic data.

Arbuckle, J. L., and Amos, A. J. (2006). 7.0 User's Guide. Chicago, IL: SPSS Inc.


lines in maize. Mol. Breed. 30, 1001–1016. doi: 10.1007/s11032-011- 9684-5


fpls-08-01039 June 20, 2017 Time: 12:35 # 15

for biofuel and biochemical production. Biotechnol. Biofuels. 4:43. doi: 10.1186/ 1754-6834-4-43


fpls-08-01039 June 20, 2017 Time: 12:35 # 16

expression in rice seedlings. Mol. Genet. Genomics 271, 468–478. doi: 10.1007/ s00438-004-0998-4


**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 Hu, Wang, Sanchez, Lipka, Liu, Yin, Blanco and Lübberstedt. 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.

# Genetic Regulation of GA Metabolism during Vernalization, Floral Bud Initiation and Development in Pak Choi (Brassica rapa ssp. chinensis Makino)

Mengya Shang1†, Xueting Wang1†, Jing Zhang<sup>1</sup> , Xianhui Qi <sup>2</sup> , Amin Ping<sup>1</sup> , Leiping Hou<sup>1</sup> , Guoming Xing<sup>1</sup> , Gaizhen Li <sup>2</sup> and Meilan Li <sup>1</sup> \*

#### Edited by:

*Yunde Zhao, University of California, San Diego, United States*

#### Reviewed by:

*Liwang Liu, Nanjing Agricultural University, China Paola Leonetti, Consiglio Nazionale Delle Ricerche (CNR), Italy*

#### \*Correspondence:

*Meilan Li 15935485975@163.com † These authors have contributed equally to this work.*

#### Specialty section:

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

Received: *07 May 2017* Accepted: *22 August 2017* Published: *30 September 2017*

#### Citation:

*Shang M, Wang X, Zhang J, Qi X, Ping A, Hou L, Xing G, Li G and Li M (2017) Genetic Regulation of GA Metabolism during Vernalization, Floral Bud Initiation and Development in Pak Choi (Brassica rapa ssp. chinensis Makino). Front. Plant Sci. 8:1533. doi: 10.3389/fpls.2017.01533* *<sup>1</sup> College of Horticulture, Shanxi Agricultural University, Taigu, China, <sup>2</sup> Vegetables Research Institute, Shanxi Academy of Agriculture Sciences, Taiyuan, China*

Pak choi (*Brassica rapa* ssp. *chinensis* Makino) is a representative seed vernalization vegetable and premature bolting in spring can cause significant economic loss. Thus, it is critical to elucidate the mechanism of molecular regulation of vernalization and floral bud initiation to prevent premature bolting. Gibberellin (GA) is the key plant hormone involved in regulating plant development. To gain a better understanding of GA metabolism in pak choi, the content of GA in pak choi was measured at different stages of plant development using enzyme-linked immunosorbent assay. The results showed that the GA content increased significantly after low-temperature treatment (4◦C) and then decreased rapidly with vegetative growth. During floral bud initiation, the GA content increased rapidly until it peaked upon floral bud differentiation. To elucidate these changes in GA content, the expression of homologous genes encoding enzymes directly involved in GA metabolism were analyzed. The results showed that the changes in the expression of four genes involved in GA synthesis (Bra035120 encoding ent-kaurene synthase, Bra009868 encoding ent-kaurene oxidase, Bra015394 encoding ent-kaurenoic acid oxidase, and Bra013890 encoding GA20-oxidase) were correlated with the changes in GA content. In addition, by comparing the expression of genes involved in GA metabolism at different growth stages, seven differentially expressed genes (Bra005596, Bra009285, Bra022565, Bra008362, Bra033324, Bra010802, and Bra030500) were identified. The differential expression of these genes were directly correlated with changes in GA content, suggesting that these genes were directly related to vernalization, floral bud initiation and development. These results contribute to the understanding of the molecular mechanism of changes in GA content during different developmental phases in pak choi.

Keywords: pak choi, vernalization, gibberellins metabolism, expression profiles, gene

# INTRODUCTION

Pak choi (Brassica rapa ssp. chinensis Makino) is a cruciferous vegetable and is representative plant that require vernalization. After exposure to low temperatures for a period of time, the plant blossoms under high temperatures and long days. A cultivar with low chilling requirements bolts easily; thus, premature bolting occurs frequently in spring and causes significant reduction in yield and quality (Liu et al., 2009). Therefore, it is crucial to study the molecular regulation of vernalization to prevent bolting in pak choi and possibly other cruciferous vegetables.

Plant hormones play an important role in vernalization and flowering (Aryal and Ming, 2014). Among the six known types plant hormones, gibberellin (GA) is particularly significant (Galvão et al., 2012). GA can partially replace long-day or low temperature to promote flowering in plants (Pearce and Dubcovsky, 2013; Hu et al., 2016). However, the phytohormone GA was shown to play duel, opposing roles in Arabidopsis. GA promoted the termination of vegetative development, but it inhibited flower formation (Yamaguchi et al., 2014). Treatment with GA has been shown to promote the flowering of pak choi (Hou et al., 2009; Yu et al., 2009). Therefore, studying the regulation of GA on flowering in pak choi is important.

The biosynthetic and deactivation pathways of GA have been studied in great detail (Liu et al., 2005). A variety of enzymes including ent-copalyl diphosphate synthase (CPS), entkaurene synthase (KS), ent-kaurene oxidase (KO), ent-kaurenoic acid oxidase (KAO), GA13-oxidase (GA13ox), GA20-oxidase (GA20ox), GA3-oxidase (GA3ox), and GA2-oxidase (GA2ox), participate in these pathways (Yu et al., 2002; Zhang et al., 2011). In Arabidopsis, CPS, KS and KO are encoded by single genes (GA1, GA2, and GA3, respectively), whereas GA20ox and GA3ox are encoded by small gene families named GA4 and GA5, respectively (Jiang et al., 2008).

The genes involved in GA metabolism have been isolated in a few plant species, including grape (Wang et al., 2012), Arabidopsis thaliana (Phillips et al., 1995), tomato (Rebers, 1999), Camellia lipoensis (Xiao et al., 2016), tomato (Shen et al., 2016), and pumpkin (Yamaguchi et al., 1996). Although these genes were found to affect the synthesis of GA, their expression patterns were different. In tomato, the expression of the GA1 homologous gene LeCPS increased during the early stage of floral bud development (Rebers, 1999). In pumpkin, GA2 was detected high transcription in the vegetative tissues (Yamaguchi et al., 1996). GA3 from Arabidopsis was expressed in inflorescence tissues (Helliwell et al., 1998). Nty, the GA4 ortholog in tobacco, was expressed in specific parts of the floral buds and floral organs (e.g., tapetum) (Yamaguchi et al., 1996; Itoh et al., 1999). GA5 was expressed in the stem and inflorescence of Arabidopsis during pod development (Phillips et al., 1995). These findings indicated that GA plays an important role in different stages of plant flowering. However, the process of GA metabolism during vernalization, floral bud initiation and development in pak choi remains poorly understood. In this study, the GA content in pak choi was measured at different growth stages, and the expression of gene encoding enzymes directly involved in GA metabolism was analyzed, which elucidated the molecular regulation mechanism of GA metabolism during vernalization, floral bud initiation and development in pak choi.

# MATERIALS AND METHODS

#### Plant Materials and Growth Conditions

Early-bolting pak choi inbred line "75# " was used in this study and was provided by the Institute of Vegetable Research, Shanxi Academy of Agricultural Sciences. The experiment included two treatments: low-temperature treatment and control. For the former (designated as V), the germinating seeds were kept at 4◦C for 20 days. In the control (designated as CK), the seeds were germinated at 25◦C. Seedlings with the same size were transplanted to trays with 50 holes full of substrate at the same time. The plants were then cultivated using traditional methods.

#### Sample Collection

After low-temperature treatment, the shoot apices were collected and designated as V0 and CK0 for the low-temperature treatment and control groups, respectively. On 10 days (vegetative growth), 15 days (immediately differentiation) and 16 days (floral bud differentiation stage 1) after transplantation, shoot apices were also collected from both groups, the low-temperature treatment and control samples collected at these times were named as V10, CK10, V15, CK15, V16, and CK16, respectively. Each 0.2 g sample was used to determine the GA content with three biological replicates. Each 0.1 g sample was used for RNA extraction. All samples were frozen in liquid nitrogen and stored at −80◦C.

#### Measurement of GA Content

The GA content at different developmental stages of pak choi were measured by enzyme-linked immunosorbent assay using the method described by Deng et al. (2008). Two types of GA (bioactive GA<sup>1</sup> and GA3) were measured.

#### Total RNA Extraction and cDNA Library Construction, Sequencing, Gene Expression Analysis, and Functional Annotation

Total RNA was extracted using an RNeasy Plant Mini Kit according to the manufacturer's instructions (QIAGEN, 74903). The samples used for sequencing were V0, CK0, V10, V15, CK16, and V16. All reads for each sequencing sample were deposited into GenBank, and the SRA accession numbers are SRP111574 (for V10), SRP064332 (for V0, CK0, V15, and CK16) (Sun et al., 2015) and SRP075755 (for V16) (Song et al., 2017; http://www.ncbi.nlm.nih.gov/sra). cDNA library construction and sequencing were performed by Biomarker Technologies Co., Ltd, Beijing, China. The methods of sequencing, expression analysis and functional annotation were previously described by Sun et al. (2015).

#### Quantitative Real-Time PCR

To validate the RNA sequencing results, quantitative real-time PCR (qRT-PCR) was performed using gene-specific primers for four randomly selected genes. Primer3 software was used to design specific primers. The ACTIN gene of pak choi was used as a reference gene. The RNA was extracted using an RNeasy Plant Mini Kit (QIAGEN, 74903). First-strand cDNA was obtained using a PrimeScript <sup>R</sup> RT reagent kit (Perfect Real Time; TaKaRa, RR037A). qRT-PCR was carried out using SYBR <sup>R</sup> Premix Ex TaqTM II (Tli RNaseH Plus; TaKaRa, RR820A) with an ABI 7500 instrument. The 25-µl reaction mixture included 20 ng cDNA. The program was as follows: 94◦C for 1 min; and 40 cycles of 94◦C for 30 s, 55◦C for 30 s, and 72◦C for 30 s. The relative expression levels were calculated using the 2−11Ct method (Stanko et al., 2014).

#### RESULTS

#### GA Contents of Pak Choi Shoot Apices at Different Growth Stages

The GA contents of the pak choi shoot apices at different growth stages were measured (**Figure 1**). After low-temperature treatment (V0), the GA content was 7.36 ng/g·FW, then decreased rapidly with vegetative growth and was approximately 6.80 ng/g·FW 10 d after transplanting (V10). Before floral bud differentiation (V15), the GA content increased rapidly to 9.34 ng/g·FW. Subsequently, the GA content increased slowly and reached 9.86 ng/g·FW at floral bud differentiation stage 1 (V16). Thus, the GA content first decreased and then increased after transplantation, peaking at floral bud differentiation stage 1. However, the GA content of the shoot apices in the control group did not show the same trend. In addition, the GA content in the low-temperature treatment group was significantly higher than in the control group, with the exception of 10 d after transplanting (V10 and CK10). This suggested that GA accumulation in the shoot apices can initiate floral bud differentiation, and low-temperature treatment can increase GA content.

# Quality Assessment of the Sequencing Results

The expression profiles of genes in the pak choi shoot apices at different growth stages were analyzed by RNA sequencing. As shown in **Table 1**, a total of 11,111,314, 11,268,910, 10,849,774, 11,072,438, 10,709,754, and 10,169,936 reads were produced from the V0, CK0, V10, V15, V16, and CK16 samples, respectively; clean reads (those remaining after filtering out the adaptor sequences, contaminating sequences and low-quality reads) accounted for 87.09, 99.86, 99.79, 99.90, 99.86, and 99.86% of the total reads, respectively, indicating that the sequencing quality was excellent.

To obtain gene expression information, high-quality clean reads from six libraries were aligned with the reference genome database. A total of 82.94, 84.80, 84.57, 85.08, 84.25, and 84.77% of the reads, including unique mapped reads and multiple mapped reads, were mapped to the reference genome for samples V0, CK0, V10, V15, V16, and CK16, respectively. The proportion



of mapped genes was high, indicating that the sequences and reference genome are suitable for further analysis. The unique mapped reads accounted for 86.17, 91.60, 93.45, 93.79, 93.83, and 93.03% of total mapped reads in the six libraries for V0, CK0, V10, V15, V16, and CK16, respectively, and could therefore be used for further analysis.

# Expression Analysis of Genes Encoding GA-Metabolizing Enzymes

Unlike the majority of GA, which is inactive, GA<sup>1</sup> and GA<sup>3</sup> have biological activity (Pan, 2008). GA is mainly synthesized in the stems and roots of plants, and a small amount is synthesized in mature leaves. It is not transported to other tissues. In cells, GA is synthesized in the plastids, endoplasmic reticulum and cytoplasmic matrix (Reinecke and Pharis, 2013). The KEGG pathway map of diterpenoid biosynthesis (ko00904) from RNA sequencing was shown in **Supplementary Figure S1**, in which the GA metabolism is marked in red. GA is synthesized in three steps from geranylgeranyl pyrophospha (GGPP) as a precursor. The first step occurs in the protoplast. GGPP forms copal pyrophosphate under the catalysis of CPS (5.5.1.13) and then forms entkaurene with catalysis by KS (4.2.3.19). The second step takes place in the endoplasmic reticulum with O<sup>2</sup> as the substrate, NADPH as the cofactor (Rademacher, 2003), and KO (1.14.13.78) as the catalyst. Ent-kaurene undergoes constant oxidation, generating ent-Kaur-16-en-19-ol, ent-Kaur-16-en-19-al, and ent-Kaur-16-en-19-oate. Finally, GA12-aldehyde is produced under the catalysis of KAO (1.14.13.79). The third step is carried out in the cytoplasm. GA12-aldehyde forms different structures of GA under the catalysis of different enzymes. During this process, under the action of GA20ox (1.14.11.12), GA3ox (1.14.11.15) and cytochrome P450 monooxygenase (P450-1, 2, 3), GA12-aldehyde generates bioactive GA<sup>1</sup> and GA3. Under the action of GA2ox (1.14.11.13), GA<sup>1</sup> undergoes 2β -hydroxylation to form inactive GA<sup>8</sup> and GA34. GA20ox and GA2ox play key roles in the entire process of GA metabolism.

Some genes encoding enzymes related to GA metabolism in pak choi were listed in **Table 2**. Seven enzymes involving 18 genes were found to participate in GA synthesis. The enzyme involved in the catabolism of GA is mainly GA2-oxidase, 11 genes were found encoding the enzyme in pak choi.

Based on the results of the gene expression profiles, the expression of genes encoding enzymes involved in GA metabolism were analyzed. The results showed that the expression of Bra000864 encoding CPS, Bra035120 encoding KS, Bra009868 encoding KO, Bra015394 encoding KAO, Bra019165 and Bra013890 encoding GA20ox were all upregulated in CK0 vs. V0, V10 vs. V15, and CK16 vs. V16; that is, the gene expression in cold treatment sample were higher than the control after cold treatment and at floral bud differentiation stage 1. Furthermore, the gene expression in shoot apices was higher immediately prior to floral bud differentiation than in the vegetative growth stage. The expression of Bra009285 encoding GA20ox was up-regulated after low-temperature treatment and immediately prior to floral bud differentiation; however, its expression was not detected in shoot apices at floral bud differentiation stage 1, indicating that low temperature promoted its expression. Bra026757 encoding GA3ox was expressed only in immediately prior to floral bud differentiation; it was not expressed in CK0, V0, V16 or CK16. Thus, the expression of Bra026757 was lower in shoot apices at the vegetative growth stage than in the shoot apices immediately prior to floral bud differentiation.

The expression of some genes that participate in GA biosynthesis were irregular. For example, the expression of Bra036239 encoding CPS, Bra005596 encoding KAO and Bra008480 encoding GA3ox were up-regulated after lowtemperature treatment, indicating that the low temperature promoted their expression. Bra028706 and Bra022565 encoding GA20ox were up-regulated after low-temperature treatment and during floral bud differentiation; that is, their expression were higher in the low-temperature treatment group compared to in the control after treatment and at floral bud differentiation stage 1. Bra028277 encoding GA20ox and Bra026122 encoding GA3ox were up-regulated in floral bud differentiation and down-regulated in CK0 vs. V0 and V10 vs. V15, indicating that their expression at floral bud differentiation stage 1 were greater than in the control. These results indicated that the high expression of Bra028706, Bra022565, Bra028277, and Bra026122 are beneficial for the flowering of pak choi.

The above results showed that the expression of genes involved in GA biosynthesis were up-regulated, indicating that GA synthesis was accelerated. The measured GA contents were higher in V0, V15, and V16 than in CK0, V10, and CK16, respectively, and the changes in gene expression were consistent with the changes in GA content.

Among the genes encoding GA2ox involved in GA catabolism, the changes in gene expression were irregular. For example, the expression of Bra030500, Bra010802, and Bra030187 were down-regulated after low-temperature treatment, indicating that low temperature inhibited their expression. The expression of Bra008362 was down-regulated after low-temperature treatment and in floral bud differentiation. The expression of Bra033324, Bra020890, and Bra038780 were down-regulated in floral bud differentiation but up-regulated in CK0 vs. V0 and V10 vs. V15. These results indicated that the expression of these genes were down-regulated, causing GA to decompose more slowly, thereby increasing GA content. This further explains the change in GA content observed in pak choi.

In addition, the genes Bra027106 encoding GA20ox, Bra008481, and Bra020909 encoding GA3ox involved in GA synthesis, and Bra005415 encoding GA2ox involved in GA decomposition were not expressed during vernalization, floral bud initiation and development.

#### Identification of Differentially Expressed Genes Related to GA Metabolism

According to the sequencing results, the expression of genes related to GA metabolism were analyzed in different periods (**Table 2**). After low-temperature treatment, 12 genes related to GA biosynthesis were up-regulated compared to the control, while two were down-regulated; three genes were up-regulated

#### TABLE 2 | Enzymes and their coding genes in GA metabolism (RPKM: reads per kb per million reads).


and four were down-regulated in the catabolic process. When the shoot apices immediately enter the floral bud differentiation, 10 genes related to GA biosynthesis were up-regulated compared to in the vegetative stage, whereas five were down-regulated; four were up-regulated, and six were down-regulated in the catabolic process. At floral bud differentiation stage 1, 10 genes related to GA biosynthesis were up-regulated compared to in the vegetative stage, while three genes were down-regulated; six gene were up-regulated and four were down-regulated in the catabolic process. In summary, after low-temperature treatment, more genes involved in GA synthesis were up-regulated than down-regulated during vegetative growth and at floral bud differentiation stage 1; for the genes taking part in GA catabolism, slightly more genes were down-regulated than up-regulated, except for at floral bud differentiation stage 1. This may have led to the higher GA contents in samples V0, V15, and V16 compared to in samples CK0, V10, and CK16. These results were coincide with the measured GA contents.

According to FDR < 0.01 and Fold Change ≥ 2, the differentially expressed genes (DEGs) were identified (**Table 3**). The DEGs related to GA metabolism were not found in floral bud differentiation stage 1 (V16) vs. the vegetative shoot apices (CK16). Four DEGs after low-temperature treatment (V0 vs. CK0) and three DEGs between V15 and V10 were identified by gene ontology (GO); all of them were involved in GA biosynthesis (GO:0009686) or catabolism (GO:0045487).

Bra005596, the KAO2 homologous gene, was up-regulated after low-temperature treatment. In the second step of GA biosynthesis, in which kaurene generates ent-Kaur-16-en-19 ol, ent-Kaur-16-en-19-al, and ent-Kaur-16-en-19-oate followed


TABLE 3 | Differentially expressed genes related to GA metabolism at different growth stages of pak choi.

by GA12-aldehyde under the catalysis of cytochrome P450 monooxygenase, KAO catalyzed ent-kaurenoic acid to form GA12-aldehyde, which belongs to the cytochrome P450 CYP88A subfamily (Helliwell et al., 2001a,b). The subcellular localization of AtKAO1 and AtKAO2 revealed that they located in the endoplasmic reticulum (Helliwell et al., 2001b), where the second stage of GA synthesis is carried out. In addition, AtKAO2 was highly expressed in germinating seeds, flowers and fruit pods (Regnault et al., 2014). The expression of KAO2 was up-regulated in pak choi, which was beneficial for early flowering.

Both Bra009285 and Bra022565, which encode GA20ox, were up-regulated after low-temperature treatment. Five homologous genes encode GA20ox in Arabidopsis: AtGA20ox1, AtGA20ox2, AtGA20ox3, AtGA20ox4, and AtGA20ox5. Bra009285 and Bra022565 corresponded to AtGA20ox3 (YAP169) and AtGA20ox2, respectively. AtGA20ox3 was expressed in dry seeds, imbibed seeds and silique (Yan et al., 2014) and was found to aid GA biosynthesis in multiple biological processes (Plackett et al., 2012). GA20ox2 promotes floral bud formation and fruit pod growth. The over-expression of GA20ox can lead to early flowering (Sun, 2008). The up-regulation of Bra009285 and Bra022565 was speculated to contribute to early flowering in pak choi.

Among the genes encoding GA2ox involved in GA catabolism, the GA2ox6 homologous genes Bra030500 and Bra033324 were down-regulated after low-temperature treatment and immediate differentiation, respectively. The ATGA2ox2 and ATGA2ox1 homologous genes were Bra010802 and Bra008362, respectively, which were down-regulated and up-regulated after transplantation, respectively. Active GA2 oxidase can be inactivated by 2β -hydroxylation, which is one of the main methods of regulating GA activity in cells (Rieu et al., 2008; Wiesen et al., 2016). In Arabidopsis, five genes encode GA2 oxidase: GA2ox1, GA2ox2, GA2ox3, GA2ox4, and GA2ox6. The over-expression of these genes in Arabidopsis causes dwarfing, reduces the GA content, and delays flowering (Wang et al., 2004). Similar results were also obtained in transgenic poplar (Populus L), rice and bean (Sakamoto et al., 2001; Busov et al., 2003; Appleford et al., 2007). In this study, the genes encoding GA2ox expressed were down-regulated after low-temperature treatment, which may have increased the content of active GA. The expression of Bra030500, Bra033324 and Bra010802 were down-regulated, which was consistent with the GA content. However, the expression of Bra008362 was up-regulated, which did not coincide with the GA measurements. Further analysis is required to explain this discrepancy.

#### qRT-PCR Verification

To verify the reliability of the gene expression profiling, Bra000393, Bra004928, Bra033324, and Bra010802 were selected for qRT-qPCR verification in six samples (CK0, V0, V10, V15, CK16, and V16). The qRT- PCR results were consistent with the sequencing results, confirming the reliability of the sequencing results (**Figure 2**).

# DISCUSSION

#### Relationship between GA and Vernalization, Floral Bud Initiation and Development

Hormones and GA in particular play an important role in plant vernalization, floral bud initiation and development. GA has been shown to induce the flowering of long-day plants under shortday conditions (Kinet, 1993; Buchanan et al., 2002). The GA content in germinating pak choi seeds treated at low temperature was approximately three times higher than that of the control (Li et al., 2002a). This suggested that vernalization treatment can promote the biosynthesis of endogenous GA, in agreement with the report of Burn et al. (1993). The results also suggest that vernalization can increase the activities of GA-biosynthetic enzymes and thus increase the content of GA. In broccoli, the GA content after low-temperature treatment was three times that in the control; the GA content then gradually decreased with green vernalization, reached a minimum at the critical period of floral bud differentiation, and then increased sharply (Jiang and Yu, 2004). The GA content was low in the vegetative growth stage of rape, increased rapidly during floral bud differentiation, and decreased sharply after floral bud differentiation (Li, 2010). For Chinese flowering cabbage, the GA content of the stem tip was highest in the vegetative period, decreased when floral bud differentiation was about to start, and then increased slightly when the floral bud began to differentiate (Li et al., 2002b). The variation of GA content in Chinese cabbage during floral

bud differentiation was similar to that observed in Chinese flowering cabbage (Li, 2002). In summary, in brassicaceous vegetables, the GA content increases after cold vernalization, decreases when floral bud differentiation is imminent, and then gradually increases after floral bud differentiation. In this study, the GA content of pak choi was significantly higher after low-temperature treatment than in the control. The GA content decreased with vegetative growth and increased after floral bud differentiation, consistent with the results of previous studies. However, right before floral bud differentiation, the GA content was inconsistent. One possible explanation for this is that pak choi needs more GA at the beginning of floral bud differentiation.

#### Expression of Some Important Genes Involved in GA Metabolism

To further clarify the cause of the variation in GA content in pak choi during different growth stages, the GA metabolic pathway was analyzed. In the synthesis of GA, the conversion of GGPP to ent-kaurene is catalyzed CPS and KS. CPS is the key enzyme in GA biosynthesis. The oxidation of entkaurene to GA12-aldehyde is catalyzed by cytochrome P450 monooxygenase. The first three steps from ent-kaurene to entkaurene acid are catalyzed by KO (Helliwell et al., 1999). The following three steps from ent-kaurene acid to GA12 aldehyde are catalyzed by KAO (Helliwell et al., 2001b). Both KO and KAO belong to the cytochrome P450 monooxygenase family. The final step in GA biosynthesis is the formation of bioactive GA catalyzed by GA20-oxidase and GA3-oxidase. GA2 oxidase catalyzes the formation of inactive GA. These enzymes play important roles in GA metabolism and affect the level of active GA in plants, thereby affecting plant growth and development.

By analyzing the expression profiles, the changes in the expression of four genes encoding enzymes related to GA metabolism were found to be consistent with the changes in GA content. These genes were Bra035120, Bra009868, Bra015394 and Bra015394, and their Arabidopsis homolog genes encode KS1, KO, KAO1, and GA20ox1, respectively. In the first stage of GA synthesis, KS catalyzes the synthesis of ent-kaurene from copal pyrophosphate. In Arabidopsis, AtKS1 was located in the chloroplast stroma by subcellular localization (Helliwell et al., 2001b). The loss of function of KS will lead to serious plant dwarfing and loss of reproductive capacity (Regnault et al., 2014). In this study, Bra035120 was up-regulated with the growth and development of pak choi, and its expression was consistent with the GA content of the shoot tips, in agreement with previous results. KO and KAO, which are members of the cytochrome P450 family of enzymes, participated in the second phase of GA biosynthesis. The process from ent-kaurene to ent-kaurene acid is catalyzed by AtKO1, which belongs to the cytochrome P450 CYP701A subfamily (Regnault et al., 2014). The up-regulation of AtKO1 in transgenic Arabidopsis resulted in an increase in GA<sup>4</sup> content (Xu et al., 2015). KAO1 is highly expressed in the seeds and flowers of Arabidopsis, while its expression is low in seedlings and vegetative tissues (Xu et al., 2015). Bra009868 and Bra015394 have been suggested to be beneficial for the synthesis of active GA, which agrees with the results of this study. In Arabidopsis, the homologous gene of Bra013890 is GA20ox1, which belongs to the GA20ox family and is highly expressed in stems and flowers (Yan et al., 2014). GA20ox1 plays an important role in the internode elongation of stamen filaments (Sun, 2008). The silencing of SlGA20ox1 does not contribute to pollen production in tomato (Olimpieri et al., 2011). Thus, Bra035120, Bra009868, Bra015394, and Bra013890 are thought to play important roles in the vernalization, floral bud initiation and development of pak choi.

Seven DEGs related to GA metabolism (Bra005596, Bra009285, Bra030500, Bra022565, Bra033324, Bra010802 and Bra008362) were identified by analyzing expression profiles. In Arabidopsis, their respective homologous genes are KAO2, YAP169 (GA20ox3), GA2ox6, GA20ox2, GA2ox6, ATGA2OX2, and ATGA2OX1, respectively. KAO mutations have been shown to cause severe stunting in a variety of plants including barley, rice, corn, peas and sunflowers (Helentjaris, 1995; Davidson et al., 2003; Sakamoto et al., 2004; Fambrini et al., 2011). KAO is thought to play an important role in the synthesis of GAs. YAP169 (GA20ox3) and GA20ox2 are members of the GA20ox family. The over-expression of GA20ox leads to early flowering and stem elongation (Sun, 2008). GA2ox6, ATGA2OX2, and ATGA2OX1 encode GA2ox, and the over-expression of GA2ox can delay plant flowering and cause dwarfing (Sun, 2008). The over-expression of SlGA2ox1 leads to reduction in endogenous GA in tomato (Shen et al., 2016). GA3ox is also a key enzyme in GA biosynthesis, and its mutation leads to dwarfing and flower hypoplasia (Sun, 2008). DEGs encoding GA3-oxidase were not found in this study. It may be that GA20-oxidase and GA2-oxidase play important roles in regulating the level of active GA in plants, while GA3-oxidase has little effect on the late stage of GA biosynthesis in pak choi (**Supplementary Figure S1**). These results improve our understanding of the molecular flowering mechanism in pak choi.

## CONCLUSION

In this study, the GA content of the shoot tips at different stages in pak choi were measured. The results showed that low-temperature treatment increased the GA content, and enhanced GA accumulation initiated floral bud differentiation. To elucidate the molecular mechanism responsible for the changes in GA content, the expression of genes encoding enzymes involved in GA metabolism were analyzed based on the RNA sequencing. The results showed that the expression patterns of most genes involved in GA metabolism, particularly those of four genes (Bra035120, Bra009868, Bra015394, and Bra013890) coincided with the observed changes in GA content. In addition, by analyzing differentially expressed genes at different growth stages, seven genes (Bra005596, Bra009285, Bra022565, Bra008362, Bra033324, Bra010802, and Bra030500) were identified, and their expression were also consistent with the measured GA contents. These results contribute to the understanding of the molecular mechanism of changes in GA content during different developmental phases in pak choi.

#### AUTHOR CONTRIBUTIONS

XW and MS performed the experiment, data analysis and prepared the manuscript. XQ and AP participated in performed the experiments. LH, GX, and GL participated in the experimental design. ML conceived the idea and participated in the interpretation of results and preparation of manuscript. JZ revised the manuscript. All authors read and approved the final manuscript.

# FUNDING

This work was supported by the Key Scientific and Technological Project of Shanxi Province [grant No. FT201402-06]; the Innovation and Utilization of Chinese Cabbage Germplasm in the Key Science and Technology Innovation Team of Shanxi Province [grant No. 2014131016]; the Natural Science Foundation of Shanxi Province [grant No. 201701D121101]; and the Agricultural Science and Technology Research Plan Project of Shanxi Province (20150311010-2).

#### ACKNOWLEDGMENTS

The authors would like to thank Prof. Yu Gao for his critical revision of this manuscript. The RNA sequencing

#### REFERENCES


were carried out by Biomarker Technologies Co., Ltd, Beijing, China.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2017. 01533/full#supplementary-material

Supplementary Figure S1 | Biosynthesis of diterpenoids (GA biosynthesis is marked in red).


**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 Shang, Wang, Zhang, Qi, Ping, Hou, Xing, 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.

# Identification, Classification, and Expression Analysis of *GRAS* Gene Family in *Malus domestica*

Sheng Fan, Dong Zhang, Cai Gao, Ming Zhao, Haiqin Wu, Youmei Li, Yawen Shen and Mingyu Han\*

*College of Horticulture, Northwest A&F University, Yangling, China*

*GRAS* genes encode plant-specific transcription factors that play important roles in plant growth and development. However, little is known about the *GRAS* gene family in apple. In this study, 127 *GRAS* genes were identified in the apple (*Malus domestica* Borkh.) genome and named *MdGRAS1* to *MdGRAS127* according to their chromosomal locations. The chemical characteristics, gene structures and evolutionary relationships of the *MdGRAS* genes were investigated. The 127 *MdGRAS* genes could be grouped into eight subfamilies based on their structural features and phylogenetic relationships. Further analysis of gene structures, segmental and tandem duplication, gene phylogeny and tissue-specific expression with ArrayExpress database indicated their diversification in quantity, structure and function. We further examined the expression pattern of *MdGRAS* genes during apple flower induction with transcriptome sequencing. Eight higher *MdGRAS* (*MdGRAS6*, *26*, *28*, *44*, *53*, *64*, *107,* and *122*) genes were surfaced. Further quantitative reverse transcription PCR indicated that the candidate eight genes showed distinct expression patterns among different tissues (leaves, stems, flowers, buds, and fruits). The transcription levels of eight genes were also investigated with various flowering related treatments (GA3, 6-BA, and sucrose) and different flowering varieties (Yanfu No. 6 and Nagafu No. 2). They all were affected by flowering-related circumstance and showed different expression level. Changes in response to these hormone or sugar related treatments indicated their potential involvement during apple flower induction. Taken together, our results provide rich resources for studying *GRAS* genes and their potential clues in genetic improvement of apple flowering, which enriches biological theories of *GRAS* genes in apple and their involvement in flower induction of fruit trees.

Keywords: apple, *GRAS*, synteny, expression analysis, flower induction

# INTRODUCTION

Transcription factors function as trans-acting factors, combining with specific cis- elements in eukaryotic gene promoter regions to regulate plant growth and development.

GRAS genes encode a family of plant-specific transcription factors. The GRAS name derives from the first letters of the first identified three members, including GAI (gibberellic acid insensitive), RGA (repressor of GA1-3 mutant), and SCR (scarecrow) (Bolle, 2004). In plants, proteins in the GRAS family consist of 400–700 amino acids with a conserved GRAS carboxyl

#### *Edited by:*

*Yong-Ling Ruan, University of Newcastle, Australia*

#### *Reviewed by:*

*Stephen Beungtae Ryu, Korea Research Institute of Bioscience and Biotechnology, South Korea Chi-Kuang Wen, Chinese Academy of Sciences, China*

> *\*Correspondence: Mingyu Han hanmy@nwsuaf.edu.cn*

#### *Specialty section:*

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

*Received: 07 January 2017 Accepted: 10 April 2017 Published: 28 April 2017*

#### *Citation:*

*Fan S, Zhang D, Gao C, Zhao M, Wu H, Li Y, Shen Y and Han M (2017) Identification, Classification, and Expression Analysis of GRAS Gene Family in Malus domestica. Front. Physiol. 8:253. doi: 10.3389/fphys.2017.00253*

**262**

terminus comprising several typical structures, including seven leucine-repeat domains (LHRI), a VHIID domain, seven leucinerepeat II domains (LHRII), and PFYRE and SAW motifs (Hirsch et al., 2009). Bacteria also harbor GRAS-like proteins, which may function as methylases (Zhang et al., 2012). In recent years, Many GRAS genes have also been identified in a variety of plant species, including Prunus mume, Populus, grape, tomato, Chinese cabbage, and castor bean (Tian et al., 2004; Song et al., 2014; Huang et al., 2015; Lu et al., 2015; Grimplet et al., 2016; Xu W. et al., 2016).The GRAS family, which has 33 members in Arabidopsis thaliana and 57 members in rice (Oryza sativa), can be divided into eight subfamilies: SCL3, SHR, PAT1, LISCL, DELLA, SCR, LS, and HAM (Hirsch et al., 2009).

Recent studies have revealed that the N-termini of GRAS proteins play important roles in their specific functions. The N-termini contain intrinsically disordered regions (IDRs); thus, GRAS proteins are also known as intrinsically disordered proteins (IDPs) (Sun et al., 2011). Several studies have focused on the structures and functions of the IDRs in the N-termini of GRAS proteins (Sun et al., 2010, 2011, 2012). Different subfamilies have different types and numbers of molecular recognition features in their N-termini (Sun et al., 2010, 2011, 2012). For instance, the DELLA subfamily has three molecular recognition features in its IDR at the N terminus: DELLA, VHYNP, and L(R/K) XI. Protein motifs at the C-terminus include the LHR I, HIID, LHRII, PFYRE, and SAW motifs (Pysh et al., 1999).

The functions of GRAS family genes have been studied in recent years; there is a large body of evidence that the GRAS genes family play important roles in various biological processes in many different plant species. They are involved in diverse processes related to root meristem development, axillary bud outgrowth, phytochrome signaling pathways, nodular signal transduction, and stress adaptation, especially in the gibberellins (GA)-signaling pathway (Ait-ali et al., 2003; Greb et al., 2003; Heckmann et al., 2006; Torres-Galea et al., 2006; Koizumi et al., 2012; Park et al., 2013). DELLA proteins are negative regulators of the GA signal, and have been shown to repress growth in many different plant species including Arabidopsis (AtGAI, AtRGA, AtRGL1, AtRGL2, and AtRGL3), O. sativa (OsSLR1), Zea mays (ZmD8), Malus domestica (MdRGL2a), and Vitis vinifera (VvGAI) (Foster et al., 2007; Vandenbussche et al., 2007; Aleman et al., 2008; Dai and Xue, 2010). Various studies have shown that the GA-GID1-DELLA complex and ubiquitin ligase E3 play important roles in GA signaling, and that SCR functions in the development of roots and aboveground organs in Arabidopsis (Zhang et al., 2011; Koizumi et al., 2012). PhHAM (Petunia hybrid), a petunia hairy meristem gene, was shown to be necessary to maintain the apical meristem in petunia, and the petunia recessive mutant phham (ham-B4281) produced fewer flowers than wild type (Stuurman et al., 2002). SCL13 was shown to positively regulate phyB, and the scl13 mutant showed impaired sensitivity in Arabidopsis (Torres-Galea et al., 2006). Several studies have shown that some GRAS genes are up-regulated under stress conditions in Arabidopsis, rice and tobacco (Day et al., 2003; Torres-Galea et al., 2006; Czikkel and Maxwell, 2007). Among all the biological processes, their flowering related characterizations and functions have been researched in some model plants. On the one hand, several GRAS genes are associated with the flowering phenotype. For example, in tomato, Ls (Later Suppressor), defined as another GRAS gene, its mutant ls showed lower numbers of inflorescence (Schumacher et al., 1999; Bolle, 2004). In Arabidopsis, the LS subfamily gene AtLAS and the SHR subfamily gene AtSHR were involved in shoot apical meristem. The Ls subfamily gene AtSCL26 and the HAM subfamily genes AtSCL6 and AtSCL27 were highly expressed in flowers (Bolle, 2004); the rga and gai mutant also showed earlier flowering in Arabidopsis (Galvao et al., 2012). On the other hand, GRAS genes can integrate or regulate flowering related genes such as SPL (SQUAMOSA PROMOTER BINDING PROTEIN-LIKE), LFY (LEAFY), and AP1 (APETALA1) as well as FT (FLOWERING LOCUS T). They all acted positive roles and intergrated various signals in flower induction, of which FT and SPL affected flowering time; while LFY and AP1 belonged to flower meristem identify genes. Several GRAS genes can be recruited and integrated with SPL proteins to regulate LFY and AP1 expression in Arabidopsis (Yamaguchi et al., 2014). Additionally, the PAT and SCL subfamilies of GRAS genes were both reported involved in photosignal in many plants, it influenced the expression of phyA (phytochrome A), a phytochrome A signal transduction gene, which caused to the change of the circadian clock and expression of FT and other flowering genes (Bolle et al., 2000; Torres-Galea et al., 2006; Li et al., 2014). However, little is known about the GRAS family of transcription factors in woody fruit trees, such as its expression in response to flower induction in apple.

Apple is one of the most widely cultivated and economically important fruit trees in the temperate regions of the world. In apple production, flower induction and flower buds formation sustain two growing seasons. Flower induction is a key stage, and always limits fruit yield (Guitton et al., 2012). Apple (Malus domestica Borkh.) cv. "Fuji" is one of the most popular cultivars, and it accounts for 65% planting areas in China. "Fuji" has trouble in flowering induction. Thus, characteristic the molecular mechanism of apple flower induction is necessary. Flower induction was affected by environmental and internal factors (Guitton et al., 2012; Xing et al., 2014, 2015; Fan et al., 2016). Sugar and hormones were involved in flower induction (Guitton et al., 2012; Xing et al., 2014, 2015; Fan et al., 2016; Li et al., 2016; Zhang et al., 2016). Among all kinds of hormones, including auxin, cytokinins, abscisic acid, GA and other hormones. GA was defined the most important hormones in regulating flower induction and GA pathway was recognized as one of the flowering pathways (Guitton et al., 2012; Porri et al., 2012; Li et al., 2016; Zhang et al., 2016). 6-BA and sucrose, which can promote flower bud differentiation, were also important and widely used (Xing et al., 2015; Li et al., 2016). Several floweringrelated genes families have been identified and characterized

**Abbreviations:** DAFB, days after full blossom; ES, early stage; MS, middle stage; LS, later stage; GA3, gibberellic acid; 6-BA, 6-benzylaminopurine.

in the apple genome, including SPL, IDD, and MADX-box (Li et al., 2013; Kumar et al., 2016; Fan et al., 2017). GRAS genes were reported to partially or entirely involve in flower induction in various plants (Hynes et al., 2003; Foster et al., 2007; Vandenbussche et al., 2007; Galvao et al., 2012; Yamaguchi et al., 2014). Little is known about MdGRAS genes and their potential involvement in apple flower induction. The recently published genome sequence of apple (Velasco et al., 2010) has made it possible to characterize the structures, functions, evolution of apple GRAS genes and their response to flower induction. Moreover, it is also available to utilize candidate MdGRAS genes for further genetic improvement aiming at flowering induction.

In this study, we performed a genome-wide survey to identify GRAS genes in the apple genome. A systematic analysis including gene classification, gene characterization, gene structures and gene phylogenies were determined. We further investigated their expression levels of all the 127 putative MdGRAS genes among different tissues and varieties during apple growth and development with ArrayExpress database. Moreover, eight MdGRAS genes were further analyzed.

We investigated their expression levels of the candidate eight genes among different tissues (stems, leaves, flowers, fruits, and buds), different flowering-related treatments (GA3, 6-BA and sugar) and different flowering varieties (Yanfu No. 6 and Nagafu No. 2). Changes in response to these flowering-related treatments indicated their potential involvement in apple flower induction. Our results provided basic clues for further analyses of MdGRAS in apple growth and development.

# MATERIALS AND METHODS

#### Identification of *GRAS* Genes in Apple

Hidden Markov Model (HMM) searches (Finn et al., 2011) were performed to identify GRAS genes in the apple genome. The HMM profiles (PF03514) were downloaded from the Pfam database (http://pfam.sanger.ac.uk). Additionally, AtGRAS protein sequences were used as queries for BLASTP search against the Apple Genome Database (https://www.rosaceae.org/) with an e-value < 1e-4 to identify further possible members of GRAS gene as previous investigation (Li et al., 2011; Jia et al., 2013; Cui et al., 2015; Xu J. N. et al., 2016). Arabidopsis GRAS genes were downloaded from TAIR (http://www.arabidopsis.org/). Finally, genes with the same accession number were removed. All the identified GRAS genes were further manually checked for the GRAS domain.

#### Gene Structures and Locations Determination

Annotations for candidate GRAS genes were obtained from the Genome Database for Rosaceae (GDR) (http://www.rosaceae.org), and chromosome locations were retrieved from these annotations. MapDraw was used to show the accurate locations of the genes on each chromosome. The online software Gene Structure Display Server (GSDS: http://gsds.cbi.pku.edu.cn) was used to generate the exon-intron structures of all the GRAS genes.

# Chemical Characteristics, Elements in the Promoters and Phylogenetic Analysis

Prosite ExPaSy (http://web.expasy.org/protparam/) was used to predict the potential chemical characteristics of the MdGRAS genes. A phylogenetic tree was generated using MEGA 6.06 with the Maximum Likelihood (ML) method and 1000 bootstrap replications. The upstream of 1.5 Kb were used for cis-elements in the promoters of the candidate MdGRAS genes. PlantCARE software (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/) was used for searching regulatory elements.

# Tandem Duplication and Synteny Analysis

Tandem duplication and synteny analysis were conducted using Circos v. 0.63 (http://circos.ca/). Tandem duplication of the MdGRAS genes was identified according to their physical locations on individual chromosomes in the apple genome. Adjacent homologous GRAS genes on the same apple chromosome with no more than one intervening gene were considered tandem duplicates. The Plant Genome Duplication Database (http://chibba.agtec.uga.edu/duplication/) was used to identify syntenic blocks.

#### Plant Material

In this study, 72 uniform 6-year-old "Fuji"/T337/Malus robusta Rehd. apple trees were randomly divided into three groups: 18 were treated with GA3, 18 were treated with sucrose, 18 were treated with 6-BA and 18 were sprayed with water as a control. These tress were grown at the experimental orchard of Northwest A& F University in Yangling (108◦ 04′ E, 34◦ 16′ N), China. And the orchard was well managed. Each group consisted of three blocks, with three replicates. The experiment was conducted from 30 to 70 days after full bloom (DAFB) in 2015. (1) GA<sup>3</sup> treatment was performed as described by Zhang et al. (2016) with a slight modification: 700 mg L−<sup>1</sup> GA<sup>3</sup> (Sigma, Deisenhofen, Germany) was sprayed once on a clear morning at 30 DAFB (May 9). (2) 300 mg L−<sup>1</sup> 6-BA (Sigma, Deisenhofen, Germany) was also sprayed on a clear morning at 30 DAFB (May 9). (3)For sugar treatment, trees were sprayed twice on clear mornings at 30 and 37 DAFB (May 9 and May 16) with 15,000 mg L−<sup>1</sup> and 20,000 mg L−<sup>1</sup> sucrose. All treatments were performed on the whole plant and were applied with a low-pressure hand-wand sprayer. Terminal buds on current-year spurs (<5 cm), which were chosen according to our previous studies (Xing et al., 2014, 2015; Li et al., 2016; Zhang et al., 2016), were collected into liquid nitrogen at 30, 40, 50, 60, and 70 DAFB, and then stored at −80◦C for use in gene expression analyses.

Additionally, buds from two apple varieties ("Yanfu No. 6" and "Nagafu No. 2") with contrasting flowering rates were collected at 30, 40, 50, 60, and 70 DAFB from 18 uniform 6-year-old trees in 2015. "Yanfu No. 6" is a spontaneous mutant of "Nagafu No. 2." It develops more flower buds than "Nagafu No. 2." "Yanfu No. 6" always exhibits a higher proportion of spurs, shorter internodes, bigger buds and more flowers. Terminal buds on current-year spurs (<5 cm) were collected as described above.

Different organs were also collected for tissue-specific expression analysis. Flowers were collected at full blossom on Fan et al. Identification and Expression of MdGASA Gene

April 9 in 2015. Stems were collected from new shoots with a diameter of 2–3 mm. Mature leaves were collected from the adjacent buds. Fruits were collected with a diameter of 2–3 cm. All tissues were immediately frozen in liquid nitrogen and stored at −80◦C for gene expression analysis. Additionally, the online ArrayExpress database (https://www.ebi.ac.uk/arrayexpress/, E-GEOD-42873) was also employed to investigate their expression patterns among different apple varieties and tissues.

#### Flowering Rate Analysis

Flowering rate was also investigated among different exogenous treatment and two different flowering varieties. One hundred twenty terminal buds on short shoots which were less than 5 cm and no fruits on them were tagged randomly at 15 DAFB. The number of flowers was counted of the tagged shorts as finally flowering rate. And it was investigated at full blossom on April 10 in 2016.

#### RNA Extraction and cDNA Synthesis

Total RNA was extracted from plant tissue samples using the cetyltrimethyl ammonium bromide (CTAB) method with slight modifications (Gambino et al., 2008). Briefly, 900µL extraction buffer (2% CTAB, 2.5% PVP-40, 2 M NaCl, 100 mM Tris-HCl [pH 8.0], 25 mM EDTA [pH 8.0], and 2% β-mercaptoethanol) was preheated at 65◦C and added to 2-mL microcentrifuge tubes just before use. Samples containing 200 mg of bud tissue stored at −80◦C were ground to a powder, added to the tubes, and mixed with extraction buffer. After shaking and inverting each tube vigorously for 5 min and incubating at 65◦C for 30 min, an equal volume of chloroform:isoamyl alcohol (24:1, v/v) was added. Each tube was shaken and inverted vigorously and then centrifuged at 12,000 × g for 10 min at 4◦C. The supernatant was collected into a new tube and re-extracted with an equal volume of chloroform:isoamyl alcohol (24:1, v/v). The resulting supernatant was then transferred into a new 2-mL tube and LiCl (3 M final concentration) was added. The mixture was incubated at −20◦C for 4 h and the RNA was selectively pelleted by LiCl after centrifugation at 18,000 × g for 20 min at 4◦C. The pellet was resuspended in 500µL of SSTE buffer (10 mM Tris-HCl [pH 8.0], 1 mM EDTA [pH 8.0], 1% SDS, and 1 M NaCl) preheated to 65◦C and an equal volume of chloroform:isoamyl alcohol. The mixture was then centrifuged at 12,000 × g for 10 min at 4◦C. The supernatant was transferred to a new microcentrifuge tube, and the RNA was precipitated with 2.5 volumes of cold ethanol at −80◦C for at least 30 min and centrifuged at 1,200 × g for 20 min at 4◦C. Finally, the pellets were washed with 70% ethanol and resuspended in diethylpyrocarbonate-treated water.

Total RNA integrity was verified by running the samples on 2% agarose gels. First-strand cDNA was synthesized from 1µg of total RNA using a PrimeScript RT Reagent kit with gDNA Eraser (Takara Bio, Shiga, Japan) following the manufacturer's instructions.

#### Expression Analysis by qRT-PCR

The expression levels of all identified MdGRAS genes were analyzed by qRT-PCR using primer pairs designed with Primer 5.0 (**Table S1**). The real-time PCR assay mix (20µL) consisted of 2-µL cDNA samples (diluted 1:8), 10µL 2× SYBR Premix ExTaq II (Takara Bio), 0.8µL of each primer (10µM) (**Table 1**), and 6.4µL distilled deionized H2O. Each PCR assay was run on an iCycler iQ Real Time PCR Detection System (Bio-Rad) with an initial denaturation at 95◦C for 3 min, followed by 40 cycles at 94◦C for 15 s, 62◦C for 20 s, and 72◦C for 20 s. The resulting fragments were subjected to melting-curve analysis to verify the presence of gene-specific PCR products. The meltingcurve analysis was performed directly after real-time PCR and included an initial step of 94◦C for 15 s, followed by a constant increase from 60◦ to 95◦C at a 2% ramp rate. The apple EF-1α gene (GenBank accession no. DQ341381) was used as an internal control to normalize all mRNA expression levels. The 2 <sup>−</sup>11Ct method was used to calculate the relative amount of template present in each PCR amplification mixture (Livak and Schmittgen, 2001).

#### Statistical Analysis

Data were analyzed using variance (ANOVA) at the 5% level with the SPSS 11.5 software package (SPSS, Chicago, IL, USA). Figures were made using Origin 8.0 (Microcal Software Inc., Northampton, MA, USA).

### RESULTS

#### Identification and Characterization Analysis of *MdGRAS* Genes

To identify members of the MdGRAS gene family, both HMM profiles and BLASTP were used with the default parameters. After manual checking, a total of 127 candidate MdGRAS genes were identified (**Table 1**). The genes were named according to their chromosomal locations. The 127 candidate MdGRAS genes were distributed on 16 chromosomes of the apple genome (**Figure 1**). Chromosome 11 contained the most MdGRAS genes (15.75%), followed by chromosome 3 (11.81%), while chromosome 16 contained no MdGRAS genes (0%) (**Figure S1**). Additionally, four genes (MDP0000950387, MDP0000319342, MDP0000640034, and MDP0000640034) could not be located on any of the chromosomes and were named MdGRAS124 to MdGRAS127. Only two genes were distributed on each of chromosomes 7 and 8. The open reading frame (ORF) lengths of the MdGRAS genes ranged from 333 to 5,787 bp, encoding peptides ranging from 110 to 1,928 amino acids. We used ExPaSy to predict the characterization of MdGRAS proteins. The predicted molecular weights of the GRAS proteins ranged from 12.50 to 218.53 kDa. The grand average of hydropathicity was less than 0 for all GRAS proteins. The instability index of most of the 127 MdGRAS proteins was greater than 40, except for MdGRAS22, MdGRAS33, MdGRAS41, MdGRAS42, MdGRAS43, MdGRAS53, and MdGRAS65, indicating that these proteins were unstable.

### Phylogenetic and Gene Structure of the GRAS Gene Family

Since Arabidopsis is the most popular model species and the functions of some GRAS genes have been well characterized. Phylogenetic analysis of GRAS genes was investigated in apple

#### TABLE 1 | List of *MdGRAS* genes and their chemical characteristics.


*(Continued)*

#### TABLE 1 | Continued


*(Continued)*



*<sup>a</sup>Gene ID in the apple (Malus* × *domestica) genome (Malus domestica Genome database v1.0);*

*<sup>b</sup>MW: Molecular weight;*

*<sup>c</sup>GRAVY: Grand average of hydropathicity;*

*<sup>d</sup>PI: Isoelectric point;*

*e II: Instability index;*

*<sup>f</sup>AI: Aliphatic index.*

and Arabidopsis. A total of 160 GRAS sequences were collected and analyzed. They were classified into eight distinct groups according to the phylogenetic tree: DELLA, SCL3, SCR, Ls, LISCL, PAT1, SHR, and HAM (**Figure 2**). 14 GRAS proteins were located in the DELLA subfamily, including five AtGRAS (AtGAI, AtRGA, AtRGL1, AtRGL2, and AtRGL3) and nine MdGRAS proteins. However, MdGRAS37 did not contain the integral DELLA and TVHYNP domains. MdGRAS37 was classified into the DELLA subfamily because of its high homology with the other proteins. To examine the gene structures, we analyzed the exon-intron structures of the 127 MdGRAS genes according to their genome sequences (**Figure S2**). Thirteen genes (MdGRAS100, MdGRAS101, MdGRAS25, MdGRAS7, MdGRAS102, MdGRAS2, MdGRAS33, MdGRAS51, MdGRAS80, MdGRAS45, MdGRAS29, MdGRAS1, and MdGRAS3) had two or more introns. Other genes had few or no introns.

#### Expansion Patterns of the MdGRAS Gene Family

Segmental and tandem duplications provide information on the expansion of the gene family. Here, we used the Circos program to evaluate the segmental and tandem duplications of the MdGRAS genes. Tandem duplications were identified from adjacent homologs on a single chromosome, while segmental duplications were identified from homologs on different chromosomes. More than 15 pairs of MdGRAS genes, such as MdGRAS122/53, MdGRAS116/50, and MdGRAS115/49, were located in duplicated genomic regions (**Figure 3**). All were located in different chromosomes. Chromosomes 11 and 3 had the most duplication, which could partly explain the larger numbers of MdGRAS genes on chromosomes 11 and 3 (**Figure S1**). Chromosomes 1, 6, 7, 8, 13, and 14 did not contain any duplicated genes. In summary, segmental and tandem

#### Comparative Analysis of the GRAS Gene Families in Different Species

To further analyze the evolutionary relationships and expansion patterns of plant GRAS genes, we compared the published GRAS genes in different species (**Table S2**). The number of GRAS genes was the largest in apple and was close to that in Populus. However, the apple genome is twice the size of the Populus genome. The number of GRAS genes was also similar between Chinese cabbage (48) and P. mume (46), and between grape (52) and tomato (53). However, their genome sizes are also very different (Chinese cabbage: 485.0 Mb vs. P. mume: 280.0 Mb; grape: 487.0 Mb vs. tomato: 760.0 Mb). These results suggested that the number of plant GRAS genes was partly related to genome size.

As Arabidopsis is a model plant species, the functions of the GRAS genes in Arabidopsis have been well researched. We generated a comparative GRAS synteny map between apple and Arabidopsis. Many pairs of syntenic orthologous genes were matched, such as AtSCL32-MdGRAS108, AtRGA-MdGRAS114,

#### Tissue-Specific Analysis of *MdGRAS* Expression Using ArrayExpress Data

To elucidate the potential roles and functions of the MdGRAS genes in apple, we downloaded expression profile data for different tissues from the ArrayExpress database (E-GEOD-42873). Eight organs and samples with two biological replicates were represented in this expression array, including seedlings, roots, stems, flowers, fruit, and leaves (**Figure 5**). The GRAS genes showed different expression patterns among different apple varieties and tissues. In the "M74" apple genotype, most of the MdGRAS genes were expressed at higher levels in the flower than in the fruit. Moreover, the expression levels of the MdGRAS genes were higher in ripe fruit (at harvest) than in unripe fruit, indicating their important roles in fruit ripening. Most MdGRAS genes were expressed at lower levels in seedlings, except for MdGRAS100, MdGRAS101, MdGRAS126, MdGRAS18, MdGRAS79, and MdGRAS27. In "M14" leaves,

MdGRAS85 showed the highest expression. However, other GRAS genes were hardly detected in leaves. Moreover, most GRAS showed lower levels in stems except MdGRAS26, MdGRAS50, and MdGRAS64.

# QRT-PCR Analysis of the Candidate Flowering Related *MdGRAS* Genes in Different Tissues in "Nagafu No. 2"

We selected eight higher expression genes (MdGRAS6, MdGRAS26, MdGRAS28, MdGRAS44, MdGRAS53, MdGRAS64, MdGRAS107, and MdGRAS122) from our transcriptome data for further analysis (**Figure S4**, **Table S3**). These eight genes were from different GRAS subfamilies, including the LS (MdGRAS6), SHR (MdGRAS26), SCL (MdGRAS28), PAT1 (MdGRAS44), DELLA (MdGRAS53, MdGRAS107, and MdGRAS122) and LISCL (MdGRAS64) subfamily. We employed qRT-PCR to investigate their expression patterns in different tissues (stems, leaves, flowers, fruits and buds). All candidate MdGRAS genes showed constitutive expression patterns in the five tissues, except that the levels of MdGRAS26 and MdGRAS28 were low in flowers (**Figure 6**). The transcription levels of MdGRAS6, MdGRAS26, MdGRAS44, MdGRAS53, MdGRAS64, MdGRAS107, and MdGRAS122 were high in buds and leaves, which indicated their important roles in mediating flower induction. Additionally, MdGRAS107 and MdGRAS28 were all showed lower levels than other candidate genes among the five tissues.

# QRT-PCR Analysis of the Candidate *MdGRAS* Genes in Different Apple Varieties

To examine the expression of the MdGRAS genes in two varieties with contrasting flowering rates, we selected the easyflowering variety "Yanfu No. 6" and the difficult-flowering "Nagafu No. 2." As shown in **Table S4**, "Yanfu No. 6" has a 1.5 fold higher flowering rate than "Nagafu No. 2." We examined the expression patterns in these varieties using qRT-PCR (**Figure 7**). All candidate genes showed different expression patterns during flower induction. MdGRAS6 expression was higher in the buds of "Yanfu No. 6" during all stages. MdGRAS44 expression was higher in "Yanfu No. 6" except at 80 DAFB, while MdGRAS122 expression was higher in "Nagafu No. 2" except at 70 DAFB. MdGRAS26 expression was lower at the first stage in "Nagafu No. 2," but was higher in all subsequent stages. The transcription level of MdGRAS107 was higher in "Yanfu No. 6" except at 30 DAFB.

# QRT-PCR Analysis of the Candidate *MdGRAS* Genes in Response to Treatments Affecting Flowering

Previous studies have shown the exogenous treatments (GA, 6-BA, and sucrose) can alter flowering rates, among which

GA always decreases the flowering rate, while 6-BA and sugar increase flowering rates (Xing et al., 2015; Li et al., 2016; Zhang et al., 2016). We used these kinds of treatments to further investigate the expression of the candidate MdGRAS genes during flowering. As shown in **Table S4**, exogenous GA<sup>3</sup> treatment inhibited flowering rate while sugar and 6-BA promoted flowering. The genes displayed various expression patterns at different time points (**Figure 8**). Some of them were down-regulated, while others were up-regulated. MdGRAS6 was up-regulated after GA<sup>3</sup> treatment except at 40 DAFB and after 6-BA treatment except at 70 DAFB, but was down-regulated after sugar treatment except at 30 DAFB. MdGRAS26 showed similar changes after GA<sup>3</sup> treatment, and was up-regulated except at 40 DAFB. However, the level of MdGRAS26 was downregulated after 6-BA and sugar treatment except at 30 DAFB. MdGRAS28 and MdGRAS44 were down-regulated after GA<sup>3</sup>

treatment except that MdGRAS28 was up-regulated at 30 and 50 DAFB. MdGRAS28 was down-regulated after sugar treatment at all time points. The expression of MdGRAS53 was promoted by exogenous GA<sup>3</sup> except at 40 and 80 DAFB, but was inhibited by exogenous 6-BA and sugar at most time points. MdGRAS64 expression was up-regulated after sugar treatment, but was not consistent in response to GA and 6-BA. MdGRAS107 was downregulated after GA<sup>3</sup> treatment except at 30 and 50 DAFB, but was up-regulated by 6-BA except at 30 and 70 DAFB. MdGRAS107 was down-regulated after sugar treatment except at 30 DAFB. The transcription level of MdGRAS122 was lower in 6-BA treated buds than control except at 40 DAFB, and it was down-regulated after sugar treatment at all time points. In order to further confirm the results, we also investigated their expression patterns in GA-treated buds in 2014 and 6-BA treated buds in 2013 based on our previous research (Li et al., 2016; Zhang et al., 2016). Samples were collected in the early stage (ES), middle stage (MS), and later stage (LS) of flower induction. As shown **Figures S5**, **S6**, their expression patterns in 2014 and 2015 were similar with our results.

# Analysis Related *cis-*Elements in the Candidate *MdGRAS* Genes

To further analyze the potential roles of MdGRAS genes in response to various circumstances, a 1.5 kb promoter

of candidate MdGRAS genes were used. The related ciselements were identified as previous study (Fan et al., 2017).The candidate MdGRAS genes all shared large light-responsive boxes, followed by stress (**Figure S7**). Additionally, hormones-related cis-elements were also found in all the candidate MdGRAS genes, including MeJA, salicylic acid, gibberellin, auxin and ethylene. These identified motifs in the promoter of each gene indicated that MdGRAS may be regulated by various cis-elements within the promoter during growth.

#### DISCUSSION

Genome-wide identification of GRAS genes has been reported in Arabidopsis, P. mume, Populus, rice, grape, castor bean, and Chinese cabbage (Tian et al., 2004; Song et al., 2014; Huang et al., 2015; Lu et al., 2015; Grimplet et al., 2016; Xu W. et al., 2016), but little in known in apple. In this study, we identified MdGRAS genes in the apple genome (Velasco et al., 2010), and analyzed their phylogenetic relationships, gene characteristics, gene structures, expansion history, and syntenic relationships. We also analyzed their tissue-specific expression patterns and their expression profiles in response to flower induction. Our comprehensive study provides a basic for further investigation of GRAS genes in apple, which can also have potential values in genetic improvement of flower induction in apple as well as some other relative species.

#### Phylogenesis of Apple *GRAS* Genes

We identified 127 putative MdGRAS genes in apple, a larger number than that reported for other plant species such as P. mume, Populus, rice, grape, and Chinese cabbage (Tian et al., 2004; Song et al., 2014; Huang et al., 2015; Lu et al., 2015; Grimplet et al., 2016; Xu W. et al., 2016). The larger number of GRAS genes in apple might be because its genome (881.3 Mb) is larger than those of P. mume (280.0 Mb), rice (372.0 Mb), grape (487.0 Mb), and Brassica rapa (283.3 Mb). The MdGRAS genes were located on almost all of the chromosomes except chromosome 16. The structures of most MdGRAS genes were similar, and most had few introns, similar to those in P. mume and Populus (Tian et al., 2004; Song et al., 2014). These findings indicated that the structures of GRAS genes are highly conserved among different plant species. A previous study identified six genes in the DELLA subfamily in apple, based on data in the EST database (Foster et al., 2007). In our study, nine putative genes belonging to the DELLA subfamily were identified from the apple genome (Xing et al., 2016) according to the phylogenetic tree (**Figure 2**). However, MdGRAS37, which was classified into the DELLA subfamily, lacked the conversed DELLA domain and other features of the DELLA subfamily (**Figure S3**), including gene length and structural domains. MdGRAS37 was probably placed in the DELLA clade because of its high sequence similarity with Arabidopsis sequences of the DELLA subfamily (Lu et al., 2015). The DELLA domain of MdGRAS37 may have gradually degenerated during the evolutionary process or its sequence may be partly similar with DELLA proteins. However, this needs to be further researched.

#### Expansion and Synteny of *GRAS* Genes

Gene duplications can be tandem, segmental, or arise from whole genome duplication. Duplicated genes are known to play important roles in plant growth and development.

To date, various kinds of genes encoding transcription factors have been identified, and most of them have shown gene duplications, including AP2, MADS, SPL, and DOF (Zahn et al., 2005; Moreno-Risueno et al., 2007; Shigyo et al., 2007; Kumar et al., 2016). Gene duplication is important for gene family expansion (Day et al., 2003; Cannon et al., 2004; Leister, 2004). The expansion and duplication of GRAS genes have also

been reported in Arabidopsis, P. mume, and Chinese cabbage (Liu and Widmer, 2014; Song et al., 2014; Lu et al., 2015). It was reported that 85% of AtGRAS genes have undergone segmental duplications, including two tandemly duplicated genes (AtSCL33 and AtSCL34). Here, we used the Circos program to identify tandemly duplicated genes in the apple genome. Synteny analysis of the duplicated blocks in the apple genome showed that some MdGRAS genes were putative segmental duplicates (**Figure 2**). The present result was consistent with previous study in Arabidopsis, which GRAS also occurred gene duplication in apple genome. Plant GRAS genes were involved in various biological processes, and the expansion of the MdGRAS might contribute to their diversity functions in regulation plant growth and developemt. Additionally, it was reported that a genomewide duplication (more than 50 million years ago) contributed to the expansion of the apple chromosome number, which led to a change to 17 chromosomes from nine chromosomes (Velasco et al., 2010). So the whole genome duplications promoted the expansion of GRAS genes in apple. This expanded by segmental duplication or whole genome duplications of GRAS genes in apple genome was also proved in grape and other species (Grimplet et al., 2016). Totally, the duplication and expansion of GRAS genes have played important roles in gene evolution, and contributed to their varied structures and functions.

To date, little functional research has been performed on rosaceous GRAS genes. Comparison with homologous GRAS genes of the model plant Arabidopsis can help us further analyze the MdGRAS genes. Comparisons of genome sequences among different species can be valuable for the reconstruction of individual gene families (Koonin, 2005). Such genome comparisons can be used to transfer annotations from a betterstudied taxon (whose genome structure and functions have been elucidated) to a less-studied taxon (Lyons et al., 2008). Based on this, we inferred the functions of the less-studied MdGRAS genes from the better-understood AtGRAS genes. We used the Circos program to compare apple and Arabidopsis. Most GRAS genes were detected in syntenic genomic regions (**Figure 4**), allowing us to predict their functions. Currently, most AtGRAS genes are well understood, including AtSCL3 (Zhang et al., 2011), AtSHR (Levesque et al., 2006), AtSCL13 (Torres-Galea et al., 2006), and AtSLR (Aleman et al., 2008). Thus, we can analyze the potential functions of MdGRAS genes based on their homologs in Arabidopsis. However, functions predicted in this way needs to be experimentally verified.

#### Expression Profiles of *MdGRAS* Genes in Different Tissues

To reveal their potential functions in apple growth and development, we employed ArrayExpress data to analyze the expression of the MdGRAS genes in different tissues and organs. These potential investigations of each GRAS can be useful to access its possible function. Their various expression patterns were consistent with their diverse gene locations, lengths and sequences. As shown in **Figure 5**, the expression patterns of the GRAS genes differed among different tissues and varieties, which is consistent with other species (Tian et al., 2004; Song et al., 2014; Huang et al., 2015; Lu et al., 2015; Grimplet et al., 2016; Xu W. et al., 2016). For example, tomato SlGRAS18 and SlGRAS38 transcript levels were high at the breaker and redripening stages, and their expression was regulated by RIN (Fujisawa et al., 2012, 2013). In another study, transcripts of 14 SlGRAS genes were detected during fruit ripening (Huang et al., 2015). Our analyses of GRAS expression profiles confirmed that most of the MdGRAS genes had higher transcript levels in fruit than in other apple tissues, indicating their important roles in fruit ripening. MdGRAS9 and MdGRAS85 expression levels

were extremely high in "M20" during the harvest stage. These results suggested that MdGRAS9 and MdGRAS85 play important roles during fruit development; this was consistent with previous study. In Arabidopsis, three AtGRAS genes (SCL6, SCL22, and SCL27) were shown to be involved in leaf development. In the apple genotype "M14," MdGRAS85 transcripts were detected at high level in leaves, suggesting that this gene may share similar functions with AtSCL6, AtSCL22, and AtSCL27 in leaf development, which was consistent with their close evolutionary relationships to MdGRAS85 (**Figure 2**). However, MdGRAS85 did not show high level in leaves of "M49"; the different expressions of MdGRAS85 may be explained by their different genotypes of "M14" and "M49." The high expression patterns of MdGRAS126, MdGRAS18, and MdGRAS79 in seeds were also consistent with PmGRAS20 and PmGRAS36 expression, and suggested that these genes may have important roles in seed development (Lu et al., 2015). These different expression patterns of GRAS indicated their various functions in apple development; further functional analysis needs to be confirmed.

We also analyzed the expression of the candidate MdGRAS genes in different tissues (stem, leaf, flower, fruit, and bud) in "Nagafu No. 2" with qRT-PCR. The MdGRAS genes all showed different expression patterns among the five tissues. Leaves and buds have been used to research flower induction in Arabidopsis and other plants (Galvao et al., 2012; Porri et al., 2012; Xing et al., 2015; Fan et al., 2016). The candidate genes MdGRAS6, MdGRAS26, MdGRAS44, MdGRAS53, MdGRAS64, and MdGRAS122 mostly showed high expression in buds and leaves, which proved their important roles in flower induction. A previous study showed that DELLA has important roles in regulating internode elongation (Bolle, 2004). We found that the expression of MdGRAS107, a DELLA subfamily member, was also higher expressed in the apple stem, which indicates it might also have functions in regulating internode elongation.

### Characterization of *MdGRAS* Expression During Flower Induction

To our best knowledge, the relationship between GRAS and flower induction was received less attention, apart from MdGRL2a. Flower induction is associated with several environmental and internal factors, including temperature, hormones and photosynthesis (Galvao et al., 2012; Porri et al., 2012; Xing et al., 2015, 2016). Hormones and sugar are of the most important factors affected apple flowering, the relationship about GRAS involved in hormones or sugar remains unknown. Up to now, GA was considered to be the most associated with flower induction; 6-BA and sugar were also involved in flower induction. And they were all well applied in orchard cultivation. Additionally, our previous studies showed that exogenous treatments can alter flowering rates; GA<sup>3</sup> can inhibit flowering, while 6-BA and sugar can increase flowering rates (Xing et al., 2015; Li et al., 2016; Zhang et al., 2016). These exogenous treatments mainly alter the nutrition or hormone balance and influence the expression of related genes, which inhibits (GA) or promotes (sugar and 6-BA) flowering (Xing et al., 2015; Li et al., 2016; Zhang et al., 2016). Based on this, we investigated the candidate MdGRAS genes in response to these floweringrelated treatments (**Figure 8**). With their multiple functions, the candidate MdGRAS genes showed different expression patterns after the treatments, and were down- or up-regulated at most time points. "Yanfu No. 6," a bud mutant of "Nagafu No. 2," has a higher proportion of spurs, shorter internodes and a higher flowering rate than "Nagafu No. 2" (**Table S4**). The differential expression of the candidate MdGRAS genes in these two varieties with contrasting flowering rates indicated their important roles in flower induction (**Figure 7**).

In recent years, although the roles of GRAS genes in regulating plant growth and development have become better understood, little is known about the roles of GRAS genes in regulating flower induction in fruit trees. Several studies have shown that GRAS genes are important in bud growth and development. Most DELLA subfamily gene regulate flower induction by the GA pathway, and it can influence the expression of some flowering control genes (Bolle, 2004; Galvao et al., 2012; Porri et al., 2012; Xing et al., 2014, 2015; Huang et al., 2015). However, genes from other subfamilies were also associated with flower induction. AtGRAS genes promoted the expression of FT and the della mutant always showed earlier flowering (Galvao et al., 2012). In Arabidopsis, DELLA proteins were shown to inhibit flower induction, and interact with SPL to control flower induction (Shigyo et al., 2007). Previous studies showed that the expression of one-fifth of the candidate genes was increased during flower development in P. mume, and DELLA gene expression was also higher in tomato buds, indicating their important roles in flower induction (Huang et al., 2015; Lu et al., 2015). In our study, we selected eight important MdGRAS genes from transcriptome data. They belong to different subfamilies, including LS (MdGRAS6), SHR (MdGRAS26), SCL (MdGRAS28), PAT1 (MdGRAS44), DELLA (MdGRAS53, MdGRAS107, and MdGRAS122) and LISCL (MdGRAS64). Our result suggested that these candidate MdGRAS genes from different subfamilies were important and they all have potential involvement in flower induction. Additionally, functions of their homologous genes have been described and verified in other species, which provided more evidence for their important response to flower induction (Fu et al., 2001; Hynes et al., 2003; Petty et al., 2003; Wang et al., 2010; Galvao et al., 2012). They can regulate flower induction directly and indirectly. In Arabidopsis, AtLAS and AtSHR were involved in shoot apical meristem. AtSCL6, AtSCL26, and AtSCL27 were highly expressed in flowers (Bolle, 2004); the Arabidopsis rga and gai mutant showed earlier flowering (Galvao et al., 2012). Additionally, GRAS genes in other species were all indicated their potential involvement in flowering induction, such as tomato (SlGRAS11 and SlGRAS18), castor beans (27568.m000253, 28650.m000187, 28650.m000190, and 28650.m000191), and grape (VviSHR1 and VviPAT6) (Huang et al., 2015; Grimplet et al., 2016; Xu W. et al., 2016). Our candidate MdGRAS genes were homologous with these wellknown GRAS genes in other plants, which is in agreement with their important functions in flower induction. However, further functional research needs to be confirmed about these MdGRAS genes. Their different expression patterns during flower induction indicated they were involved in GA-mediated, 6-BA mediated or sugar-mediated flowering in apple trees. Previous transgenic Arabidopsis of MdRGL2a (one of the DELLA members in apple) showed delayed flowering phenotype, which proved their functions in regulating flowering. Additionally, the predication ofcis-elements within their promoters indicated their potential roles in response to hormone mediated signal during apple growth and development (**Figure S7**). However, with the large number of GRAS gene in Arabidopsis and apple genome, we cannot capture more genes aiming at flower induction except the candidate eight MdGRAS. The other left MdGRAS genes need to be further confirmed. And more GRAS genes need to be functionally researched.

All in all, extensive cross-talk about GRAS genes has been identified in herbaceous and woody plants; extensive research effort has also been devoted to characterization GRAS genes in plant growth and development. However, the GRAS genes of woody economically important tree species have received less attention. In this study, we identified 127 putative GRAS genes from the apple genome, which were located on 16 chromosomes. We classified these genes and performed systematic phylogenetic, structural and synteny analyses. We also analyzed their expression in different tissues (stems, leaves, flowers, fruits and buds). The spatiotemporal expression patterns indicated they potentially have multiple functions in regulating plant growth and development. Additionally, some candidate MdGRAS genes were investigated under treatments with several flowering-related compounds, which indicated the candidate MdGRAS genes are involved in the complex flower induction process. These results may provide useful strategies for further improvement of flower induction in apple.

# AUTHOR CONTRIBUTIONS

MH, SF, and DZ conceived and designed the experiment. SF, CG, and MZ performed the experiment. SF, HW, YL, and YS analyzed the data. MH and SF wrote the manuscript.

#### ACKNOWLEDGMENTS

This work was financially supported by China Apple Research System (CARS-28), National Natural Science Foundation of China (31672101), Science and Technology Innovative Engineering Project in Shaanxi province, China (2016KTZDNY01-10), China Postdoctoral Science Foundation (2014M56806), Science Foundation from the Northwest

### REFERENCES


A&F University (2452015291), Yangling Subsidiary Center Project of the National Apple Improvement Center and Collaborative Innovation of the Center for Shaanxi Fruit Industry Development.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fphys. 2017.00253/full#supplementary-material

Figure S1 | Distribution of *MdGRAS* genes on the apple chromosomes.

Figure S2 | *MdGRAS* gene structures. *MdGRAS* gene exon–intron composition. Red boxes and black lines represent exons and introns, respectively.

Figure S3 | Alignment of DELLA subfamily members in apple and *Arabidopsis.*

Figure S4 | Heat map showing the transcript levels of *MdGRAS* genes during flower induction. Transcript levels of *GRAS* genes in the buds of "Nagafu No. 2" and "Yanfu No. 6" were investigated during flower induction. Samples were collected three times at the early stage (ES), middle stage (MS) and later stages (LS) of flower induction.

Figure S5 | Expression of candidate *MdGRAS* genes in GA-treated buds. Samples were collected three times at ES, MS and LS in 2014, which was consistent with 40, 50, and 70 DAFB in 2016. Each value represents the mean ± standard error of three replicates.

Figure S6 | Expression of candidate *MdGRAS* genes in 6BA-treated buds. Samples were collected three times at ES, MS and LS in 2013, which was consistent with 30, 50, and 80 DAFB in 2016. Each value represents the mean ± standard error of three replicates.

Figure S7 | Predicted *cis*-elements in the promoter regions of the candidate *MdGRAS* genes. The upstream 1.5 kb promoter sequences were analyzed.

Table S1 | Primer sequences of *MdGRAS* and reference genes in quantitative reverse transcription PCR.

Table S2 | Comparison of *MdGRAS* genes in different species.

Table S3 | Expression levels of *MdGRAS* genes in Nagafu No. 2 and Yanfu No. 6. Data were collected from the transcriptome.

Table S4 | Investigation of flowering rate in 2016.

in the Malus domestica genome. Mol. Genet. Genomics 290, 1435–1446. doi: 10.1007/s00438-015-1004-z


**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 Fan, Zhang, Gao, Zhao, Wu, Li, Shen and Han. This is an openaccess 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.

Edited by: Zuhua He,

Sciences, China Xuelu Wang,

\*Correspondence: Jiayang Li jyli@genetics.ac.cn Bing Wang

bingwang@genetics.ac.cn †Present address: Lei Wang,

Agricultural Genome Institute, Chinese Academy of Agricultural Sciences, Shenzhen, China

Max Planck Institute of Molecular Plant Physiology, Potsdam, Germany ‡These authors have contributed

Liang Jiang,

equally to this work.

Specialty section: This article was submitted to

Plant Physiology, a section of the journal Frontiers in Plant Science Received: 10 August 2017 Accepted: 26 October 2017 Published: 09 November 2017

China

Shanghai Institutes for Biological Sciences (CAS), China Reviewed by: Haiyang Wang,

Biotechnology Research Institute, Chinese Academy of Agricultural

Huazhong Agricultural University,

fpls-08-01935 November 7, 2017 Time: 16:47 # 1

# DWARF14, A Receptor Covalently Linked with the Active Form of Strigolactones, Undergoes Strigolactone-Dependent Degradation in Rice

Qingliang Hu1,2‡ , Yajun He1,2‡ , Lei Wang<sup>1</sup>† , Simiao Liu<sup>1</sup> , Xiangbing Meng<sup>1</sup> , Guifu Liu<sup>1</sup> , Yanhui Jing<sup>1</sup> , Mingjiang Chen<sup>1</sup> , Xiaoguang Song<sup>1</sup> , Liang Jiang<sup>1</sup>† , Hong Yu<sup>1</sup> , Bing Wang<sup>1</sup> \* and Jiayang Li1,2 \*

<sup>1</sup> State Key Laboratory of Plant Genomics and National Center for Plant Gene Research (Beijing), Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing, China, <sup>2</sup> University of Chinese Academy of Sciences, Beijing, China

Strigolactones (SLs) are the latest confirmed phytohormones that regulate shoot branching by inhibiting bud outgrowth in higher plants. Perception of SLs depends on a novel mechanism employing an enzyme-receptor DWARF14 (D14) that hydrolyzes SLs and becomes covalently modified. This stimulates the interaction between D14 and D3, leading to the ubiquitination and degradation of the transcriptional repressor protein D53. However, the regulation of SL perception in rice remains elusive. In this study, we provide evidences that D14 is ubiquitinated after SL treatment and degraded through the 26S proteasome system. The Lys280 site of the D14 amino acid sequence was important for SL-induced D14 degradation, but did not change the subcellular localization of D14 nor disturbed the interaction between D14 and D3, nor D53 degradation. Biochemical and genetic analysis indicated that the key amino acids in the catalytic center of D14 were essential for D14 degradation. We further showed that D14 degradation is dependent on D3 and is tightly correlated with protein levels of D53. These findings revealed that D14 degradation takes place following D53 degradation and functions as an important feedback regulation mechanism of SL perception in rice.

Keywords: Oryza sativa, plant hormones, strigolactones, hydrolase, DWARF14, proteolysis, signal transduction

# INTRODUCTION

Citation:

Hu Q, He Y, Wang L, Liu S, Meng X, Liu G, Jing Y, Chen M, Song X, Jiang L, Yu H, Wang B and Li J (2017) DWARF14, A Receptor Covalently Linked with the Active Form of Strigolactones, Undergoes Strigolactone-Dependent Degradation in Rice. Front. Plant Sci. 8:1935. doi: 10.3389/fpls.2017.01935

Strigolactones (SLs), a group of carotenoid-derived terpenoid lactones produced by plants, were initially characterized as signals that enable parasitic plants to detect their host (Cook et al., 1966), and also as signals recognized by arbuscular mycorrhizal (AM) fungi in the rhizosphere to build the symbiotic association with host plants (Akiyama et al., 2005). Besides the roles of SLs in the rhizosphere, SLs have been identified as endogenous phytohormones that are transported from roots to shoots and suppress shoot branching by inhibiting the outgrowth of axillary buds (Beveridge et al., 1996; Foo et al., 2001; Booker et al., 2005; Gomez-Roldan et al., 2008; Umehara et al., 2008). In addition, SLs have profound effects in many aspects of plant development, including

**281**

internode elongation, leaf shape and senescence, shoot gravitropism, stem secondary thickening, root architecture, and the drought tolerance (Al-Babili and Bouwmeester, 2015; Waters et al., 2017).

The key components required for SL biosynthesis and signaling have been identified from genetic characterizations of highly branched mutants, including ramosus (rms) in pea (Pisum sativum), more axillary growth (max) in Arabidopsis thaliana, decreased apical dominance (dad) in petunia (Petunia hybrida), and dwarf (d) or high-tillering dwarf (htd) in rice (Oryza sativa). In the SL biosynthetic pathway, SLs are derived from all-trans-β-carotene, which is converted to 9 cis-β-carotene by the isomerase DWARF27 (D27) (Lin et al., 2009; Alder et al., 2012), and subsequently catalyzed into carlactone by carotenoid cleavage oxygenase 7 (CCD7) and CCD8 (Beveridge et al., 1996; Sorefan et al., 2003; Booker et al., 2004; Seto et al., 2014). The subsequent catalytic reactions are diverse in different species. In Arabidopsis, the cytochrome P450 enzyme MAX1 converts carlactone to carlactonoic acid (CLA), which is further converted to a methyl carlactonoate (MeCLA) by an unknown enzyme (Abe et al., 2014). Subsequently, LATERAL BRANCHING OXIDOREDUCTASE (LBO) converts MeCLA to an unknown SL-like product (MeCLA+16 Da) (Brewer et al., 2016). In rice, the MAX1 homolog Os01g0700900 is responsible for the oxidation of carlactone to 4-deoxyorobanchol (4DO), while another MAX1 homolog, Os01g0701400, functions as an orobanchol synthase that converts 4DO to orobanchol (Zhang et al., 2014). In sorghum, a newly identified sulfotransferase LOW GERMINATION STIMULANT1 (LGS1) is responsible for a change of the dominant SL in root exudates from 5-deoxystrigol to orobanchol via an unknown mechanism and regulates the Striga resistance (Gobena et al., 2017).

In SL signaling, three key components have been identified from genetic screen of SL-insensitive mutants in rice, Arabidopsis, pea, and petunia, which include the α/β-fold hydrolase D14/AtD14/RMS3/DAD2, the F-box protein D3/MAX2/RMS4/PhMAX2A, and the repressor proteins D53/D53-Like SMXLs (Stirnberg et al., 2002, 2007; Ishikawa et al., 2005; Arite et al., 2009; Gao et al., 2009; Liu et al., 2009; Hamiaux et al., 2012; Waters et al., 2012; Jiang et al., 2013; Nakamura et al., 2013; Zhou et al., 2013; Soundappan et al., 2015; Wang et al., 2015). Perception of SLs depends on a novel mechanism involving the formation of a covalently linked intermediate molecule (CLIM) from the binding of a SL molecule to the receptor D14 and being hydrolyzed by D14. This reaction promotes a conformational change of D14, leading to the interaction between D14 and D3 and triggering the SL signal transduction (Yao et al., 2016). S Ls can also induce the interaction between D14 and D53, leading to ubiquitination and degradation of D53 in a D3 and D14-dependent manner (Jiang et al., 2013; Zhou et al., 2013). D53 contains three ethylene-responsive element binding factor-associated amphiphilic repression (EAR) motifs, which are essential to recruit the transcriptional co-repressor TOPLESS (TPL) and TPL-related proteins (TPRs) and could potentially repress the activities of target transcription factors (Jiang et al., 2013). Recently, Ideal Plant Architecture1 (IPA1), a key regulator of plant architecture in rice, has been identified as one of the long-speculated transcription factors involved in SL signaling. IPA1 can physically interact with D53 and plays an essential role in the feedback regulation of SL-induced D53 expression (Song et al., 2017).

Although the "de-repression activation" mechanism in SL signaling is similar to the signaling pathways of auxin, gibberellic acid (GA) and jasmonic acid (JA) (Dharmasiri et al., 2005; Kepinski and Leyser, 2005; Ueguchi-Tanaka et al., 2005; Chini et al., 2007; Thines et al., 2007; Jiang et al., 2013; Zhou et al., 2013), the SL receptor D14 functions as a non-canonical receptor (Yao et al., 2016). Therefore, the molecular mechanism underlying the inactivation of SL signaling triggered by the D14-SCFD3-D53 complex becomes an important open question. The degradation of signal receptors by the 26S proteasome system is critical for fine-tuning on signal transduction of plant hormones and environmental signals, such as JA, abscisic acid (ABA), ethylene and the blue light (Chen et al., 2007; Kevany et al., 2007; Yan et al., 2013; Kong et al., 2015; Tao et al., 2015; Liu et al., 2016). In Arabidopsis thaliana, AtD14 has been shown to undergo 26S proteasome-dependent degradation, but the molecular mechanism of AtD14 degradation remains elusive (Chevalier et al., 2014). In rice, whether D14 undergoes degradation and the regulatory mechanism underlies SL perception remain unknown. Due to the profound effects of SLs on tillering, a key agronomic trait in cereal crops, investigating the regulation of SL perception is important for improving plant architecture and grain yield of rice. In this study, we show that SLs stimulate the ubiquitination and degradation of D14 through 26S proteasome in rice. A point mutation at Lys280 of D14 could greatly impair D14 degradation, but has little effect on SL signal transduction. We also show that the hydrolase activity of D14 and the intact functions of D3 and D53 are both required for SL-induced D14 degradation. These discoveries have paved a way for elucidating of the inactivation mechanism of the SL perception in rice.

# MATERIALS AND METHODS

## Plant Materials

Rice (Oryza sativa L. spp. japanica) mutants used in this study were d53 and d14 from Nipponbare and d3 from Zhonghua 11 (ZH11) as described previously (Jiang et al., 2013; Sang et al., 2014). Rice plants were cultivated in the experimental field of the Institute of Genetics and Developmental Biology at Beijing in the summer and Hainan in the winter. For quantitative PCR with reverse transcription and transient expression analysis, the seedlings of wild-type and mutants were grown in a growth chamber with a 16-h light at 28◦C and 8-h dark at 25◦C photoperiod with approximately 200 µM m−<sup>2</sup> s <sup>−</sup><sup>1</sup> photon density and 70% humidity. For calli treatment assays, rice calli were cultured on selection medium (Gamborg et al., 1968; Chu et al., 1975) for 7 days at 28◦C in a light-avoided environment in greenhouse.

### Chemicals and Reagents

fpls-08-01935 November 7, 2017 Time: 16:47 # 3

The synthetic SL analog GR24, a racemic mixture (rac-GR24) comprising amounts of GR245DS and GR24ent-5DS, was the product from Chiralix, MG132 from Calbiochem, the complete protease inhibitor cocktail (Cat#04693132001) and anti-GFP antibody (Cat#11814460001) from Roche, anti-actin antibody (Cat#M20009L) from Abmart, anti-Flag antibody (Cat#M20008) from Abmart, GFP-Trap <sup>R</sup> A beads (Cat#120716001A) from Chromotek, TRIzol kit (Cat#15596018) and Superscript III RT kit (Cat#18080-051) from Invitrogen, TURBO DNA-freeTM Kit (Cat#AM1907) from Ambion, SsoFast EvaGreen supermix (Cat#172-5201AP) from Bio-Rad, and Glutathione Sepharose 4 Fast Flow (Cat#17-5132-01) from GE healthcare. The antiubiquitin and anti-D53 antibodies are generated as described previously (Jiang et al., 2013; Tian et al., 2015).

# Plasmid Construction

The plasmids of Actin:D14-GFP, Actin:D14S147A-GFP and Actin:D14 H297Y-GFP were generated in previous studies (Jiang et al., 2013). To construct the 35S:D14-GFP plasmid, the coding sequence of D14 was amplified with primers of pBI221-D14- F/R (Supplementary Table 1) and recombined into pBI221-GFP vector. To generate the 35S:D14K280E-GFP plasmid, the fulllength of D14K280E coding sequences were derived from 35S:D14- GFP plasmid by site-directed mutagenesis using the primers of 35S-D14K280E-F/R (Supplementary Table 1), and subsequently cloned into the pBI221-GFP vector. To construct the plasmid of D14-GFP, the coding sequence of D14 was amplified with primers of Ubi-D14-F/R (Supplementary Table 1) and cloned into pTCK303. To generate the plasmids of D14K33E-GFP, D14K55E-GFP, D14K166E-GFP, D14K246E-GFP and D14K280<sup>E</sup> -GFP, the full-length of the coding sequences of D14K33E , D14K55E , D14K166E , D14K246E and D14K280E were derived from 35S:D14- GFP by site-directed mutagenesis using primers of D14K33E - F/R, D14K55E-F/R, D14K166E-F/R, D14K246E-F/R and D14K280E - F/R (Supplementary Table 1), respectively, and subsequently recombined into the binary vector pTCK303.

# Chemical Treatment of Rice Calli

To examine whether SLs could induce the degradation of D14- GFP, D14K33<sup>E</sup> -GFP, D14K55<sup>E</sup> -GFP, D14K166<sup>E</sup> -GFP, D14K246<sup>E</sup> - GFP, and D14K280<sup>E</sup> -GFP, calli of these transgenic lines were cultured on the selection medium (Gamborg et al., 1968; Chu et al., 1975) for 7 days at 28◦C and then transferred into a liquid medium. After the treatment with rac-GR24 at the indicated concentration for various hours, calli were collected and frozen at −80◦C. Total proteins were extracted with the extraction buffer (50 mM sodium phosphate buffer, pH 7.0, 150 mM NaCl, 10% (v/v) glycerol, 0.1% NP-40, and 1× complete protease inhibitor cocktail). The supernatant was boiled with 5× SDS buffer at 100◦C for 10 min and immunoblotting was performed with anti-GFP and anti-Actin antibodies.

#### In Vivo Ubiquitination Assay

The seedlings of wild-type were cultured in greenhouse for 2 weeks. Protoplasts were prepared from shoot tissues and

FIGURE 1 | Strigolactones (SLs) promote ubiquitination and degradation of D14 in rice. (A) Protein levels of D14-GFP in calli of the Act:D14-GFP/d14 transgenic line at different time points after 20 µM rac-GR24 treatment with DMSO as a control. D14-GFP was detected by immunoblotting with an anti-GFP monoclonal antibody. Relative protein levels were determined by densitometry and normalized to loadings determined by Ponceau staining (red) in the immunoblotting analyses. (B) Relative expression levels of D14 and D53 in 2-week-old seedlings after 5 µM rac-GR24 treatment. Values represent means ± SEM, n = 3. (C) D14-GFP protein levels in calli of the D14-GFP/d14 transgenic line at different time points after rac-GR24 treatment in the presence or absence of MG132. Calli are pretreated with 50 µM MG132 for 1 h and then treated with 20 µM rac-GR24 or DMSO for 3 h. D14-GFP was detected by immunoblotting with an anti-GFP monoclonal antibody. Actin1 were used as loading control in the immunoblotting analyses. (D) Ubiquitination analysis of D14-GFP in rice protoplasts. Rice (Nipponbare) protoplasts were transformed with 35S:D14-GFP plasmids and incubated for 12 h, then pretreated with 50 µM MG132 for 1 h and immediately treated with 20 µM rac-GR24 or DMSO for 3 h. Proteins were extracted for affinity purification with an agarose-immobolized anti-GFP monoclonal antibody and followed by immunoblotting analysis with an anti-ubiquitin (upper panel) or anti-GFP (lower panel) monoclonal antibody.

transformed with 35S:D14-GFP as described (Bart et al., 2006). After incubation at 28◦C for 12 h in W5 solution (154 mM NaCl, 125 mM CaCl2, 5 mM KCl, 2 mM MES pH 5.7), the protoplasts were collected and pretreated with 50 µM MG132 for

1 h and then treated with 20 µM rac-GR24 or DMSO at 28◦C for 3 h. Total proteins were extracted with the extraction buffer (50 mM sodium phosphate buffer, pH 7.0, 150 mM NaCl, 10% (v/v) glycerol, 0.1% NP-40, 50 mM MG132, and 1× complete protease inhibitor cocktail). The lysates were centrifuged at 18,000 g for 20 min at 4◦C. The supernatant was then taken for immunoprecipitation and subsequent immunoblot analysis using anti-ubiquitin and anti-GFP antibodies. Under the guidance of the supplier's instruction, 20 µL of GFP-Trap <sup>R</sup> A beads were added into 1.2 mL totally extracted proteins and incubated at 4◦C for 3 h with gentle rotation. The beads were washed three times with the washing buffer without NP-40 and then boiled with 50 µL SDS-PAGE sample buffer for protein blotting. Mouse antiubiquitin monoclonal antibody was used at a 1:2,000 dilution and mouse anti-GFP polyclonal antibodies at a 1:3,000 dilution.

#### Gene Expression Analysis

The seedlings of wild-type were hydroponic-cultured (pH 5.5) in greenhouse for 2 weeks. After 5 µM rac-GR24 treatment, the shoot base (0.5 cm) of the seedlings were harvested at different time points, total RNAs were extracted using a TRIzol kit according to the manufacturer's manual and then treated with the TURBO DNA-freeTM Kit and used for complementary DNA synthesis with the Superscript III RT kit. About 12.5 µg total RNA was added into a 20-µL TURBO DNase mixture reaction system and incubated at 37◦C for 30 min, and then added 2 µL DNase inactivation reagent and centrifuged at 12,000 g for 10 min, and finally 4 µL of the supernatants was used for complementary DNA synthesis. The quantitative PCR with reverse transcription experiments were performed with gene-specific primers of D14- RT-F/R and D53-RT-F/R (Supplementary Table 1) on a CFX 96 real-time PCR detection system (Bio-Rad). Each reaction volume was set as 10 µL, consisting of 5 µL SsoFast EvaGreen supermix, 0.5 µL sense primer (5 µM), 0.5 µL antisense primer (5 µM), 2.0 µL diluted cDNA, and 2 µL ddH2O. Rice UBIQUITIN (LOC\_Os03g13170) gene was used as the internal control.

# Co-IP Assay

The seedlings of the wild type were hydroponic-cultured (pH 5.5) in greenhouse for 2 weeks. Protoplasts generated from shoot tissues were transformed with 35S:D14-Flag, 35S:D3-GFP, or 35S-GFP (Bart et al., 2006). After incubation at 28◦C for 12 h in W5 solution (154 mM NaCl, 125 mM CaCl2, 5 mM KCl, 2 mM MES, pH 5.7), proteins were extracted from the collected protoplasts using the extraction buffer (50 mM sodium phosphate buffer, pH 7.0, 150 mM NaCl, 10% (v/v) glycerol, 0.1% NP-40, 50 µM MG132, 1× complete protease inhibitor cocktail) by centrifugation at 20,000 g for 20 min at 4◦C and then the supernatant was taken out for Co-IP experiments. Then 20 µL of GFP-Trap <sup>R</sup> A beads was added into 1.2 mL extracted total proteins and incubated at 4◦C for 3 h with gentle rotation in the presence or absence of 10 µM rac-GR24. The beads were washed three times with the washing buffer lack of NP-40 and then boiled with 50 µL SDS-PAGE sample buffer for protein blot.

FIGURE 2 | Characterization of key amino acid residues in D14 for its degradation. (A) The lysine sites in the D14 amino acid sequence and their predicted scores from the BDM-PUB website. (B) Schematic diagrams showing the constructs of D14 with point mutation of candidate ubiquitination sites. (C) Protein levels of D14-GFP, D14K33E-GFP, D14K55E-GFP, D14K166E-GFP, D14K246E-GFP and D14K280E-GFP in rice protoplasts transformed with plasmids shown in (B) after treatment with 10 µM rac-GR24 or DMSO for 2 h. Protein levels were detected by immunoblotting with an anti-GFP monoclonal antibody. Relative protein levels were normalized to the loading control determined by Ponceau staining (red).

The proteins of D3-GFP and GFP were detected by mouse anti-GFP antibody at a 1:3,000 dilution, and the D14-Flag proteins by mouse anti-FLAG antibody at a 1:2,000 dilution.

#### Microscopy Analyses

fpls-08-01935 November 7, 2017 Time: 16:47 # 5

Protoplasts were prepared from shoot tissues of 2-week-old seedlings and transformed with the plasmid of 35S:GFP, 35S:D14- GFP or 35S:D14K280E-GFP. The SV40NLS-mCherry plasmid, which bears a strong nucleus localization signal peptide, was co-transformed into the protoplasts to label the nucleus (Ye et al., 2012). After incubation for 14 h at 28◦C in the dark, the protoplasts were collected to observe the GFP and mCherry signals with confocal microscope at the excitation wavelengths of 488 and 559 nm, respectively (FluoView FV1000; Olympus).

#### RESULTS

#### SLs Stimulate the Ubiquitination and Degradation of D14

To explore whether the SL receptor undergoes a feedback regulation in rice, we first investigated the D14 protein levels after SL treatment. In Actin:D14-GFP transgenic calli treated with 20 µM rac-GR24, the D14-GFP fusion protein amount begun to decrease at 1 h, and dramatically reduced at 4 h (**Figure 1A**). We then examined the expression levels of D14 and D53 upon 5 µM rac-GR24 treatment in 2-week-old seedlings, and found that D14 transcripts were unaffected within 6 h treatment, while D53 transcripts were strongly induced after 2 h treatment (**Figure 1B**). These results indicate that SLs could induce the degradation of the D14 protein but have no effect on D14 transcription.

To investigate whether D14 is degraded through the ubiquitin-26S proteasome system, we detected D14 protein levels in the Actin:D14-GFP transgenic calli treated with 20 µM rac-GR24 in the presence or absence of MG132, and found that the SL-induced D14 degradation was strongly inhibited by MG132, indicating that the 26S proteasome pathway is involved in the degradation of D14 (**Figure 1C**). We further examined the polyubiquitination of D14 upon rac-GR24 treatment in rice protoplasts transformed with 35S:D14-GFP. After 3 h treatment, the transiently expressed D14-GFP recombinant protein was polyubiquitinated, but no polyubiquitination signal was detected in any other negative control (**Figure 1D**). Collectively, these data demonstrate that SLs stimulate the ubiquitination and degradation of D14 via the ubiquitin-26S proteasome system in rice.

#### Identification of Key Amino Acids Responsible for SL-Induced D14 Degradation

To identify the potential ubiquitination sites of D14, we analyzed the protein sequence of D14 using BDM-PUB<sup>1</sup> . Five lysine sites, K33, K55, K166, K246, and K280, were

localizations of GFP, D14-GFP and D14K280E-GFP in rice protoplasts. The 35S:SV40NLS-mCherry plasmid was cotransformed to label the nucleus. BF, bright field. Bar = 10 µm. (B) In vivo interaction between D14-Flag and D3-GFP revealed by co-IP assay in rice protoplasts. After transformation and incubation for 12 h, protoplasts were treated with 10 µM rac-GR24 for 1 h, and then the supernatant extracted from the protoplasts was incubated with an agarose-conjugated anti-GFP monoclonal antibody at 4◦C for 3 h in the presence or absence of 10 µM rac-GR24, following which the D14 recombinant protein was detected with an anti-Flag monoclonal antibody, while D3-GFP and GFP were detected with an anti-GFP monoclonal antibody. Input refers to the total protein lysate before immunoprecipitation. (C) D53 protein levels in calli of Act:D14-GFP/d14, Act:D14K280E-GFP/d14 and d14 treated with 10 µM GR24. D53 was detected by immunoblotting with anti-D53 polyclonal antibodies. Rice Actin1 contents were used as loading controls in the immunoblotting analyses.

predicted as potential candidate ubiquitination sites of D14, of which K280 has the highest score (**Figure 2A**). We then tested whether the degradation of D14 is affected when each ubiquitination site was mutated through creating the pointmutation constructs containing D14K33E-GFP, D14K55E-GFP, D14K166E-GFP, D14K246E-GFP or D14K280E-GFP and expressed them in rice protoplasts, respectively (**Figure 2B**). After

<sup>1</sup>http://bdmpub.biocuckoo.org/results.php

treatment with 10 µM rac-GR24 for 2 h, these recombinant proteins showed degradation at different extents. Compared with the wild-type D14-GFP, the degradation of D14K280E-GFP was severely impaired and the degradation of D14K33E-GFP was moderately decreased upon SL treatment, but the degradation of D14K55E-GFP, D14K166E-GFP or D14K246E-GFP was relatively unaffected (**Figure 2C**), suggesting that K280 might be the key amino acid for SL-induced D14 degradation.

AtD14 has been reported to be degraded after SL treatment in Arabidopsis thaliana (Chevalier et al., 2014). We then compared the amino acid sequences of D14 with its orthologues including AtD14 from Arabidopsis, RMS3 from pea and DAD2 from petunia, and found that the K280 site of D14 is conserved in pea and petunia, but changes to Arginine in Arabidopsis (Supplementary Figure 1), suggesting that the mechanism of D14 degradation might be different between rice and Arabidopsis. Further analysis on the destabilizing effects of the K280E mutation indicate that no obvious conformational change occurs in D14K280E based on structural prediction and comparison (Supplementary Figure 2). Taken together, these results indicate that the ubiquitination and degradation of D14 may involve multiple amino acids, of which K280 appears to play a major role.

# D14K280E Mutation Does Not Affect SL-Induced D53 Degradation

We then evaluated whether the K280 mutation influence the function of D14 in SL signaling. It has been reported that AtD14 is localized in the nucleus and cytoplasm (Chevalier et al., 2014). We therefore investigated the subcellular localization of D14- GFP and D14K280E -GFP in a transient expression system using the SV40NLS-mCherry as a nuclear marker (Wang et al., 2015). As shown in **Figure 3A**, both D14-GFP and D14K280E-GFP proteins were localized in cytoplasm and nucleus, suggesting that K280 mutation does not affect the subcellular localization of D14. We then tested the interaction between D14K280E and D3 in wildtype protoplasts using the co-IP assay and found that D14K280E could interact with D3 after rac-GR24 treatment (**Figure 3B**). Furthermore, we compared the SL-induced degradation of D53 in calli of d14, D14-GFP/d14 and D14K280E-GFP/d14. The protein levels of D14 and D14K280E were comparable in different transgenic lines and displayed no obvious decrease after rac-GR24 treatment for 30 min (Supplementary Figure 3). Consistent with our previous study (Jiang et al., 2013), D53 degradation upon rac-GR24 treatment is dramatically impaired in d14 mutant. Importantly, both D14-GFP and D14K280E -GFP rescue the defect of D53 degradation in d14, indicating that D14K280E - GFP could trigger SL-induced D53 degradation and potential SL signaling (**Figure 3C**). Collectively, the D14K280E mutation doses not disturb the process of SL perception, which requires D14-D3 interaction and D53 degradation.

#### SL-Induced D14 Degradation Requires the Hydrolase Activity of D14

D14 encodes a member of the α/β hydrolase superfamily, which features the canonical triad catalytic center (Hamiaux et al., 2012; Kagiyama et al., 2013). The 147th serine and the 297th histidine sites are essential for the hydrolase activity and SL perception of D14 (**Figure 4A**) (Zhao et al., 2013; Yao et al., 2016). To address whether the hydrolase activity of D14 is required for its degradation, we mutated these key amino acids and generated the D14S147A-GFP and D14H297Y - GFP overexpression transgenic lines in the d14 background. Compared to the Actin:D14-GFP/d14 transgenic calli, the SL-stimulated D14 degradation was severely inhibited in Actin:D14S147A-GFP/d14 and Actin:D14H297Y-GFP/d14 calli (**Figure 4B**), indicating that the hydrolase activity of D14 is essential for the SL-induced D14 degradation. More importantly, the Actin:D14S147A-GFP/d14 and Actin:D14H297Y-GFP/d14 transgenic plants exhibited dwarf and high tillering phenotypes as the d14 mutant (**Figure 4C**), demonstrating that the S147 and H297 sites are both indispensable for the signal perception of SLs in vivo.

### SL-Induced D14 Degradation Is Dependent on D3 and Coupled to D53 Degradation

D3 encodes an F-box protein, which is a subunit of the Skp-Cullin-F-box (SCF) complex and is responsible for substrate recognition (Smith and Li, 2014). SLs trigger the interaction between D14 and D3, which in turn leads to the ubiquitination and degradation of D53, and probably relieves the repression on gene expression (Jiang et al., 2013; Zhou et al., 2013). To check whether the degradation of D14 is depended on D3, we first identified a loss-of-function mutant of d3 in the ZH11 background, which contains a premature translation mutation due to a 1-bp (base pair) deletion in the first exon and displayed dwarf and high-tillering phenotypes (Supplementary Figure 4). We then treated the Actin:D14-GFP transgenic calli in the wildtype or d3 background with 20 µM rac-GR24. Consistent with the role of D3 in targeting D53 for ubiquitination and degradation, we found that D53 degradation upon SL treatment was strongly inhibited in the d3 mutant. More importantly, D14 degradation was also blocked in the d3 mutant (**Figure 5A**), suggesting that D3 may directly involves in the proteolysis of D14 or D3 regulates the expression of unknown SL-responsive genes and consequently control D14 degradation.

Furthermore, we compared the amounts of D14-GFP and D53 in Actin:D14-GFP transgenic calli in the wild-type and d53 background (**Figure 5B**). The d53 mutant shows typical SLdeficient phenotypes including dwarf and high tillering, and is insensitive to exogenous SL treatment. These developmental defects are caused by a significant attenuation of D53 degradation due to an in-frame deletion of amino acid 813-817 and an amino acid substitution at 812 of D53 protein (Jiang et al., 2013). Thus the SL signaling is impaired in the d53 mutant. After rac-GR24 treatment for 1 h, endogenous D53 protein almost disappeared in wild type (left) but remained stable in the d53 mutant (right), consistent with results in **Figure 3C** and our previous observation that D53 was almost disappeared within 30 min but remained stable in d53 (Jiang et al., 2013). Meanwhile, the protein levels of D14-GFP in wild type began to decrease

(left) but remained stable in d53 (right). With extended SL treatment for 3 and 6 h, endogenous D53 in the d53 mutant began to degrade and reached about half of the value at zero time. The D14-GFP also degraded in d53 (right) but the degradation degree was significantly impaired compared to that of wild type (left) (**Figure 5B**). These results indicate that D14 degradation is coupled to D53 degradation, which is closely related to the status of SL signaling.

# DISCUSSION

The degradation of receptors exists as a fine-tune mechanism in the signaling of hormones, such as ABA, JA, brassinolide and ethylene (Kevany et al., 2007; Wu et al., 2011; Yan et al., 2013; Kong et al., 2015). In this study, we found that D14 degradation was observed from 1 h and reached maximum level at about 3–4 h (**Figures 1A,C**, **4B**, **5**), while D53 protein is degraded at 10 min and almost disappeared at 30 min after GR24 treatment (**Figure 3C**), indicating that the SL receptor D14 is degraded through the 26S proteasome following the degradation of D53 in rice. SLs are recognized and hydrolyzed into a D-ring-derived molecule by D14 and form a covalently link bridge with the catalytic sites of D14 (Yao et al., 2016). This process stimulates a conformational change of D14 and the formation of a SCFD3-D14-D53 complex, leading to the ubiquitination and degradation of D53 (Jiang et al., 2013; Zhou et al., 2013). Subsequently, the repression of D53 on SL signaling is relieved, and the downstream components are activated to regulate plant development. Meanwhile, SLs could in turn trigger the degradation of D14 through the 26S proteasome pathway, leading to the inactivation of SL perception (**Figure 6**).

The molecular mechanisms by which D14 was ubiquitinated and degraded are still open questions in higher plants (Chevalier et al., 2014; Wang et al., 2017). We found that D14 degradation was blocked in the loss-of-function mutant of d3 that displays completed obstructed D53 degradation, and was impaired in the dominant mutant d53 that displays significantly attenuated D53 degradation (**Figure 5**). These phenomena raised two possibilities for D14 degradation in vivo. The F-box protein D3 may target D53 and then D14 for ubiquitination and degradation. In this case, the over-accumulated D53 protein in the d53 mutant may hold more D3 protein and disturb the ubiquitination of D14. But another scenario also fit the available information. An unknown E3 ligase downstream of SL signaling may involve in the ubiquitination and degradation of D14. When SL perception is blocked (as in d3 mutant) or attenuated (as in d53 mutant), the expression or modification of this E3 ligase may change accordingly and thus influence D14 degradation (**Figure 6**). Crucially, further analysis of D14 ubiquitination and degradation

FIGURE 5 | The SL-triggered D14 degradation in d3 and d53 mutants. (A) D14-GFP and D53 protein levels in Act:D14-GFP transgenic calli in the wild-type (ZH11) or d3 background after 20 µM rac-GR24 or DMSO treatment. (B) D14-GFP and D53 protein levels in transgenic calli of wild-type (Nipponbare) or d53 background after 20 µM rac-GR24 or DMSO treatment at different time points. D14-GFP and D53 levels were detected by immunoblotting with an anti-GFP monoclonal antibody and anti-D53 polyclonal antibodies, respectively. Relative amounts of proteins were determined by densitometry and normalized to actin1 and expressed relative to the value at zero time. Actin1 contents were used as loading controls in all the immunoblotting analyses.

based on in vitro reconstitution of SCFD3 E3 ligase activity and the structural analysis of D14-D3-D53 complex will give direct evidence of whether D3 direct target D14 for ubiquitination and degradation.

Feedback regulation is an important and universal mechanism to maintain the homeostasis of a signaling pathway (Schwechheimer, 2008). In plants, the biosynthesis and signal transduction of hormones are adjusted by several transcriptional and post-translational feedback regulation (Benjamins and Scheres, 2008; Kendrick and Chang, 2008; To and Kieber, 2008; Vlot et al., 2009; Jaillais and Chory, 2010; Jaillais and Vert, 2016; Yu et al., 2016; Carvalhais et al., 2017). For examples, auxins

trigger the degradation of repressor proteins AXIN/INDOLE-3-ACETIC ACIDs (Aux/IAAs) and subsequently release the AUXIN RESPONSE FACTOR (ARF) transcription factors to regulate signal transduction. However, ARF could also promote the expression of Aux/IAA genes to repress the auxin signaling (Benjamins and Scheres, 2008). In the SL biosynthetic pathway, SLs repress the transcription of MAX3 and MAX4, which are key biosynthetic genes in Arabidopsis (Mashiguchi et al., 2009; Wang et al., 2015). In addition, in max and d mutants, the expression levels of MAX4 and its rice orthologous gene D10 were up-regulated (Jiang et al., 2013; Zhou et al., 2013; Wang et al., 2015). In the SL signaling pathway, SLs trigger rapid degradation of D53 to promote SL signaling in rice, but in turn activate the transcription of the D53 gene to limit SL signaling (Jiang et al., 2013; Zhou et al., 2013). This feedback regulation is mediated by the transcription factor IPA1, which could interact with D53 and activate D53 expression through directly binding the D53 promoter (Song et al., 2017). In Arabidopsis, AtD14 has been reported to undergo SL-induced and MAX2-dependent degradation through the proteasome system (Chevalier et al., 2014). In this work, we showed that SLs also promote the D14 degradation in rice, which would form a negative feedback loop of SL signaling. Intriguingly, SLs could trigger D53 degradation from 5 min, then downregulate MAX3 and MAX4 expression levels while upregulate the D53 expression from about 1–2 h. The SL-induced D14 degradation was observed from 1 h and reached maximum level at about 3–4 h, suggesting that the precise feedback regulation loops in different time dimension would effectively modulate the duration and intensity of SL signaling.

Parasitic plants are the largest biotic cause of reduced crop yields throughout the Africa (Rodenburg et al., 2016; Gobena et al., 2017). The germination of parasitic plants, mainly including Striga and Orobanche, is triggered by SLs secreted from crop roots (Waters et al., 2017). Currently, several strategies have been proposed to control parasitic plants, including induction of Striga germination through treating the unplanted fields with SL-like compounds and kill the parasitic plants before planting crops, and breeding of new crop varieties with altered SL biosynthesis or secretion without harming their normal growth (Nakamura and Asami, 2014; Toh et al., 2015; Fernández-Aparicio et al., 2016; Holbrook-Smith et al., 2016). Recent studies on highly Strigaresistant sorghum lines indicated that the SL profiles of root exudates from the lgs1 variants displayed reduced 5-deoxystrigol, a highly active Striga germination stimulant, but enhanced orobanchol, an SL required for normal growth. This change did not bring disadvantageous effect on plant development, but could inhibit the stimulation of Striga germination effectively (Gobena et al., 2017). In this study, we found that the mutation of D14K280E did not affect the normal function of D14 and SL signaling, but inhibited the D14 ubiquitination and degradation following SL treatment (**Figures 2**, **3**). It is speculated that overexpressing the D14K280E may lead to the hypersensitivity to SLs but this may also need other amino acids working cooperatively with K280, for example K33, which moderately regulate D14 degradation (**Figure 2C**). Furthermore, when the mutated D14 protein resistant to ubiquitination and degradation is overexpressed, tiller number in rice is speculated to decrease due to an enhanced SL perception. Overexpression of this kind of D14 protein may suppress the high tillering phenotypes d27 and d17 that display impaired SL biosynthesis. This may present an alternative approach to cultivate rice varieties that perform decreased SL biosynthesis and reduced germination of Striga without harming the internal SL signaling pathway and regular plant development. It is rational that genetic modifications on the biosynthesis and perception of SLs are applicable to control the parasitic plant growth in crops in the future.

# AUTHOR CONTRIBUTIONS

QH, YH, BW, and JL conceived this project and designed all experiments. QH, YH, LW, SL, XM, GL, YJ, MC, XS, and LJ performed some of the experiments. BW, QH, YH, HY, and JL analyzed the data and wrote the paper. All authors commented on the article.

# FUNDING

This work was supported by grants from the National Key Research and Development Program of China (Grant 2016YFD0100901), National Natural Science Foundation of China (Grant 91635301), and the Strategic Priority Research Program "Molecular Mechanism of Plant Growth and Development" (Grant XDPB0401).

# ACKNOWLEDGMENTS

We thank Prof. Yonghong Wang and Linzhou Huang (Institute of Genetics and Developmental Biology, Chinese Academy of Sciences) for providing the rice d3 mutant, Prof. Yijun Qi (Tsinghua University) for providing the plasmid of SV40NLS-mCherry, Prof. Qi Xie (Institute of Genetics and Developmental Biology, Chinese Academy of Sciences) for technical assistance in ubiquitination analysis, and lab members (Dr. Jingbo Duan, Huihui Liu) for their suggestions and technical assistance.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2017.01935/ full#supplementary-material

### REFERENCES

fpls-08-01935 November 7, 2017 Time: 16:47 # 10


through inhibiting auxin biosynthesis. Proc. Natl. Acad. Sci. U.S.A. 111, 11199–11204. doi: 10.1073/pnas.1411859111


**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 Hu, He, Wang, Liu, Meng, Liu, Jing, Chen, Song, Jiang, Yu, Wang 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.

fpls-08-01935 November 7, 2017 Time: 16:47 # 11

# Accumulation and Transport of 1-Aminocyclopropane-1-Carboxylic Acid (ACC) in Plants: Current Status, Considerations for Future Research and Agronomic Applications

#### Lisa Vanderstraeten and Dominique Van Der Straeten\*

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

1-aminocyclopropane-1-carboxylic acid (ACC) is a non-protein amino acid acting as the direct precursor of ethylene, a plant hormone regulating a wide variety of vegetative and developmental processes. ACC is the central molecule of ethylene biosynthesis. The rate of ACC formation differs in response to developmental, hormonal and environmental cues. ACC can be conjugated to three derivatives, metabolized in planta or by rhizobacteria using ACC deaminase, and is transported throughout the plant over short and long distances, remotely leading to ethylene responses. This review highlights some recent advances related to ACC. These include the regulation of ACC synthesis, conjugation and deamination, evidence for a role of ACC as an ethyleneindependent signal, short and long range ACC transport, and the identification of a first ACC transporter. Although unraveling the complex mechanism of ACC transport is in its infancy, new questions emerge together with the identification of a first transporter. In the light of the future quest for additional ACC transporters, this review presents perspectives of the novel findings and includes considerations for future research toward applications in agronomy.

Keywords: 1-aminocyclopropane-1-carboxylic acid, ACC, agriculture, conjugation, deaminase, ethylene, signal, transport

### ACC, THE DIRECT PRECURSOR OF THE PLANT HORMONE ETHYLENE

1-aminocyclopropane-1-carboxylic acid (ACC) is a three-membered ring non-protein amino acid which is the direct precursor of the plant hormone ethylene. This gaseous plant hormone was identified as a regulator of plant growth in 1901 by Neljubov. Decades of dedicated research revealed a myriad of plant responses to ethylene (Abeles et al., 1992). This two-carbon atom molecule controls several processes linked to vegetative plant growth but is also a major player in seed germination, fruit ripening, leaf and flower senescence and abscission (Lin et al., 2009; Bakshi et al., 2015). During the above-mentioned processes ethylene production increases significantly in comparison to the relatively low basic levels. The regulation of seedling growth is one of the best characterized ethylene responses. First reported by Neljubov, and later confirmed by Knight and Crocker (1913), was the ethylene response of dark-grown seedlings, known as the triple response including (1) shortening of the hypocotyl and the root, (2) radial swelling of the hypocotyl, and (3) exaggeration of the apical hook.

#### Edited by:

Chi-Kuang Wen, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, China

#### Reviewed by:

Jin-Song Zhang, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, China Chuanli Ju, Capital Normal University, China

#### \*Correspondence:

Dominique Van Der Straeten dominique.vanderstraeten@ugent.be

#### Specialty section:

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

Received: 29 November 2016 Accepted: 09 January 2017 Published: 24 January 2017

#### Citation:

Vanderstraeten L and Van Der Straeten D (2017) Accumulation and Transport of 1-Aminocyclopropane-1-Carboxylic Acid (ACC) in Plants: Current Status, Considerations for Future Research and Agronomic Applications. Front. Plant Sci. 8:38. doi: 10.3389/fpls.2017.00038

(Auxin); CK, cytokinin; BR, brassinosteroid; JA, jasmonic acid; ABA, abscisic acid.

Lieberman and Mapson (1964), Lieberman et al. (1965), Murr and Yang (1975), Adams and Yang (1977), and Adams and Yang (1979) made major contributions to our understanding of ethylene biosynthesis. A simplified overview of the pathway is presented in **Figure 1**. In higher plants, ethylene is produced from the amino acid methionine (Lieberman and Mapson, 1964; Lieberman et al., 1965), which is converted to S-adenosyl-Lmethionine (SAM) (Burg, 1973) by SAM-synthetase (SAMS; also known as L-methionine-S-adenosyltransferase) (Murr and Yang, 1975; Adams and Yang, 1977, 1979). Subsequently, the threemembered ring amino acid 1-aminocyclopropane-1-carboxylic acid (Adams and Yang, 1979) is formed in a reaction catalyzed by the enzyme ACC synthase (ACS) (Boller et al., 1979). ACS is part of a family of PLP dependent enzymes, which require pyridoxal-5<sup>0</sup> -phosphate (PLP) as a cofactor. After binding of PLP to its catalytic site, ACS cleaves a 5<sup>0</sup> -methylthioadenosine (MTA) molecule from SAM, and induces the formation of the cyclopropane ring characteristic for ACC (Yu et al., 1979; Liang et al., 1992). MTA is recycled back to methionine through a set of different reactions known as the Methionine Salvage Pathway or Yang-cycle (Murr and Yang, 1975; Miyazaki and Yang, 1987; Burstenbinder et al., 2010). By recycling the sulfur and methyl group from SAM through the Yang-cycle, the plant is capable of producing high levels of ethylene without depleting the sulfurcontaining methionine pool (Miyazaki and Yang, 1987). This process is essential as sulfur is a limiting compound for plants. In a final biosynthetic step, ACC is converted to ethylene in a reaction catalyzed by the enzyme ACC-oxidase (ACO), which was identified and characterized by John et al. (1985); Ververidis and John (1991).

# CONTROL OF THE ACC POOL: ACC ACCUMULATION, ACC REDUCTION, AND REGULATION THEREOF

The levels of ethylene biosynthesis throughout the plant are relatively low during vegetative development, but are increased in response to a wide variety of developmental conditions as well as by several hormonal signals and environmental cues (**Figure 1**). Firstly, ethylene biosynthesis can be altered in response to developmental processes such as germination, fruit ripening, leaf and flower senescence, and abscission (Yang and Hoffman, 1984; Abeles et al., 1992). Secondly, it is regulated by plant hormones auxins, cytokinins, brassinosteroids, jasmonic acid, and abscisic acid (ABA) (Yang and Hoffman, 1984; Abeles et al., 1992; Vogel et al., 1998b; Woeste et al., 1999). Thirdly, ethylene can affect its own biosynthesis through positive and negative feedback loops. (Riov and Yang, 1982a,b; Kende, 1993; Nakatsuka et al., 1997, 1998). Finally, ethylene biosynthesis can be enhanced by biotic and abiotic stress signals such as flooding, wounding, drought, and pathogen attack (Morgan and Drew, 1997; Pierik et al., 2007; Kazan, 2015).

The majority of the regulatory mechanisms of ethylene biosynthesis act at the level of ACC production by ACS. However, there are additional regulatory mechanisms. Under conditions of high ethylene production, the pathway can also be regulated at the level of the conversion of ACC into ethylene by ACO. Conjugation and deamination of ACC regulates the pool of available ACC. In addition, the ACC pool is also indirectly altered by action of VAS1 (REVERSAL OF SAV3 PHENOTYPE 1). The vas1 mutant, identified by Zheng et al. (2013) as a genetic

suppressor of the auxin biosynthesis mutant shade avoidance3 (sav3), links ethylene with auxin biosynthesis. VAS1 is an aminotransferase catalyzing the transamination of the auxin biosynthetic intermediate indole-3-pyruvic acid into the amino acid L-tryptophan. For this transamination reaction VAS1 uses the ethylene biosynthetic precursor methionine as an amino donor. VAS1 activity was linked to shade avoidance and it was proposed that the reduction in auxin and ethylene by VAS1 restricts elongation growth, providing a mechanism to prevent plants from over-reacting to shade.

#### Regulation of the ACC Pool: ACC Production by ACS

The conversion of SAM to ACC by ACS is the major regulatory step in ethylene biosynthesis, hence all conditions which ultimately lead to ethylene formation cause an accumulation of ACC (Yang and Hoffman, 1984). As mentioned previously, ACS is a member of a superfamily of proteins requiring pyridoxal-5<sup>0</sup> phosphate (PLP) as a cofactor, known as PLP-dependent proteins. In most plant species, ACS is encoded by a multigene family, the members of which show distinct but overlapping expression patterns. The Arabidopsis genome contains eight ACS genes encoding functionally active ACS enzymes (ACS2, ACS4-9, and ACS11), and a ninth catalytically inactive member, ACS1 (Liang et al., 1992; Van Der Straeten et al., 1992; ACS1 in the latter reference corresponds to ACS2 in the former). ACS mutant complementation by Tarun and Theologis (1998) suggested that the enzyme functions as a homodimer whose active site is formed through interaction of shared residues from the monomeric subunits. Capitani et al. (1999) confirmed this mode of action by study of the quaternary structure of an apple ACS protein. Besides functioning as homodimers, it was suggested that ACSs are also capable of forming heterodimers (Tarun and Theologis, 1998), which was corroborated by Tsuchisaka and Theologis (2004a). The latter study revealed that the ACS proteins can only form enzymatically active heterodimers among members of the same phylogenetic branch, while all functional ACS are homodimers with shared active sites. The capability of acting as homo- or hetero-dimers is a characteristic also present in other PLP-dependent enzymes.

In most plant species, the members of the ACS gene family are differentially regulated at the transcriptional level, in an organspecific, tissue-specific and/or cell-type specific manner (Kende, 1993; De Paepe and Van Der Straeten, 2005). To date, most gene expression studies have been performed on an organ or tissue specific basis. Rodrigues-Pousada et al. (1993) performed a complete analysis of ACS1 (termed ACS2 in Liang et al., 1992) gene expression during Arabidopsis development using a GUSreporter construct. Ishiki et al. (2000) investigated the ACS gene expression of etiolated melon seedlings (root, hypocotyl, and cotyledons) and melon fruit, Peng et al. (2005) investigated the Arabidopsis ACS gene expression in the root and the shoot, and Xue et al. (2008) determined ACS expression patterns in different rose floral tissues (sepals, petals, stamens, gynoecia, and receptacles). However less abundant, a number of examples can be found of cell type-specific gene expression analyses. Geisler-Lee et al. (2010) studied the ACS gene expression in different cell types in maize roots. Tsuchisaka and Theologis (2004b) looked both at the organ specific and cell-type specific Arabidopsis ACS expression patterns. In seedlings, similar but not identical expression patterns were found in the light and in darkness. Interestingly, no expression of ACS9 is observed in both conditions, while in the light ACS8 is the only gene expressed in the root tip. In mature plants, ACS1 is mainly expressed in vascular tissue, ACS2, 4, 5, 6, and 8 are expressed in roots, inflorescence stem, siliques and younger leaves, ACS11 is expressed in roots, inflorescence stem, younger leaves, cauline leaves, and ACS9 is barely expressed. In their cell type-specific expression studies they analyzed tissues from cotyledons, the hypocotyl and the root. In cotyledons and the hypocotyl, expression of all genes, except ACS1 and ACS9, is restricted to the epidermal cell layer, the vascular bundles and the guard cells. In the roots, three different zones were examined: lateral root cap, cell division, and cell expansion. ACS8 is uniquely expressed in the lateral root cap zone. In the other two zones, most ACS expression is restricted to the endodermis, pericycle, and stele. However, IAA treatment enhanced the expression of ACS7 and ACS8 in the epidermis and of ACS11 in all cell types. Using data from GeneAtlas and the AREXdb, these expression patterns were confirmed by Dugardeyn et al. (2008); however, they also revealed expression in the root cap for ACS2, 4, 8, 10, and 12. This discrepancy with the analysis of Tsuchisaka and Theologis (2004b) probably results from limitations of the GUS reporter. Although little attention has been given to the exact mechanisms at the origin of these differences in expression levels, it is established that ACSs are differentially regulated in response to a variety of developmental or environmental signals and in response to plant hormones.

A first example illustrates how ACS genes in tomato are differentially regulated during fruit ripening. Similar to Arabidopsis, tomato ACS genes are part of a multigene family with differential expression patterns regulated by developmental, biotic and abiotic signals. When ripening starts in tomato or other climacteric fruits, the regulation of ethylene biosynthesis switches from auto-inhibitory to auto-stimulatory. In these plant species, two systems of ethylene production have been proposed (Barry et al., 2000; Alexander and Grierson, 2002). System 1 operates during vegetative growth, during which ethylene inhibits its own biosynthesis (auto-inhibition), while system 2 operates upon fruit ripening, during which ethylene induces its own biosynthesis (auto-catalysis). Barry et al. (2000) observed that the transcripts corresponding to four ACS genes, LE-ACS1A, LE-ACS2, LE-ACS4, and LE-ACS6, were detected in tomato fruit. A detailed expression analysis using the wild type and ripening mutants revealed that each of these ACS genes is regulated in a unique way. System 1 ethylene biosynthesis is regulated by the expression of LE-ACS1A and LE-ACS6 in green fruit, followed by a transition period in which the MADS-Box Transcription Factor RIN (RIPENING INHIBITOR) plays an important role in increasing LE-ACS1A expression and inducing LE-ACS4 expression (**Figure 2**). The increased ethylene production triggers the expression of LE-ACS2. Subsequently, autocatalytic system 2 is initiated. The elevated ethylene production results in a negative feedback on the system 1 pathway, reducing LE-ACS1A and

LE-ACS6 expression. System 2 is maintained by expression of LE-ACS2 and LE-ACS4 (Lincoln et al., 1993; Barry et al., 2000).

The next few examples illustrate that the expression of ACS genes show different susceptibilities to other plant hormones, and that these susceptibilities can differ between different cell types. Rodrigues-Pousada et al. (1999) provided the first evidence for transcriptional regulation of ACS1 by auxin and cytokinin. Yamagami et al. (2003) demonstrated that indole-3-acetic acid (IAA) induces six ACS genes (ACS2, 4, 5, 6, 8, 11) in 7 days old etiolated and light grown seedlings. These results were confirmed by Tsuchisaka and Theologis (2004b), who also observed an effect of IAA on the expression of ACS7. The effect of IAA on different ACS genes differed between cell types. For example, IAA induced expression of ACS8 and ACS11 in all cell layers of the cell expansion zone of the root, while it activated ACS5 expression only in the endodermis. Dugardeyn et al. (2008) revealed that, with the exception of ACS4, the same genes are upregulated by gibberellins (GAs). Nemhauser et al. (2006) showed that ACS6 is up-regulated by ABA and brassinolide (BL), and that ACS10 is down-regulated by IAA, gibberellic acid (GA), and ABA, as well as by methyl jasmonate (MeJA). Finally, Wang et al. (2005) showed that ACS4 and ACS5 are also up-regulated by ABA, and that ACS7 is upregulated by GA and ABA.

While the transcriptional regulation of ACS genes is of great importance, post-translational regulation of ACS by protein degradation also plays a major role in the regulation of ethylene biosynthesis (see **Figure 3**). The C-termini, however, unimportant for enzyme activity, were found to play a crucial regulatory role in enzyme stability for proteasomal degradation. The Arabidopsis ACSs are classified into three groups based on the presence or absence of phosphorylation sites in these C-termini (Liang et al., 1992; Chae and Kieber, 2005). The type I ACS enzymes, ACS2 and ACS6, have a C- terminus containing phosphorylation sites for both MPKs (mitogenactivated protein kinases; Liu and Zhang, 2004) and CDPKs (calcium-dependent protein kinases; Tatsuki and Mori, 2001). The type II ACS enzymes, ACS4, ACS5, ACS8, ACS9, and ACS11, have a C-terminus containing only phosphorylation sites for CDPKs. The type III ACS enzyme ACS7 has a C-terminus without any phosphorylation sites. Phosphorylation and/or dephosphorylation of ACS proteins has a severe impact on ethylene biosynthesis. An increased phosphorylation stabilizes the ACS protein, while dephosphorylation leads to an increased proteasomal degradation (Spanu et al., 1994; Liu and Zhang, 2004; Skottke et al., 2011; Xu and Zhang, 2014).

1-aminocyclopropane-1-carboxylic acid synthase stability is controlled by two types of kinases, Ca2+-dependent protein kinase (CDPK) and mitogen-activated protein kinases. Types I and II ACS proteins are phosphorylated by CDPKs, as in both groups CDPK phosphorylation motifs were identified in their C-termini (Tatsuki and Mori, 2001; Sebastia et al., 2004). Type I ACS proteins are dephosphorylated by protein phosphatase 2A (PP2A) and Protein phosphatase 2C (PP2C), decreasing their protein stability (Skottke et al., 2011; Ludwikow et al., 2014). Additionally, the activity of the type I ACSs can be increased by phosphorylation by the MAP kinases MPK3 and MPK6, which have been shown to be pathogen/stress-activated (Liu and Zhang, 2004; Joo et al., 2008; Han et al., 2010). Although there are no CDPK phosphorylation sites found in the C-terminus of the type III ACS7 protein, it was recently shown that ACS7 can be phosphorylated in vitro in its catalytic domain (Huang et al., 2013). The E3 ubiquitin ligase targeting

phosphorylated state. These proteins are ubiquitinated by ETO1 and EOL1/2 ubiquitin ligases. Cytokinin (or brassinosteroid) treatment has been suggested to block the ETO or EOL1/2 mediated targeting of type II ACSs. The type III ACS7 protein is potentially phosphorylated by CDPKs at its catalytic site, and is ubiquitinated by

type-I ACSs for proteasomal degradation has not been identified yet. The degradation of the type-II ACS enzymes is mediated by ETHYLENE-OVERPRODUCER1 (ETO1) and ETO1-like 1/2 (EOL1/2), substrate-specific adaptor proteins of a Cullin4-based E3 ubiquitin ligase complex (Chae et al., 2003; Wang et al., 2004; Christians et al., 2009). The current model is that the phosphorylation of the type II proteins blocks the ability of the ETO1 and EOL1/2 proteins to bind, inhibiting the degradation of these ACSs. The degradation of the type-III ACS7 enzyme and the type-II ACS4 enzyme is mediated by XB3 ORTHOLOG 2 IN ARABIDOPSIS THALIANA (XBAT32), a RING-type E3 ligase (Prasad et al., 2010; Lyzenga et al., 2012).

the E3 ligase XBAT32. P, phosphate group; S, serine residue.

Cytokinin is one of the hormones capable of regulating ethylene biosynthesis, as reflected by the elevated ethylene biosynthesis after treatment of etiolated and light grown seedlings with cytokinin (Vogel et al., 1998a,b; Woeste et al., 1999). Cytokinin does not elevate ACS transcript levels, but decreases the rapid degradation of ACS4 and ACS5 (Vogel et al., 1998b; Woeste et al., 1999; Chae et al., 2003). Brassinosteroids elevate ethylene biosynthesis in a similar fashion by increasing the stability of ACS5 and ACS9, and its effect is additive with that of cytokinin (Hansen et al., 2009). The stabilization of these ACSs by cytokinins and brassinosteroids includes the inhibition of the C-terminus-dependent targeting by ETO1 or EOL1/2; however, a C-terminus-independent regulation was also suggested. Hence, ACSs are regulated by other plant hormones through regulatory signals that can act together to continuously adjust ethylene biosynthesis in the different plant tissues and in response to signals from the environment.

# Regulation of the ACC Pool: Ethylene Production by ACO

While the conversion of SAM into ACC by ACS is known to be the major rate limiting step in ethylene biosynthesis, the enzyme converting ACC into ethylene (ACO) can become rate limiting under conditions of high ethylene production such as fruit ripening (Barry et al., 1996; Nakatsuka et al., 1997; Van de Poel et al., 2012; Rudus et al., 2013). In Arabidopsis, ACO is encoded by a small gene family of five members (ACO1, ACO2, ACO3, EFE, and ACO5). Dugardeyn et al. (2008) investigated the expression patterns of the Arabidopsis ACO genes using Geneatlas and AREXdb. On an organspecific basis, ACO1 is present at low levels in all plant organs, while ACO2 is highly expressed allover, and EFE shows a strong expression in leaves. With respect to cell type- specific expression, all three genes are strongly expressed in the columella, while ACO1 is also strongly represented in the lateral root cap, and the expression of ACO2 is the highest in the vascular tissue of the fast elongation and differentiation zone. Though appearing in almost all vegetative and reproductive tissues, there are differences in accumulation of the members of the ACO gene family during different developmental and physiological processes (Barry et al., 1996; Rudus et al., 2013). In contrast to ACS, much less is known about the transcriptional and post-translational regulation of ACO genes. Similar to the ACSs, ACO expression can also be regulated by plant hormones, as reported for salicylic acid, auxins, ABA, and gibberellins (Leslie and Romani, 1988; Huang et al., 1993; Ogawa et al., 2003; Zhang et al., 2009).

#### Regulation of the ACC Pool: Conjugation and Deamination

Additional to the regulation of ethylene biosynthesis through ACS and ACO, ethylene biosynthesis is also regulated by the capability of forming ACC derivates (see **Figure 4**). Three types of conjugates have been identified so far; however, knowledge on the importance and function of these conjugates remains poor. The first ACC conjugate, N-malonyl-ACC (MACC), was isolated and identified independently by two research teams (Amrhein et al., 1981; Hoffman et al., 1982; Peiser and Yang, 1998). The conjugation of ACC into MACC is catalyzed by the enzyme ACC-N-malonyl transferase (AMT), a reaction requiring malonyl-coenzyme-A (Martin et al., 1995; Peiser and Yang, 1998). MACC can be translocated between the cytosol and the vacuole by ATP-dependent tonoplast carriers (Bouzayen et al., 1988, 1989; Tophof et al., 1989) suggesting that MACC formation and storage in the vacuoles might be important to control the pool of available ACC. This hypothesis was further strengthened by the observation from Liu et al. (1985) and Martin et al. (1995) that the production of MACC by AMT can be induced by exogenous ethylene treatments during the ripening of preclimacteric tomato (Liu et al., 1985; Martin et al., 1995). Furthermore, Jiao et al. (1986) and Hanley et al. (1989) demonstrated that MACC can be reconverted into ACC. A second conjugate that can be formed from ACC is γ-glutamyl-ACC (GACC; Martin et al., 1995). The formation of GACC is catalyzed by the enzyme γ-glutamyl transpeptidase (GGT), a reaction requiring glutathione (GSH; Martin et al., 1995; Peiser and Yang, 1998). A third ACCconjugate is jasmonyl-ACC (JA-ACC), which is formed through the activity of the enzyme jasmonic acid resistance 1 (JAR1; Staswick and Tiryaki, 2004). This linkage between the precursors of ethylene and the active form of JA, isoleucinoyl-JA, might be a mechanism to control both hormones.

The pool of available ACC can also be reduced by the irreversible deamination of ACC (See **Figure 4**). Importantly, this reaction is not only plant-borne but also carried out by certain plant growth-promoting rhizobacteria (PGPR). Honma and Shimomura (1978) identified an ACC deaminase (ACD) in Pseudomonas sp. strain ACP. Plants are capable of releasing ACC into the rhizosphere to attract these PGPR which use ACC as a carbon and nitrogen source (Glick et al., 1998; Penrose and Glick, 2001; Penrose et al., 2001). As a consequence, the available ACC pool, and thus also plant ethylene production; is reduced; hence, plant growth is promoted through the plant– bacterium interaction (Glick et al., 2007a,b). In a similar manner, the reduction of the ACC pool by PGPR helps the plant to cope with stresses such as flooding (Grichko and Glick, 2001a), pathogen attack (Robison et al., 2001), or salinity (Ali et al., 2014). McDonnell et al. (2009) identified the first plant-encoded ACC deaminase and demonstrated its importance in regulating the ethylene balance.

### ACC TRANSPORT: ACC AS A WATER SOLUBLE, MOBILE SIGNAL

#### Long and Short Distance ACC Transport

For many hormones, the site of synthesis does not always coincide with the site of action. The same has been observed for ethylene. Because ethylene is a gaseous molecule, it diffuses

rapidly through the plant tissues inducing mainly local responses, with the exception of aerenchyma through which long distance transport can be conducted, as observed in conditions of waterlogging. In contrast, long distance ethylene signaling between different plant tissues mostly occurs by the transport of ACC.

Because of its role as a stress hormone, ethylene signaling in response to biotic and abiotic stresses has been investigated thoroughly, often including a role for ACC transport. Root to shoot transport of an ethylene signal in response to waterlogging or submergence is a prime example of the importance of ACC transport and has been studies multiple times (Bradford and Yang, 1980; Metraux and Kende, 1983; Zarembinski and Theologis, 1993; Else et al., 1995; Vartapetian and Jackson, 1997; Else and Jackson, 1998; Grichko and Glick, 2001b; Almeida et al., 2003; Vriezen et al., 2003; Jackson, 2008; Dawood et al., 2016).

When plants are waterlogged or submerged, the available oxygen levels in roots drop rapidly, triggering a myriad of effects on plant metabolism. First of all, it induces a transcriptional cascade targeting genes involved in anaerobic metabolism and survival (Gibbs et al., 2011). Secondly, it induces the expression of several ACSs in the roots (Olson et al., 1993, 1995; Zarembinski and Theologis, 1993; Van Der Straeten et al., 1997; Shiu et al., 1998; Rieu et al., 2005). At the same time the conversion of ACC into ethylene is suppressed, because ACO requires oxygen. After root to shoot transport, ACC promotes the expression of ACO genes, stabilizes ACO mRNA, and increases the activity of ACOs already present, resulting in an increased ethylene production in the shoots (English et al., 1995). Differential ACS and ACO expression patterns in response flooding or hypoxia have been observed in Arabidopsis (Peng et al., 2005) and maize (Geisler-Lee et al., 2010). Furthermore, waterlogging and submergence leads to an overall accumulation of MACC and GACC in both roots and shoots (Amrhein et al., 1981).

The importance of ACC transport in response to waterlogging or submergence can be illustrated by research conducted in tomato (Bradford and Yang, 1980; Amrhein et al., 1981; Else et al., 1995; English et al., 1995; Else and Jackson, 1998) and in rice plants (Metraux and Kende, 1983; Zarembinski and Theologis, 1993; Kende et al., 1998; Almeida et al., 2003; Vriezen et al., 2003). Bradford and Yang (1980) showed that ACC is synthesized in tomato plant roots and then transported from hypoxic roots through the xylem to the shoots, where it is rapidly converted to ethylene inducing leaf epinasty. Furthermore, they observed that drainage results in a simultaneous decrease in ACC flux and ethylene production, and that the petioles are dependent upon the ACC flux for the high rates of ethylene synthesis. Besides epinasty, some of the other responses of plants to waterlogging and submergence, such as reduced root permeability, closure of stomata, development of aerenchyma and adventitious roots, and premature fruit drop might be directly or indirectly related to the enhancement of ACC or derivatives, besides ethylene (Vartapetian and Jackson, 1997; Grichko and Glick, 2001b; Sasidharan and Voesenek, 2015). Amrhein et al. (1982) observed that ACC can also be transported through the phloem. Based on these findings, Morris and Larcombe (1995) assessed phloem transport by foliar application of radioactive ACC to the leaves of 21 day-old cotton plants. Total radioactivity was measured over time in the stem, the hypocotyl and the apex and the patterns were consistent with phloem-mediated transport of the radioactive ACC. The authors suggested that part of the ACC which accumulates in leaves after transport through the xylem from stressed roots (Bradford and Yang, 1980) may be re-exported to other organs through the phloem. Despite the observed overall increase in the formation of the ACC conjugates MACC and GACC (Amrhein et al., 1981), leaves are apparently not capable of exporting MACC to the phloem (Morris and Larcombe, 1995). This could be a consequence of translocation of MACC to the vacuoles of leaf cells (Bouzayen et al., 1988; Tophof et al., 1989).

1-aminocyclopropane-1-carboxylic acid transport is not only important during stress conditions. Several studies evaluated the gene expression patterns of ACS and/or ACO during different developmental processes and in different organs or cell types. Woltering and associates investigated the transport of ACC in Cymbidium orchid flowers for the coordination of senescence (Woltering, 1990). They revealed that both endogenous and exogenously applied ACC is rapidly transported from the site of production/application to other flower parts. Jones and associates analyzed ethylene production and ACS and ACO expression patterns in different floral organs of carnation flower (Dianthus caryophyllus L.) in response to pollination. Ethylene production was observed in the styles, petals, and ovaries. High ACS expression levels were found in petals and styles, while ACO expression was highest in the ovaries (Jones and Woodson, 1997, 1999). In Arabidopsis, Dugardeyn et al. (2008) observed little ACO expression in the root meristematic zone, while several ACS genes showed high expression in this zone (ACS3, 4, 10, and 12). The epidermis and endodermis of the elongation/differentiation zone also showed little ACO expression, while different ACS genes were induced. Gallie et al. (2009) investigated the celltype specific gene expression patterns of maize root cells and found that ACS was only expressed in the root cortex, while ACO was mainly expressed in the protophloem sieve elements, suggesting ACC transport between these cell types. In all of the above-mentioned studies, the spatial differentiality in ACS and ACO expression within a given organ suggests that ACC is transported from the site of synthesis to the site of conversion to ethylene.

Shortly after the identification of ACC, the occurrence of intercellular ACC transport over short distances and intracellular compartmentalization of ACC in the vacuole was proven. Lurssen (1981) investigated ACC uptake in soybean leaf disks by measuring ethylene production after incubation with ACC. Pretreatment with ACC, followed by treatment with non-polar amino acids resulted in a reduced ethylene production, suggesting that ACC is transported by an amino acid transport system. Saftner and Baker (1987) observed the uptake of radiolabeled ACC and α-amino-iso-butyric acid (αAIB, structural analog of ACC) in tomato pericarp slices. Finally,

both Tophof et al. (1989) and Saftner and Martin (1993) demonstrated the transport of ACC across the tonoplast in mesophyll cells in three different plant species (barley, wheat, and maize).

## Identification of a First ACC Transporter, LHT1

In a recent study, a first plant ACC transporter was found. Shin et al. (2015) identified an Arabidopsis thaliana mutant, are2 (ACC-resistant2), showing a dose-dependent resistance to exogenously applied ACC. In the light, the wild type shows an elongated hypocotyl and shortened root growth in the presence of 1 µM exogenous ACC (Smalle et al., 1997), while the are2 mutant exhibited a reversed phenotype, with a shorter hypocotyl and more elongated roots. When grown in darkness in the presence of 1 µM ACC the hypocotyls and the roots of the are2 mutant were not significantly inhibited as compared to the wild type. However, at elevated concentrations of ACC (10 µM or above) a triple response was observed, indicating sensitivity of the are2 mutant (**Figure 5**). Interestingly, the mutant displayed normal sensitivity to ethylene, and the synthesis of ethylene was not impaired, neither did the insensitivity to ACC affect sensitivity of the seedlings to other plant hormones. The are2 mutant corresponded to a null allele of LHT1 (LYSINE HISTIDINE TRANSPORTER 1), a plasma membrane (PM)-localized amino acid transporter. LHT was first identified by Chen and Bush (1997), and further characterized by Hirner et al. (2006). To confirm that are2 and LHT1 are allelic, Shin et al. (2015) performed an allelism test with the lht1-5 mutant. The double mutant and the single mutants showed the same ACC resistance and early leaf senescence phenotypes. An uptake experiment using radiolabeled ACC and protoplasts from wild type and are2/lht1-5 mutant leaf mesophyll cells, showed a reduction of 40% in ACC uptake in both mutants, supporting a function of LHT1 in ACC uptake transport. In Arabidopsis, LHT1 is part of a family of amino acid transporters with 10 members (AtLHT1-10), all of which are localized to the plasma membrane and transport a broad spectrum of amino acids from the cell wall space inward (Hirner et al., 2006). The LHT genes show distinct and overlapping expression patterns, indicating that LHT1 homologs may function in ACC uptake under certain developmental or environmental conditions (Foster et al., 2008). Based on unpublished results, Shin et al. (2015) suggest that at least one LHT1 homolog could complement the lht1 mutant.

### FUNCTIONS OF ACC AS AN ETHYLENE-INDEPENDENT SIGNAL

The secretion of ACC into the rhizosphere to attract and interact with PGPR, as discussed in the previous section, is an example of an ethylene-independent function for ACC. Recent studies

present evidence that ACC, or one of its derivates, might also have other signaling roles independent from ethylene (Yoon and Kieber, 2013).

First, the leucine-rich repeat receptor-like protein kinases (LRR-RLKs) FEI1 and FEI2 appear to define an ACCmediated signaling pathway that regulates cell wall function and anisotropic cell expansion (Xu et al., 2008). The kinase domains of both FEI isoforms interact with the type II ACS proteins ACS5 and ACS9 (Chae and Kieber, 2005) without affecting their catalytic activity (Xu et al., 2008). They found that mutation of both FEI genes perturbs the biosynthesis of cell wall polymers. Phenotypic analysis of light-grown seedlings showed that although single fei1 and fei2 mutants are indistinguishable from the wild type, the fei1fei2 double mutant has short, radially swollen roots, and a significant, but less pronounced swelling of the hypocotyl, indicating a defective anisotropic expansion. This expansion defect was suppressed by inhibition of ACS with aminooxyacetic acid (AOA), but not by disruption of the ethylene response pathway with 1-methylcyclopropene (1-MCP) or silver thiosulfate, neither by mutation of ETR1 (an ethylene receptor) or EIN2 (a central regulator of ethylene signaling). All together, these results suggest that the FEIs do not act through ethylene, but rather through ACC. Finally, Xu et al. (2008) also proposed that FEI proteins might act as a scaffold to localize ACS or assemble ACS in a protein complex.

Further evidence for a signaling role for ACC came from a detailed study of multiple loss of function of ACS genes (Tsuchisaka et al., 2009). Strong alleles of octuple mutants displayed embryo lethality as well as reduced branching, none of which are observed in ethylene signaling mutants (Alonso et al., 1999; Tsuchisaka et al., 2009; Tsang et al., 2011). These phenotypic differences between ethylene biosynthesis and signaling mutants once again suggest that ACC or one of its derivates might act as a signal independent from ethylene.

A third study relates ACC signaling to primary root elongation in response to pathogen-associated molecular patterns (PAMPs). Tsang et al. (2011) studied the acute response of plants to perturbation of cell wall integrity using inhibitors of cellulose biosynthesis. Application of AOA, aminoethoxyvinylglycine (AVG), or 2-anilino-7-(4-methoxyphenyl)-7,8-dihidro-5(6H) quinazolinone (7303; ACS inhibitors) or AIB (ACO inhibitor) could revert the suppression of root cell expansion caused by isoxaben, a blocker of cellulose synthesis. Similar experiments using silver ions or norbornadiene, interfering with ethylene perception, did not result in a significant effect on elongation. Additionally, they present that not only isoxaben, but also ACC, has an effect on root elongation both in the ethylene insensitive ein3eil1 mutant and the wild type. These results suggest that ACC has a short-term effect on root elongation that is partially independent of its conversion to ethylene or ethylene signaling, and that it is responsible for the rapid reduction of root elongation triggered by PAMPs.

Considering the above-mentioned observations, a signaling role for ACC independent from ethylene is highly likely. Whether it is ACC itself serving as a signal, or one of its derivatives, is still a matter of debate and needs further investigation.

# CONSIDERATIONS FOR FUTURE RESEARCH

#### Co-regulation of Ethylene Biosynthesis Genes and ACC Transporters?

The identification of LHT1 as an ACC transporter is a cornerstone in unraveling the mechanism and regulation of ACC transport in plants. Giving the complexity of ethylene biosynthesis and signaling, and the evidence that ACC can be transported both over long distances between plant organs and over short distances within or between cells, one can assume that ACC transport involves several proteins, presumably even different transport systems. It is thus quite possible that many more ACC transporters are yet to be identified. Performing uptake experiments of the are2 mutant using radiolabeled ACC, Shin et al. (2015) observed an incomplete reduction of ACC uptake, supporting the presence of additional ACC transporters.

Aiming at a more profound understanding of ACC transport, it is tempting to look at the organ and cell-type specific gene expression patterns of the ethylene biosynthesis genes ACS and ACO as a main determinant of ACC accumulation. In tissues with high ACS expression, where, however, little or no ACO expression is found, ACC export might be important. In contrast, in tissues with low ACS expression, though high ACO expression, ACC import might be important. For an overview of the detailed expression patterns for both enzyme families, we extracted celltype specific and developmental stage specific absolute expression values from the Toronto Arabidopsis eFP browser (BAR). The relative expression percentages, as presented in **Figures 6** and **7,** were calculated for each gene, relative to the expression value of the cell-type with the highest absolute value.

**Figure 6** presents an overview of the cell type-specific expression levels in the roots, the hypocotyl and the leaves. Little ACO expression is found in the root epidermal atrichoblasts, hypocotyl xylem, and leaf guard cells, while some ACSs are highly expressed in these cell-types. This suggests that the ACC produced is most likely transported to other tissues (with higher ACO expression) where it is converted into ethylene. It has to be acknowledged, however, that these expression patterns do not give much information about long distance or intracellular transport of ACC. **Figure 7** presents an overview of the relative expression levels for certain tissues of different developmental stages. These expression patterns show in which developmental tissues ethylene production is important, but they also indicate between which tissues the translocation of ACC might occur. For example, in young flowers the ACOs are barely expressed, while in mature flowers increased expression levels are observed for both ACO2 and EFE. This means that ethylene production is most likely higher in mature flowers. In addition, there is little difference in the expression levels of the ACSs between these two developmental stages. This might suggest that the increased ethylene production in mature flowers is dependent on the supply of ACC from other tissues. In pollen, high levels of ACS expression can be found, while there is almost no ACO expression. This discrepancy corresponds with the observation that high levels of ACC are present in mature pollen, and that

this pollen-ACC may be an important mediator of the early response of flowers to pollination (Whitehead et al., 1983; Reid et al., 1984). This was confirmed by (Reid et al., 1984), with the observation that application of ACC to the stigmas of carnation flowers (Dianthus caryophyllus), causes an initial increase in gynoecium and petal ethylene production similar to that reported for pollinated flowers, and that application of [2-14C]ACC to the stigmas resulted in radioactive ethylene production both by the gynoecia and the petals. This report provides additional evidence that during pollination (and post-pollination) ACC is exported from the pollen, and transported toward different flower tissues, where it is converted into ethylene.

The ACS and ACO gene expression patterns as described above can be compared with those of known transporters to find novel ACC transporters. There are two major paths that can be followed here. On the one hand, one can look at members of general transporter families such as the family of ATP binding cassette (ABC) transporters, of which some members are important for the transport of IAA, ABA, CK, and SL (Xu et al., 2014; Borghi et al., 2015). One the other hand one can look at known amino acid transporters. The Arabidopsis genome contains 60 or more predicted amino acid transporters, belonging to at least two large families.

One of these amino acid transporter families is the LHT family. Previous research already showed that LHT1 is expressed in roots and mesophyll cells, and that LHT2 expression is localized to the tapetum of anthers (Chen and Bush, 1997; Lee and Tegeder, 2004). **Figures 6** and **7** present an overview of the cell type-specific and development specific expression patterns for the Arabidopsis LHTs. LHT1 expression is high in epidermal atrichoblasts and guard cells, two cell types in which ACC transport was suggested to be important (see above). This is in part a confirmation that in these tissues ACC transport could be supported by LHT1 action. LHT2, LHT5, and LHT7 also show high expression in guard cells, while LHT8 and LHT10 show high expression in the hypocotyl xylem. In the pollen, where ACC export might be essential for the induction of pollination, LHT2, 5, 8, and 9 are highly expressed. These results reveal that certain

other members of the LHT family might also be linked with ACC transport. However, as they are also amino acid transporters, these expression patterns might not necessarily mean that this is the case.

# ACC Transport by Other Amino Acid Transporters?

Because ACC is a non-protein amino acid, it seems evident that ACC transporter(s) would have the characteristics of an amino acid transporter, or be transported by a protein that serves in transport of other amino acids as well. In an assay measuring ethylene upon incubation of leaf disks with ACC, Lurssen (1981) tested competition of other amino acids with ACC. Uptake of ACC was inhibited by L-amino acids with non-polar side chains such as L-methionine, L-tyrosine, and L-phenylalanine, but not by polar amino acids. These were the first indications that ACC might be transported by an amino acid transporter. Saftner and Baker (1987) investigated ACC transport in tomato pericarp slices, and observed that the uptake was inhibited by neutral amino acids, with the highest inhibition from methionine. Given that ACC uptake can be inhibited by the competition of other amino acids, it might be of interest to investigate whether other known amino acid transporters are also capable of transporting ACC, for instance by a systematic analysis using heterologous expression in yeast or in protoplasts. ACC transport could be assayed using radiolabeled ACC or by measuring ethylene production.

Besides comparing gene expression patterns of amino acid transporters, as was suggested in the previous section, another approach is to look at the amino acid substrates of the different transporters. As discussed earlier, ACC transport can be suppressed by neutral and/or non-polar amino acids (Lurssen, 1981; Saftner and Baker, 1987). Amino acid transporters having a high affinity for one or more of these amino acids might be considered potential candidates for ACC transport.

# Identification of Additional ACC Transporters

The studies discussed in the previous two sections assume that novel ACC transporters can be found in families of known transporters. It is, however, quite possible that some ACC transporters remain uncharacterized to date. For identification of novel transporters several approaches could be taken. Most of our current knowledge on plant transporters has been

obtained by use of the yeasts Saccharomyces cerevisiae and Schizosaccharomyces pombe (Dreyer et al., 1999). Using one of these heterologous expression systems, cDNA libraries can be screened to identify novel ACC transporters. An alternative approach is a forward genetic screen using the chemical mutagen ethyl methanesulfonate (EMS). For this forward genetic screen, the phenotypic analysis might include the screening for ACC resistance (dose-dependent or not) as seen by Shin et al. (2015) for are2/lht1-5.

# Long-Distance Transport of ACC Conjugates and Their Role in Ethylene Biology

Several studies explored the possibility of transport of ACC conjugates. Fuhrer and Fuhrer-Fries (1985) investigated the formation and the transport of ACC and MACC in pea plants after wounding, which normally induces an increase of ethylene biosynthesis. They concluded that MACC could be transported from the shoot to the root and that the roots act as an MACC sink. However, further investigation is needed to rule out the possibility that the MACC increase at the roots may result from de novo MACC production after ACC transport from the shoot to the roots. Furthermore, the MACC in the shoot could be reconverted into ACC. Finlayson et al. (1991) investigated the transport and metabolism of ACC in sunflower seedlings. This study presents evidence for the presence of MACC in xylem sap of seedlings treated with radiolabeled ACC. However, the presence of MACC in the xylem sap could not be confirmed by GC-MS analysis of untreated seedlings. These results suggest that MACC can be transported through the xylem, but that this transport might only occur during certain stress conditions. Morris and Larcombe (1995) investigated the possibility of MACC transport through the phloem in cotton. In contrast to the previously discussed studies, they found no evidence for MACC transport.

Overall it is clear that additional evidence is needed to confirm a more general nature of MACC transport, and add to our understanding of its biological function in plants. Moreover, it is also important to further investigate whether the other ACC conjugates can be transported, or whether they rather function as a storage forms of ACC to regulate ethylene production within the organ or tissue where they are formed.

#### POSSIBLE APPLICATIONS IN AGRICULTURE

With the rapidly increasing world population and concomitant food demand, more efforts are being made to increase crop yield and crop product quality all over the world. As discussed previously, ethylene regulates a wide variety of vegetative and developmental processes in plants, all being key processes in agricultural and horticultural context.

Several chemicals regulate these ethylene-dependent characteristics either through the inhibition of ethylene synthesis or perception [aminoethoxyvinylglycine (AVG; Byers et al., 2005; Drake et al., 2006), aminooxyacetic acid (AOA; Broun and Mayak, 1981), diazocyclopentadiene (DACP; (Blankenship and Sisler, 1992), silver thiosulphate (STS; Hansen et al., 1996; Hoyer, 1998) and 1-methylcyclopropene (1-MCP; Drake et al., 2006; Zhu et al., 2015)], or through the release of ethylene from 2-chloroethylphosphonic acid (Ethephon; Logendra et al., 2004; Drake et al., 2006). Treatment with inhibitors of ethylene biosynthesis can reduce postharvest senescence of leafy vegetables (Able et al., 2003; Lomaniec et al., 2003) and increase shelf life of cut flowers and potted plants (In et al., 2015; Silva and Finger, 2015). In climacteric fruit, preharvest treatment with inhibitors of ethylene biosynthesis can delay the initiation of ripening to increase fruit quality and reduce early fruit abscission of for example apple (Hofman et al., 2001; Drake et al., 2006), pear (Villalobos-Acuna et al., 2010), avocado (Hofman et al., 2001; Salazar-Garcia et al., 2006), mandarin (Nawaz et al., 2008), papaya (Hofman et al., 2001), and mango (Hofman et al., 2001). Conversely, postharvest treatment can reduce loss of fruit quality due to over-ripening (Drake et al., 2006; Zhu et al., 2015). Preharvest treatment with inducers of ethylene synthesis reduces the harvest window and increases crop uniformity (Logendra et al., 2004; Ampa et al., 2016). However, because these inducers of ethylene synthesis are not tissue specific, they will promote senescence and abscission in other plant organs making them unsuitable for use on perennial plants which produce clusters of fruit sequentially over several months or years (Logendra et al., 2004).

The use of chemical blockers or inducers of ethylene biosynthesis is expensive, impractical in natural soil and their effect on human health and/or the environment is still ambiguous. The latter led to the banishment of some of these products from agricultural practices in several countries, and more may follow (Scariot et al., 2014; Tacuri et al., 2016). This means that there is a need for better alternatives. With the elucidation of the ACC transport mechanisms in plants, ethylene production could be fine-tuned with an improved specificity. The transport of ACC toward certain plant organs could be induced/blocked without affecting development of other organs. Possible applications can be envisaged for all processes that are known to be dependent on ACC transport. These include the above-mentioned conditions where chemical ethylene blockers/releasing agents are currently used, such as the regulation of the timing of fruit ripening and/or abscission, flower abscission and/or senescence, as well as leaf senescence. Furthermore, it will also be possible to synchronize harvest time in perennial crop species in which the use of chemical ethylene releasing compounds is not possible. Overall, it will allow these perennials to recuperate better and faster after each subsequent harvesting period. With the increasing interest for the use of these perennial crops, this application will have significant impact on the agricultural sector.

In addition to the above-mentioned fine-tuning of ACC transport as alternative for ethylene blocking/releasing compounds, it also has an application potential in mitigation of the effects of environmental stresses. Exploration of new options to reduce the vulnerability of our agricultural systems to these stresses becomes critical. As discussed previously,

conditions such as drought, flooding, heat, and salinity have been linked with long distance root to shoot transport of ACC and the subsequent elevation of ethylene production (Morgan and Drew, 1997; Schachtman and Goodger, 2008; Sasidharan and Voesenek, 2015; Tao et al., 2015). Consequently, under these stress conditions, plant growth and developmental processes are influenced severely, leading to significant yield losses, the burden of which can only increase as a result of global change. By blocking long distance ACC transport from the root to the shoot in response to environmental stress signals, ethylene-induced negative effects on crop yield could be minimized. Appropriate constructs with tissue/cell type specific promotors could be envisaged once specific transporters are identified. However, it has to be noticed equilibrated impact on ACC transport might be needed, in order to reach the optimal balance between the adaptive response of the plant (needing ethylene as a signal) and the rescue of its normal growth (impeded by ethylene). Another path to be explored relates to the beneficiary effect of PGPR (Barea et al., 2005), a subset of which can stimulate plant growth as a result of conversion of ACC exported by the plant toward the rhizosphere. PGPR colonize the roots of monocots and dicots, and influence many aspects of plant life including nutrition, growth, morphogenesis and response to biotic and abiotic stress by direct and indirect mechanisms. Hence, the identification of specific ACC exporters in the root epidermis could open several avenues for applications in the area of biostimulants.

To tune ACC transport in order to control a specific response, an approach based on site-specific genome editing would be most promising. Of all current site-specific genome editing technologies, CRISPR/Cas9 shows the greatest potential for crop improvement (Belhaj et al., 2015). Cas9 is an RNA-guided DNA endonuclease which can be targeted to specific genomic sequences by expression of a customized guide RNA with which it is capable of forming a complex. If the double-strand breaks (DSBs) resulting from the action of Cas9 are imperfectly repaired by the plant's endogenous non-homologous end joining (NHEJ) repair mechanism, gene function can be disrupted resulting in knock-down or knock-out lines. The technique has already successfully been applied in a variety of plant species, with efficiency rates higher than 90% in Arabidopsis (Feng et al., 2014) and rice (Miao et al., 2013). In addition, this tool allows multiplex editing, meaning that multiple sites can be targeted at the same time (Li et al., 2013). A second possible approach for the regulation of ACC transport is the use of tools altering the epigenetic regulation, i.e., without altering the primary DNA sequence (incl. DNA methylation, histone modification) of ACC transporters. Two examples of such epigenetics tools are RNA-directed DNA methylation (RdDM) and RNA interference (RNAi). RdDM induces transcriptional gene silencing by methylation of promoter sequences (Chinnusamy and Zhu, 2009), thus allowing epigenetic modification of gene expression in crop species (Hartung and Schiemann, 2014). In contrast to the genome editing technique CRISPR/Cas9, however, these epigenetic changes are only stable for a few generations (Hartung and Schiemann, 2014). RNAi is a commonly used technique allowing the transcriptional and post-transcriptional silencing of a gene of interest by small interfering RNAs (siRNAs) (Small, 2007). These siRNAs are capable of binding with a complementary sequence in a messenger RNA (mRNA) molecule, inducing its subsequent cleavage. In addition, siRNAs have been linked with changes in heterochromatin formation inducing transcriptional gene silencing (Djupedal and Ekwall, 2009).

# CONCLUSIONS AND PERSPECTIVES

Since the discovery of ethylene as a plant hormone and ACC as its direct precursor, accumulated evidence supports a central role for ACC in many aspects of ethylene synthesis, but also beyond ethylene biology. In order to advance our understanding of ACC as a pivotal molecule in plant growth and development, a few key issues need to be addressed. First, a better understanding of the transcriptional or post-translational control of ACO is necessary. Foremost, a profound understanding of ACC transport, including identification of genes encoding different types of transporters, together with a detailed organ-specific metabolomics study of ACC and its conjugates, would help to draw a map of ACC traffic within the plant body. As for ACC itself, besides the mechanisms of its transport, its possible role as a signaling molecule is an additional challenging route with a lot of potential.

For the identification of novel ACC transporters we suggest three potential approaches. The first focusses on a gene expression study assuming that ACC transporters show high expression at sites of differential ACS and ACO expression, the second concentrates on analysis of the affinities of known amino acid transporters, while the third aims at the identification of yet uncharacterized transporters using a heterologous expression system or a forward genetic screen. With the information gained from detailed ACC transporter analyses, agricultural applications can be explored, including the regulation of ACC transport as an alternative for ethylene releasing/blocking chemicals, or as a way to alleviate the increasing negative effects of environmental stresses on crop yield. Fine tuning of ACC transport could be obtained by using novel techniques for gene modification. These include CRISPR/Cas9 genome editing technology for the site specific mutagenesis of a gene of interest and epigenetic tools altering the transcriptional and post-transcriptional regulation of a gene of interest.

# AUTHOR CONTRIBUTIONS

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

# FUNDING

DS acknowledges the Research Foundation Flanders (FWO) for financial support (G065613N and G032717N). LV acknowledges the Research Foundation Flanders (FWO) for a Ph.D. fellowship strategic basic research (1S17917N).

# REFERENCES



Biochem. Biophys. Res. Commun. 104, 765–770. doi: 10.1016/0006-291X(82) 90703-3





**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 Vanderstraeten 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) 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.

# N-Terminus-Mediated Degradation of ACS7 Is Negatively Regulated by Senescence Signaling to Allow Optimal Ethylene Production during Leaf Development in Arabidopsis

#### Edited by:

Chi-Kuang Wen, Institute of Plant Physiology and Ecology, Shanghai Institutes for Biological Sciences (CAS), China

#### Reviewed by:

Long-Chi Wang, National Chung Hsing University, Taiwan Jos Schippers, RWTH Aachen University, Germany

> \*Correspondence: Ning Ning Wang wangnn@nankai.edu.cn

#### †Present address:

Li Xiong, College of Life Sciences, Sichuan University, Chengdu, China

‡These authors have contributed equally to this work.

#### Specialty section:

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

Received: 30 June 2017 Accepted: 17 November 2017 Published: 06 December 2017

#### Citation:

Sun G, Mei Y, Deng D, Xiong L, Sun L, Zhang X, Wen Z, Liu S, You X, Nasrullah, Wang D and Wang NN (2017) N-Terminus-Mediated Degradation of ACS7 Is Negatively Regulated by Senescence Signaling to Allow Optimal Ethylene Production during Leaf Development in Arabidopsis. Front. Plant Sci. 8:2066. doi: 10.3389/fpls.2017.02066 Gongling Sun‡ , Yuanyuan Mei‡ , Dewen Deng, Li Xiong† , Lifang Sun, Xiyu Zhang, Zewen Wen, Sheng Liu, Xiang You, Nasrullah, Dan Wang and Ning Ning Wang\*

Department of Plant Biology and Ecology, College of Life Sciences, Nankai University, Tianjin, China

Senescence is the final phase of leaf development, characterized by key processes by which resources trapped in deteriorating leaves are degraded and recycled to sustain the growth of newly formed organs. As the gaseous hormone ethylene exerts a profound effect on the progression of leaf senescence, both the optimal timing and amount of its biosynthesis are essential for controlled leaf development. The rate-limiting enzyme that controls ethylene synthesis in higher plants is ACC synthase (ACS). In this study, we evaluated the production of ethylene and revealed an up-regulation of ACS7 during leaf senescence in Arabidopsis. We further showed that the promoter activity of ACS7 was maintained at a relatively high level throughout the whole rosette development process. However, the accumulation level of ACS7 protein was extremely low in the light-grown young seedlings, and it was gradually restored as plants aging. We previously demonstrated that degradation of ACS7 is regulated by its first 14 N-terminal residues, here we compared the phenotypes of transgenic Arabidopsis overexpressing a truncated ACS7 lacking the 14 residues with transgenic plants overexpressing the full-length protein. Results showed that seedlings overexpressing the truncated ACS7 exhibited a senescence phenotype much earlier than their counterparts overexpressing the full-length gene. Fusion of the 14 residues to SSPP, a PP2C-type senescencesuppressed protein phosphatase, effectively rescued the SSPP-induced suppression of rosette growth and development but had no effect on the delayed senescence. This observation further supported that N-terminus-mediated degradation of ACS7 is negatively regulated by leaf senescence signaling. All results of this study therefore suggest that ACS7 is one of the major contributors to the synthesis of 'senescence ethylene'. And more importantly, the N-terminal 14 residue-mediated degradation of this protein is highly regulated by senescence signaling to enable plants to produce the appropriate levels of ethylene required.

Keywords: leaf senescence, ethylene, ACS7, signal transduction, post-transcriptional regulation

# INTRODUCTION

fpls-08-02066 December 4, 2017 Time: 15:0 # 2

Senescence is the last phase of leaf development characterized by the mobilization and recycling of a great majority of nutrients to newly formed organs (Lohman et al., 1994; Himelblau and Amasino, 2001). This process is highly regulated and must be appropriately integrated into whole-plant developmental processes. For this reason, senescence mutants with precocious or delayed senescence are often characterized by abnormal growth and seed development (Uauy et al., 2006; Xu et al., 2011; Wu et al., 2012). Although the optimal timing of initiation and progression of leaf senescence are prerequisites for controlled plant growth and development, the mechanisms underlying these processes remain largely unknown.

The gaseous phytohormone ethylene is a widely acknowledged positive regulator of leaf senescence. The progression of this developmental process can therefore be significantly altered by manipulating biosynthesis or signal transduction of ethylene. Both the exogenous ethylene treatment (Zacarias and Reid, 1990) and over-expression of genes encoding key components like EIN3 in ethylene signaling pathway (Li et al., 2013) induce early senescence in Arabidopsis. When ethylene biosynthesis or signaling is blocked, for example via antisense suppression of the ACC oxidase in tomato or mutations of the ethylene receptor ETR1 in Arabidopsis, delayed leaf senescence can be induced (John et al., 1995; Resnick et al., 2006). We have previously shown that the leucine-rich repeat-receptor-like kinase AtSARK acts as a positive regulator of leaf senescence in Arabidopsis. Overexpression of this kinase led to significant induction of both ethylene biosynthesis and subsequent responses. At the same time, inhibition of ethylene biosynthesis alleviated the senescence induced by AtSARK while mutation of the key ethylene signaling component EIN2 completely reversed AtSARK-mediated senescence (Xu et al., 2011). We have also shown that overexpression of the PP2C protein phosphataseencoding gene SSPP, an interactor and negative regulator of AtSARK in Arabidopsis, resulted in a significant suppression of ethylene biosynthesis and responses, while senescence was greatly delayed in the SSPP-overexpressing transgenic plants (Xiao et al., 2015). These observations suggest that both optimal timings and levels of ethylene production are required to control leaf development.

The ethylene biosynthetic pathway in higher plants has been well-documented (Yang and Hoffman, 1984); the rate-limiting step is known to be the conversion of S-adenosylmethionine (SAM) to 1-aminocyclopropane-1-carboxylic acid (ACC) catalyzed by ACC synthase (ACS). Regulation of this enzyme is therefore essential for controlling the rate and level of ethylene production. In most higher plants, ACS is encoded by a multigene family. Based on their C-terminal sequences, ACS proteins can be divided into three main groups. The type 3 ACSs possess a very short C-terminal extension that lacks the conserved sites for CDPK or MAPK6 phosphorylation found in types 1 and 2 ACS proteins. There are nine ACS genes in Arabidopsis of which ACS7 is the only type 3 ACS (Yamagami et al., 2003).

To fulfill the requirement for stringent regulation of ethylene biosynthesis, the expression of ACS members is highly synchronized at both transcriptional and post-transcriptional levels. At the transcriptional level, distinct ACS subsets are expressed in response to different developmental, environmental and hormonal signals (Tsuchisaka and Theologis, 2004; Wang et al., 2005). Ethylene biosynthesis is also tightly controlled by both autocatalytic and autoinhibitive regulations (Kende, 1993). We have previously shown that the transcription of ACS7 is significantly enhanced via exogenous ethylene application and can be markedly suppressed in the ethylene receptor mutant etr1-1 (Wang et al., 2005; Tang et al., 2008).

At the post-transcriptional level, stability of type 1 ACS isoforms can be enhanced via phosphorylation by either MAPKs or CPKs (Liu and Zhang, 2004; Han et al., 2010; Luo et al., 2014) while negatively affected via dephosphorylations by protein phosphatases 2A and 2C, respectively (Skottke et al., 2011; Ludwików et al., 2014). Type 2 ACS proteins have a unique C-terminal cis-acting sequence called TOE domain that is recognized specifically by ETO1/EOL1/EOL2 proteins in ubiquitin-dependent proteolysis (Yoshida et al., 2006; Prasad et al., 2010). It is also possible that ACS proteins of this type can be phosphorylated on the CDPK motif although the detailed mechanisms of this process remain unclear (Hernández Sebastià et al., 2004). At the same time, ACS7, the only type 3 ACS in Arabidopsis that lacks C-terminal regulation sites, can also be degraded via the ubiquitin-26S proteasome pathway that requires XBAT32 E3 ligases (Lyzenga et al., 2012). We have recently demonstrated that the 1–14 residues on its N-terminus are involved in the post-translational regulation of ACS7 (Xiong et al., 2014). In addition, ACS proteins are also known to form homo- and hetero-dimers, 25 of which exhibit catalytic activity and render the post-transcriptional regulation of this gene family even more sophisticated (Tsuchisaka and Theologis, 2004). Taken together, all these regulatory mechanisms enable the optimal ethylene production required by different tissues, developmental stages and environmental conditions.

Great achievements so far have been made on elucidating the mechanisms underlying transcriptional and post-transcriptional regulations of ACS genes. However, few studies aimed to determine the specific members of this gene family that contribute significantly to the ethylene biosynthesis during leaf senescence, let alone to unravel the underlying mechanisms on how the expressions of ACS members are coordinated at both transcriptional and post-transcriptional levels to meet the requirement of plant development and senescence progression. Here we demonstrated that ACS7 is a major contributor to ethylene production during leaf senescence in Arabidopsis. And the results of this study also show that the N-terminus-mediated ACS7 degradation is highly regulated by senescence signals to enable optimal ethylene production at the appropriate time during leaf development.

#### MATERIALS AND METHODS

#### Plant Materials and Growth Conditions

Arabidopsis thaliana ecotype Columbia-0 was used throughout the study apart from the mutant assay where Arabidopsis

ecotype Wassilewskija-4 was used for both the wild-type and the T-DNA insertion line of ACS7. The sources of the two materials in the latter case were described in Dong et al. (2011). The T-DNA insertion line of ACS7 was further confirmed to have a single insertion site by southern blot analysis (**Supplementary Figure S1**) in this study following standard protocols (Sambrook and Russell, 2001). The generation of transgenic Arabidopsis GVG:ACS7∆1−<sup>14</sup> -Flag, GVG:ACS7- Flag, ACS7:GUS and ACS7:N7(1−54) -GUS were described in Xiong et al. (2014). In order to generate 35S:ACS7-eGFP and 35S:ACS7∆1−<sup>14</sup> -eGFP, ACS7 and ACS7∆1−<sup>14</sup> were inserted into the binary vector pCABMBIA1301 and fused with a GFP gene to create the recombinant transcription units, ACS7-eGFP and ACS7∆1−<sup>14</sup> -eGFP, respectively. Similarly, in order to generate 35S:N7(1−14) -SSPP-HA, a fusion construct was obtained via PCR by initially using the 35S-N7-SSPP-F2 and 35S-N7-SSPP-R2 primer pair and the SSPP gene (Xiao et al., 2015) as template. The diluted PCR product was then used as a template while 35S-N7-SSPP-F1 and 35S-N7-SSPP-R1 were used as the second primer pair to accomplish construction. Constructions of this fusion gene have been patented by our group in the State Intellectual Property Office of the People's Republic of China with number 201510050322.1 and we have also applied for the international patent with application number PCT/CN2016/070504. The fusion gene was subsequently transferred into the pBI121 expression vector. The construction of SSPP-HA was generated on the basis of N 7(1−14) -SSPP-HA by removing the N 7(1−14) part through double digestion. All the primer sequences used in this study are listed in Supplementary Table S1. All recombinant constructs were transferred into Arabidopsis by Agrobacteria-mediated floral dip method (Clough and Bent, 1998). Recombinant constructs in transgenic plants were confirmed by PCR genotyping and homozygous plants were used for experiments.

Seeds were surface sterilized in 10% (v/v) sodium hypochlorite (Tianjin Chemicals, 559) for 2 min, washed at least 10 times with sterilized water and geminated on one-half-strength Murashige Skoog (MS) medium (Duchefa Biochemie, M0222) containing 0.8% (w/v) agar (Solarbio, A8190), pH5.7, 1% (w/v) Suc (Jiangtian Chemicals, 11411), supplemented with or without antibiotics, stratified at 4◦C for 2 days in the dark, and then grown in plant growth chamber at 22/19◦C with cycles of 16 h light and 8 h darkness under 100 to 150 µmol m−<sup>2</sup> s −1 light intensity. The 10-day-old seedlings were then transferred to soil and grown under the same conditions for further experiments. Where dexamethasone (DEX, Sigma, D1756) spray was applied, the compound was dissolved in ethanol and the final concentration was at 30 µM.

### Measurements of Ethylene Emission and Chlorophyll Content

Ethylene emission of the 41-day-old wild-type Arabidopsis was determined in 10 different leaves of one rosette by gas chromatography (Agilent 7890A) as described in Li and Mattoo (1994). Approximately two leaves of each were separately collected depending on their sizes and vial spacing. Detached leaves were subsequently incubated in vials under light before analysis. The chlorophyll content in mesophyll cells was spectrophotometrically measured as described in Arnon (1949). At least six independent biological replicates were carried out.

### Histochemical GUS Staining

GUS (β-glucuronidase) histochemical staining of transgenic Arabidopsis was carried out as described previously in Wang et al. (2005). The images of blue-colored whole plants were recorded with a scanner (Epson V30). Three to six biological replicates for each transgenic line were performed in this study. In total five and seven independent transgenic lines were generated for ACS7:GUS and ACS7:N7(1−54) -GUS, respectively, and data shown were the typical results of lines with approximately same levels of GUS gene expression.

# RNA Extraction and RT-PCR Analysis of Gene Expression

RNA extraction, cDNA synthesis and RT-PCR analysis were performed as described previously in Liu et al. (2010). The real-time RT-PCR was done using SYBR Green Perfect Mix (Tiangen, FP205-02) on iQ5 (Bio-Rad) machine following the manufacturer's instructions. All reactions were performed under the following conditions: 95◦C for 2 min, 40 cycles of 95◦C for 10 s, and 56◦C or 58◦C in case of SSPP amplification for 30 s. TIP41-like gene was used as an internal control. The calculation of relative gene expression levels was as described in Xiong et al. (2014). At least three independent replicates were performed to give typical results shown here. All primers used in RT-PCR analysis were listed in Supplementary Table S1.

## Protein Extraction and Immunoblot Analysis

Seedling or leaf samples were ground in liquid nitrogen and total soluble protein was extracted as described in Wang et al. (2005). The concentration of protein extract was measured using the Bradford reagent (USB, DH078) and total protein was separated by 12% SDS-PAGE gel and detected by antibodies after being transferred to a polyvinylidene fluoride (PVDF) membrane (GE Healthcare, 10600023). The antibodies used in this study include rabbit anti-GFP (Abcam, ab290), mouse anti-Flag (Sigma, F3165), rabbit anti-HA (Abcam, ab9110), mouse anti-actin (Abmart, 26F7), mouse anti-Tubulin (Abmart, 10B1), goat anti-rabbit IgG H&L (HRP) (Abcam, ab6721) and rabbit anti-mouse IgG H&L (HRP) (Abcam, ab6728). Protein bands were visualized using an ECL western blotting reagent pack (GE Healthcare, RPN2232) flowing the manufacturer's instructions. Images were recorded by a chemiluminescence imaging system (CELVIN's, Biostep). Quantification of the signals was performed in Image J software as described in Xiong et al. (2014).

# RESULTS

### The Peak of Ethylene Emission Occurs after the Initiation of Leaf Senescence

To evaluate the spatiotemporal distribution of ethylene biosynthesis and the correlations between ethylene production and leaf senescence, we measured changes in ethylene emissions and chlorophyll contents in the developing leaves of 41-day-old Arabidopsis (**Figure 1**). The results showed that the content of chlorophyll in mesophyll cells was gradually decreased during leaf senescence and the most significant decline was between leaves ten and nine. At the same time, however, the level of ethylene production was gradually increased from the mature 10th leaf, reaching a peak in the seventh. In the older leaves, from the sixth to the fourth, both the chlorophyll content and ethylene emission were continuously dropped (**Figure 1**). Indeed, both chlorophyll content and ethylene production were hardly measurable in the much more senescent leaves, including the third, second and first (data not shown).

#### The Expression of ACS7 Is Up-regulated in Both Natural and Dark-Induced Leaf Senescence Processes

The quantitative RT-PCR assay was performed to determine the involvement of ACS7 in Arabidopsis leaf senescence. Results showed that during natural senescence, the transcription levels of ACS7 remained relatively low in young leaves but increased gradually as they developed to late senescence stages (**Figure 2A**). It was also shown that 4-day dark treatment resulted in a significant decrease in the chlorophyll content in the wild-type Arabidopsis plants at stage 5.10 (Boyes et al., 2001). In the meantime, however, an obvious increase in the level of ACS7 transcript was revealed during the dark-induced senescence process (**Figure 2B**).

and quantitative RT-PCR was performed using TIP41-like as internal control. The experiments were repeated three times with similar results. Error bars represent

# High-Level Accumulation of ACS7 Protein Promotes Leaf Senescence in Arabidopsis

standard deviations (SD).

As we have previously reported, the N-terminal 1–14 residues of ACS7 negatively regulate the protein stability. Consistently, the truncated form of ACS7 lacking the 14 residues is more stable than the full-length form when transgenically expressed in the light-grown Arabidopsis seedlings (Xiong et al., 2014). We therefore used previously generated transgenic Arabidopsis plants expressing Flag-tagged full-length ACS7 as well as truncated ACS7 with 1–14 residues deleted under the control of a DEX-inducible promoter, namely GVG:ACS7-Flag and GVG:ACS7∆1−<sup>14</sup> -Flag (Xiong et al., 2014) in this study to examine the detailed effects of ACS7 on leaf senescence. Quantitative RT-PCR revealed comparable transcript levels of ACS7 in these two transgenic materials (**Supplementary Figure S2**). It was shown that the GVG:ACS7-Flag transgenic Arabidopsis exhibited normal growth and development compared to the GVG:GUS control plants after 4 days following DEX spray. In contrast, the GVG:ACS7∆1−<sup>14</sup> -Flag transgenic plants displayed prominent leaf yellowing and had significantly smaller rosettes (**Figure 3A**), as initially noted by Xiong et al. (2014). We further sampled the fifth and sixth leaves of GVG:GUS, GVG:ACS7-Flag and GVG:ACS7∆1−<sup>14</sup> -Flag after DEX treatment and performed quantitative RT-PCR to determine the transcription levels of several known senescenceassociated marker genes including two transcriptional factors NAP1 (Guo and Gan, 2006) and WRKY6 (Robatzek and Somssich, 2002), as well as one PP2C family protein phosphatase gene SAG113 (Zhang et al., 2012). No significant increases in the transcription levels of these marker genes were found in the DEX-treated GVG:GUS and GVG:ACS7-Flag plants while the expressions of these senescence-associated markers in the DEX-treated Arabidopsis expressing ACS7∆1−<sup>14</sup> -Flag were greatly up-regulated (**Figure 3B**). Moreover, western blot assay using anti-Flag antibody revealed a much higher level of ACS7 protein accumulation in the fifth leaves of DEX-treated GVG:ACS7∆1−<sup>14</sup> -Flag plants than in their GVG:ACS7-Flag counterparts (**Figure 3C**, for original images of the blots please see **Supplementary Figure S6**). These results confirmed that the induced high-level accumulation of ACS7 protein led to precocious leaf senescence in Arabidopsis.

### The acs7-1 Mutant Exhibits Delayed Natural Senescence

In order to further confirm the function of ACS7 in leaf senescence, we examined the phenotype of a loss-of-function mutant of ACS7, acs7-1 (Tsuchisaka et al., 2009; Dong et al., 2011). As was shown in **Figures 4A,B**, the acs7-1 mutant exhibited delayed leaf senescence compared to its wildtype control (Wassilewskija-4). In line with the phenotypic observations, the chlorophyll content in the acs7-1 mesophyll cells was significantly higher than in the wild-type control when measured at 36 days after emergence (DAE) (**Figure 4C**). The acs7-1 mutants exhibited larger rosettes, faster growth, and reached greater heights (**Supplementary Figure S3**). We also

observed an early flowering phenotype within these mutant plants, consistent with a previous report (Tsuchisaka et al., 2009), as well as a shorter blooming period in comparison to their wild-type control (**Supplementary Figure S3**).

#### The Accumulation of ACS7 Protein Is Extremely Low in Light-Grown Young Seedlings but Is Gradually Restored As Plants Aging

To further explore the underlying mechanisms of ACS7 regulated leaf senescence, we generated transgenic Arabidopsis expressing the GUS gene fused to the N-terminal 1–54 residues of ACS7, namely N 7(1−54) -GUS, or a GUS gene alone driven by ACS7 native promoter. Two lines that exhibited no significant differences in levels of transcriptional expression of the GUS genes were selected for further study (**Supplementary Figure S4**) and photographed at six different developmental stages including stage 0.5, stage 0.7, stage 1.02, stage 1.04, stage 5.10, and stage 6.00 as previously described by Boyes et al. (2001) (**Figure 5**, upper row). In view of the fact that N-terminal 1–14 residues play an important role in the post-translational regulation of ACS7, the activities of N7(1−54) -GUS are thus indicative of changes in the accumulation levels of internal ACS7 protein during leaf development (Xiong et al., 2014). The histochemical GUS staining assay revealed that ACS7 promoter activity and the accumulation of ACS7 protein were both at very high levels in the late stages of seed germination. With the growth of seedlings, the promoter activity of ACS7 was maintained at a very high level while its protein accumulation varied from stage to stage (**Figure 5**). For instance, the blue coloring of ACS7:N7(1−54) - GUS was significantly decreased at vegetative stages such as stages 1.02 and 1.04 but was gradually restored at stages 5.10 and 6.00 along with the initiation of leaf senescence. A closer look revealed that the blue colors in ACS7:N7(1−54) -GUS at the last two stages were generally observed in mature and senescent leaves while almost no traces were seen in young leaves (**Figure 5**, middle row). All results presented here suggested that the N-terminal mediated ACS7 degradation was restricted during the process of seed germination, and that the restriction was released at vegetative stages but was tighten again with the progressing of leaf senescence.

### N-Terminus-Mediated ACS7 Protein Degradation Is Negatively Regulated by Senescence Signals

In order to confirm the role of senescence signals in the regulation of ACS7 stability, we generated transgenic Arabidopsis expressing GFP-tagged full-length ACS7 (ACS7-eGFP) or truncated ACS7 lacking N-terminal 1–14 residues (ACS7∆1−<sup>14</sup> -eGFP) under the control of 35S promoter.

To begin with, we evaluated expression levels of the ACS7 transgenes using quantitative RT-PCR. The results showed that the transcript levels of ACS7 in line 7 of 35S:ACS7-eGFP and in line 16 of 35S:ACS7∆1−<sup>14</sup> -eGFP were about 500 and 400 times

FIGURE 4 | The acs7-1 mutant showed delayed leaf senescence. (A) The acs7-1 mutant plants and their wild-type control (Wassilewskija-4) were photographed at 44 DAE stage. (B) Leaves from acs7-1 and WT plants at 39 DAE stage were laid out in order of emergence. (C) The chlorophyll contents in the fifth and sixth leaves of both acs7-1 and WT plants were measured at 36 DAE stage. Data represent means ± SD. The experiments were repeated at least three times to give the typical results shown here. Asterisk indicates statistically significant differences in student's t-test (α = 0.01).

higher, respectively, than that of the internal control TIP41-like. In line 3 of 35S:ACS7-eGFP and line 12 of 35S:ACS7∆1−<sup>14</sup> -eGFP, the expressions of the transgenes were at approximately the same but relatively much lower levels (**Figure 6A**).

The phenotypic analysis of etiolated seedlings showed that all of the four lines mentioned above exhibited prominent triple responses, implying that expression of ACS7∆1−<sup>14</sup> -eGFP and ACS7-eGFP both led to over-production of ethylene at this

FIGURE 6 | Effects of over-expression of the full-length (35S:ACS7-eGFP) or truncated ACS7 (35S:ACS7∆1−<sup>14</sup> -eGFP) on the development of Arabidopsis seedlings. (A) Quantitative RT-PCR analysis was performed to determine the expression levels of ACS7 in line 3 and line 7 of 35S:ACS7-eGFP as well as in line 12 and line 16 of 35S:ACS7∆1−<sup>14</sup> -eGFP transgenic Arabidopsis. RNA was extracted from 7-day-old seedlings and RT-PCR was performed using TIP41-like as an internal control. Data presented by means ± SD were the typical results from three independent biological replicates. (B) Phenotypic presentations and ACS7 protein accumulations in etiolated and light-grown seedlings of 35S:ACS7-eGFP and 35S:ACS7∆1−<sup>14</sup> -eGFP transgenic lines. The accumulations of ACS7-eGFP and ACS711−14-eGFP proteins were detected using anti-GFP antibody and the detection of anti-actin served as loading control. Numbers below the blots indicate the intensity ratio of the ACS7 band to the β-actin control band in each lane. (C) Young seedlings of the line 3 of 35S:ACS7-eGFP and line 12 of 35S:ACS7∆1−<sup>14</sup> -eGFP as well as their wild-type control at the same developmental stage were transferred from germination plates into soil at 7 DAE and photographed at 17, 29, and 40 DAE, respectively. (D) ACS7 protein accumulations in the fifth and sixth leaves of line 3 of 35S:ACS7-eGFP and line 12 of 35S:ACS7∆1−<sup>14</sup> -eGFP transgenic lines were determined at 17, 29, and 40 DAE. Total proteins loaded were 40, 40, and 20 µg for WT, line 3 of 35S:ACS7-eGFP and line 12 of 35S:ACS7∆1−<sup>14</sup> -eGFP, respectively. ACS7-eGFP and ACS711−14-eGFP proteins were detected using anti-GFP antibody and the detection of anti-actin served as loading control. Numbers below the blots indicate the intensity ratio of the ACS7 band to the β-actin control band in each lane. In total six and eight independent transgenic lines were obtained for 35S:ACS7-eGFP and 35S:ACS7∆1−<sup>14</sup> -eGFP, respectively, and data shown here were representatives of the typical results from at least three independent biological replicates.

developmental stage (**Figure 6B**, upper panel). Immunoblot analysis also revealed approximately equally high levels of ACS7 accumulation in the transgenic etiolated seedlings of line 3 of 35S:ACS7-eGFP and line 12 of 35S:ACS7∆1−<sup>14</sup> -eGFP, however, the light-grown line 3 seedlings of 35S:ACS7-eGFP exhibited a much lower accumulation level of ACS7 protein in comparison to line 12 of 35S:ACS7∆1−<sup>14</sup> -eGFP (**Figure 6B**, for original images of the blots please see **Supplementary Figure S6**), even though the transcripts of ACS7 in both lines were at similar levels (**Figure 6A**). In addition, line 7 of 35S:ACS7 eGFP, which had extremely high levels of ACS7 transcript and protein accumulation in etiolated seedlings (**Figure 6B**, upper panel), also displayed a remarkable decrease in the accumulation level of ACS7 protein in light-grown seedlings compared to the line 16 of 35S:ACS7∆1−<sup>14</sup> -eGFP (**Figure 6B**, lower panel, for original images of the blots please see **Supplementary Figure S6**). These results implied that the accumulation of fulllength ACS7 was inhibited and further supported our previous findings that the truncated ACS7 protein lacking the N-terminal 1–14 residues was more stable than the full-length one in the vigorously growing seedlings under normal condition (Xiong et al., 2014).

To further evaluate the phenotypic differences in leaf senescence, line 12 of 35S:ACS7∆1−<sup>14</sup> -eGFP and line 3 of 35S:ACS7-eGFP that had approximately same levels of ACS7 transcript as well as wild type Arabidopsis were grown under light-grown conditions and photographed at three different stages including 17, 29, and 40 DAE (**Figure 6C**). The results revealed no significant differences between 35S:ACS7-eGFP and the wild-type control until 40 DAE when the transgenic

plants exhibited slightly earlier senescence and smaller rosette, indicating a restoration of ACS7 protein level in late stages. In contrast, transgenic plants overexpressing the truncated ACS7 with the N-terminal 14 residues deleted exhibited a number of significant ethylene response phenotypes including smaller rosette since 17 DAE and much earlier senescence at 40 DAE. In line with this, the accumulation of ACS7 protein in the fifth and sixth leaves of line 3 of 35S:ACS7-eGFP was at very low level when the plant was at vegetative stages such as 9 and 17 DAE, however, it was increased when the plant approached its firstflower opening stage at 29 DAE and reached a much higher level at 40 DAE as plant aging. By contrast, the ACS7 proteins in the fifth and sixth leaves of line 12 of 35S:ACS7∆1−<sup>14</sup> -eGFP were maintained at a high level throughout all these time points and did not show any significant response to senescence signaling (**Figures 6B,D**). All these results demonstrated again that the N-terminus mediated ACS7 degradation was negatively regulated by senescence signals.

#### Fusing ACS7 N-Terminal 14 Residues to SSPP Effectively Rescues the SSPP-Induced Growth Suppression but Imposes No Effects on the Delayed Senescence

In order to further confirm the role of senescence signaling in the regulation of N-terminus mediated ACS7 protein degradation, we generated transgenic Arabidopsis plants over-expressing HAtagged SSPP, a negative regulator of Arabidopsis leaf senescence, or HA-tagged SSPP fused with the N-terminal 1–14 residues of ACS7 at its N-terminus under the control of 35S promoters [namely 35S:SSPP-HA and 35S:N7(1−14) -SSPP-HA, respectively]. Two transgenic lines that had approximately the same transcript levels of SSPP were used in this study (**Supplementary Figure S5**) and photographed at 10, 21, 48, and 61 DAE, respectively. In line with our previous findings from the 35S:SSPP plants (Xiao et al., 2015), the 35S:SSPP-HA transgenic lines displayed a remarkable decrease in rosette size. However, all the 35S:N7(1−14) -SSPP-HA transgenic plants developed normally before the onset of leaf senescence, regardless of the similar transcript levels of SSPP (**Figure 7A**). And more interestingly, after the initiation of leaf senescence, the 35S:N7(1−14) -SSPP-HA transgenic plants also exhibited significantly delayed leaf senescence in comparison to the wild-type control at 48 DAE and much later stages such as 61 DAE (**Figure 7A**). We further carried out western blot assay to evaluate SSPP accumulations in both 35S:N7(1−14) -SSPP-HA and 35S:SSPP-HA transgenic plants at different developmental stages. As expected, while the accumulation levels of SSPP in the 35S:SSPP-HA plants were constantly maintained at a relatively high and stable level from 10 to 48 DAE stages, the SSPP accumulation in the 35S:N7(1−14) -SSPP-HA seedlings was very low at 10 DAE, but it was increased at 21 DAE and reached a higher level at 48 DAE as plants aging (**Figure 7B**). These observations suggested that the N-terminus-mediated degradation of SSPP was also suppressed by senescence signaling. The senescence-delayed effect of N 7(1−14) -SSPP fusion gene was further confirmed by the quantitative analysis of several senescence-associated marker genes. As shown in **Figure 8**, the expressions of SAG12, WRKY6, NAP1, and NAC1 were significantly down-regulated in both 35S:N7(1−14) -SSPP-HA and 35S:SSPP-HA transgenic plants in comparison to WT at 32 DAE.

#### DISCUSSION

Both the onset and progression of leaf senescence are tightly regulated by genetic programming in order to ensure an

adequate supply of assimilative products required for plant growth and development and, at the same time, guarantee a continuous and effective export of nutrients to newly formed organs. As an important senescence-promoting phytohormone, the biosynthesis of ethylene is tightly regulated during leaf development. Jing et al. (2005) noted that the ability of ethylene to induce leaf senescence is dependent on age-related changes. Our results also show that ethylene production in the rosette leaves of Arabidopsis was mostly maintained at a relatively low level and the peak of ethylene emission occurred only after the initiation of leaf senescence (**Figure 1**). In addition, we observed continuous decreases in both ethylene and chlorophyll contents in much older leaves from the sixth to the fourth. This is in line with the previous idea that ethylene is needed for chlorophyll breakdown and nutrient remobilization at relatively earlier time points during senescence, but not for the cell death that occurs at the final stage of senescence, which is more under the control of salicylic acid (Morris et al., 2000; Breeze et al., 2011). Nevertheless, how plants coordinate their ethylene biosynthesis with leaf senescence remains an area for further work. Given the number of ACS genes present in Arabidopsis genome, it will also be essential to determine which members specifically play major roles in the biosynthesis of 'senescence ethylene' and elucidate the underlying mechanisms.

In higher plants, the expression and activities of ACS family members can be highly regulated at both transcriptional and post-transcriptional levels to enable the optimal levels of ethylene production required in different organs and tissues at distinctive developmental stages. Given the fact that the peak of ethylene production occurred in the later stage of leaf senescence (**Figure 1**) and the biosynthesis of ethylene can be both autocatalyzed and auto-inhibited (Wang et al., 2005), it can be speculated that the key ACS genes that play major roles in ethylene biosynthesis during leaf senescence, namely 'senescence ACSs,' should all share the following characteristics. In the first place, these genes should be up-regulated at both transcriptional and post-transcriptional levels by senescence signals and autocatalyzed by ethylene itself. Secondly, the regulations of these genes should correspond with the senescence progress to ensure that the level of ethylene production remains relatively low at vegetative stages but can be further enhanced by the positive feedback mechanism in later stages.

We previously found that the promoter activity of ACS7 could be positively regulated by ethylene (Wang et al., 2005). And in AtSARK-mediated leaf senescence, the expression levels of ACS members that formed active dimers with ACS7, including ACS4, ACS6, ACS9, and ACS7 itself, were all up-regulated, whereas those of the partners that formed inactive dimers such

as ACS2, ACS5, and ACS11, were down-regulated (Xu et al., 2011). All these implied that ACS7 may play important roles in the biosynthesis of 'senescence ethylene'. In the present study, we demonstrated that expression of ACS7 was up-regulated during both natural and dark-induced leaf senescence (**Figure 2**). Consistently, we further showed that high accumulation of ACS7 protein led to precocious leaf senescence as well as greatly upregulated expressions of several critical senescence-associated marker genes (**Figure 3**). At the same time, the T-DNA insertion mutant acs7-1 exhibited delayed leaf senescence (**Figure 4**). All of these results further confirmed our aforementioned assumption that ACS7 is one of the major contributors to the synthesis of 'senescence ethylene.' However, it should be noted that over-expression of full-length ACS7 did not induce as severely precocious leaf senescence as expected (**Figure 3**), suggesting post-transcriptional regulation plays more important roles in this process.

As noted above, ACS7 is the only type 3 ACS in Arabidopsis and the N-terminus-mediated ACS7 degradation has been suggested to be regulated by both developmental and environmental signals (Xiong et al., 2014). In the present study, we have investigated the effects of senescence signaling on the regulation of ACS7 stability. It was shown that despite a constantly high level of ACS7 promoter activity as indicated by GUS histochemical staining in the ACS7:GUS transgenic Arabidopsis, the accumulation of ACS7 protein, i.e., the GUS activities of N7(1−54) -GUS, was extremely low in light-grown young seedlings but was gradually restored as plants aging (**Figure 5**). In line with this, even though the transcript and protein accumulation levels of ACS7 in etiolated Arabidopsis over-expressing the full-length or truncated ACS7 were both remained at approximately the same levels, respectively, ACS7 protein accumulated at a significantly lower level in the young light-grown 35S:ACS7-eGFP seedlings compared to in their 35S:ACS7∆1−<sup>14</sup> -eGFP counterparts (**Figures 6A,B**). Similarly, the transgenic Arabidopsis expressing the truncated version of ACS7 exhibited much more severe precocious leaf senescence (**Figure 6C**). As our previous research has excluded the involvement of light/dark signal in the degradation of ACS7 protein (Xiong et al., 2014), we tempt to speculate that the N-terminus-mediated ACS7 degradation was negatively regulated by senescence signaling. The fact that the transgenic Arabidopsis expressing full-length ACS7 also exhibited earlier senescence than the wild-type control implied a recovery of the accumulation of ACS7 protein at the late developmental stages of rosette leaves. This was further corroborated by measurements of the accumulation levels of ACS7 protein in the 35S:ACS7 eGFP and 35S:ACS7∆1−<sup>14</sup> -eGFP plants during leaf development (**Figure 6D**).

We next fused the N-terminal 14 residues of ACS7 to SSPP, a negative senescence regulator previously identified in our research (Xiao et al., 2015). Compared to the SSPP over-expressing Arabidopsis, the N 7(1−14) -SSPP over-expressing seedlings exhibited normal growth rather than restricted rosette sizes at vegetative stages, suggesting that the N 7(1−14) fragment also exerted a negative effect on the accumulation of SSPP protein. However, after the initiation of leaf senescence, both the SSPP and N 7(1−14) -SSPP over-expressing plants exhibited significant delays in leaf senescence, as well as remarkable downregulations of several critical senescence-associated marker genes (**Figures 7**, **8**). This is consistent with the western blot results that SSPP accumulation was significantly increased in the senescing 35S:N7(1−14) -SSPP plants as shown in **Figure 7B**. All these observations strongly support the idea that the N-terminus-mediated protein degradation is tightly controlled by development and/or senescence signals, independently of ACS7.

Leaf senescence is a complex and highly regulated process that involves degradation of macromolecules such as nucleic acids, proteins, lipids and so on (Lim and Nam, 2007; Park et al., 2007; Morita et al., 2009). Modulation of the protein stabilities of key regulators is one important mechanism to ensure normal leaf senescence process. For instance, enhanced stabilities of thylakoid membrane proteins were observed in a wheat staygreen mutant tasg1 (Tian et al., 2013). In addition, the stability of transcription factor WRKY53 was reported to be regulated by E3 ubiquitin ligase UPL5 to ensure that senescence is executed in the correct time frame in Arabidopsis (Miao and Zentgraf, 2010). Here we show, for the first time, that the stability of an ACS family member, specifically ACS7, can be regulated by senescence signaling to allow optimal ethylene production during leaf development.

It is also noteworthy that ACS7 may not be the only senescence-associated member of this family present in Arabidopsis. As discussed above, several other ACSs have also been observed to be up-regulated in both SARK-induced and natural leaf senescence. However, whether, or not, these ACSs merely facilitate ACS7-mediated ethylene biosynthesis by forming active dimers with ACS7 or perform additional unique roles in leaf senescence will require further research. The functions of other ACS members that form inactive dimers with ACS7 and exhibit down-regulated expressions during leaf senescence also remain to be clarified. Previous work has shown that loss of ACS6 in maize, a possible type 3 ACS, resulted in a reduction of up to 90% of ethylene production and led to a substantial delay in leaf senescence. At the same time, loss of another member ACS2 resulted in an up to 45% reduction in foliar ethylene emissions and a moderate delay in leaf senescence (Young et al., 2004). It is clear that the mechanisms underlying post-transcriptional regulation vary among different ACS members (Lee et al., 2017), additional research on the roles of senescence signaling in the posttranscriptional regulation of other ACSs will provide more insights on the regulatory mechanisms underlying ACS turnover during leaf senescence.

#### AUTHOR CONTRIBUTIONS

NNW conceived and designed the study, supervised the experiments, and compiled and finalized the article. GS, DD, LX, LS, XZ, ZW, SL, XY, N, and DW performed the experiments. NNW and YM analyzed the data. YM drafted and wrote the manuscript. NNW drafted and revised the manuscript. All authors read and approved the final manuscript.

#### FUNDING

fpls-08-02066 December 4, 2017 Time: 15:0 # 12

The authors acknowledge the financial supports from the Major Technological Program on Cultivation of New Varieties of Genetically Modified Organisms (Grant No. 2016ZX08010002- 007), the National Natural Science Foundation of China (Grant Nos. 31570293 and 31370285), the 111 Project (Grant No. B08011) and the Tianjin Research Program of Applied Basic and Cutting-Edge Technologies (Grant No. 16JCQNJC09300).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2017.02066/ full#supplementary-material

FIGURE S1 | Southern blot analysis reveals a single-copy T-DNA insertion in the acs7-1 genome. (A) Schematic map of the binary vector pGKB5 used in generation of the acs7-1 T-DNA insertion mutant. GUS, beta-glucuronidase gene; nos ter, nopaline synthase terminator; ocs Ter, ocs terminator; nptII, kanamycin resistance gene; nor Pro, nopaline synthase promoter; 35S Pro, 35S promoter; Bar, phosphiothrici resistance gene; g7 Ter, g7 terminator. LB and RB, left border and right border. Positions of four restriction enzymes, EcoR V (V), EcoR I (I), Pst I (P), and BstP I (B), were indicated with dark short vertical lines. Position of the probe in the right border was labeled by dark line below the structure. The arrows indicate direction of the transcription. (B) Southern blot analysis of the acs7-1 and WT genomic DNA that were digested with the indicated four different restriction enzymes and hybridized with the right border probe. Primers used to amplify the probe sequence were listed in Supplementary Table S1. Red arrows pointed the hybridized bands in each case in comparison to the wild-type control. The DIG-labeled DNA marker was on the leftmost lane (M).

FIGURE S2 | Determination of the transcription levels of ACS7 in the GVG:ACS7-Flag and GVG:ACS7∆1−<sup>14</sup> -Flag transgenic Arabidopsis. The 9-day-old transgenic seedlings of GVG:ACS7-Flag and GVG:ACS7∆1−<sup>14</sup> -Flag were treated with 10 µM DEX (+DEX) or its solvent, ethanol (mock) for 24 h, and then harvested for RNA extraction and subsequent quantitative RT-PCR analysis

# REFERENCES


of ACS7 transgene. TIP41-like was used as an internal control. All primers used were listed in Supplementary Table S1. Data presented were the typical results from three biological replicates with at least three technical repeats. Error bars represent SD.

FIGURE S3 | Effects of ACS7 loss-of-function mutation on rosette development (A), plant height (B), flowering time (C) and blooming period (D) of the Arabidopsis plants. (A) The rosette sizes of acs7-1 mutant and its wild-type control (WT) were recorded with time. (B) The heights of both acs7-1 and WT were measured at 44 DAE. (C) The flowering time was recorded for both acs7-1 and WT. (D) The blooming period was calculated for both acs7-1 and WT. Bars represent means ± SD. The experiments were repeated at least three times to give the typical results shown here. Asterisk indicates statistically significant differences in student's t-test (α = 0.05).

FIGURE S4 | Quantitative RT-PCR analysis of GUS expressions in the ACS7:GUS and ACS7:N7(1−54) -GUS transgenic Arabidopsis. The 9-day-old transgenic seedlings of ACS7:N7(1−54) -GUS and ACS7:GUS were harvested for RNA extraction and subsequent quantitative RT-PCR analysis of GUS. TIP41-like was used as an internal control. (A) The comparison on GUS expressions in the transgenic etiolated seedlings of ACS7:GUS and ACS7:N7(1−54) -GUS. (B) The comparison on GUS expressions in the transgenic light-grown seedlings of ACS7:GUS and ACS7:N7(1−54) -GUS. Three biological replicates with at least three technical repeats were performed. Error bars represent SE.

FIGURE S5 | Determination of the transcription levels of SSPP in the 35S:SSPP-HA and 35S:N7(1−14) -SSPP-HA transgenic Arabidopsis. The 9-day-old seedlings of 35S:SSPP-HA and 35S:N7(1−14) -SSPP-HA were harvested for RNA extraction and subsequent quantitative RT-PCR analysis of SSPP. TIP41-like was used as an internal control. Three biological replicates with at least three technical repeats were performed in each case. Error bars represent SE.

FIGURE S6 | The original images of the grouped blots in Figures 3, 6. (A) The original blot image shown in Figure 3C indicating accumulation levels of ACS7-Flag and ACS7∆1−14-Flag proteins in 25-day-old GVG:GUS, GVG:ACS7∆1−<sup>14</sup> -Flag and GVG:ACS7-Flag transgenic Arabidopsis plants sprayed with either 30 µM DEX (+DEX) or its solvent, ethanol (mock). (B) The original images of grouped anti-GFP and anti-actin blots shown in Figure 6B for the detection of ACS7 protein accumulations in etiolated seedlings and 9-day-old light-grown seedlings of both 35S:ACS7-eGFP and 35S:ACS7∆1−<sup>14</sup> -eGFP.


by directly repressing miR164 transcription in Arabidopsis. Plant Cell 25, 3311–3328. doi: 10.1105/tpc.113.113340


competence contribute to drought stress resistance in the tasg1 wheat staygreen mutant. J. Exp. Bot. 64, 1509–1520. doi: 10.1093/jxb/ert004


**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 Sun, Mei, Deng, Xiong, Sun, Zhang, Wen, Liu, You, Nasrullah, Wang and Wang. 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.

fpls-08-02066 December 4, 2017 Time: 15:0 # 13

# Possible Interactions between the Biosynthetic Pathways of Indole Glucosinolate and Auxin

Siva K. Malka<sup>1</sup> and Youfa Cheng1,2 \*

<sup>1</sup> Key Laboratory of Plant Molecular Physiology, Institute of Botany, Chinese Academy of Sciences, Beijing, China, <sup>2</sup> School of Life Sciences, University of Chinese Academy of Sciences, Beijing, China

Glucosinolates (GLS) are a group of plant secondary metabolites mainly found in Cruciferous plants, share a core structure consisting of a β-thioglucose moiety and a sulfonated oxime, but differ by a variable side chain derived from one of the several amino acids. These compounds are hydrolyzed upon cell damage by thioglucosidase (myrosinase), and the resulting degradation products are toxic to many pathogens and herbivores. Human beings use these compounds as flavor compounds, anti-carcinogens, and bio-pesticides. GLS metabolism is complexly linked to auxin homeostasis. Indole GLS contributes to auxin biosynthesis via metabolic intermediates indole-3-acetaldoxime (IAOx) and indole-3-acetonitrile (IAN). IAOx is proposed to be a metabolic branch point for biosynthesis of indole GLS, IAA, and camalexin. Interruption of metabolic channeling of IAOx into indole GLS leads to high-auxin production in GLS mutants. IAN is also produced as a hydrolyzed product of indole GLS and metabolized to IAA by nitrilases. In this review, we will discuss current knowledge on involvement of GLS in auxin homeostasis.

Keywords: glucosinolate, auxin, metabolism, development, Arabidopsis

# INTRODUCTION

Glucosinolates (GLS) are a group of secondary metabolites found almost exclusively in Brassicaceae (Agerbirk and Olsen, 2012). GLS are nitrogen and sulfur rich compounds, forming a twocomponent defense system (mustard oil bomb) with myrosinases against herbivores and microorganisms (Bones and Rossiter, 1996; Rask et al., 2000). Upon tissue damage by herbivores, the mustard oil bomb comes into action, where myrosinases hydrolyze GLS into different degradation products that are toxic to the enemy (Rask et al., 2000; Chen and Andreasson, 2001). Apart from plant defense, some of the GLS degradation products are physiologically significant in plant nutrition and growth regulation (Hull et al., 2000; Rask et al., 2000; Kutz et al., 2002; Grubb et al., 2004; Katz et al., 2015). GLS breakdown products are part of human consumption and health. For instance, some of the GLS metabolites give characteristic flavors to Brassica vegetables (cabbage, cauliflower, broccoli, etc.) and condiments (mustard, horseradish, wasabi, etc.); and some others such as sulforaphane, indole-3-carbinol and phenethyl isothiocyanate act as cancer-preventive agents (Zhang et al., 1994; Hecht, 2000; Nakajima et al., 2001; Keck and Finley, 2004; Hayes et al., 2008). Moreover, Brassica crops are used for crop rotation and/or biofumigation as certain GLS metabolites exhibit natural biopesticide properties (Gimsing and Kirkegaard, 2009).

#### Edited by:

Anna N. Stepanova, North Carolina State University, United States

#### Reviewed by:

Stephan Pollmann, Centre for Plant Biotechnology and Genomics, Spain John J. Ross, University of Tasmania, Australia

> \*Correspondence: Youfa Cheng yfcheng@ibcas.ac.cn

#### Specialty section:

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

Received: 31 July 2017 Accepted: 30 November 2017 Published: 14 December 2017

#### Citation:

Malka SK and Cheng Y (2017) Possible Interactions between the Biosynthetic Pathways of Indole Glucosinolate and Auxin. Front. Plant Sci. 8:2131. doi: 10.3389/fpls.2017.02131

**323**

Glucosinolates are evolutionarily younger and evolved from cyanogenic glucosides. Cyanogenic glucosides are widespread in planta, whereas GLS are restricted to the order Capparales and the genus Drypetes of Euphorbiaceae (Johnson et al., 2009). Cyanogenic glucosides and GLS do not coexist in plants, except for one species, Carica papaya that produces both phenylalaninederived cyanogenic glucosides and GLS (Bennett et al., 1997). Evolutionary link between cyanogenic glucosides and GLS is supported by having similarities in their biosynthesis such as amino acids as precursors and CYTOCHROME P450s (CYPs) as aldoxime metabolizing enzymes (Bak et al., 1998, 2001; Hansen et al., 2001a; Naur et al., 2003).

Glucosinolates (GLS) are characterized by having a thioglucose moiety, a sulfonated oxime, and a side-chain derived from aliphatic, aromatic, or indole amino acids (**Figure 1**). Currently, more than 130 different GLS structures have been identified (Agerbirk and Olsen, 2015). GLS are biosynthesized from amino acids and stored in the vacuoles of specific laticiferlike sulfur-rich cells called S-cells localized in the phloem cap along the vasculature and the leaf margins (Koroleva et al., 2010). GLS hydrolyzing myrosinases are localized in myrosin cells and are spatially separated from GLS (Thangstad et al., 1991; Xue et al., 1993; Andreasson et al., 2001; Husebye et al., 2002). Biosynthesis and long-distance transport of GLS are critical for spatio-temporal distribution of the GLS in plants (Nour-Eldin et al., 2012; Andersen and Halkier, 2014; Jørgensen et al., 2015). In Arabidopsis, transport of GLS compounds is mediated by transporter proteins GTR1/NPF2.10 and GTR2/NPF2.11 (Nour-Eldin et al., 2012).

Several lines of evidence suggest that indole GLS are metabolically linked to auxin homeostasis. Interruption of GLS metabolism leads to severe defects in plant growth and development similar to high-auxin phenotypes (Boerjan et al., 1995; Mikkelsen et al., 2000, 2004; Bak and Feyereisen, 2001; Hansen et al., 2001b; Reintanz et al., 2001; Chen et al., 2003;

Tantikanjana et al., 2004; Skirycz et al., 2006; Ueda et al., 2006). The high-auxin phenotypes of GLS mutants were an effect of blocking the indole GLS pathway downstream of indole-3 acetaldoxime (IAOx), which resulted in overflow of IAOx to indole-3-acetic acid (IAA) (Halkier and Gershenzon, 2006; Nafisi et al., 2006). In this review, we discuss current knowledge on potential involvement of GLS especially indole GLS metabolism in auxin homeostasis.

# STRUCTURE AND CLASSIFICATION OF GLS

The typical chemical structure of GLS consists of β-Dglucopyranose residue linked via a sulfur atom to a (Z)-Nhydroximinosulfate ester plus a precursor amino acid derived R group (**Figure 1**). Based on the precursor amino acid and the types of modification to the variable R group, GLS can be classified into aromatic (phenylalanine or tyrosine), aliphatic (alanine, leucine, isoleucine, methionine, or valine), and indole GLS [tryptophan (TRP)] (Fahey et al., 2001; Agerbirk and Olsen, 2012). A list and structures of GLS can be found in an excellent review (Clarke, 2010).

# AN OVERVIEW OF BIOSYNTHESIS OF GLS AND CAMALEXIN

Glucosinolates biosynthetic pathway comprises three steps (**Figure 2**): amino acid chain-elongation, core structure formation, and secondary modifications (Chen and Andreasson, 2001; Halkier and Gershenzon, 2006). In the amino acid chain-elongation step, certain aliphatic and aromatic amino acids are elongated by insertion of methylene groups into their side chains. These reactions are mediated by the METHYLTHIOALKYLMALATE SYNTHASE (MAM) 1-3 and MAM-like (MAML) genes (Kroymann et al., 2001; Textor et al., 2007). The amino acid moiety of either chain elongated or not, is converted to a core GLS structure in a series of reactions. During this process, amino acids are converted to their corresponding aldoximes by CYP79s (**Figure 2**). CYP79A2 catalyzes phenylalanine (Wittstock and Halkier, 2000), CYP79F1 and CYP79F2 metabolize chain-elongated methionine (Hansen et al., 2001b; Chen et al., 2003), whereas CYP79B2/B3 convert TRP (Hull et al., 2000) to their aldoximes. The enzyme catalyzing homophenylalanine is unknown. Aldoximes are metabolized to S-alkylthiohydroximates by CYP83s. Methionine-derived aldoximes are catalyzed by CYP83A1, whereas aromatic- and indole-acetaldoximes are catalyzed by CYP83B1/SUPERROOT2 (SUR2) (Bak and Feyereisen, 2001; Hemm et al., 2003). S-alkylthiohydroximates are cleaved into thiohydroximates by a C-S lyase/SUPERROOT1 (SUR1) (Mikkelsen et al., 2004), followed by S-GLUCOSYLTRANSFERASE (UGT) mediated glycosylation (Petersen et al., 2001). Finally, sulfonation of the desulfo-GLS is carried out by sulfotransferases (Piotrowski et al., 2004). Later, GLS core structure undergoes several secondary modifications at the side chain and glucose moiety

(Hopkins et al., 2009). Side chain of aliphatic GLS is modified by oxygenation, hydroxylation, alkenylation, and benzoylation, whereas side chain of indole GLS is modified by hydroxylation and methoxylation (Sønderby et al., 2010b).

Camalexin is a major phytoalexin found in specific group of Cruciferous plants including Arabidopsis (Glawischnig, 2007; Rauhut and Glawischnig, 2009; Bednarek et al., 2011). Camalexins are synthesized in response to fungal pathogens and play positive role in their resistance (Pedras et al., 2011). In camalexin biosynthetic pathway, CYP71A13/12 convert IAOx into indole-3-acetonitrile (IAN) (Nafisi et al., 2007; Müller et al., 2015) which is then oxidized and conjugated to glutathione by glutathione-S-transferase GSTF6 to produce GSH (IAN) (Su et al., 2011). This GSH (IAN) is metabolized to Cys(IAN) by γ-glutamyl peptidases GGP1 and GGP3 (Geu-Flores et al., 2011) which is metabolized to camalexin by CYP71B15/PAD3 (Zhou et al., 1999; Schuhegger et al., 2006; Böttcher et al., 2009).

# AN OVERVIEW OF AUXIN BIOSYNTHESIS

IAA is proposed to be biosynthesized from two pathways, TRP-independent and TRP-dependent pathway (Woodward and Bartel, 2005; Chandler, 2009; Normanly, 2010). TRP-dependent IAA biosynthesis is considered as the main route of IAA biosynthesis in plants (**Figure 3**).

# TRP-Independent Pathway

In the TRP-independent pathway, indole-3-glycerol phosphate or indole is the likely precursor, while the genes and enzymes involved in this pathway are still largely unknown (Ouyang et al., 2000; Zhang et al., 2008; Wang et al., 2015). Studies on TRP auxotrophic mutants that were unable to synthesize TRP revealed the existence of TRP-independent pathway (Baldi et al., 1991; Wright et al., 1991; Normanly et al., 1993; Ouyang et al., 2000; Tivendale et al., 2014). When TRP auxotrophic mutants in maize and Arabidopsis were fed with isotope-labeled anthranilate and TRP, IAA was more enriched than TRP, and the incorporation of the label into IAA from TRP was low, indicating occurrence of TRP-independent IAA biosynthesis (Wright et al., 1991; Normanly et al., 1993). The TRP biosynthetic mutants trp3 and trp2, defective in TRP SYNTHASE A and TRP SYNTHASE B subunits, respectively, accumulated higher levels of IAA than the wild-type despite containing lower TRP levels (Last et al., 1991; Normanly et al., 1993; Radwanski et al., 1996; Ouyang et al., 2000). Recently, INDOLE SYNTHASE was suggested to catalyze indole-3-glyceralphosphate in TRP-independent pathway and play essential role in embryo development (Zhang et al., 2008; Wang et al., 2015).

# TRP-Dependent Pathway

In the TRP-dependent pathway, IAA is synthesized via indole-3-pyruvic acid (IPA), indole-3-acetamide (IAM), tryptamine (TAM) and/or IAOx as intermediates (Woodward and Bartel, 2005; Chandler, 2009; Normanly, 2010; **Figure 3**). The IPA pathway is considered as a predominant auxin biosynthesis pathway in plants (Mashiguchi et al., 2011; Zhao, 2014).

#### The IAM Pathway

The IAM pathway is well established in bacteria where TRP MONOOXYGENASE (iaaM) converts TRP to IAM, which is further converted to IAA by IAM HYDROLASE (iaaH) (Patten and Glick, 1996). AMIDASE 1 (AMI1) was identified as a homolog of iaaH in Arabidopsis and Nicotiana (Pollmann et al., 2002, 2006; Nemoto et al., 2009). However, no iaaM homologs

have been found in plants. IAM was identified as an endogenous compound in many plant species including Arabidopsis, maize, rice, and tobacco (Sugawara et al., 2009; Mano and Nemoto, 2012; Novak et al., 2012). Additionally, AMI1 enzyme activity was detected in various plants (Kawaguchi et al., 1991, 1993; Arai et al., 2004; Pollmann et al., 2009; Sánchez-Parra et al., 2014), suggesting existence of this pathway in the plants.

#### The IPA Pathway

In the IPA pathway, the TRP aminotransferase TAA1/TAR (TRP AMINOTRANSFERASE OF ARABIDOPSIS) gene family converts TRP to IPA (Stepanova et al., 2008; Tao et al., 2008; Yamada et al., 2009; Zhou et al., 2011; Tivendale et al., 2012), which is subsequently converted to IAA by YUCCA (YUC) flavin monooxygenase (Mashiguchi et al., 2011; Stepanova et al., 2011; Won et al., 2011). Homologs of TAA1 were identified in maize (Chourey et al., 2010; Phillips et al., 2011), pea (Tivendale et al., 2012), and several other species (Chourey et al., 2010; Liu et al., 2012). There are 11 YUCs in Arabidopsis, overexpression of YUCsresults in auxin-overproduction phenotypes in Arabidopsis (Zhao et al., 2001; Woodward et al., 2005; Cheng et al., 2006; Kim et al., 2007; Hentrich et al., 2013) and various other plants (Zhao, 2014). Conversely, loss-of-function yuc mutants displayed low auxin synthesis with developmental defects (Cheng et al., 2006, 2007; Zhao, 2014), which can be rescued by adding auxin to growth media (Chen et al., 2014) or by expressing bacterial auxin biosynthetic gene iaaM under the control of a YUC promoter (Cheng et al., 2006). Overexpression of TAAs does not cause any obvious developmental phenotypes. However, low auxin phenotypes of sav3 and wei8 caused by mutations in TAA1 were rescued by overexpressing iaaM or by using the synthetic auxin picloram (Stepanova et al., 2008; Tao et al., 2008).

#### The TAM Pathway

In the TAM pathway, TRP DECARBOXYLASE (TDC) converts TRP to TAM. Moreover, TDCs were functionally characterized to participate in indole alkaloid and serotonin biosynthesis. For instance, transgenic tobacco plants overexpressing the TDC gene of Catharanthus roseus accumulated very high levels of TAM, whereas IAA levels were unaffected (Songstad et al., 1990). Hence, the TAM pathway is not considered as a major player of auxin biosynthesis.

#### The IAOx Pathway

Indole-3-acetaldoxime is biosynthesized from TRP by CYP79B2/B3 (Hull et al., 2000). IAOx was first isolated from

Brassica oleracea (Kindl, 1968), later conclusively identified from extracts of Brassica campestris (Ludwig-Müller and Hilgenberg, 1988). Earlier studies suggested that a cytochrome P450-like activity or plasma membrane-associated peroxidases might mediate conversion of TRP to IAOx (Kindl, 1968). Arabidopsis CYP9B2/B3 were identified in a yeast screen for cDNAs conferring resistance to 5-fluoroindole (the precursor of a toxic TRP derivative) and demonstrated their ability to specifically convert TRP to IAOx in vitro (Hull et al., 2000; Mikkelsen et al., 2000). YUC monooxygenase was assumed to take part in IAOx synthesis, however, endogenous IAOx levels were not significantly changed in yuc quadruple mutants demonstrating GLS metabolism as the main contributor of IAOx synthesis (Sugawara et al., 2009; Zhao, 2014). Endogenous IAOx has not been detected in non-GLS plants such as tomato, pea, rice, maize, or tobacco (Cooney and Nonhebel, 1991; Quittenden et al., 2009; Sugawara et al., 2009), correspondingly, CYP79B2/B3 have only been identified in GLS plants (Bak et al., 1998; Sugawara et al., 2009), suggesting IAOx synthesis is specific to GLS plants.

It has been proposed that IAOx is channeled into biosynthesis of IAA via IAN (Hull et al., 2000; Nafisi et al., 2006; Sugawara et al., 2009). It is well demonstrated that CYP71A13 catalyzes IAOx to IAN, but this gene is mainly induced in response to pathogen infection to produce camalexin (Glawischnig, 2007). IAN levels were increased in CYP79B2 overexpressing plants, and cyp79B2/B3 double mutants showed reduced IAN content (Zhao et al., 2002). Metabolite feeding studies showed that IAM and IAN are likely produced from IAOx (Sugawara et al., 2009). When the IAOx-deficient cyp79b2/b3 double mutants supplied with <sup>13</sup>C6-labeled IAOx, <sup>13</sup>C<sup>6</sup> atoms were efficiently incorporated into IAM, IAN, and IAA, indicating that IAA can be produced from IAOx via IAM and IAN. In consistent with this, wild-type plants supplied with IAM and IAN showed auxin-overproduction phenotypes (Sugawara et al., 2009). When Arabidopsis CYP79B2 or CYP79B3 genes was ectopically expressed in tobacco, IAOx and IAN were identified as endogenous compounds in the transgenic plants together with elevated levels of IAA compared to their controls (Nonhebel et al., 2011).

In addition to being metabolized from IAOx, IAN is also produced from indole GLS by myrosinases (Halkier and Gershenzon, 2006). A tendency of IAN accumulation in accordance with turnover of glucobrassicin was observed in Arabidopsis (Müller and Weiler, 2000; Reintanz et al., 2001; Zhao et al., 2002). It has been reported that low concentration of IAN can induce a high-auxin phenotype in the Arabidopsis seedlings (Normanly et al., 1997). In maize, IAN was identified as an endogenous compound at lower magnitude than Arabidopsis, exogenously supplied IAN inhibited the root growth (Thimann, 1953; Park et al., 2003; Kriechbaumer et al., 2007). The altered auxin response in the root tips of Arabidopsis myrosinase double mutants tgg4 tgg5 was likely due to no IAN production from indole GLS under myrosinase disruption (Fu et al., 2016). In vivo conversion of IAN to IAA was observed in root tissue of Arabidopsis (Müller et al., 1998).

Nitrilases (NITs) are proposed to catalyze IAN to IAA (Bartling et al., 1992; Bartel and Fink, 1994; Normanly et al., 1997; Pollmann et al., 2002). Early studies assayed NIT enzyme activities from members of plant families including Cruciferae, Gramineae, and Musaceae (Mahadevan and Thimann, 1964; Thimann and Mahadevan, 1964). Arabidopsis genome contains four NITs named NIT1–NIT4, categorized into NIT4 and NIT1-subfamilies. The members of NIT1-subfamily, NIT1-3, were suggested to be emerged from phylogenetically older NIT4-subfamily by gene duplication events and subsequent neo-functionalization (Piotrowski, 2008). Transgenic plants overexpressing each of NIT1-3 were more sensitive to exogenously supplied IAN (Schmidt et al., 1996; Dohmoto et al., 2000a,b), whereas nit mutants were tolerant (Normanly et al., 1997). Increased NIT activity was appeared to alter levels of IAN and free IAA in NIT1 overexpressing transgenic Arabidopsis plants (Lehmann et al., 2017). In maize, loss-offunction of ZmNIT2, a homolog of AtNIT4, caused the mutants less sensitive to IAN with significantly lower amounts of total IAA in kernels and roots of young seedlings compared to wild-type plants (Kriechbaumer et al., 2007). NIT3 was proposed to catalyze IAN to IAA in sulfur deprived Arabidopsis roots (Kutz et al., 2002). In Brassica plants, development of root galls caused by Plasmodiophora brassicae infection appeared to be mediated by IAN-derived IAA (Grsic et al., 1999; Grsic-Rausch et al., 2000; Ishikawa et al., 2007).

Apart from several lines of supporting evidence, the role of NITs in IAA biosynthesis is still arguable (Piotrowski, 2008). It was shown that these enzymes have lesser substrate preference to IAN than the compounds including phenylpropionitrile, allylcyanide, phenylthio acetonitrile, and methylthio acetonitriles raising doubt on the role of these enzymes in IAA biosynthesis (Vorwerk et al., 2001). The substrate preference of NITs would be different if the enzymes were challenged with IAN as a predominant substrate (Pollmann et al., 2002). For instance, in the roots of sulfur-deprived plants, expression of NIT3 was strongly induced in response to intensified turnover of IAN precursor glucobrassicin and was suggested to metabolize IAA (Kutz et al., 2002). Additionally, in planta NITs were shown to much more efficient than the recombinant ones. For example, Arabidopsis NIT2 ectopically overexpressed in tobacco was able to convert IAN supplied at micromolar concentrations, 15 fold below the enzyme's apparent in vitro K<sup>m</sup> (Schmidt et al., 1996). In the absence of exogenous IAN, neither the nit mutants nor NIT overexpressors showed severe auxin phenotypes, and endogenous IAN and IAA levels were not clearly distinguishable (Normanly et al., 1997). NIT1 overexpressors displayed strong reduction in their primary root length with clearly elevated levels of free IAA and IAN, while nit1-3 mutant displayed wild-type like roots with reduced total IAA levels (Lehmann et al., 2017). The cyp79b2/b3 mutants that are deficient in glucobrassicin have strongly reduced levels of IAN (Zhao et al., 2002), however, are not affected in infection rates of P. brassicae and consequent root gall development (Siemens et al., 2008). The role of NITs in the development of clubroot was also questioned, as cyp79b2/b3 mutants with low levels of IAN showed normal clubroot symptoms (Siemens et al., 2008). The involvement of indole GLS and NITs in the development of clubroot disease is seemingly more complicated. For instance, the transcripts of

BnCYP83B1 and BnNit4 were up-regulated in the infected root of Brassica napus (Xu et al., 2016), whereas CYP83B1 and other GLS biosynthesis genes CYP79B2/B3 and CYP83A1 were differentially downregulated during different stages of infection in Brassica macrocarpa (Zhang et al., 2016). Hence, further studies are needed to understand this phenomenon. The role of IAN as a direct metabolite of IAOx and the involvement of NITs in IAA biosynthesis was argued, because the high-auxin phenotype of sur2 is not rescued in the nit1 genetic background (Bak et al., 2001). The mutant impaired with all NITs would exclude the possible redundancy of NIT activity and gives conclusive idea on the contribution of NITs in IAA biosynthesis.

# REGULATION OF GLS BIOSYNTHESIS

## Transcriptional Control of GLS Biosynthesis

Various Transcription Factors (TFs) are known to regulate GLS biosynthesis at transcriptional level. Of which, some TFs have been shown to control GLS production at global level. For instance, IQD1 positively affected production of both aliphatic and indole GLS. The gain- and loss-of-function of IQD1 led to increased and decreased accumulation of GLS, respectively (Levy et al., 2005). SLIM1, an EIN3-like TF, involved in sulfur deficiency response, was shown to repress the expression of GLS biosynthetic genes (Maruyama-Nakashita et al., 2006; Frerigmann and Gigolashvili, 2014). AtDOF1.1 was reported to promote GLS production (Skirycz et al., 2006). Finally, TFL2, an Arabidopsis homolog of HETEROCHROMATIN PROTEIN1, was shown to affect GLS biosynthesis (Kim et al., 2004; Bennett et al., 2005).

R2R3-MYBs constitute the largest MYB gene family in plants, characterized by possessing two repeats of DNA binding domains named R2 and R3 at the N-terminal end, and an activation or repression domain usually located at the C-terminus (Stracke et al., 2001). These TFs involve in various processes including development, metabolism and stress response (Chezem and Clay, 2016). Members of sub-group 12 R2R3-MYB are specific regulators of GLS biosynthesis: MYB34, MYB51, and MYB122 control indole GLS production, whereas MYB28, MYB29, and MYB76 regulate aliphatic GLS biosynthesis (Gigolashvili et al., 2009).

MYB28 is characterized as a dominant regulator of aliphatic GLS, whereas MYB29 and MYB76 are suggested to play additional accessory role. Overexpression of these MYBs has been shown to induce aliphatic GLS biosynthetic genes and aliphatic GLS content (Gigolashvili et al., 2007b, 2008; Hirai et al., 2007; Sønderby et al., 2007). Consistently, loss-of-function of MYB28 affected production of both short- and long-chain derived aliphatic GLS. However, myb29 and myb76 were defective in accumulation of only short-chain derived aliphatic GLS albeit to a lesser extent in myb76 (Hirai et al., 2007; Sønderby et al., 2007; Beekwilder et al., 2008; Gigolashvili et al., 2008), indicating dominance of MYB28 over other two MYBs in controlling aliphatic GLS production. In line with this, expression of aliphatic GLS biosynthetic genes was greatly affected in myb28 than in myb29 (Gigolashvili et al., 2007b, 2008; Hirai et al., 2007; Sønderby et al., 2007). Additionally, disruption of both MYB28 and MYB29 showed complete reduction of aliphatic GLS levels in myb28 myb29 double mutants (Sønderby et al., 2007; Beekwilder et al., 2008). This shows that MYB28 controls both short- and long-chain derived aliphatic GLS, while MYB29 and MYB76 regulate only the short-chain derived aliphatic GLS. However, Sønderby et al. (2010a) showed that MYB29 and MYB76 were able to regulate the production of long-chain derived aliphatic GLS in the genetic backgrounds of myb28 myb76 and myb28 myb29, respectively, suggesting interplay of MYBs in controlling GLS biosynthesis. Aliphatic biosynthetic genes were differentially transactivated by these MYBs, though MYB28 showed highest transactivation potential over the two MYBs (Gigolashvili et al., 2007b, 2008). For instance, MAML was strongly transactivated by MYB28 than MYB29; CYP79F2 was greatly transactivated by MYB28 and MYB29 but to a less extent by MYB76 (Gigolashvili et al., 2008). Additionally, the transcript levels of aliphatic biosynthetic genes were uncoupled from the levels of GLS metabolites in the myb28, myb29, and myb76 knockouts (Sønderby et al., 2010a). These reports suggest that a complex network of MYB28, MYB29, and MYB76 controls the aliphatic GLS biosynthesis specifically and coordinately.

MYB34/ATR1 was initially identified as a regulator of TRP biosynthesis as it controls the expression of TRP biosynthetic gene ASA1 (Bender and Fink, 1998). The expression of indole GLS biosynthetic genes CYP79B2/B3 and CYP83B1/SUR2 were altered in myb34 mutants, and transcript levels of MYB34 were reduced in atr4/cyp83B1/sur2, indicating its involvement in GLS biosynthesis (Smolen and Bender, 2002). The plants constitutively overexpressing MYB34 accumulated 10-fold higher indole GLS compared to their control plants. Conversely, myb34 knockouts displayed low indole GLS with reduced transcript levels of indole GLS biosynthetic genes (Celenza et al., 2005). Similar to MYB34, both MYB51 and MYB122 positively regulated indole GLS production. Metabolite analysis showed increase of indole GLS levels in the plants overexpressing MYB51 and MYB122, and decrease in myb51 and myb122 knockouts (Gigolashvili et al., 2007a). Thus, MYB34, MYB51, and MYB122 positively regulate indole GLS production.

Because CYP79B2/B3 can convert TRP to IAOx, and the expression of CYP79B2/B3 is regulated by MYB34, MYB51, and MYB122, it is possible that production of other IAOx derived metabolites including IAA, camalexin and indole-3-carboxylic acids may also be regulated by these MYB genes. Indeed, it was reported that MYB34, MYB51, and MYB122 could show a conditional and probably indirect impact on the biosynthesis of camalexin and indole-3-carbolic acids (Frerigmann et al., 2015, 2016). Moreover, elevated IAA levels were found in MYB34, and MYB122 overexpression lines (Celenza et al., 2005; Gigolashvili et al., 2007a), suggesting a potential role of these TFs in auxin homeostasis.

Recently, it has been shown that bHLH04, bHLH05, and bHLH06 genes of sub-group IIIe of bHLH TF family take part in GLS biosynthesis together with R2R3-MYBs (Schweizer et al., 2013; Frerigmann et al., 2014). bHLH06/MYC2 was shown to negatively regulate indole GLS biosynthesis, as levels of indole

GLS were increased in bhlh06/myc2 mutants (Dombrecht et al., 2007). However, it was later shown that bHLH06, bHLH04, and bHLH05 positively regulate indole GLS biosynthesis, as the triple mutants had reduced levels of indole GLS (Schweizer et al., 2013; Frerigmann et al., 2014). bHLH06 bound directly to the G-box motif in the promoters of GLS biosynthetic genes (Schweizer et al., 2013). Moreover, MYB-bHLH interactions can play essential role in controlling GLS biosynthesis. For instance, the reduction of indole GLS levels was more pronounced in myb51 bhlh05 plants than bhlh05 single mutants. In line with this, double gain-of-function mutants myb34-1D bhlh05D94 had 20-fold more indole GLSs than their single mutants and wildtype plants (Frerigmann et al., 2014). These findings suggest that MYB and bHLH TFs play critical roles in regulating indole GLS biosynthesis.

#### Hormonal Control of GLS Biosynthesis

Glucosinolates biosynthesis is regulated by various phytohormones, including jasmonic acid (JA), ethylene, salicylic acid (SA), and brassinosteroids (BRs) (Mikkelsen et al., 2003; Guo et al., 2013). JA is a well characterized stress signaling molecule known to integrate plant response to various environmental cues, and is involved in a wide variety of plant processes (Turner et al., 2002; Koo and Howe, 2009). JA and its precursors and derivatives collectively called as jasmonates (JAs). They are shown to induce various TFs that are involved in secondary metabolite production (Memelink et al., 2001). JAs signaling involves perception of JAs by F-box protein CORONATINE INSENSITIVE1 (COI1) protein of Skp–Cullin–F-box protein complex (SCFCOI1) that facilitates ubiquitin-26S proteasome mediated degradation of transcriptional repressors called JAZ (Jasmonate ZIM domain). JAZ proteins interact with and repress a variety of TFs; this repression is released upon JAZ degradation via JAs signaling. As JAZ proteins are known to repress sub-group IIIe of bHLH TFs, to inhibit interaction between MYB-bHLH (Chini et al., 2007), these proteins play critical role in controlling GLS biosynthesis in a JAs dependent manner. It was found that the interaction strength between bHLH-MYB proteins could affect the interaction between bHLH-JAZ proteins. An amino acid substitution in bHLH05 weakens its interaction with JAZ protein (Frerigmann et al., 2014) allowing the bHLH05 to induce indole GLS biosynthesis. Exogenous JAs treatment induced GLS biosynthetic genes and GLS content in various plant species (Brader et al., 2001; Gigolashvili et al., 2008; De Geyter et al., 2012). Therefore, MYB-bHLH interactions may play a crucial role in JAs responsive GLS biosynthesis.

Salicylic acid differentially regulates GLS biosynthesis. SA treatment has been shown to induce nearly all types of GLS, of which, 2-phenylethyl GLS showed highest accumulation in oilseed rape (Kiddle et al., 1994). In Arabidopsis, 4-methoxyglucobrassicin was reported to be induced by SA, while the contents of glucobrassicin and neoglucobrassicin were decreased (Mikkelsen et al., 2003). Increased SA production in mpk4 and cpr1 induced 50% more GLS accumulation in the mutants compared to wild-type plants (Mikkelsen et al., 2003).

Glucosinolates production is negatively regulated by BR. Microarray analysis of BR-responsive genes showed that CYP79B2 was down-regulated by BR treatment in Arabidopsis (Goda et al., 2002). BR treatment reduced accumulation of both aliphatic and indole GLS. The inhibitory role of BR was confirmed by significantly higher accumulation of GLS content in BR-deficient mutant cpd, whereas transgenic plant overexpressing BR biosynthetic gene DWF4 showed dramatically reduced GLS levels (Guo et al., 2013). In another study, binding sites of BZR1 were identified in the promoters of MYB34 and MYB51 (Sun et al., 2010). Hence, it has been suggested that BR induced inhibition of GLS biosynthesis may be mediated by BR signaling TFs BZR1 and BES1 by binding directly to GLS biosynthetic genes or indirectly by interacting with MYB factors (Guo et al., 2013).

Further, ethylene has also been shown to induce the expression of GLS biosynthetic genes and their regulators (Mikkelsen et al., 2003; Frerigmann and Gigolashvili, 2014). Broccoli florets treated with ethylene were found to have higher quantities of specific indole GLS (Villarreal-Garcia et al., 2016). It was reported that abscisic acid (ABA) can also induce indole GLS biosynthesis (Frerigmann and Gigolashvili, 2014).

The distinct roles of indole GLS biosynthesis regulators MYB34, MYB51, and MYB122 in response to the phytohormones have been reported (Frerigmann and Gigolashvili, 2014). MYB34 was found to mediate ABA- and JA- induced indole GLS production. ABA and JA treatments strongly induced indole GLS biosynthesis in myb51 and myb122 but this tendency was not observed in myb34 mutants. Ethylene/SA induced accumulation of indole GLS was highly affected in myb51 than in myb34 and myb122, indicating a potential role of MYB51 in indole GLS synthesis in response to the treatment of these two hormones. MYB122 has been suggested to play a minor role in ethylene/JA induced GLS biosynthesis (Frerigmann and Gigolashvili, 2014; Frerigmann, 2016).

Hormonal cross-talks in controlling GLS biosynthesis were also suggested. For example, methyl-JA induced specific indole GLS accumulation was significantly decreased in SAoverproducing mutant cpr1 than in wild-type, indicating suppression of methyl-JA induced GLS biosynthesis by SA (Mikkelsen et al., 2003). In Brassica rapa, SA antagonistically affected methyl-JA induced GLS accumulation in the root regardless of the site of elicitation. Similar effect was found in the leaves when the roots were elicited, however, the effect was synergetic if the leaves were elicited (Zang et al., 2015).

# GLS METABOLISM IS A MODULATOR OF AUXIN HOMEOSTASIS

Some of the GLS mutants were isolated from the genetic screens aimed to identify genes involved in auxin homeostasis in Arabidopsis. For instance, sur1 was isolated in a screen designed to identify mutants with high-auxin phenotypes including small and epinastic cotyledons, an elongated hypocotyl, excess adventitious and lateral roots, and a reduced number of leaves (Boerjan et al., 1995). The different alleles of sur1, alf1, rty, and hsl3, which encodes a C-S lyase, identified in independent root morphology screens, also showed high-auxin related abnormal

root morphology (Celenza et al., 1995; King et al., 1995; Lehman et al., 1996). Later, sur2 and rnt1, loss-of-function mutants of CYP83B1, were found to display the phenotypes similar to sur1 (Delarue et al., 1998; Winkler et al., 1998; Barlier et al., 2000; Bak et al., 2001). UGT74B1 encodes a UDP-glucose:thiohydroximate S-glycosyltransferase. Insertional mutations in UGT74B1 resulted in phenotypes reminiscent of auxin overproduction, such as epinastic cotyledons, elongated hypocotyls in light grown plants (Grubb et al., 2004).

The GLS mutants with high-auxin phenotypes were found to have altered levels of IAA along with impaired GLS content. In sur1/rty, free and conjugated IAA levels were over accumulated (Boerjan et al., 1995; King et al., 1995) with undetectable levels of all types of GLS (Mikkelsen et al., 2004). Similarly, the mutants of UGT74B1 having excess free and conjugated IAA were associated with reduction in all types of GLS compared to their control plants (Grubb et al., 2004). Indole GLS production was reduced to 50% in sur2 plants (Bak et al., 2001) compared to wild-type, whereas free IAA levels were increased at all developmental stages tested (Delarue et al., 1998). cyp79f1/bushy1/sps mutants showed extremely bushy phenotype (Reintanz et al., 2001; Tantikanjana et al., 2001). cyp79f1 mutant showed decreased levels of shortchain derived GLS but increased levels of indole-3-ylmethyl-GLS, IAA (Reintanz et al., 2001), and cytokinin (Tantikanjana et al., 2001). Because the cytokinin responsive reporter ARR5::uidA and auxin responsive reporter DR5::uidA in the cyp79f1 mutant showed that increased levels of cytokinin, but not auxin, correlate well with a root-specific expression pattern, the bushy phenotype might be caused by increased level of cytokinin (Tantikanjana et al., 2004). Both auxin and cytokinin can influence the hormone levels of each other. Increased cytokinin levels in cyp79f1 might induce accumulation of auxin. Alternatively, increased indole GLS production in cyp79f1 likely increased IAA biosynthesis (Mikkelsen et al., 2003). CYP79B2/B3 were up-regulated in stressed plants, resulting in increased indole GLS and IAA (Mikkelsen et al., 2003). Therefore, it is also possible that sps/cyp79f1 mutants were stressed because of the perturbation of cytokinin homeostasis, which in turn up-regulates CYP79B2/B3 genes (Tantikanjana et al., 2004).

The altered IAA levels found in the GLS mutants were proposed to be synthesized from IAOx (Hull et al., 2000), a common precursor for indole GLS, camalexin, and IAA (**Figure 3**). The CYP71 clade genes SUR2/CYP83B1 and CYP71A13/12 channel IAOx into biosynthesis of indole GLS and camalexin, respectively. SUR2 catalyzes IAOx into an indole-3-Salkyl-thiohydroximate and is subsequently metabolized to indole GLS (Bak et al., 2001).

Indole-3-acetaldoxime channeling into production of either IAA, or secondary metabolites indole GLS or camalexin must be tightly controlled. In CYP79B2 overexpressing plants, elevated IAOx has been found to be channeled into biosynthesis of indole GLS (Mikkelsen et al., 2000) and IAA (Zhao et al., 2002). In response to the increased production of IAA in CYP79B2 overexpressors, transcripts of IAA-inducible genes including IAA/AUX, SAUR, and GH3s were induced (Zhao et al., 2002). Consistently, disruption of CYP79B2/B3 abolished production of indole GLS, and affected rate of IAA biosynthesis in the mutants (Zhao et al., 2002). CYP79B2 was shown to highly express in response to silver nitrate treatment that induces camalexin synthesis. Consistently, cyp79B2 single and cyp79B2/B3 double mutants were unable to synthesize camalexin under induced conditions (Glawischnig et al., 2004; Ljung et al., 2005). In the absence of pathogen attack, IAN levels were not altered in cyp71A13 knockout mutants (Sugawara et al., 2009), indicating fine-tuned regulation of IAOx metabolic channeling into the corresponding pathways. Loss-of-function mutations of SUR2 restricted IAOx flux into biosynthesis of indole GLS, resulting in decreased indole GLS production (Bak et al., 2001) and increased free IAA levels (Delarue et al., 1998). In agreement with the elevated free IAA levels, sur2 disruption induced transcription of early auxin responsive genes such as Aux/IAAs and GH3s (Mikkelsen et al., 2009; Morant et al., 2010). Conversely overexpression of SUR2 led to increased production of indole GLS (Bak et al., 2001). It was shown that TAM can competitively inhibit SUR2 that resulted in conversion of IAOx into IAA (Bak and Feyereisen, 2001). These studies indicate that IAOx plays important roles in plant development and defense responses as a branch point for biosynthesis of indole GLS, camalexin, and IAA.

Differential activation of IAOx pathway resulted in altered auxin homeostasis in post-acetaldoxime mutants or overexpressors of CYP79B2 (Nafisi et al., 2006). For instance, in sur2, increased endogenous IAA levels were associated with up-regulation of CYP79B2, IAA conjugation genes such as GH3s (Morant et al., 2010), and subsequent accumulation of IAA catabolites such as IAA-aspartate and oxindole-3-acetic acid (Barlier et al., 2000). Consistently, IAA-leucine resistant 1-like family of amidohydrolases ILL1 and ILL2 which release IAA from amide conjugates were down-regulated indicating that increased IAA levels were catalyzed to irreversible conjugates in sur2 plants (Morant et al., 2010). Similar to sur2 plants, overexpression of CYP79B2 induced expression of auxin responsive genes and increased accumulation of IAN (Zhao et al., 2002). It appears that impaired aliphatic GLS production indirectly affects production of IAOx derived indole GLS and IAA levels, as demonstrated by upregulation of CYP79B2/B3 genes in CYP79F1 co-suppressed plants (Hansen et al., 200lb), and increased accumulation of indole-3-ylmethyl-GLS, IAA in cyp79f1 mutant (Reintanz et al., 2001). Nevertheless, the increased IAA levels in cyp79f1 mutants may not be responsible for the bushy phenotype.

It was shown that cyp79b2/b3 double mutants displayed wildtype like IAA levels under normal growth conditions, but showed a modest decrease of free IAA levels under high temperature (Zhao et al., 2002; Sugawara et al., 2009). Overexpression of CYP79B2 significantly elevated levels of indole GLS and IAN but with normal IAA levels (Zhao et al., 2002). It was suggested that CYP79B2/B3 were primarily responsible for production of secondary metabolites GLS and camalexin (Glawischnig et al., 2004). These observations question whether IAOx pathway can contribute to basal IAA production, and its role in regulating plant growth and development.

Besides this, potential involvement of IAOx pathway in certain circumstances is well documented. It has been proposed that root growth under sulfur starvation is initiated by extra IAA

produced from IAN (Kutz et al., 2002). IAA produced from IAOx and IPA-pathways has been shown to involve in PIF4 mediated hypocotyl elongation in response to high temperature (Franklin et al., 2011). The expression of TAA1 and CYP79B2 genes were induced in response to high temperature, however, their expression was greatly reduced in pif4-101 mutants (Franklin et al., 2011). Similarly, IAOx and IPA pathways were shown to be hyperactive during high temperature induced microsporogenesis as demonstrated by tremendous increase of transcripts of NIT2 and TAA1 (Rodríguez-Sanz et al., 2015). Recently, it has been demonstrated that miR10515 promotes IAA biosynthesis via IAOx pathway under high temperature by suppressing SUR1 (Kong et al., 2015). Overexpression of miR10515 partially phenocopied sur1 phenotype with repressed SUR1 expression and elevated IAA concentration. NIT3 expression was strongly induced or repressed in the miR10515 overexpressing and silenced plants, respectively (Kong et al., 2015). Low auxin phenotype of cyp79b2/b3 double mutants appeared only in high temperature grown plants (Zhao et al., 2002). These reports suggest that IAOx pathway may provide extra auxin in response to environmental stresses.

It was reported that upregulation of the IAOx pathway can compensate defects in the IPA pathway (Stepanova et al., 2008). The elevated IAA in sur2 plants had attenuated the meristem maintenance and lateral root formation defects of TAA1 and TAR2 double mutants wei8 tar2. Additionally, dwarf phenotypes of sur2/rty1 were alleviated in wei8 tar2 sur2 and wei8 tar2 rty1 genetic backgrounds suggesting an existence of a functional overlap between the IPA- and IAOx-dependent routes of auxin biosynthesis (Stepanova et al., 2008). Developmental defects in ASA1/WEI2 mutants were suppressed by excess IAA levels accumulated in sur1 and sur2 mutants, respectively (Stepanova et al., 2005). These findings indicate that IAOx pathway may be operative under normal growth conditions as well; however, further studies are needed to confirm this idea. Regardless of

#### REFERENCES


how significant this pathway is in controlling plant growth and development, endogenous IAOx and its metabolizing enzymes CYP79B2/B3 were found only in GLS plants. Thus, IAOx pathway is considered as a species-specific pathway (Sugawara et al., 2009).

### CONCLUSION

Biosynthetic pathways of indole GLS, camalexin and IAA are metabolically connected by their common metabolic intermediate IAOx. Disruption of indole GLS production leads to altered auxin homeostasis via differential activation of IAOx pathway. Physiological role of NITs in IAA biosynthesis is so far not conclusive. Our current knowledge on IAOx pathway suggests that this pathway is likely operative under special circumstances. Identification of the enzyme responsible for conversion of IAOx to IAA, and its functional characterization under normal and induced conditions would help us to better understand the role of this pathway in plant development.

### AUTHOR CONTRIBUTIONS

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

#### ACKNOWLEDGMENTS

This work was supported by the projects from the National Basic Research Program of China (2014CB943400), National Natural Science Fund of China (31171389; 91217310; 91017008; 31270330) and One-hundred Talent Project, and the President's International Fellowship Initiative of the Chinese Academy of Sciences.

P450 in the biosynthesis of the cyanogenic glucoside dhurrin. Plant Mol. Biol. 36, 393–405. doi: 10.1023/A:1005915507497


metabolism in Arabidopsis thaliana relatives. New Phytol. 192, 713–726. doi: 10.1111/j.1469-8137.2011.03824.x




catalyzes the conversion of indole-3-acetaldoxime in camalexin synthesis. Plant Cell 19, 2039–2052. doi: 10.1105/tpc.107.051383



nitrilase subfamily encoded by the NIT2/NIT1/NIT3-gene cluster. Planta 212, 508–516. doi: 10.1007/s004250000420


ssp. pekinensis) as a systemic response to methyl jasmonate and salicylic acid elicitation. J. Zhejiang Univ. Sci. B 16, 696–708. doi: 10.1631/jzus.B1400370


**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 Malka and Cheng. 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.

# The Mammalian Peptide Adrenomedullin Acts as a Growth Factor in Tobacco Plants

Rafael Peláez 1 †, María Niculcea1 † and Alfredo Martínez 1, 2 \*

<sup>1</sup> Biomass Booster Ltd, Logroño, Spain, <sup>2</sup> Oncology Area, Center for Biomedical Research of La Rioja, Logroño, Spain

Growth factors are extracellular signals that regulate cell proliferation and total body mass. Some animal growth factors can work on plant tissues and vice versa. Here we show that the mammalian growth factor adrenomedullin (AM) induces growth in tobacco plants. Addition of synthetic AM resulted in a dose-dependent growth of tobacco calluses. Furthermore, AM transgenic plants showed enhanced survival and significant increases in stem diameter, plant height, leaf length, weight of all organs, and a reduction in the time to flowering when compared to plants transformed with the control vector. These differences were maintained when organs were dried, resulting in a mean total biomass increase of 21.3%. The levels of soluble sugars and proteins in the leaves were unchanged between genotypes. AM transgenic plants had a significantly higher expression of cyclin D3 and the transcription factor E2FB than controls, suggesting that cell cycle regulation may be part of the intracellular signaling of AM in plants. In summary, mammalian AM increases vascular plants' survival and biomass with no apparent detriment of plant's morphological and/or biochemical properties, thus this strategy could be useful for crop productivity improvement.

#### Edited by:

Chi-Kuang Wen, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, China

#### Reviewed by:

Guodong Wang, Shaanxi Normal University, China Georg Groth, University of Düsseldorf, Germany

#### \*Correspondence:

Alfredo Martínez amartinezr@riojasalud.es

† These authors have contributed equally to this work.

#### Specialty section:

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

Received: 22 November 2016 Accepted: 27 March 2017 Published: 12 April 2017

#### Citation:

Peláez R, Niculcea M and Martínez A (2017) The Mammalian Peptide Adrenomedullin Acts as a Growth Factor in Tobacco Plants. Front. Physiol. 8:219. doi: 10.3389/fphys.2017.00219 Keywords: adrenomedullin, Nicotiana tabacum, growth factor, plant survival, biomass increase, cell cycle

# INTRODUCTION

Growth factors are signaling molecules produced by living organisms to stimulate their cellular growth rate, healing, cellular differentiation, and final size. Chemically, growth factors are very heterogeneous and can be constituted by peptides, such as growth hormone or insulin-like growth factors in mammals (Pfaffle, 2015) or root meristem growth factor in plants (Shinohara et al., 2016); lipids such as corticosteroids (Loke et al., 2015) or strigolactones (Pandey et al., 2016); secondary metabolites such as auxins or gibberellins (Gantait et al., 2015; Li et al., 2016); or even gaseous mediators such as nitric oxide or ethylene (Napoli et al., 2013; Street et al., 2015). It is commonly understood that plants and animals are two very distinctive systematic groups and they are supposed to rely on completely different signaling molecules, but there is evidence that some molecules may have activity in both kingdoms. For instance, addition of human epidermal growth factor to sorghum seedlings (Dyer, 1980) or to corn primary roots and mesocotyls (Kato et al., 1995) resulted in a significant growth increase for these structures. In addition, the reverse situation has been reported as well, with plant peptides being able to bind and activate growth factor receptors in human cells (Treggiari et al., 2015; Molesini et al., 2017). A common signaling molecule is nitric oxide, which reduces growth in plants (Arora and Bhatla, 2017) and has a miscellaneous effect on mammalian cells, depending on nitric oxide concentration and cell type (Napoli et al., 2013).

**337**

Adrenomedullin (AM) is a 52 amino acid peptide with an internal disulphide bond and a terminal amide group, whose tridimensional structure is characterized by a central alpha-helix (Lopez and Martinez, 2002; Perez-Castells et al., 2012). AM is produced from a 185 amino acid precursor termed preproadrenomedullin, which undergoes post-translational processing by endo- and exo-peptidases, generating a series of chemical intermediaries. The last intermediate before fully mature AM is Gly-extended AM (or Gly-AM), which usually lacks physiological activity (Martinez et al., 2006). AM homologs have been found in all vertebrates, from humans to fish (Martinez et al., 2006), and immunoreactive epitopes were detected in invertebrates such as echinoderms (Martinez et al., 1996). No apparent homology is found between AM and any of the described plant peptide hormones. Among other physiological roles, AM has been shown to act as a potent growth factor for normal (Isumi et al., 1998) and tumor cells (Miller et al., 1996). In addition, the abrogation of the gene coding for this peptide results in 100% embryo lethality (Caron and Smithies, 2001), underscoring the relevance of this molecule for normal development in mammals.

Given the broad distribution of AM among animals and its demonstrated potency as a growth factor, we decided to investigate whether this peptide may have some influence on the in vitro and in vivo growth of vascular plants, choosing tobacco as a model organism.

# MATERIALS AND METHODS

#### In vitro Culture and Treatment of Tobacco Calluses

Seeds of Nicotiana tabacum brasiliensis, cv Paraguay, were obtained from University of Navarra germplasm collection and cultured under standard conditions. Calluses were prepared from explants of young tobacco leaves placed in callus-producing solid medium [Murashige and Skoog (MS) complete medium (Murashige and Skoog, 1962) plus 100 mg/L 2-4D and 100 mg/L kinetin] and incubated in the dark at 24◦C. When the calluses were large enough, they were divided in small fragments that were weighed and transferred to fresh callus-producing solid medium containing different amounts of AM or Gly-AM synthetic peptides (Phoenix Pharmaceuticals, Burlingame, CA), as indicated. Synthetic peptides had a purity >95%. Cultures were grown in the dark at 24◦C for 30 days. At this time, individual calluses were weighed again. Growth rate was calculated as the weight at the end of the procedure divided by the weight of the same fragment at the beginning. All values were normalized to the untreated control that was given the value 100. Calluses were dried in an oven (Selecta, Barcelona, Spain) at 80◦C and weighed again.

# Production of a Molecular Vector for Transgenesis

Mammalian AM was cloned from a cDNA library obtained from human lung carcinoma cell line A549, which has been previously shown to express high levels of AM (Miller et al., 1996). PCR was performed with primers AM sense: ATG AAG CTG GTT TCC GTC GCC CTG and AM antisense: CTA AAG AAA GTG GGG AGC ACT TCC ACT CG, thus cloning a fragment containing mature AM, surrounded by restriction sites XbaI and SacI (underlined). PCR products were subcloned into pCRII vector (ThermoFisher, Madrid, Spain) and the identity of the cloned product was confirmed through direct sequencing. This plasmid was named pCRII-AM. The A. tumefaciens Tibased vector pBI-121 was purchased from Clontech (Palo Alto, CA). Both pBI-121 and pCRII-AM were excised with XbaI and SacI and the AM fragment was inserted in the pBI vector in place of the GUS gene. In this vector, gene expression is driven by the 35S promoter. This new vector was named pBI-AM and its sequence fidelity was also confirmed through direct sequencing.

# Generation of Transgenic Tobacco Plants

Agrobacterium tumefaciens bacteria were transformed with plasmids pBI-121 and pBI-AM by electroporation and grown in YEP medium (10 g/L yeast extract, 10 g/L peptone, 5 g/L NaCl, pH 7.0) without antibiotics at 30◦C. Three hours later, bacteria were transferred to YEP medium containing 100µg/ml streptomycin and 50µg/ml kanamycin and incubated overnight at 30◦C. Optical density of the bacterial culture was adjusted to 0.8 and small tobacco leaf fragments were immersed in this medium for 30 min and then transferred, still with the bacteria, to solid co-cultivation MS medium (with 0.6 g/L MES and 1 mg/ml BAP) without antibiotics, following a protocol slightly modified from Horsch et al. (1986). After 72 h, leaf fragments were transferred into shoot MS regeneration medium with 0.6 g/L MES, B5 vitamins, 100 mg/ml carbenicillin, 100 mg/ml cefotaxime, and 50 mg/ml kanamycin and cultured bi-weekly to the same medium. In vitro plant culture was performed inside an environmental chamber (FitoClima 1200 Bio, Aralab, Rio de Mouro, Portugal) in dim light (60–80 mE/m.s.) and 18/6 photoperiods at 24◦C. When shoots and leaves reached a convenient size, they were transferred to root development medium (MS/B5/MES medium and antibiotics) and when the radicular system was mature enough, the plantlets were placed in small pots containing regular soil. At this point young plants of both genotypes were moved into the greenhouse and regular horticultural care was applied. Periodically, several morphological parameters such as plant height, stem diameter, number of leaves, leaf length, and internodal length were measured in all plants by researchers blinded to the actual genotypes. After the first floral buds opened but before complete flower maturation, all plants were uprooted and weighed, measuring total weight and the weight of the roots, stems, leaves, and flowers. After drying the plants, the dry weight for all organs was also recorded.

# Genotyping of Transgenic Plants

DNA was obtained from fresh leaf punches with the Phire Plant Direct PCR kit (Thermo Scientific). PCR was performed with primers AM sense: CGC CAG AGC ATG AAC AAC T; AM antisense: CGA CGT TGT CCT TGT CCT TA; GUS sense: TGC TGT CGG CTT TAA CCT CT; and GUS antisense: GGC ACA GCA CAT CAA AGA GA. AM primers render a PCR amplicon of 121 bp while GUS primers produce a 332 bp band. PCR was run in a GeneAmp PCR System 9700 (Applied Biosystems, Foster City, CA) for 30 cycles with an annealing temperature of 60◦C.

#### Gene Expression Quantification

Leaf tissue samples were homogenized with TRIzol (Invitrogen, Madrid, Spain) and RNA was isolated with RNeasy Mini kit (Qiagen, Germantown, MD). Three micrograms of total RNA were treated with 0.5µl DNAseI (Invitrogen) and reversetranscribed into first-strand cDNA using random primers and the SuperScript III kit (Invitrogen). Reverse transcriptase was omitted in control reactions, where the absence of PCR-amplified DNA confirmed lack of contamination from genomic DNA. Resulting cDNA was mixed with SYBR Green PCR Master Mix (Invitrogen) for quantitative real time polymerase chain reaction (qRT-PCR) using 0.3µM forward and reverse oligonucleotide primers (**Table 1**). Quantitative measures were performed using a 7300 Real Time PCR System (Applied Biosystems, Carlsbad, CA). Cycling conditions were an initial denaturation at 95◦C for 10 min, followed by 40 cycles of 95◦C for 15 s and 60◦C for 1 min. At the end, a dissociation curve was implemented from 60 to 95◦C to validate amplicon specificity. Gene expression was calculated using absolute quantification by interpolation into a standard curve. All values were divided by the expression of the house keeping gene Tubulin α1.

#### Protein Extraction and Western Blotting

Small leaf fragments were homogenized in RIPA buffer (Thermo Scientific) containing protease (EDTA-free complete, Roche, Basilea, Switzerland) and phosphatase (PhosStop, Roche) inhibitors, 2 mM dithiothreitol, and 8 M urea. Homogenates were sonicated three times (Soniprep 150, MSE, London, UK), centrifuged for 30 min at 15,000 × g and the supernatants collected. Protein concentration was determined with the

#### TABLE 1 | Primers used for qRT-PCR.


Annealing temperature for all primers was 60◦C.

BCA kit (Pierce, Rockford, IL), with bovine serum albumin as standard, using a spectrophotometer (POLARstar Omega, BMG Labtech, Ortenberg, Germany). Then, 100µg of each sample were mixed with 4x sample buffer (Invitrogen) and heated for 10 min at 70◦C. Samples were run on 4– 12% SDS—polyacrylamide gels. Seeblue plus 2 Prestained Standards (Invitrogen) were used as molecular weight markers. For Western blot analysis, proteins were transferred onto 0.2-µm polyvinylidene difluoride (PVDF) membranes (iBlot system, Invitrogen). For protein identification, membranes were incubated overnight at 4◦C with a rabbit polyclonal anti-AM antibody made in house and previously characterized (Martinez et al., 1995) at a 1:2,000 dilution. To visualize immunoreactivity, membranes were incubated with peroxidaselabeled goat anti-rabbit IgGs (Jackson Immunoresearch, West Grove, PA), developed with a chemoluminiscence kit (GE Biosciences, Miami, FL), and exposed to X-ray films (GE Biosciences).

#### Histological Processing

Fragments of roots, stems, and leaves were immersed in FEA fixative (5% formalin, 45% ethanol, 5% acetic acid) for 24 h, dehydrated, and embedded in paraffin. Sections (3µm-thick) were stained with hematoxylin-eosin. Photographs were taken with an Eclipse 50i microscope (Nikon, Barcelona, Spain) equipped with a DXM 1200C digital camera (Nikon).

#### Biochemical Composition of Transgenic Plants

To investigate whether the weight differences were reached in detriment of the biochemical characteristics of the transgenic plants, we measured the concentration of starch, soluble sugars, and proteins in mature leaves of both genotypes. Soluble sugars were estimated following established methods (Yemm and Willis, 1954). Briefly, frozen leaves were reduced to powder and 250µl of this extract were mixed with 3.0 ml anthrone reagent, stirred, and incubated at 80◦C for 10 min. After cooling, absorbance was measured at 620 nm. Concentration of soluble proteins was evaluated with a BCA protein assay kit (Thermo Scientific, Rockford, IL). Starch concentration was measured as reported (Jarvis and Walker, 1993). Briefly, leaves were dissociated in 5 ml 95% ethanol and centrifuged at 28,710 × g for 10 min at 4 ◦C. Two ml 1 M KOH were added to the supernatant and the solution was incubated in the dark for 24 h to solubilize the starch. Then, the solution was neutralized with 2 ml 1 M HCl and 0.5 ml iodine reagent were added and allowed to react for 15 min in the dark. The resulting solution was measured at 565 nm.

#### Statistical Analysis

Statistical analysis was performed with SPSS v18 software package. Data sets were tested for normalcy (Kolgomorov-Smirnov) and homoscedasticity (Levene). Normally distributed data were compared using Student's t-test or ANOVA followed by post-hoc tests (Dunnet). When data did not follow a normal distribution they were compared with Mann-Whitney's U nonparametric test. Survival rates were analyzed using a χ 2 -test. A p-value lower than 0.05 was considered statistically significant.

# Search for AM Orthologs in Tobacco's Genome

The genome of N. tabacum is available (Sierro et al., 2014) and a search of tobacco's predicted proteome was performed looking for peptides with some sequence homology to any of the available vertebrate AM sequences (Martinez et al., 2006) using BLAST approaches, Prosite, and the EggNOG v3.0 database, using threshold default values (Karlin and Altschul, 1990), as described (Kuzniar et al., 2008). We also performed a search for the peptide sequences Cys-X-X-X-X-Cys and Tyr-Gly-Arg-Arg-Arg-Arg-Arg, which define the motifs essential for maintaining AM's function (Lopez and Martinez, 2002).

# RESULTS

#### AM Induces Tobacco Callus Growth In vitro

Small fragments of tobacco calluses were weighed and exposed to increasing concentrations of AM synthetic peptide. After growing in the dark for 30 days (**Figure 1A**), calluses were weighed again and their growth rate for that period of time was calculated. One-way ANOVA indicated that the treatment was significantly efficacious (p < 0.05). The higher dose of AM (10 nM) induced a 36.46% weight increase (p < 0.01, Dunnet post-hoc test) in comparison with the untreated control group (**Figure 1B**). The AM precursor, Gly-AM, had no effect on callus growth at any concentration (**Figure 1B** shows only the highest concentration tested). To investigate whether this AM-induced weight increase was due to water retention or to a real biomass increase, calluses were dried and the statistically significant weight difference (22.71%) was maintained (**Figure 1C**), thus demonstrating that AM elicits a real growth of the tobacco calluses.

### Transgenic Tobacco Plants Expressing AM Grow Larger than Controls

Transgenic plants were generated using plasmid pBI-AM, or pBI-121 as a control plasmid. All plants were genotyped by PCR with specific primers for AM and β-glucuronidase, GUS (**Figure 2A**). To confirm that AM was expressed in the transgenic plants, qRT-PCR (**Figure 2B**), and Western blotting (**Figure 2C**) were performed on leaf extracts showing a high expression of both the mRNA and the peptide in transformed plants. For all experiments, only primary transformants were used, representing independent plasmid insertion events. A first physiological observation was the improved survival of AM transgenic plants in the critical step of the transfer from agarosebased medium to real soil. Survival for AM transgenic plants was 13.25% higher than for pBI-121 control plants (**Figure 2D**). Furthermore, several morphological parameters were measured at different times during the plant's life cycle. Stem diameter, plant height, and leaf length were significantly larger (14.18, 17.72, and 8.28%, respectively) in AM transgenic plants than in their pBI-121 transgenic counterparts at all times (**Figures 3A–C**) whereas there was no difference in internodal length or in the total number of leaves per plant (results not shown). Another parameter where statistically significant differences were found was in the flowering time, which occurred earlier in AM transgenic plants. For instance, at 5 weeks after transplantation, many more (755.00%) AM plants had flowered than controls (**Figure 3D**).

After the first floral buds opened but before complete flower maturation, the plants' total weight and the weight of particular organs were measured. The weight of AMexpressing transgenic plants and their particular organs:

genotypes. χ 2 -test, p = 0.0014.

roots (29.89%), stems (21.66%), leaves (14.01%), and flowers (187.09%), was significantly heavier than this of the pBI-121 controls (**Figures 4A–D**). These differences were maintained (roots 37.18; stems 20.68; leaves 17.46; flowers 271.13%) after drying the different organs (**Figures 4E–H**), demonstrating a true increase in plant biomass. Interestingly, when the dry/fresh weight ratio was calculated it was the same between genotypes for the stems, leaves, and flowers (**Figures 4J–L**) but in the case of the roots, the ratio was significantly higher (8.80%) in the AM transgenic plants (**Figure 4I**), indicating that AM aids in the development of a larger and denser root system. The mean general increase in dry biomass for the whole plant was 21.31% when compared with pBI-121 controls.

#### Transgenic Tobacco Plants Expressing AM Possess a Larger Root System but Maintain a Normal Histological Structure

Given the differences in organ weight, we investigated the morphology of these organs in more detail. The aspect of the leaves and stems, although larger (as described above), were morphologically similar between AM transgenic plants

and normal controls, but the root system was strikingly different containing more numerous and thinner roots than the controls (**Figure 5A**). To check whether there were changes in the histological structure of these plants, sections of roots (**Figures 5B,C**), stems (**Figures 5D,E**), and leaves (**Figures 5F,G**) were studied. Apart from differences in diameter, the general structure of these organs was similar between genotypes.

# Transgenic Tobacco Plants Expressing AM Do Not Lose Basic Biochemical Components

To test whether the growth increase observed in AM-expressing plants resulted in a detrimental decrease of biochemical components, we measured the levels of soluble sugars, proteins, and starch in the leaves of both genotypes at the end of the plant's life cycle. No differences were observed for soluble sugars and proteins (**Figures 6A,B**) but there was a 15.22% significant increase in starch in the leaves of AM transgenic tobacco plants when compared to their controls (**Figure 6C**).

### Overexpression of AM Co-opts Key Intracellular Cell Cycle Controllers

To better understand AM's mechanism of action on plant cells, we investigated the gene expression of several key cyclins and transcription factors related to cell cycle modulation on plants of both genotypes. There was a statistically significant 81.29% increase of cyclin D3 (Cyc D3) (**Figure 7A**) and a 109.04% increase of transcription factor E2FB (**Figure 7C**) expression in the transgenic plants when compared with the pBI-121 controls. Other tobacco intracellular signals, such as GTPase (**Figure 7B**), auxin responsive-like gene (AuxResp) (**Figure 7D**), cyclindependent kinase B1 (CDK B1) (**Figure 7E**), and transcription factor Transparent Testa Glabra 2 (TTG2) (**Figure 7F**) were not significantly different between genotypes.

### A Search of the Tobacco Genome Did Not Identify an AM Ortholog

We performed a search of tobacco's predicted proteome looking for peptides with some sequence homology to any of the available vertebrate AM sequences and none was found using the default value threshold parameters of the software packages. Since the more relevant features of AM are the intramolecular 6-amino acid ring: Cys-X-X-X-X-Cys, and the terminal amide group represented by the sequence Tyr-Gly-Arg-Arg-Arg-Arg-Arg (Lopez and Martinez, 2002), we also performed a search for molecules containing these motifs. Although some proteins contained these sequences, none of their predicted post-translational processing or intracellular

FIGURE 4 | Weight parameters after flowering. The fresh (A–D) and dry (E–H) weight, as well as the dry/fresh ratio (I–L) was determined for the root system (A,E,I), the stems (B,F,J), the leaves (C,G,K), and the flowers (D,H,L) of AM transgenic plants (AM) and those transformed with plasmid pBI-121 as a control. Bars represent the mean ± SD of all specimens (n = 36 for controls; n = 38 for AM). Asterisks represent statistically significant differences (Student's t-test). \*p < 0.05; \*\*p < 0.01; \*\*\*p < 0.001.

FIGURE 5 | Morphology of transgenic plants. Macroscopic view of the root systems (A) for a representative AM transgenic plant (right) and a pBI-121 control plant (left). Histological appearance of roots (B,C), stems (D,E), and leaves (F,G) of AM transgenic plants (C,E,G) and those transformed with plasmid pBI-121 as a control (B,D,F). e, epidermis; f, phloem; p, parenchyma; pp, pith parenchyma; x, xylem. Bar in (A) = 5.0 mm. Bar for (B–G) = 400µm.

distribution was compatible with a secreted peptide hormone similar to mammalian AM.

### DISCUSSION

In this study we have shown that mammalian AM increases tobacco's biomass both in tissue culture and in transgenic plants. These observations are in line with previous publications where another animal peptide, epidermal growth factor, was shown to induce the growth of corn mesocotyls in vitro (Kato et al., 1995). The search of the tobacco's genome for an ortholog of AM in plants did not produce any obvious candidate, thus we have to hypothesize that the action of mammalian AM on plant's tissues must be mediated by the co-option of an unknown receptor system involved in the plant's growth. Future studies should focus in identifying this receptor system and its intrinsic agonists.

In this study we have analyzed primary transformant plants. Transgenesis is a technique that relies on the random insertion of the vector into the plant's genome. Therefore, a phenotype could be the result of altering the expression or epigenetic status of neighboring genes rather than of the expression of the inserted gene (Kohli et al., 2006). In that regard it is very important to study as many founders as possible to ensure that the common phenotype is independent of the insertion locus. This is why we believe that, at this stage, it is better to study primary transformants rather than downstream generations. Future studies should analyze transgenic lines to demonstrate that the transgene is stable and that the gene expression can be transferred from generation to generation.

Our experiments on tobacco calluses demonstrated that AM acts as a proliferation factor in this setting. In mammals, AM growth-related actions are mediated by intracellular elevations of cAMP and activation of mitogen-activated protein kinase cascades, among others (Larrayoz et al., 2014). These pathways are also present in vascular plants and both of them regulate plant cell proliferation (Xu and Zhang, 2015; Sabetta et al., 2016). Cell cycle, both in animals and plants, is driven by several regulatory proteins designated as cyclins and their cyclin-dependent kinases (CDK), whose expression often depends on plant hormones, growth conditions, and developmental stages (Espinosa-Ruiz et al., 2004). Cell cycle-dependent phosphorylation of RB-related protein by CDKs results in the release of active E2F transcription factors that induce a wave of transcriptional activity in the cell (Nakagami et al., 2002). In our transcription factor expression study we have found that AM transgenic plants had a higher expression of the CycD3 and the transcription factor E2FB. Cyclins of the D group control the onset of cell division and responses to extracellular signals during G1 phase and CycD3 is expressed only in organs and tissues where cell division is taking place (Qu and Wang, 2007). E2FB stimulates cell division by promoting both G1-to-S and G2-to-M transitions, leading to shorter duplication times. E2FB abundance and stability is increased by exogenously applied auxin (Magyar et al., 2005). Our present results show that these signaling molecules are also activated by AM. On the other hand, other genes which are also related to the auxin pathway, such as TTG2 (Zhu et al., 2013), auxin response-like gene (Gonzalez-Perez et al., 2014), or the GTPase did not respond to AM overexpression. This panorama suggests a specific intracellular pathway for AM signaling that shares some components but not others with classic growth promoters such as auxin. More studies are needed to draw the complete circuitry of AM's intracellular mediators.

Nevertheless, a factor that induces mitosis in plants may result in disorganized growth and the generation of tumorlike tissue. This is the well-known behavior of the Ti plasmid of A. tumefaciens which results in the uncontrolled growth of plant cells and the apparition of crown gall disease in infected plants (Gohlke and Deeken, 2014). This is why it was important to generate AM transgenic plants and investigate whether the overexpression of this peptide influences growth in the whole plant and whether it has an undesirable impact on plant morphology and development. In our case, the mammalian peptide generated larger and heavier plants but preserving plant architecture and biochemical composition, suggesting that this molecule works as a bona fide growth factor for plants, not influencing pattern formation. This is important if we want to apply this factor to improve commercially relevant crops.

differences (Mann-Whitney's U-test). \*p < 0.05; \*\*p < 0.01.

Another important feature was the improved survival of plantlets in the critically stressful step of transplantation from in vitro culture to real soil condition. Many growth factors act as survival and stress-reducing factors, as well, due to their activity modulating apoptosis and other intracellular preservation pathways (Collins et al., 1994). In mammals, AM has been shown to exert a survival function at the level of individual cells (Cuttitta et al., 2002) and also at the organism level (Caron and Smithies, 2001), and to be involved in stress reduction (Fernandez et al., 2008). Future studies would address whether AM transgenic plants are more resistant to typical plant stressors such as drought or infectious pathologies. In this regard, we should remember that AM has antimicrobial properties due to its alpha-helical structure and an ability to insert itself in the pathogen's membrane, opening up hydrophilic pores that result in bacterial demise (Allaker et al., 2006; Martinez et al., 2006). This feature could be an added bonus for AM transgenic crops.

An interesting observation was that the roots of AM transgenic plants were more numerous and had more dry matter proportion than their control counterparts whereas all other organs had similar percentages. This suggests that AM has a special impact on the root, making it not only to grow larger but also increasing the amount of solid biological tissue. Other plant growth factors and intracellular signals have preferential activity on specific organs. For instance, the HHO2 (hypersensitivity to low phosphate-elicited primary root shortening 1 homolog 2) transcription factor regulates root growth without influencing other plant organs (Nagarajan et al., 2016). Our observation that AM may share part of the auxin transcription factor machinery suggests a special regulation of the root system. Perhaps AM could be used to facilitate the development of a robust root system for plants, which is essential for successful cultivation and crop improvement.

Although AM-transgenic crops are not intended for human or animal consumption, if some of these plants find their way accidentally into the food chain, the presence of AM should be harmless. AM is naturally present at high concentration in human and mammalian milk, where it contributes to the healthy development of the infant's gut (Pio et al., 2000). In addition, dietary AM is rapidly degraded by pancreatic exo- and endo-peptidases into individual amino acids, thus destroying its signaling properties.

Preliminary experiments ongoing in our laboratory show similar results in the growth of other plant species, including

#### REFERENCES


vascular plants as well as unicellular microalgae, suggesting that AM could be used as a general growth promoter for increasing plant biomass through biotechnological manipulation of different crops.

In conclusion, we have shown that mammalian AM improves tobacco plant survival and growth without modifying its general architecture and biochemical composition. These observations suggest broad applications for improving crop productivity through biotechnology.

# AUTHOR CONTRIBUTIONS

RP and MN performed experiments and analyzed data; AM designed and supervised the experiments and wrote the manuscript; all authors read and approved the final version of the manuscript.

#### FUNDING

This study was funded in part by a grant from Agencia de Desarrollo Económico de La Rioja (ADER), exp n◦ 2014-I-DPT-00047.

#### ACKNOWLEDGMENTS

We are very grateful to Ms Judit Narro (CIBIR, Logroño, Spain) for excellent technical assistance with histological processing.


a first step toward the analysis of its interactions with receptors and small molecules. Biopolymers 97, 45–53. doi: 10.1002/bip.21700


**Conflict of Interest Statement:** RP and MN are former employees of Biomass Booster, Inc. AM is inventor of several patents regarding the beneficial effects of adrenomedullin on crop production and owns stock on Biomass Booster, Inc.

Copyright © 2017 Peláez, Niculcea and Martínez. 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.

# miRNA and Degradome Sequencing Reveal miRNA and Their Target Genes That May Mediate Shoot Growth in Spur Type Mutant "Yanfu 6"

Chunhui Song1 †, Dong Zhang1 †, Liwei Zheng<sup>1</sup> , Jie Zhang<sup>2</sup> , Baojuan Zhang<sup>1</sup> , Wenwen Luo<sup>1</sup> , Youmei Li <sup>1</sup> , Guangfang Li <sup>1</sup> , Juanjuan Ma<sup>1</sup> and Mingyu Han<sup>1</sup> \*

*<sup>1</sup> College of Horticulture, Yangling Subsidiary Center Project of National Apple Improvement Center, Collaborative Innovation Center of Shaanxi Fruit Industry Development, Northwest A&F University, Yangling, China, <sup>2</sup> Tongchuan Fruit Tree Experiment Station, Tongchuan, China*

#### Edited by:

*Yunde Zhao, University of California, San Diego, USA*

#### Reviewed by:

*Shaohua Zeng, South China Institute of Botany (CAS), China Yi Li, University of Connecticut, USA*

\*Correspondence:

*Mingyu Han hanmy@nwsuaf.edu.cn*

*† These authors have contributed equally to this work.*

#### Specialty section:

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

Received: *24 November 2016* Accepted: *14 March 2017* Published: *30 March 2017*

#### Citation:

*Song C, Zhang D, Zheng L, Zhang J, Zhang B, Luo W, Li Y, Li G, Ma J and Han M (2017) miRNA and Degradome Sequencing Reveal miRNA and Their Target Genes That May Mediate Shoot Growth in Spur Type Mutant "Yanfu 6." Front. Plant Sci. 8:441. doi: 10.3389/fpls.2017.00441* The spur-type growth habit in apple trees is characterized by short internodes, increased number of fruiting spurs, and compact growth that promotes flowering and facilitates management practices, such as pruning. The molecular mechanisms responsible for regulating spur-type growth have not been elucidated. In the present study, miRNAs and the expression of their potential target genes were evaluated in shoot tips of "Nagafu 2" (CF) and spur-type bud mutation "Yanfu 6" (YF). A total of 700 mature miRNAs were identified, including 202 known apple miRNAs and 498 potential novel miRNA candidates. A comparison of miRNA expression in CF and YF revealed 135 differentially expressed genes, most of which were downregulated in YF. YF also had lower levels of GA, ZR, IAA, and ABA hormones, relative to CF. Exogenous applications of GA promoted YF shoot growth. Based on the obtained results, a regulatory network involving plant hormones, miRNA, and their potential target genes is proposed for the molecular mechanism regulating the growth of YF. miRNA164, miRNA166, miRNA171, and their potential targets, and associated plant hormones, appear to regulate shoot apical meristem (SAM) growth. miRNA159, miRNA167, miRNA396, and their potential targets, and associated plant hormones appear to regulate cell division and internode length. This study provides a foundation for further studies designed to elucidate the mechanism underlying spur-type apple architecture.

Keywords: apple, shoot, miRNAs, internode, SAM

# INTRODUCTION

Apple is an important fruit crop in temperate regions of the world. The use of dwarfing rootstocks and high density planting of apple trees is common in commercial orchards as it increases yields per unit area and reduces the cost of management practices, such as pruning. The use of spur-type scion varieties is one of effective methods used to facilitate high-density planting, along with the use of dwarfing rootstocks.

"Fuji" is the major cultivar planted in China, accounting for more than 65% of apple plantings. Several spontaneous "Fuji" spur-type mutants have been identified and utilized in the breeding and planting of "Fuji" apples. The spur-type growth habit in apple is characterized by a reduced number of vegetative shoots and a corresponding increase in the number of fruiting spurs, short distances between nodes, compact and dense growth, and dark green and relatively thick leaves (Lapins and Lapins, 1969). Spur-type apple trees exhibit a high rate of bud break but have weak branching, characterized by rosette foliage and overall reduced levels of vegetative growth. The anatomical structure of spur-type shoots is characterized by vessel elements that are shorter and narrower than in standard-type shoots. The spur-type vessel elements have longer ends (tails), and thicker secondary cell walls (Yang et al., 2000).

Several earlier studies have reported that spur-type apple trees have lower levels of GA than standard-type trees, are less sensitive to applications of GA, and have reduced rates of conversion of GA19 to GA20 under high temperature conditions in summer (Steffens and Hedden, 1992; Song et al., 2012). The majority of shoot elongation is the result of cell expansion, with gibberellins playing a dominant role in this process. GAs promote internode elongation and overall shoot elongation by regulating cell proliferation, differentiation, and expansion. The "green revolution," semi-dwarf varieties of cereal crops, are defective in GA biosynthesis and/or signaling. Auxin is recognized as the major hormone involved in shoot development (Gallavotti, 2013). Auxin also influences stem elongation and regulates the formation, activity, and fate of meristems. Cytokinins (CK) inhibit the differentiation of cells in the SAM and instead promotes cell division. CK promotes SAM enlargement partly through WUSCHEL, and defects in CK synthesis or perception reduces the size of the SAM size (Leibfried et al., 2006; Kyozuka, 2007). Except for GA, however, the levels of plant hormones in spur-type apples have not been reported or are unclear.

MicroRNAs (miRNAs) are a class of endogenous 20–24 nt non-coding RNAs. In plants, miRNAs are transcribed by RNA polymerase II, 5′ capped, spliced, and polyadenylated at the 3′ end to create pri-miRNAs. DICER-LIKE (DCL) protein processes the stem-loop structure in pri-miRNAs, creating pre-miRNA, and subsequently miRNA/miRNA<sup>∗</sup> duplexes. A single mature miRNAs is generated by the removal of the miRNA<sup>∗</sup> . The mature miRNA is recruited into a RISC complex (RISC) which degrades mRNA targets or suppresses their translation. Plant miRNAs play a critical role in growth, organ development, phase changes, senescence, fruit ripening, stress response, disease resistance, stress tolerance, and mineral assimilation and translocation.

Some miRNAs in plants have been shown to regulate plant architecture. A point mutation in OsSPL14 perturbs the cleavage of OsSPL14 transcripts by OsmiR156, which affects tiller number and lodging resistance, as and enhances grain yield (Jiao et al., 2010). OsmiR397 down regulates a laccase-like protein OsLAC involved in brassinosteroid sensitivity, and thereby promotes panicle branching (Zhang et al., 2013). OsMIR444 participates in a miRNA/MADS/TCP/D14 (miMTD) regulation module, in which D14 functions in strigolactone (SL) signaling to control tillering, by suppressing OsMADS57 expression (Guo et al., 2013). In pear trees, miR6390 is involved in the transition from endormancy to ecodormancy by targeting PpDAM transcripts, degrading them and stimulating the release of PpFT2 (Niu et al., 2015). Overexpression of the MbDRB1, a gene involved in miRNA biogenesis in Malus baccata, increased the level of adventitious rooting, curly leaf, and columnar-like architecture (You et al., 2014).

The molecular mechanisms responsible for spur-type architecture is unclear. Few studies have been conducted on the role of miRNAs in the architecture of fruit trees. In the present study, miRNA and degradome sequencing was used to identify miRNA involved in shoot development in Malus domestica "Nagafu 2" and spur bud mutation "Yanfu 6." Morphological indicies, plant hormone levels, miRNA, and target gene expression, and the application of various hormone treatments were also measured and/or used to compare and contrast shoot development in "Yanfu 6" vs. "Nagafu 2." This comprehensive analysis was conducted in order to identify and determine the potential role of miRNA-mediated regulatory network in shoot growth and involvement of miRNA in the architecture of fruit trees and other woody, perennial plants.

# MATERIALS AND METHODS

#### Plant Materials

Experiments were conducted in 2013 and 2014 using standardtype "Nagafu 2" and spur-type mutant "Yanfu 6" apple (Malus domestica) trees that were grafted on Malling "M.26" rootstocks and planted in 2009. Trees were located in the experimental plots of the Yangling National Apple Improvement Center, Yangling, China (34.31◦ N, 108.04◦ E), trained and managed using standard horticultural practices. The meteorological conditions of the experiment site in 2013 and 2014 are described in **Figure S1**. Three biological replicates consisting of five trees of each genotype were used, for a total of 15 trees.

Apical portions (5–8 mm below and including the shoot meristem) of YF and CF shoots (without any leaves or petiole tissue), were collected at 45 days after bud break (DABB) in 2013 and used for deep miRNA sequencing. Similar apical portions of shoots were also collected at 65, 85, 105, 125, and 145 DABB in 2013 for miRNA and mRNA expression analysis. Approximately 30 shoot samples of each biological replicate of YF and CF were collected. Leaves near the shoot tips were also collected for plant hormone measurements. All samples were immediately frozen in liquid nitrogen and stored at −80◦C for subsequent RNA and hormone extraction.

#### Measurement of Tree Growth

Six individual shoots on each of six trees were marked in each genotype to assess shoot elongation on a weekly basis in 2013 and 2014. The number of internodes present on each shoot were also recorded. Internode lengths were estimated by dividing the overall shoot length by the number of internodes.

#### Hormone Measurements

Approximately 0.5 g of leaf samples were used to measure IAA, GA, ZR, and ABA. The plant hormones were extracted using the method described by Dobrev and Kamınek (2002) and Dobrev et al. (2005), quantifed by the High Pressure Liquid Chromatography (Waters, USA) and enzyme-linked immunosorbent assay.

# GA Treatment

YF and CF trees were sprayed with 500 mg·L <sup>−</sup><sup>1</sup> GA4+<sup>7</sup> at the end of April 2014. Shoot tips, as previously described, were subsequently sampled at 0, 2, 4, 6, and 8 days after spraying. Both shoot length and internode length were also recorded.

#### Small RNA and Degradome Sequencing

Total RNA was isolated using the CTAB method (Chang et al., 1993). After ligating adaptors to the 5′ and 3′ RNA, the purified RNAs were reverse transcribed using primers with partial complementarity to the adaptors. The DNA pool amplified from the first-strand cDNA was subsequently sequenced using a Hiseq 2000 (Illumina, San Diego, CA) sequencer located at the Beijing Genomics Institute (BGI), Shenzhen.

The YF sample for miRNA sequencing was also subjected to degradome sequencing in order to better determine the miRNA/mRNA pairing relationship. For the construction of the degradome libraries, a 5′ RNA adapter with a Mme I recognition site was ligated to the 5′ terminus of purified polyA RNA. The 5′ ligated products were reverse transcribed using five PCR cycles, digested with Mme I, and ligated to a 3′ double DNA adapter. Lastly, the ligated products were amplified by PCR and gel-purified for deep sequencing.

The miRNA and degradome sequencing raw data were available at NCBI SRA (BioProject ID: PRJNA361540).

#### Bioinformatic Analysis of miRNA

The original sequence data from the sequencing machine were used for basic analysis. Filtering the low quality reads and contaminants sequence (adaptor, polyA, reads shorter than 18 nt, insert tag, reads without 3′ primer) and discarding the remaining reads with lengths smaller than 18 nt, the clean reads were used for the advanced analysis. The length distribution of the clean reads were summarized to analyze the compositions of small RNA sample. Unique tags and total tags existence among the sample were summarized to verify whether the uniformity of the two samples on the whole of sequencing was good.

The small RNA tags were mapped to the apple reference genome at GDR (http://www.rosaceae.org) (Jung et al., 2014) using SOAP software to analyze their distribution in the genome and their expression levels. All unique, clean reads, particularly the non-redundant reads, were subjected to further analysis. The clean tags were queried against the known apple miRNAs in the mirBASE 18.0 (http://www.mirbase.org) database in order to identify known microRNAs based on the following criteria: (1) the tags aligned to a miRNA precursor in the mirBASE database with no mismatch, and; (2) based on the first criteria, the tags aligned to mature miRNA in the mirBASE database with at least 16 nt overlap, allowing offsets. The unmatched unique reads in each sample were screened in Genbank in order to annotate non-coding RNA sequences (rRNA, scRNA, snoRNA, snRNA, and tRNA). Differentially expressed miRNAs were identified by comparing the sequencing reads between CF and YF and determining statistically significant differences (P < 0.05 and P < 0.01). miRNAs with more than one normalized read were analyzed by calculating their fold-change and p values. miRNAs with p < 0.05 and fold changes (log2) less than −1 or greater than 1 were considered to have significantly altered expression.

# Prediction of miRNA Targets and Functional Annotation

The psRNATarget server (http://plantgrn.noble.org/psRNATarget/) with default parameters was used to predict putative targets of the identified apple miRNAs. All of the identified apple miRNA were used as a query against "Malus domestica (apple), predicted consensus gene set CDS sequences, version 1" database located on the psRNATarget server.

For the degradome sequencing raw data about 50 nt length obtained from HiSeq sequencing, go through steps likely adaptor, pollution, low quality trimmed to obtain "clean" sequence. After a series of annotation of the degradated sequence, the degradation mRNA obtained go through comparison with miRNA to find mRNA-miRNA pairing.

### GO Enrichment and KEGG Pathway Analysis

GO enrichment and KEGG pathway analysis for all of the annotated miRNAs and their targets, using the apple reference genome background, were conducted in order to predict the function of the miRNA and their targets. Blast2GO was employed to store information from the GO and KEGG pathway databases. All of the miRNA targets were queried against the GO protein database (http://www.geneontology.org/) using BLASTX. A combined query was used in order to complete the GO annotation and pathway analysis against the GO and KEGG databases.

used for plant hormone IAA, GA, ZR, and ABA measurements. \*Indicates *P* < 0.05, Student's *t*-test; Error bars indicate SE.

# Quantitative Real-time PCR Validation of microRNA and mRNA Expression

Stem-loop RT-qPCR method was used to quantify miRNA expression (Chen et al., 2005). mRNAs of the target genes were also quantified. 5S rRNAs were used as an internal reference of miRNA expression. MdActin and MdEF were used as internal references for mRNA expression. The primers utilized are listed in **Tables S12**, **S13**. miRNA and mRNA relative expression were calculated using the comparative ∆∆CT method (Livak and Schmittgen, 2001).

# Statistical Analysis

The data were analyzed using Data Processing System (DPS, version 7.05; Zhejiang University, Hangzhou, China) software. Significant differences were determined using a Student's t-test at (p ≤ 0.05).


FIGURE 4 | Clustering of conserved and novel miRNAs into families.

Frontiers in Plant Science | www.frontiersin.org

#### RESULTS

#### Growth Characteristics of YF and CF

The shoot length in YF was significantly shorter than in CF (**Figures 1A,E**). YF also had shorter internodes, that were approximately 66% shorter than CF (**Figures 1C,D**). Most YF shoots stopped growing between 80 and 105 DABB, which was earlier than CF shoots. After 105 DABB, any continued growth of YF shoots extended their length with a section of shorter internodes.

#### Plant Hormone Analyses

IAA in YF leaves near the shoot tips was higher than CF leaves at 45 DABB. In contrast, IAA levels in YF were significantly lower than CF at 125 and 145 DABB (**Figure 2A**). The GA content in YF leaves remained at a consistently low level, relative to CF, throughout the entire growth period (**Figure 2B**). The ABA content in YF was significantly lower than CF at 65 and 85 DABB (**Figure 2C**). The ZR content in YF was significantly lower than CF at 45, 65, 125, and 145 DABB (**Figure 2D**). The changes in the levels of the growth promoting hormones, IAA, GA, and ZR, were in accordance with the growth of shoot, namely, exhibiting a high content in the level during the rapid growth period, and a low level as the growth period waned (**Figure 2**). The stress hormone ABA exhibited the opposite trend, exhibiting a low level during the active growth period and a high level as growth slowed and terminated.

#### Small RNA Sequencing Profiles in CF and YF

Two cDNA libraries of small RNAs from CF and YF were sequenced in order to identify miRNAs involved in apple shoot development. The sequencing produced a total of 13,264,874 raw reads from CF and 19,773,007 from YF (**Table S1**). After

removal of the adaptor sequences, filtering out low-quality tags, and omitting contamination resulting from adaptor–adaptor ligations, 13,083,017 (98.95% of the total) CF and 19,580,358 (99.41%) YF clean reads remained, comprising 5,872,194 and 7,680,257 CF and YF unique sequences, respectively (**Table S2**). The unique sequences were mapped to the apple genome assembly using SOAP, resulting in 3,674,249 (62.57%) and 4,473,261 (58.24%) genome-matched reads for CF and YF, respectively. A total of 143,629 (2.44%) and 278,103 (2.54%) exon RNAs (sense and antisense), 267,025 (5.08%) and 321,431(4.18%) intron RNAs (sense and antisense), 130,796 (2.23%) and 204,833 (2.67%) rRNAs, 1,973,617 (33.61%) and 2,370,059 (30.86%) repeat regions, 4733 (0.08%) and 6054 (0.08%) snRNAs, 1,238 (0.02%) and 1,451 (0.02%) snoRNAs, 11,364 (0.19%), and 15,596 (0.20%) tRNAs, were removed from the CF and YF reads, respectively, This left 3,338,736 (56.86%) and 4,564,590 (59.43%) unannotated CF and YF RNAs, respectively, and a total of 1,056 (0.02%) and 1,149 (0.01%) miRNA reads for CF and YF, respectively. These were utilized as miRNA candidates in subsequent analyses. The read lengths in the two small RNA libraries were assessed (**Figure 3A**).The lengths ranged of the designated small RNAs ranged from 21 to 24 nt and accounted for more than 90% (approximately 92.9% and 94.6% of the CF and YF sequences, respectively) of all the clean reads in each library. The 24 nt small RNA was the most abundant small RNA length (about 50%). Similar results were reported in apple leaves, flowers, fruit, and roots (Xia et al., 2012), peach (Prunus persica) (Zhu et al., 2012), pear (Pyrus bretschneideri) (Wu et al., 2014), tomato (Lycopersicon esculentum) (Gao et al., 2015), and chickpea (Jain et al., 2014). The next most abundant small RNAs were 21 nt > 22 nt > 23 nt. There was a greater number of 23 nt and 24 nt small RNA in CF than in YF. The

number of 21 and 22 nt small RNA were greater in YF than in CF.

The clean reads in each library were queried against the Malus domestica microRNA database in miRBase in order to identify the known miRNAs that were present in CF and YF apple shoot tips. Collectively, one million reads from the two libraries were perfectly matched to the Malus domestica microRNA database. A total of 202 unique sRNAs sequences were assigned to 41 microRNA families (**Table S3**, **Figure 3B**).

A comparison of family members revealed that the number of miRNA members varied among the different miRNA families (**Figure 3C**). The top five miRNA families, mdm-miR156, mdm-miR172, mdm-miR171, mdm-miR167, and mdm-miR399, had more than 10 members. For example, the mdm-miR156 family had 31 members. mdm-miR1511, mdm-miR391, mdmmiR7125, mdm-miR7126, mdm-miR827, and mdm-miR858 only had one member. miRNA156 is conserved in embryophytes (Zhang et al., 2006; Taylor et al., 2014) and represents the largest microRNA family in pear (Wu et al., 2014), which like apple, is a member of the Rosaceae. The differences in the number of members in the different microRNA families may be due to the different evolutionary rates in the microRNA families and to genome duplication events. Two known microRNAs families, mdm-miR7123, and mdm-miR7128, were not found to be present in the libraries (**Figure 3C**), therefore, these microRNAs may not be expressed in apple shoot tips. Statistical analysis of the read counts for the known miRNA families indicated significant differences in the level of expression among the different miRNAs (**Table S3**). Read counts of some miRNAs in a family were nearly the same, but some were obviously different. For example, all of the mdm-miR395 family members had the same number of counts. In contrast, mdm-miR172a, mdm-miR172o, and mdmmiR172g-h, members of the mdm-miR172 family (15 members), had a high read count, while mdm-miR172 a-c, mdm-miR172ik, and mdm-miR172 m-o had a moderate read count, and mdm-miR172l had a low count. Perhaps some of the microRNA clustered in the same area of the genome, and thus had the same cis-regulation elements.

The read counts of about 30% of the known miRNAs were greater than 1,000 (**Figure 3D**). Another 30% of the known miRNA had read counts between 100 and 1,000. Due to high level of expression of miRNA in the apple shoot tip tissue, these miRNAs could be accurately identified. Approximately 20% of the known miRNA had read counts between 0 and 20.

Among the miRNA in YF and CF shoot tips, mdmmiRNA166 families exhibited the highest level of expression, with approximately 10,000 standardized reads (**Figure 3E**). This was followed by mdm-miR482a-5p, which had about 4000 standardized reads. The juvenile to adult phase transition related microRNAs, mdm-miR156 and mdm-miR172, and the flowering related, mdm-miR535, mdm-miR168, and mdm-miR167, also showed high expression levels in apple shoot tips.

After the identification of the known miRNAs, the unannotated tags were analyzed using MIREAP software in order to identify novel miRNAs based on the characteristics of the secondary hairpin structure of miRNA precursors. Using the hairpin structures of the precursors, 498 novel miRNAs were identified. The 23 nt length novel miRNAs represented an overwhelming amount 76.8% of the 306 total novel miRNAs identified (**Table S4**).

The novel miRNA was used a query to blast the miRBASE 18.0. Results indicated that 11 of the 498 novel miRNAs possessed a high significant similarity to known miRNAs from rice, wheat, and other plant species (**Figure 4**). Some varied only in nucleotide length, which may be due to differential cleavage sites in the pre-miRNA. These miRNAs represent new members of conserved miRNA families, and as a result, were designated as novel miRNA isoforms. The novel miRNAs would be speciesspecific Malus domestica miRNA.

Lower negative minimal folding free energy (MFE) value for RNA secondary structure, indicates greater stability (Bonnet et al., 2004). The average negative MFE was −51.03 kcal/mol (**Table S4**). This MFE value is similar to the data for Arabidopsis thaliana miRNA precursors (−57 kcal/mol) (Debat and Ducasse, 2014), chickpea (−57.58 kcal/mol) (Jain et al., 2014), maize (−61.15 kcal/mol) (Li et al., 2013), and pear (−51.48 kcal/mol) (Wu et al., 2014).

The GC content of miRNA has been shown to be related to biological function (Mishra et al., 2009). The GC content of 90% of the novel miRNA ranged between 30 and 70% (**Table S4**) with an average value of 42%. This average GC is similar to Arabidopsis (44%), pear (45%), and chickpea (44%), but less than grape (50%) and the GC content of the known miRNA of apple (50%).

The U residue is the most prevalent nucleotide at the first 5′ nucleotide site of miRNA, especially 21 nt miRNA (Chen, 2005). About 46% of the 5′ end nucleotides of the 21 nt novel miRNAs were U (**Figure S2**).

#### Differentially Expressed miRNAs

The abundances of miRNA reads in the CF and YF libraries were compared in order to identify miRNAs involved in shoot development and to determine their expression pattern in the two different genotypes with contrasting phenotypes. A total of 42 miRNAs, belonging to 12 known miRNA families, were identified whose expression differed significantly between the CF and YF shoot tips (**Figure 5A**). Among these, only mdm-miR5225c and mdm-miR7125 were upregulated in YF. In contrast, five members of mdm-miR164, three members of mdm-miR171, six members of mdm-miR172, three members of mdm-miR393, nine members of mdm-miR395, two members of mdm-miR396, six members of mdm-miR399, two members of mdm-miR5225, two members of mdm-miR7124, and mdm-miR858 were all downregulated in YF shoot tips.

Among of the 498 novel miRNAs, the expression of 93 differed significantly between CF and YF. A total of 34 upregulated and 59 were downregulated in YF (**Figure 5B**). A total of 49 novel miRNAs were expressed only in CF while 28 novel miRNAs were expressed only in YF (**Table S4**).

#### Target Prediction of miRNAs

Plant miRNAs pair with their complementary target and inhibit translation or degrade the target mRNA. The complementary pairing of the miRNA/mRNA duplex provides a useful approach

#### TABLE 1 | Identification potential target genes of known miRNA by the degradome sequencing.


*(Continued)*

#### TABLE 1 | Continued


*(Continued)*

#### TABLE 1 | Continued


for target prediction. In order to better understand the role of apple miRNAs in shoot development, psRNATarget software was used to predict potential miRNA target genes. Using previously described criteria, 689 potential targets for 168 of the known miRNAs, and 3,663 targets for the 466 novel miRNAs were predicted (**Tables S5**, **S6**). The number of predicted targets for the various miRNAs varied from one to as many as 20, suggesting that these miRNAs may have diverse biological functions. On the other hand, a single gene may also be targeted by several miRNAs. MDP0000237964 encodes a LAC gene that is targeted by mdmmiR397 at two different sites, and two similar CuRO\_LCC\_plant domains (**Figure 6A**). miRNA828 and miRNA858 target the MYB family at the R3 domian (Xia et al., 2012; **Figure 6B**). MDP0000147309 is likely to be targeted by mdm-mi159 and mdm-mi319 (**Figure 6C**).

Degradome sequencing was conducted in order to better understand the role of miRNA-target regulation in shoot tips. Approximately 23 million total clean reads were generated from the degradome sequencing. About 49.5% of the unique reads mapped to apple mRNA (**Table S7**). The degradome sequencing results indicated that 246 genes are potentially regulated by 164 miRNAs, including both known and novel miRNAs. The main miRNAs and their target genes verified by degradome sequencing are listed in **Tables 1**, **2**. Two-thirds of the degradome validated known miRNA target genes also identified in the target gene prediction software.

#### GO and KEGG Analysis of the Degradome-Predicted Target Genes in Shoot Tips

BLASTX querying against the protein database, Gene Ontology (GO), and KEGG pathway analysis were used to annotate the target genes identified in the degradome analysis in order to determine the potential functions of the miRNA target genes. A total of 69 degradome-identified target genes of 21 known miRNA were associated with 238 GO terms (**Table S8**). The first three enrichment GOs were GO:0005634, GO:0003700, and GO:0006351. GO:0005634, involved with the nucleus in the cellular component had 49 associated target genes. GO:0003700 involved in transcription factor activity in the molecular function component had 45 associated target genes. GO:0006351, involved in transcription, DNA-dependent in the biological process component had 41 associated target genes.

KEGG pathway analysis indicated that the target genes were involved in 21 pathways, in which, Transcription factors, ko03000; Metabolic pathways, ko01100, and; Plant hormone signal transduction, ko04075 were significantly enriched (**Table S9**).

Twenty-nine of the degradome identified genes targeted by 15 of the novel miRNAs were assigned to 176 GO terms (**Table S10**). Like the known miRNA, the GO enrichment in novel miRNA were transcription regulation related GO terms: 13 genes in GO:0005634; 9 genes in GO:0006355; 9 genes in GO:0003677, and; 7 genes in GO:0003700. KEGG pathway analysis indicated that the target genes were involved in 35 pathways, in which, Transcription factors, ko03000, and Metabolic pathways: ko01100 were significantly enriched. (**Table S11**).

YF exhibited a shorter duration of shoot elongation (**Figure 1B**). The duration of shoot growth is regulated by the shoot apical meristem (SAM). Based on the GO annotation of the target genes and with the KEGG pathway analysis, the potential targets of miR164, miR166, miR171, miR172, and miR482 are involved in meristem development (**Table 3**). In

#### TABLE 2 | Identification target of novel miRNA by the degradome sequencing.


*(Continued)*

#### TABLE 2 | Continued


particular, miR164 participates in the formation of meristem-toorgan boundaries in the SAM (Laufs et al., 2004). The potential targets of miR166, miR159, and miR156 are involved in meristem phase transition. The potential targets of miR167, miR160, and miR159 play a role in auxin response and the auxin signaling pathway may also participate in SAM development.

Internode length is determined by cell division and cell elongation. Based on the GO annotation of the target genes and the KEGG pathway analysis, miRNAs related to cell division and cell differentiation were identified (**Table 4**). The potential targets of miR159, miR166, miR167, miR171, miR172, miR393, miR858, and miR828 are involved in cell growth. Gibberellin also plays an important role in regulating internode length. The potential targets of miR159 are involved in the gibberellic acid mediated signaling pathway and gibberellin biosynthetic process. While the potential targets of miRNA160 and miRNA167 regulate auxin response to control cell division. Thus, the potential targets of miR167, involved in response to brassinosteroid stimulus and the auxin-mediated signaling pathway, may play a role in internode development. Secondary metabolism affects primary growth by competing for common substrates. The potential targets of miR858, miR828, and miR156 participate in the fatty acid biosynthetic process. Cell walls also affect cell elongation and cell size. The potential targets of miR166 and miR167 are involved in cell wall biogenesis and cell wall organization. Mineral nutrition affects plant physiology and growth and the potential targets of miR169, miR166, miR7125, miR393, and miR395 are involved in mineral element response.

#### RT-qPCR of Differentially Expressed miRNAs and Their Targets Involved in SAM Development

The miRNAs and their targets putatively involved in SAM development that were differentially expressed in YF vs. CF are illustrated in **Figure 7**. The expression analysis of miRNAs and their associated target genes revealed inversed expression patterns. The expression of miRNA164 decreased from 45 to 145 DABB in both YF and CF while expression of the potential target gene, MdNAC, increased from 45 to 55 DABB, decreased from 85 to 105 DABB, and then increased rapidly after 125 DABB. The expression of miRNA164 in YF was significantly lower than in CF at 45, 65, and 85 DABB. The expression of miRNA166 in CF and YF peaked at 105 DABB. The expression of miRNA166 in YF was significantly higher than in CF at 45, 105, and 125 DABB. The expression of miRNA171 in CF and YF increased from 45 to 65, decreased from 85 to 125, and then increased to a high level at 145 DABB. The potential target of miRNA171, MdHAM, fluctuated in its expression level from 45 to 105 DABB, and was then downregulated after 125 DABB. The expression of miRNA171 in YF was significantly lower than CF from 65 to 145 DABB.

#### TABLE 3 | SAM development relative miRNAs.


*(Continued)*

TABLE 3 | Continued


#### RT-qPCR of Differentially Expressed miRNAs and Their Targets Involved in Internode Elongation

MiRNAs and their targets, which are putatively involved in internode elongation, that were differentially expressed in YF vs. CF are shown in **Figure 8**. The expression of miRNA159 in YF was significantly lower at 65, 85, and 125 DABB than it was in CF. The potential target, MdMYB65, was downregulated from 45 to 145 DABB. The highest expression level of miRNA167 in both CF and YF was observed during the slow growth period at 80 DABB. The potential target, MdARF6, was downregulated from 45 to 125 DABB, and upregulated at 145 DABB. The expression of miRNA167 in YF was significantly higher than in CF at 65, 85, and 105 DABB. The expression of miRNA396 in YF peaked at 105 DABB. The expression of miRNA396 in YF was significantly higher than in CF at 105, 125, and 145 DABB. The miRNA396 potential target gene, MdGRF, exhibited a high level during the later growth stage. The expression of miRNA159, miRNA167, and miRNA396 in YF and CF shoot tips exhibited their highest level of expression during the period of slow shoot growth. These data suggest that these miRNAs inhibited cell division in developing shoots.

### Expression of Cell Cycle- and Cell Elongation-Related Genes

A strong correlation has been observed between final internode length and cell number (Brown and Sommer, 1992). Therefore, the expression of cell cycle and cell elongation related genes were examined (**Figure 9**). The expression of the cell cycle related genes, MdCYCD, MdCYCB1.2, MdCYCD1.3, and MdRBR1 exhibited high levels of transcript abundance during the period of rapid growth after bud break, and were downregulated during the period of slow growth, followed by a slight upregulation at the later stages of sampling. MdCYCD expression in YF was significantly lower than in CF from 45 to 125 DABB. Expression levels of MdCYCB1;2, MdCYCD1;3, and MdRBR1 genes in YF were significantly lower than in CF during the period of rapid growth just after bud break. The cell elongation genes, MdTCH4 and MdXTH23, exhibited high expression levels during the period of slow growth and in the autumn. MdTCH2 and MdXTH23 expression in YF exhibited a lower expression level than in CF during these periods.

### Expression of Hormone-Related Genes

Based on the observed effect of plant hormones on growth, the expression of hormone-related genes was also examined (**Figure 10**). The expression of the GA2ox gene, which catalyzes the catabolism of bioactive GA or their precursors, was significantly higher in YF than in CF. In addition, the GA signal transduction gene, MdRGA, which negatively regulates GA response, was also more highly expressed in YF than in CF, especially at 80 and 105 DABB. The expression of the cytokinin synthesis gene, MdIPT9, was significantly lower in YF than in CF during the period of rapid growth at 45, 65, 125, and 145 DABB. The MdCKX5 gene, which is involved in the breakdown and catabolism of cytokinin, was upregulated to a higher level in YF than in CF at 45 and 65 DABB. The expression of the ABA synthesis gene, MdNCED3, was significantly lower in YF than in CF at 65 and 85 DABB, but higher in YF than in CF at 105 DABB.

#### Effect of GA on Shoot Growth and miRNA Expression

Auxin and GA4+<sup>7</sup> was applied to YF and CF trees. Results indicated that the application of GA promoted shoot growth and internode elongation in YF trees (**Figure 11A**). The terminal shoot started to grow rapidly 1 week after the GA treatment was applied (**Figure 11B**). In total, internode length in YF trees increased by about 33% after the GA treatment, however, this was still less than the internode length in CF trees (**Figure 11A**). The auxin application did not stimulate shoot growth in either YF or CF trees (data not shown).

The GA treatment dramatically downregulated miRNA166 expression (**Figure 12**). The potential target of miRNA166,

#### TABLE 4 | Internode growth related miRNAs.


*(Continued)*

#### TABLE 4 | Continued


FIGURE 7 | Differentially expressed miRNAs and their targets involved in SAM development between CF and YF. \*Indicates *P* < 0.05, Student's *t*-test; Error bars indicate SE.

MdHD-ZIP, was also downregulated during the first 2 days after the GA treatment. After the 6th day, however, MdHD-ZIP was upregulated. During the first 4 days after the GA treatment, the expression of miRNA164 in CF and YF trees was slightly upregulated. Then miRNA164 was downregulated in CF, but upregulated in YF. The expression of miRNA171 in YF trees exhibited little response to the GA treatment. The expression of miRNA171 in CF trees was upregulated on the 8th day after the application of the GA treatment. The expression of miRNA159 in both YF and CF trees showed little response to the GA treatment during the first 6 days, however, on the 8th day, the miRNA159 was upregulated in CF trees but downregulated in YF trees. Relative to the expression levels in CF trees, miRNA396 expression in YF trees slowly went down after the application of the GA treatment. The expression of miRNA167 exhibited little response during the first 2 days after the GA application but was downregulated on the 6th day, and then subsequently upregulated.

#### DISCUSSION

YF is a natural spur mutant. YF trees have short internodes and reduced shoot length. Whether or not spur-type apple trees are also a fruiting characteristics, has been a source of debate, especially since the spur type phenotype of siblings derived from the spur-type apple exhibiting great variability (Fideghelli et al., 2003). Regardless, spur-type apple cultivars undoubtedly moderately reduced tree size as compared to standard-type trees (Costes et al., 2010). Unlike dwarfing rootstocks that reduce the proportion of lateral buds that develop into long shoots during the early development of trees (Lauri and Trottier, 2006; Seleznyova et al., 2008), YF produces a dwarf tree architecture by reducing the duration of shoot vegetative growth and shortening internode length.

In the present study, YF trees were demonstrated to contain lower levels of GA, ZR, IAA, and ABA. An application of GA promoted cell division in the shoot apical meristem and an increase in internode elongation in YF trees (**Figures 11A,B**). In contrast, YF trees exhibited no response to IAA. Previous studies have reported that the application of 6-benzylaminopurine and GA to apple nursery trees increased shoot elongation to a much greater extent than applying 6-benzylaminopurine alone (Zhang et al., 2011; Doric et al., 2015). Hence, among the hormones, GA, ZR, IAA, and ABA, it appears that GA plays a major role in YF trees.

The miRNA sequencing conducted in the present study confirmed the presence of 202 known miRNAs expression in apple shoot tips and discovered an additional 498 novel miRNAs. A total of 42 known miRNAs and 93 novel miRNAs were differentially expressed in the shoot tips of CF and YF trees. The differential expression of the miRNAs appeared to be related to the differences in shoot development observed in the two genotypes. This supports the premise that miRNAs are involved in the regulation of shoot architecture in YF trees.

Internode length and the number of nodes are two essential components that determine plant height. Compared to internode length, however, the impact of shoot apical meristem activity

has a greater effect on shoot development. For example, the shoot apical meristem is responsible for terminating cell division when trees enter dormancy or when the vegetative meristem is converted to a floral meristem. The number of nodes is also determined by continued cell divisions in the shoot apical meristem. Therefore, factors affecting SAM development would also affect the number of nodes produced on a shoot during an annual growth cycle. Our results indicate that miRNA164, miRNA166, and miRNA171 are potentially involved in SAM development and differentially expressed in YF vs. CF apple shoot tips. A degradation of the NAC-domain transcription factors, CUC1 and CUC2 (Mallory et al., 2004; Raman et al., 2008) by miRNA164 has been reported to constrain the expansion of the boundary domain by degrading the NAC-domain transcription

factors, CUC1 and CUC2 (Mallory et al., 2004; Raman et al., 2008). In Arabidopsis, CUC1, CUC2, and CUC3 function redundantly in initiating SAM and establishing organ boundaries (Aida et al., 1997). Abolishing miR164 regulation of CUC2 resulted in progressive enlargement of the boundary domain (Raman et al., 2008). The shoot tips in YF trees exhibited abnormal shoot tips, with the dome of the apex appearing a little larger than the apical dome in CF trees. This phenotype is also present in apples with columnar architecture and in some dwarfing rootstocks (Petersen and Krost, 2013). The mir164abc mutant frequently displayed extremely short internodes, small floral organs, and reduced fertility. Thus, miRNA164 may be involved in abnormal shoot and internode development in YF trees.

Among all of the miRNAs, miRNA166 had the highest level of expression in the shoot tips. miRNA166 potentially inhibited SAM development by targeting a set of HD-ZIP transcription factors that are important for maintaining pluripotency in shoot apical meristems (Williams et al., 2005). AGO10 binds miRNA165/166 to prevent its association with the broadly expressed AGO1 and thus prevents the degradation of HD-ZIP transcription factors in the SAM (Zhou et al., 2015). YF trees have a high proportion of spur shoots. Many of the lateral shoots in YF trees stopped elongating soon after bud break. Thus, miRNA166 may be involved in apple regulating the spur shoot habit in apple trees.

miRNA171 potentially targets the HAM subfamily genes, SCL6-II, SCL6-III, and SCL6-IV, within the GRAS gene family. miRNA171 affects the maintenance of SAM indeterminacy by regulating the shoot apical meristem WUS-CLV feedback loop (Schulze et al., 2010). The expression level of miRNA171 was significantly lower in YF trees than in CF trees. The height of Arabidopsis plants overexpressing miR171c was significantly higher than wild-type plants (Long et al., 2010). Upregulation of OsMIR171c in a delayed heading mutant rice genotype exhibited pleiotropic phenotypic defects, including a prolonged vegetative phase and a delayed heading date (Fan et al., 2015). These observations correspond to the reduced growth in YF shoots,

relative to shoot growth in CF trees, and the earlier date in which growth ceased in YF trees, relative to CF trees.

As a result, it appears that the final internode phenotype results from an interplay between cell division and elongation that progresses in an inverse directions. Cell number has a more predominant role in regulating the growth and development of internodes because of cell length increases only two to three times, but cell numbers increase by 10 to 30 times (Pallardy, 2008). Analysis of the expression pattern of cell cycle genes further supports the premise that the SAM in YF trees has a lower rate of cell division than in CF trees (**Figure 12**). The expression of MdCYCD, MdCYCB1;2, MdCYCD1;3, and MdRBR1 was significantly lower in YF shoot tips than in CF shoot tips. The period of high expression of the cell cycle genes corresponded to the period of rapid shoot growth; suggesting that cell division may play an instrumental role in shoot growth, and more so than cell elongation. Ripetti et al. (2008) also found that genetic variation in internode length could be primarily attributed to cell numbers, while cell length played more of a secondary role. The miRNA sequencing and GO annotation in the present study identified several miRNAs related to cell division.

miRNA159 potentially deregulates its target genes, MYB33 and MYB65, in vegetative tissues, and inhibit growth by reducing cell proliferation (Alonso-Peral et al., 2010). The expression of miRNA159 was significantly higher in CF shoot tips than in YF shoot tips. The miRNA159 loss of function mutant, mir159ab, has an enlarged shoot apical meristem and smaller lateral organs (Alonso-Peral et al., 2010). miRNA159 and its targets have also been reported to be involved in GA-mediated flowering (Gocal et al., 2001; Achard et al., 2004). The expression of miRNA159 in YF shoot tips exhibited little response to the application of GA during the first 6 days after spraying GA. Interestingly, miR159a and miR159b also remained unchanged after the application of GA in wild-type Arabidopsis (Alonso-Peral et al., 2010). The potential target gene of miR159a, MYB65, however, was downregulated after the application of GA in Malus domestica. Thus, GA maybe regulate the expression of the target of miRNA159 to produce the miRNA159 phenotype.

miRNA396 negatively regulates its potential target gene, GROWTH RESPONDING FACTOR (GRF), to affect cell division in shoot meristems, leaves, and roots. miRNA396 appears to act as a strict repressor of the GA and CK pathways, and negatively regulates cell proliferation by decreasing cell division activity and the expression of cell cycle-related genes (Wang et al., 2010). The potential downregulation of the GA pathway genes by miR396 corresponds to the dwarf phenotype observed in miR396 overexpressing lines (Liu et al., 2009). The high level of expression of miRNA396 in YF shoot tips may inhibit cell division and internode elongation. miRNA396 was downregulated after GA application. Thus, GA may affect downstream GA response by downregulating miRNA396.

miRNA167 regulates vegetative and reproductive growth by controlling the expression of auxin response factors 6 and 8 (ARF6/8) in plants. Overexpression of miRNA167a results in twisted leaves, short inflorescences, and arrested flower development; thereby fully producing the mature phenotype of arf6arf8 plants (Wu et al., 2006). The expression of miRNA167 was significantly higher in YF shoot tips than in CF shoot tips. GA treatment downregulated miRNA167 at first, and then subsequently induced it; resulting in upregulating. But GA also increased the expression of the target gene, ARF6. DELLA has been reported to interact with ARF6 and inhibit ARF6 DNA-binding to modulate ARF6 gene expression (Oh et al., 2013). Hence, GA may promote cell elongation mechanisms in YF trees that involve miRNA167-ARF-mediated responses.

Based upon the findings of the current study, we present a model to explain how plant hormones and miRNA are involved in regulating the shoot development in spur-type YF trees (**Figure 13**). Low levels of IAA, GA, and CK are associated with reduced shoot growth in YF trees. GA appears to play a major role in regulating shoot development. Low levels of GA in YF shoots may reduce the number of cell divisions in the SAM, relative to the SAM in CF shoots, thus causing shoot growth to cease much earlier in the YF trees as compared to in the CF trees. In the model, low expression of miRNA164, miRNA171, and high expression of miRNA166 and their potential target genes affected SAM development and reduced the duration of shoot growth

in YF trees. Of the three miRNAs, the regulation of miRNA166 by GA appeared to play a major role in shoot tips elongation in YF trees. The high levels of expression of miRNA167 and miRNA396, and the low expression of miRNA159, in YF trees appears to have inhibited cell division and reduced internode length. The results of the present study, and the hypothetical model presented, contributes to a better understanding of shoot development in perennial plants; especially those that exhibit a spur-type growth architecture.

#### AUTHOR CONTRIBUTIONS

CS, DZ, JM, and MH participate in the experimental design. CS, JZ, LZ, and WL participate in material sampling, field measurements and the laboratory data measurement. CS, BZ, YL, and GL participate in the laboratory data measurement. CS, DZ, JM, and MH participate in the paper writing and discussion. All authors reviewed the manuscript.

#### ACKNOWLEDGMENTS

This study was sponsored by the National Apple Industry Technology System of Agriculture Ministry of China (CARS-28), Science and Technology Innovative Engineering Project in Shaanxi province of China (2015NY114), Yangling Subsidiary Center Project of National Apple Improvement Center, Collaborative Innovation of Center Shaanxi Fruit Industry Development, Evaluation and Screening of Apple Rootstock and Scion-rootstock combination in Different

#### REFERENCES


Areas of Weibei Region (K3310216040) and Science and Technology Co-ordination Innovation Project of Shaanxi Province (2016KTZDNY01-10).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2017. 00441/full#supplementary-material

Figure S1 | The meteorological condition of the experiment site.

Figure S2 | Size distribution of novel miRNAs and the identity of the first nucleotide.

Table S1 | Quality evaluation of CF and YF sequencing data.

Table S2 | Distribution of small RNAs among different categories in CF and YF libraries.

Table S3 | Character of known miRNA.

Table S4 | Character of novel miRNA.


Table S7 | Distribution of different RNAs categories in degradome library.

Table S8 | GO functional analysis of identified targets of known miRNAs.

Table S9 | Known miRNA KEGG pathway analysis.

Table S10 | GO functional analysis of identified targets of novel miRNAs.


Table S13 | Primer of mRNA for real time Q-PCR.


Pallardy, G. S. (2008). Physiology of Woody Plants. San Diego, CA: Elsevier Inc.


zone descendants in Arabidopsis shoot meristems. Plant J. 64, 668–678. doi: 10.1111/j.1365-313X.2010.04359.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 Song, Zhang, Zheng, Zhang, Zhang, Luo, Li, Li, Ma and Han. 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.

# Single-Molecule Fluorescence Methods to Study Plant Hormone Signal Transduction Pathways

Song Song, Jian Chang, Chongjun Ma and Yan-Wen Tan\*

State Key Laboratory of Surface Physics, Department of Physics, Fudan University, Shanghai, China

Plant-hormone-initiated signaling pathways are extremely vital for plant growth, differentiation, development, and adaptation to environmental stresses. Hormonal perception by receptors induces downstream signal transduction mechanisms that lead to plant responses. However, conventional techniques—such as genetics, biochemistry, and physiology methods—that are applied to elucidate these signaling pathways can only provide qualitative or ensemble-averaged quantitative results, and the intrinsic molecular mechanisms remain unclear. The present study developed novel methodologies based on in vitro single-molecule fluorescence assays to elucidate the complete and detailed mechanisms of plant hormone signal transduction pathways. The proposed methods are based on multicolor total internal reflection fluorescence microscopy and a flow cell model for gas environment control. The methods validate the effectiveness of single-molecule approaches for the extraction of abundant information, including oligomerization, specific gas dependence, and the interaction kinetics of different components.

Keywords: plant signaling pathway, plant hormone, intrinsic molecular mechanism, in vitro, single-molecule fluorescence

# INTRODUCTION

Plant hormones are naturally occurring organic substances that influence physiological processes at low concentrations (Davies, 1995). In addition to the five classical phytohormones, namely auxins, abscisic acid (ABA), cytokinin (CK), gibberellins (GAs), and ethylene, additional compounds have been recognized as hormones, including brassinosteroids (BRs), jasmonate, salicylic acid, nitric oxide, and strigolactones (Santner and Estelle, 2009). These hormones govern every aspect of plant biological processes, including developmental processes, signaling networks, and responses to biotic and abiotic stresses (Bari and Jones, 2009). Auxins regulate diverse processes, such as cell enlargement, cell division, tropic responses, root initiation, and vascular tissue differentiation (Woodward and Bartel, 2005). ABA modulates stomatal closure, inhibits shoot growth, and induces storage protein synthesis in seeds (Davies, 2010). GAs regulate stem growth, fruit setting, and fruit growth and induce seed germination (Davies, 2010). Different hormones regulate distinct members of protein families; however, they may also share components and modulate similar processes (Nemhauser et al., 2006).

The past decades have witnessed the remarkable progress of plant hormone research. Molecular models of plant hormone action, including receptors and their downstream signal transduction components, have been identified and studied using genetic, analytical, biochemical, and physiological approaches. The complete sequencing of the Arabidopsis genome, the systematic identification of knockout mutations in CK response, and genetic crosses have revealed the

#### Edited by:

Chi-Kuang Wen, Institute of Plant Physiology and Ecology, Shanghai Institutes for Biological Sciences (CAS), China

#### Reviewed by:

Yuda Fang, Institute of Plant Physiology and Ecology, Shanghai Institutes for Biological Sciences (CAS), China Caren Chang, University of Maryland, College Park, United States Georg Groth, Heinrich Heine Universität Düsseldorf, Germany

> \*Correspondence: Yan-Wen Tan ywtan@fudan.edu.cn

#### Specialty section:

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

Received: 29 June 2017 Accepted: 18 October 2017 Published: 02 November 2017

#### Citation:

Song S, Chang J, Ma C and Tan Y-W (2017) Single-Molecule Fluorescence Methods to Study Plant Hormone Signal Transduction Pathways. Front. Plant Sci. 8:1888. doi: 10.3389/fpls.2017.01888

**373**

functions of the CK family (Ferreira and Kieber, 2005). Mass spectrometry combined with liquid chromatography (LC-MS) or gas chromatography (GC-MS) can be used for the separation, quantification, and analysis of small compounds such as ABA (Müller and Munnébosch, 2011), auxins (Barkawi et al., 2010), and BRs (Shigeta et al., 2011). Plant hormones are usually transduced by protein–protein interactions. NRT1/PTR FAMILY proteins are the transporters of nitrate, ABA, GAs, and auxins. Modified yeast two-hybrid (Y2H) systems with NRT1/PTR FAMILY have been developed to investigate GA and ABA transport activities (Chiba et al., 2015). By using a pulldown assay coupled with immunoaffinity chromatography, a direct functional association between ABA signaling and RNA processing was established through the interactions of ABA, FCA, and FY with 3′ -end RNA processing machinery (Razem et al., 2006). Isothermal titration calorimetry quantified the affinity and kinetics of peptide hormone iminodiacetate and its signaling compounds (Santiago et al., 2016).

Plant hormone signaling pathways are complicated. Although GC- and LC-MS are the preferred methods for the quantitative analysis of plant hormones, they are unsuitable for unstable and highly polar molecules (Metz et al., 2007). Coimmunoprecipitation assay and Western blotting are usually used to analyze moderately stable and strong protein–protein interactions (Persani et al., 2007). However, specific antibodies or protein tags are required, and artefactual aggregation may trigger false positive signals. Similarly, Y2H systems can screen stable, strong, and direct protein–protein interactions (Brückner et al., 2009). Isothermal titration calorimetry has acceptable sensitivity for interaction and thermodynamic studies and kinetic measurements (Rajarathnam and Rösgen, 2014); however, large quantities of proteins are required. In addition, transient interactions involving low-affinity proteins or an interaction complex with more than two protein components cannot be detected. Furthermore, these methods cannot be used for conducting time-course studies on plant hormones and elucidating the dynamic interactions of signaling components.

Single-molecule fluorescence methods have provided many new insights into the biological processes dominated by macromolecules. In contrast to ensemble-averaged measurements, single-molecule measurements not only describe the real-time conformational dynamics of individual molecules but also identify details of protein–protein or protein–environment interactions that are undetected in ensemble-averaged experiments. For example, studies on the G protein-coupled receptors of the mammalian signal transduction pathway have benefited from various single-molecule methods, facilitating the demonstration of the oligomerization state and mobility dynamics of G protein-coupled receptors within the cell membrane and observation of the conformational transitions of the distinct ligand binding–related states. This crucial information is buried in ensemble-averaged measurement and cannot be effectively extracted from traditional experimental data (Tian et al., 2017). Similar to mammalian G proteincoupled receptors, plant hormone–related receptors in signaling pathways have also attracted increased attention; however, single-molecule experiments were found to be relatively challenging because of the thick plant cell walls and the lack of appropriate imaging techniques (Wan et al., 2011). Currently, single molecule–based quantification methods for plant hormone–related receptors are emerging rapidly (Wang et al., 2015; Bücherl et al., 2017).

Previous studies have mainly focused on plasma membrane receptors, and several details of the complete signaling pathway inside living cells, particularly regarding the downstream signal transduction components, remain unknown (Clouse, 2011). Alternatively, in vitro single-molecule methods are feasible for identifying the key points in signaling pathways because in vitro experiments suppress the long-distance transportation of signal transduction components between different organelles in living cells and restrict the target motion range near the slide surface. The most common in vitro single-molecule assays are usually performed using conventional total internal reflection fluorescence microscopy (TIRFM).

Various single-molecule methods have been developed for the detection of different fluorescent observables. The oligomerization state of functional molecules and stoichiometry of different components within a complex can be quantified using the stepwise photobleaching of fluorescence trajectories (Ulbrich and Isacoff, 2007; Jiang et al., 2011). When this technique is combined with super-resolution microscopy, a single fluorophore can be localized with a precision of down to 10 nm, which can determine the spatial alignment of different subcomponents within one dense structure and the true size of the fine structures blurred by the diffraction limit (Hell and Wichmann, 1994; Betzig et al., 2006; Rust et al., 2006).

To ascertain intramolecular conformational changes or intermolecular interactions, single-molecule Förster resonance energy transfer (smFRET) (Förster, 1948; Clegg, 1992) may be a suitable method. Energy transfer efficiency represents the distance between FRET donor and acceptor, resulting in a smFRET resolution of 3–10 nm, which is even less than the spatial resolution achieved using super-resolution microcopy (Ha et al., 1996). Alternating laser excitation (ALEX) is an auxiliary technique used to distinguish whether low smFRET efficiency is caused by a large distance or a missing acceptor. In principle, ALEX uses two excitations to alternately excite the donor and acceptor. Three observables, namely the respective emissions of two fluorophores and the smFRET signal, yield the stoichiometry ratio and FRET efficiency simultaneously (Kapanidis et al., 2004). In plant hormone signal transduction pathways, the membrane or transmembrane receptors exhibit specificities and diversities. The receptors are the portals of these pathways and play a crucial role in distinguishing different hormones and transducing signals into the cell. However, the exact mechanisms through which receptors recognize their ligands and subsequently switch on or off their functions are under investigation. Furthermore, the mechanism mediating the conformational changes in receptors after associating with plant hormones has remained unknown. By detecting nanoscale conformational changes as well as ligand binding events, the smFRET dynamics of macromolecular machines can be revealed (Juette et al., 2014).

Macromolecules in the natural state are always in motion. Some studies have suggested that the experimental outcomes of immobilization procedures or single-molecule trapping, such as anti-Brownian electrokinetic (ABEL) trap (Cohen and Moerner, 2006; Fields and Cohen, 2011), may not represent the true behaviors of individual molecules. To avoid any unfavorable effects, the characteristics of freely diffusing molecules are being measured. Real-time single-particle tracking with widefield fluorescence microscopy (Saxton and Jacobson, 1997) is the most commonly used method for characterizing diffusion. Diffusion parameters can be obtained by analyzing tracked molecules within an adjacent image frame series. Fluorescence correlation spectroscopy (FCS) with confocal microscopy has also been frequently used for diffusion measurements (Haustein and Schwille, 2007). It allows the correlation analysis of fluorescence intensity fluctuations, which may be derived from many fluorophore factors within the detection volume, including diffusion and concentration. Therefore, experimental data analysis can provide the diffusion coefficients of individual tagged molecules.

In this paper, we describe novel single-molecule fluorescence methods for exploring plant hormone signal transduction pathways. The paper explains the experimental design and data analysis and validates the method in several manners: using the stoichiometry of a light-dependent plant signaling protein, through a test of the fluorescent protein's fluorescence state under different oxygen environments, and identifying the competitive association with substrates among proteins in the BR signaling system.

## MATERIALS AND METHODS

#### Chemicals and Reagents

For plasmid construction, universal or specific primers were obtained from Sunny Biotechnology Company (China). DNA polymerase KOD Plus-Neo was purchased from TOYOBO (Japan). DNA restriction enzyme, T4 ligase, and DNA marker were purchased from Takara (China). Top10 competent cells were purchased from TianGen (China). Agar powder (>99.5%, AR) and 50× TAE buffer (>99.5%, AR) were purchased from Sangon (China). Plasmid miniprep, DNA gel extraction, and polymerase chain reaction clean kits were purchased from Axygen (China). To enable protein expression and purification, tryptone (>99.5%, AR) and yeast extract (>99.5%, AR) were purchased from BBI (China). Escherichia coli BL21 (DE3)-PLysS and E. coli BL21-CodonPlus (DE3)- RIPL competent cells were purchased from 2nd lab (China). Antibiotics, such as ampicillin, were purchased from Solarbio (China). Affinity and Mono-Q columns were purchased from GE health (Sweden). Organic dyes, such as Alexa Fluor 555 C2 Maleimide, were purchased from Invitrogen (A20346, America). SulfoLink coupling resin was purchased from Thermo Fisher Scientific (20401, America). For slide cleaning and passivation, Na2S2O4, concentrated sulfuric acid, 30% H2O2, and acetone with guaranteed grades (>99.9%) were purchased from SCRC (China). 3-Aminopropyl-triethoxysilane (APES) was purchased from Aladdin (A107147, China), and Methoxy poly(ethylene glycol) succinimidyl carboxymethyl ester (mPEG-SCM) and biotin p-(ethylene glycol) succinimidyl carboxymethyl ester (biotin-PEG-SCM) were purchased from Biomarik (5675-5K and 5472-4K, respectively, China). Streptavidin was purchased from Amresco (E497, America), and the anti-His-tag antibody was obtained from Rockland (600-406-382, America). For flow cell construction, cover slips (25 mm × 25 mm, Fisher) and a 0.3-mm dual adhesive tape (3M, 467) were used. Poly(dimethylsiloxane) (PDMS) was obtained from Dow Corning (Sylgard-184, America). BODIPY-FL-ATP was purchased from Thermo Fisher Scientific (A12410, America).

#### Microscopy

The single-molecule methods are based on TIRFM. We used a homebuilt TIRFM that was based on a previously reported design (Friedman and Gelles, 2012) for multicolor imaging. Two excitation lasers (wavelengths 488 and 532 nm) shaped by a beam-expander, dichroic, doublet lens are focused on the back focal plane of the objective (Olympus, APON 60XOTIRF, Japan) using a tiny mirror (Edmund, 2-mm Diameter 45◦ Rod Lens Aluminum-Coated, 54-092, America). The focused beam passes through the objective and illuminates the sample immobilized on the slide. The reflected laser beam is directed out of the beam path using another tiny mirror placed symmetrically. The signals are collected by the same objective without the obstruction of the tiny mirror pair. The images are separated into two channels using DV2 (Photometrics, DV2) and finally projected onto an electron multiplying charge-coupled device (EMCCD, Andor, DU897E-13CS0-#BV, England). A homemade airtight flow cell was mounted on the TIRFM for air-controlled experiments.

#### Expression Plasmid Construction

For in vitro TIRF experiments, at least one of the proteins of interest must be immobilized on the cover slip. The most common technique for achieving this is immobilization through streptavidin–biotin linkage (van Oijen et al., 2003) For monomeric proteins, the linkage was completed using anti-His-tag antibodies, and vectors such as pET-28a were used to express, purify, and immobilize through His-tag. For other proteins investigated in the same experiment, pMAL-C2X or pET-42a vector was used to express and purify through MBP-tag or GST-tag. For fusion proteins, 6–10 amino acid linkers were inserted between their sequences. To assess protein dimerization and stoichiometry, antibodies should not be used for linkage; instead, vectors with StrepII-tag or avidity vectors, such as pET-52b or pAN/C vector, must be used. One-to-one stoichiometry linkage was achieved using bivalent streptavidin, the preparation of which is described in subsequent sections.

#### Protein Expression and Purification

Vectors containing His-MBP-BES1, MBP-BIN2, His-BIN2-eGFP, MBP-BKI1C, and GST-JKC were transformed into E. coli BL21-CodonPlus (DE3)-RIPL competent cells for expression. Differences in codon usage preference among organisms lead to various problems related to heterologous gene expression. Therefore, rare codon analysis must be performed prior to vector selection. Target proteins, whose sequences contain numerous rare codons, are better expressed in E. coli BL21-CodonPlus (DE3)-RIPL cells (**Figure 1**). A culture of transformed E. coli BL21-CodonPlus (DE3)-RIPL cells was grown at 37◦C in Luria Bertani (LB) medium until an optical density of 0.4–0.6 was achieved at 600 nm. The cell culture was then induced with 400µM isopropyl β-D-1-thiogalactopyranoside (IPTG), and grown for an additional 10–12 h at 20◦C. After expression, purification had to be performed as soon as possible to maintain sample freshness. For purification, lysates were dissolved in a buffer containing 20 mM Tris-HCl (pH 8.0) and 20 mM NaCl and were purified in an affinity chromatography column (amylose, GST, and Ni columns for MBP-tag, GST-tag, and His-tag, respectively) and gel filtration column.

Vectors containing MBP-14-3-3κ/A206K were transformed into E. coli BL21 (DE3)-plysS competent cells for expression. The culture of transformed E. coli BL21 (DE3)-plysS cells was grown at 37◦C in LB medium until an optical density of 0.4–0.6 was achieved at 600 nm. Subsequently, the cell culture was induced with 400µM IPTG, and grown for an additional 6 h at 37◦C. For purification, lysates were dissolved in a buffer containing 20 mM

compared with BL21(DE3)plysS in the same condition. M, protein marker; –,

without IPTG; +, with 1mM IPTG. \*, His-MBP-BES1 band.

Tris-HCl (pH 8.0) and 20 mM NaCl and were purified in amylose and gel filtration columns.

Homemade vectors with aCry-mOrange2 were transformed into E. coli BL21 (DE3) cells. Recombinant aCry-mOrange2 proteins were expressed as StrepII-fusion and His-fusion proteins in E. coli at 20◦C. For the purification of the recombinant proteins, lysates were purified through Ni2<sup>+</sup> affinity, anion-exchange, and size-exclusion chromatography. The flavin adenine dinucleotide redox states in aCry-mOrange2 proteins were monitored and maintained in oxidized states after each purification step.

Expression and purification method of His-AK can be found in the literature (Tan et al., 2009).

After purification, the proteins labeled with fluorescent proteins were subjected to cover slip immobilization. For proteins designed to be labeled by organic dyes, the following additional steps (Tan et al., 2014) were performed. The concentrated target proteins (50–100 µM/100 µL) were mixed with 1 µL of dye (1 mg of dye powder dissolved in 100 µL of dimethylsulfoxide) in the dark at room temperature for approximately 3 h. The labeled and unlabeled proteins were separated from free dye molecules in a gel filtration column in the dark. Because of minor differences in the molecular weights of labeled and unlabeled proteins, gel filtration was unable to separate these proteins. Therefore, SulfoLink coupling resin was used, which absorbs unlabeled proteins by forming disulfide bonds with free Cysteine (Cys) residues in a protein mixture. The sample was mixed with 100 µL of resin in a slow rotating mixer at 40 rpm in the dark at room temperature. The sample was washed using a protein buffer, and the resin and sample were separated through centrifugation at 13,000 rpm. The washing and centrifugation steps were repeated three times, and the protein supernatant was then collected. After this step, only a low percentage of unlabeled proteins was present. After purification, microscopic observations had to be conducted as soon as possible to maintain the proteins in the freshest state. Usually, microscopic observation must be completed within 3 days or even 24 h, which is before the degradation of unstable proteins after purification. The methods used to probe the functional integrity or degradation of purified proteins were specific to each protein under investigation. For example, the functional integrity of a phosphatase-like BIN2 was probed using the level of BES1 phosphorylation.

#### Cover Slip Passivation

Cover slip cleaning and surface passivation were performed as follows (Tan et al., 2014). Cover slips were incubated in a mixture of concentrated H2SO<sup>4</sup> and 30% v/v H2O<sup>2</sup> at the ratio 3:1 for 30 min, followed by thorough rinsing using ultrapure water. Subsequently, the cleaned cover slips were incubated in acetone containing 3% 3-aminopropyltriethoxysilane for 30 min, followed by thorough rinsing using ultrapure water and drying using clean nitrogen gas. The PEG solution was prepared by dissolving 4.5 mg of mPEG-SCM and 0.5 mg of biotin-PEG-SCM powder in 200 µL of 0.1 M NaHCO3. Finally, 50 µL of the PEG/biotin-PEG solution was incubated between two cover slips for 3 h at room temperature. After cover slip passivation, the slide surfaces were coated with a PEG layer with randomly distributed functional biotin terminals.

#### Flow Cell Construction

**Figure 2A** presents an assembled flow cell model. Double-sided tape, glass slides, and PDMS were prepared before assembling the flow cell. Double-sided tape 1 (**Figure 2C**) was designed to adhere the cover slip and glass slide and simultaneously leave an elliptical hole as the flow cell chamber. The detailed sizes are displayed in **Figure 2B**; the two 1-mm diameter circles are holes that were punched and drilled on double-sided tape 2 and the glass slide, respectively. In addition, a piece of 25 by 25 by 10 mm<sup>3</sup> PDMS with a flat smooth bottom surface was prepared to mount Teflon pipes. The glass slide and cover slip were cleaned and passivated in advance. Because the glass slide was not designed to immobilize proteins, a PEG layer was coated on its surface. The flow cell was assembled in five steps. First, the two pipes were mounted using PDMS (**Figure 2C**, step 1). The position of the two pipes exactly matched the two holes on the glass slide. The pipes mounted with PDMS are reusable. Subsequently, the cover slip and two-hole glass slide were tightly attached using doublesided tape 1 (**Figure 2C**, steps 2 and 3). Finally, the glass slide and PDMS were tightly held using double-sided tape 2 (**Figure 2C**, steps 4 and 5).

#### Protein Immobilization and Stoichiometry Regulation

A passivated cover slip embellished with PEG and biotin-PEG was incubated with 0.5 µL of streptavidin (stock solution concentration, 10 mg/mL) dissolved in 0.5 mL of buffer for 15 min. In addition, 0.5 mL of His-tag fusion protein (concentration, approximately 50 pM−1 nM) was incubated with 1 µL of biotinylated His antibody (stock solution concentration, 1µM) in a sample tube for 15 min. After thoroughly rinsing the cover slip with a buffer solution, the protein/antibody solution was transferred onto the cover slip and incubated for another 15 min, followed by thorough rinsing using 1 mL of protein buffer. When using a flow cell, the aforementioned procedure can be performed using a syringe pump and piping. Moreover, a simple CoverWell profusion chamber can be used with a drop (approximately 30 µL) of protein buffer on the cover slip (Tan et al., 2014). A sample is ready for microscopic observation after these procedures. If protein dimerization or stoichiometry is under investigation, the proteins can be fused with StrepII-tag or Avi-tag for direct linkage without the biotinylated His-tag antibody. Furthermore, bivalent streptavidin (Sun et al., 2014) were prepared to eliminate the possibility of additional binding sites on regular streptavidins by using the following procedure: the mixtures of streptavidin and biotinylated DNA were incubated for 0.5 h at room temperature at a molar ratio of 1:2. Streptavidins

FIGURE 2 | Flow cell model and assembly diagram. (A) Flow cell model. (B) Detailed geometry of double-sided tape 1, 2, with respect to the holes drilled on glass slide. The elliptical hole is the flow cell chamber. (C) Flow cell is assembled in five steps. Step 1, mount Teflon pipes on PDMS. Step 2–5, successively and securely adhere double-sided tape 1 with coverslip, glass slide, double-sided tape 2, PDMS. The position of two pipes, holes in double-sided tape 2, and glass slides should be precisely aligned.

with different valences were separated through Mono-Q anionexchange chromatography under alkaline conditions (optimal pH 8–9; **Figure 3**).

### RESULTS AND DISCUSSION

#### Experimental Design General Considerations

In the employed TIRFM-based single-molecule fluorescence technique, if the target proteins are moving freely in the buffer, the dwell time of the target protein flowing through the focal plane is shorter than common camera frame rates. Therefore, little information can be effectively captured. To maintain focus on the target proteins and their interaction events, the specimens had to be immobilized on a glass slide (**Figure 4A**; van Oijen et al., 2003). For systems involving more than one protein, such as a phosphatase associating with its substrate or three proteins interacting in a certain sequence to relay signals, a general protein construct design is required, as shown in **Figure 4B**. Usually, one protein (protein A) is immobilized on a slide and another protein (protein B) is added, followed by incubation or immediate observation. On the basis of the expected interaction cascade, a protein related to all other proteins is most suitable for immobilization. Moreover, a stable signal from protein B can only be observed when the interaction occurs between protein B and surface-tethered protein A. To study different systems, experiments can be operated in different manners. For example, to investigate the interaction strength of two proteins regulated by a certain ion, protein B can be added on the slide embellished with protein A under the condition of ion titration to observe the binding percentage or dynamics of proteins A and B. The interaction readout can be in the format of colocalization single-molecule spectroscopy (CoSMoS) (Friedman and Gelles, 2012), FRET (Förster, 1948; Clegg, 1992), or BiFC (Bimolecular fluorescence complementation) (Hu et al., 2002).

#### Labeling Strategy

Most proteins are nonfluorescent; therefore, they must be labeled by fluorophores for single-molecule fluorescent interaction or dynamic assays. The selection of appropriate fluorophores as labels is a critical step in target protein construct design. Fluorescent proteins and organic dyes are the most popular labeling methods in single-molecule fluorescence microscopy. For proteins with few Cys residues or a known structure, organic dyes are preferred because of their small size and flexibility. The labeling sites should be carefully selected when using organic dyes. The most common linkage is the conjugation of the maleimide derivative of organic dyes with the thiol group of proteins (Allewell et al., 2013). Therefore, the number and sites of Cys residues are critical factors. To assess protein–protein interaction stoichiometry, one protein molecule should only be labeled by one dye molecule. Mutations should be introduced to eliminate additional Cys residues such that only one Cys residue in each molecule is available for labeling; this residue must be accessible from the buffer solution to the target protein structure. The mutations of Cys residues are typically selected on the basis of homologous sequence alignment and structure analysis.

When the structure of the protein of interest is unknown or the protein sequence has many Cys residues, fluorescent proteins must be used as probes to avoid generating many mutations that disrupt the native protein function. The sequence of fluorescent proteins can be fused to the N terminal or after the C terminal of the target proteins. This labeling method has two advantages: one-to-one labeling efficiency and compatibility with in vivo usage. With regard to their size, fluorescent proteins may affect the function of the corresponding terminal or block sites of interaction with other proteins. Therefore, a linker between the target and fluorescent proteins is necessary. In addition, fluorescent proteins must not be linked to the functional terminal. Control experiments are required to assess whether the fusion proteins retain their original function. If neither of these labeling methods are successful, unnatural amino acids with amber codons or a chemoenzymatic labeling strategy (Juillerat et al., 2003; Gautier et al., 2008; Tian et al., 2017) is an alternative method.

#### Determination of Fluorophore Colors

The colors of fluorophores are mainly determined by the microscope used for the experiment. For a dual-color-labeled system, we require lasers of two colors for excitation and

between three proteins can be observed.

two signal channels for detection. The fluorophore excitation and emission spectra must cover the laser wavelength and signal channel detection range, respectively. In addition, the emission spectra of two fluorophores must not substantially overlap because this would contribute to the background noise and reduce the signal-to-noise ratio (SNR). For example, for a microscope equipped with 488- and 532-nm excitation lasers, the fluorophore pair of GFP/mOrange2 or alexa488/alexa555 can be used.

#### Method Validation

#### Protein Sample Preparation for Detection

TIRFM-based single-molecule experiments are extremely susceptible to environmental contamination. Because of the use of ultrasensitive detectors, even tiny nonfluorescent dust particles on the slide surface can generate troublesome background noise through light scattering. Therefore, cover slips should be thoroughly cleaned before microscopic observation. In addition, impurities can be generated during the surface passivation step. Therefore, before each set of data is acquired, a clean slide without protein samples should be tested. The total number of impurity spots in the camera frame must be restricted to within 5% of the molecule number observed in a regular experiment.

To achieve protein immobilization, purification, and fluorophore labeling, tags or mutations are introduced to modify the target proteins. These operations may cause structural change or hinder functional activity; therefore, control experiments must be conducted to assess target protein activity.

#### Single-Molecule Monomer Regulation and Dimer Detection

Limited by diffraction, the resolution of the TIRF microscope is approximately 200 nm. This is almost 10 times regular protein sizes. Therefore, for distinguishing individual proteins using TIRFM, their spacing should be at least larger than 200 nm; otherwise, massive overlapping spots will be observed. Consequently, for single-molecule experiments, labeled proteins are usually diluted to pM–nM concentrations before microscopic observation. However, achieving such low concentrations across different batches of protein samples is difficult. In addition, each streptavidin has four biotin-binding sites; therefore, more than one target protein may be linked to one streptavidin, and multiple and overlapping bright spots may obscure data interpretation. This complication is particularly evident in smFRET experiments because their readouts are based on relative intensities. To resolve this problem, modified streptavidin with only two binding sites can be used. This modification is performed using anionexchange columns that prepare multivalent streptavidin (Sun et al., 2014). Streptavidin with two dsDNA can be used for linkage because of the prior blocking of the two biotin-binding sites.

The animal-like cryptochrome (aCry) of Chlamydomonas reinhardtii is a dimerization-prone plant signaling protein (Oldemeyer et al., 2016). In the present study, we used a single-molecule assay to demonstrate that, with careful sample handling, aCry molecules can be maintained in their monomeric state by using divalent streptavidin immobilization. In this experiment, aCrys with an N-terminal StrepII-tag fused with C-terminal monomeric mOrange2 was immobilized on a slide through divalent streptavidin and then monitored under a TIRF microscope. Analysis of the fluorescence trajectories of aCrymOrange2 revealed that nearly all aCry-mOrange2 proteins underwent one bleaching step (**Figure 5A**). Less than 2% of the overall trajectories exhibited two bleaching steps (**Figure 5B**, blue trajectory), indicating the presence of a dimer or more than two aCry-mOrange2 molecules within the bright spot. These results

maleimide derivative of alexa555: ∼ 1,250 (The structure of alexa555 hasn't been announced yet). The relative intensity of peak 29073.83, which represents unlabeled 14-3-3κ, is 56.9%. The relative intensity of peak 29985.57, which represents labeled 14-3-3κ, is 30.3%. The labeling efficiency E = 30.3%/(56.9% + 30.3%) = 34.7%.

indicated that the aforementioned procedure can be used to prepare monomeric aCry molecules.

The 14-3-3 proteins form a large protein family that has been extensively studied in both plant and other eukaryotic systems (Aitken et al., 1992). They can associate with phosphorylated client proteins to modulate their function (Denison et al., 2011). In sodium dodecyl sulfate–polyacrylamide gel electrophoresis, 14-3-3 proteins appear in a dimerized protein band (Aitken et al., 1992). In the present study, single-molecule methods were used to observe the active form of 14-3-3κ when bound to phosphorylated BES1 (pBES1). In this experiment, BES1 with a His-tag, which was incubated with BIN2 for 30 min at 37◦C in a phosphorylation buffer (Ryu, 2013), was immobilized on a slide. Subsequently, alexa555-labeled 14-3-3κ/C103S was added, and the interaction between 14-3-3κ/C103S-alexa555 and pBES1 was observed under a TIRF microscope. In the 188 observed trajectories, one, two, or three bleaching steps were recorded (**Figure 5C**). Because the labeling efficiency of 14-3-3κ with alexa555 was 34.7% (from matrix-assisted laser desorption/ionization–MS; **Figure 5D**), the likelihood of 14-3-3κ functioning as a dimer was the highest among all probable models. Molecules with three bleaching steps may be due to two or three molecules appearing under the same pixel. This single-molecule experiment revealed the active dimeric form of 14-3-3κ.

#### Flow Cell for Plant Hormone Signal Transduction Pathway Investigation

In plant hormone signal transduction pathways, gas-phase molecules such as ethylene or some reactive oxygen species are also very important factors for regulating and controlling signal transduction (Apel and Hirt, 2004; D'Autréaux and Toledano, 2007). Under different gas concentrations, cells may have different responses. However, in the conventional experimental setting, it is difficult to control gaseous conditions in vivo or in vitro. In our method, we constructed a flow cell to maintain an airtight environment during microscopic observation. All buffers and samples were sealed in the flow cell system to prevent air exchange. **Figure 2** presents a diagram of the easy homemade flow cell model. The edge of the cover slip and glass slide substrate were attached using 0.3-mm double-sided tape, leaving an elliptical center for buffer flow and an observation chamber. The holes for liquid handling were drilled on the glass slide substrates. The buffer was fed through Teflon piping, with inlets mechanically supported by PDMS, entered through the elliptical chamber, and exited through the other hole. The amount of gas dissolved in the buffer solution could be controlled at the buffer reservoir and monitored in real time. Using this system, the gas concentration could be fixed or even titrated during observation. In addition, for systems involving multiple protein interactions, proteins could be added consecutively to control reaction sequences, conduct small-scale stop-flow experiments, and observe interactions in real time.

The activation or intensity of some fluorescent proteins is affected by the environmental oxygen (Tsien, 1998). The present study used a single-molecule assay with a flow cell to determine whether eGFP exhibits different responses under variable oxygen concentrations (Vordermark et al., 2001). By using a flow cell, we depleted O<sup>2</sup> from 7.21 to 0.53 ppm and enumerated the fluorescence spots and intensity of eGFP/A206K in the same view (**Figure 6**). Because the original eGFP can form a weak dimer, the monomeric form of eGFP/A206K was used as a test sample. The results indicated that 106/116 = 91% of spots remained bright when the oxygen concentration was decreased from 7.21 to 0.53 ppm. This slight decrease may be attributed to photobleaching during the observation period. This experiment demonstrated that a normal environmental oxygen concentration does not affect the fluorescence intensity or florescent state of eGFP/A206K.

#### Study of the Competitive Association of Phosphorylated BKI1 and pBES1 with 14-3-3κ Using CoSMoS

CoSMoS (Friedman and Gelles, 2012) is usually used to investigate the binding and dissociation times and kinetics of multiple components or to map the distribution of two or more components in cells. Unlike traditional biological methods, CoSMoS can capture real-time protein interaction dynamics. For example, if two labeled molecules, molecules 1 and 2, are extremely close to each other and associating directly or binding with the same molecule, the spots of these molecules coincide when observed through an EMCCD, indicating a direct

or indirect interaction between them. However, limited by diffraction, CoSMoS is unsuitable for observing molecules at high concentrations. Moreover, only when the locations of two or more molecules are superpositioned at an extremely low density, such as <1 molecule/µm, can an interaction be considered to have occurred among the molecules.

The present study used the BR signaling pathway (Wang et al., 2012) as an example. The existence of an alternative and possibly faster BR signaling transduction pathway has always been debated. After BR binds to its receptor BR insensitive1 (BRI1), BRI1 phosphorylates BKI1 at its C terminal, and this phosphorylated BKI1 (pBKI1) can also associate with 14-3-3κ (Wang et al., 2011). However, evidence of this direct competition is rare. If pBKI1 can competitively strip 14-3-3κ off the 14-3-3κpBES1 complex, pBES1 may be directly imported into nuclei to perform its function as a transcription factor with the help of protein phosphatase 2A. This signaling pathway could respond quickly to BR binding because it involves fewer components. To verify the existence of this faster signaling pathway or whether 14-3-3κ is competitively replaced by pBKI1 on pBES1, allowing pBES1 to perform its transcription function, we used CoSMoS to measure the binding percentage of 14-3-3κ with pBES1 before and after adding pBKI1. **Figure 7A** presents the experimental design. BES1 and BKI1C (BKI1 C terminal) were phosphorylated by BIN2 and JKC (the intracellular domain of BRI1 containing a juxtamembrane segment, kinase domain, and cytoplasmic portion), respectively, in a phosphorylation buffer (Ryu, 2013) for 1 h at 37◦C. The results showed that during this hour, the binding percentage to 14-3-3κ was altered only when the concentration of BKI1 was 1,000-fold higher than that of pBES1 (**Figure 7B**). Considering that such a wide concentration gap is impossible under normal physiological conditions, the binding of 14-3-3κ with pBES1 is comparably stable. However, unbound 14-3-3κ can readily interact with both pBKI1 and pBES1. Therefore, pBKI1 cannot compete with the 14-3-3κ-pBES1 complex to initiate a faster signaling pathway.

FIGURE 7 | The competitive association of 14-3-3κ- with pBES1 and pBKI1 observed by CoSMoS. (A) Protein BES1 with His-tag was immobilized on slide surface. BES1 and BKI1C are phosphorylated by BIN2 and JKC, respectively, in phosphorylation buffer for 1 h at 37◦C. The binding percentages of 14-3-3κ with pBES1 were observed before and after incubating with pBKI1C. (B) Binding percentage of 14-3-3κ with pBES1 before and after pBKI1C treatment. The percentage only decreased by half, from (18 ± 5)% to (9 ± 5)%, even though we used a very high concentration pBKI1C (>1,000 times larger than that of pBES1) and incubated for 1 h.

#### Data Analysis

#### Extraction of Intensity–Time Trajectories

In single-molecule fluorescence methods, the raw data collected by an EMCCD take the form of monochrome video files. The brightness of each pixel corresponds to the observed light intensity. The first step of data processing is screening for effective spots and their intensity–time trajectories in each video. Effective spots are defined as bright spots formed by light emitted from the labeled target proteins instead of background noise or light emitted from impurities on the slide. For regular fluorophores, photobleaching is a typical feature during microscopic observation, which is defined by oxidationinduced irreversible changes in the fluorophore structure. In contrast to photoblinking, photobleaching leads to a permanent nonfluorescent state. A stepwise sharp photobleaching-caused decrease in the intensity in an intensity–time trajectory is the signature of a single molecule (**Figure 8E**). Before photobleaching and without photoblinking (Moerner, 1997), the fluorescence intensity is typically stable without drastic fluctuations.

Intensity–time trajectory extraction can be completed in four major steps (**Figure 8**): (1) Estimate the mean background intensity by using at least two images in which all spots are nearly photobleached; (2) Set a threshold above which any bright spot is considered an effective molecule. Use the threshold to estimate the SNR by the background intensity and screen the effective spots. The threshold need not be constant and can be adjusted with time and the microenvironment around the fluorophores; (3) Obtain output intensity–time trajectories. When the SNR of the video is low, the overall data quality can be improved by summing the intensities of the nearest neighboring pixels. **Figure 8E** shows the trajectories formed by a single pixel or five adjacent pixels; and (4) Apply additional criteria to the effective trajectory pool. For example, identify an abrupt intensity drop indicating single-fluorophore bleaching, rejecting fluorophore blinking or features such as the stability of intensity.

#### Model Building and Parameter Fitting

The single-molecule protein interaction assay used herein is unique because it allows the determination of parameters such as reaction rate constants and stoichiometry. For studying protein– protein or protein–ligand interactions in plant hormone signal transduction pathways, an appropriate model must describe the interaction kinetics. For instance, the reaction of an enzyme associating with or dissociating from its fluorophore-labeled ligand can be described as:

$$\begin{aligned} k\_1 \\ E + L &\rightleftharpoons E \cdot L \\ k\_{-1} \\ &\downarrow k\_2 \\ E + L^b &\rightleftharpoons E \cdot L^b \end{aligned}$$

In this system, fluorophore-labeled enzymes are immobilized on slides and dye-labeled ligands of a certain concentration are added to the buffer to obtain fluorescence information in two channels (two color labels for enzymes and ligands). In the stated equation, k<sup>1</sup> and k−<sup>1</sup> are association and dissociation rate constants, respectively. Considering the photobleaching of labeling fluorophores, a rate constant k<sup>2</sup> was added to this model. L b is the photobleached ligand. Because enzyme E is immobilized on a slide, the observed fluorescence time does not indicate any information of its association or dissociation. In this study, we did not consider the photobleaching of E. In principle, the photobleaching of ligands' probe will not interfere with the reaction between the enzyme and its ligand. In this system, E·L and E·L b are the bound states, and the ligand can only be observed in the bound state before photobleaching.

In this kinetic model, k<sup>2</sup> is a parameter that depends only on the fluorophore photostability and microscopic setup. It can be fitted separately by observing the fluorophore alone. The unbound time, represent the probability of an L association event. The bound time was affected by bleaching rate constant k<sup>2</sup> and the dissociation rate constant k−1. By combining the ligand concentration L, fitted k2, and observed unbound and bound time distributions, the rate constants k<sup>1</sup> and k−<sup>1</sup> can be obtained.

To illustrate how to acquire the rate constants k1, k−1, and k2, we use the enzyme Adenylate Kinase (AK) as an example. Adenylate Kinase (AK) is a well-studied kinase which can catalyze the reaction Mg2+· ATP + AMP ⇋ Mg2+· ADP + ADP. The aforementioned method was applied to measure the rate constants of AK and ATP association. Alexa555 labeled AK was immobilized on the slide surface via His-tag, 10 nM ATP labeled with BODIPY FL was added through AK buffer, and the interactions are observed on TIRFM. In this case, we can revise the kinetic model into:

$$\begin{aligned} AK + ATP &\rightleftharpoons AK \cdot ATP \\ k\_{-1} \\ &\quad \downarrow k\_2 \\ AK + ATP^b &\rightleftharpoons AK \cdot ATP^b \\ k\_{-1} \end{aligned}$$

A typical fluorescent intensity trajectory is shown in **Figure 9A**. BODIPY FL labeled ATP kept binding and dissociating with AK consecutively. Here, toff , the unbound time, represent the probability of an ATP association event. The dissociation distribution is:

$$P\_{on}\left(t\right) = [ATP]k\_1e^{-[ATP]k\_1t} \tag{1}$$

Here, [ATP] = 10 nM. The bound time, ton, was affected by the bleaching rate constant k<sup>2</sup> of BODIPY FL and the dissociation rate constant k−1. The binding distribution (for details, please see Supplementary Information) is:

$$P\_{\text{off}}\left(t\right) = \left(k\_2 + k\_{-1} + k\_2 k\_{-1}\right) e^{-\left(k\_2 + k\_{-1}\right)t} \tag{2}$$

The bleaching rate constant k2, can be measured independently by observing BODIPY FL-ATP on TIRFM alone. Its distribution

spectral range: 500–600 nm. (B–D) Applying a signal to background threshold, effective spots (red circles) can be selected. (C,D) are enlarged images of the areas marked by green and yellow squares in (B). (E) The intensity-time trajectory of a spot analyzed by two different algorithms. Red curve: trajectory of only 1 pixel (pixel 1 in D); blue curve: trajectory of intensity summation of five pixels (pixel 1–5 in D).

function can be written as:

$$P\_b\left(t\right) = k\_2e^{-k\_2t} \tag{3}$$

The histograms of ton, toff , t<sup>b</sup> , and their fitting results were shown in **Figures 9B–D**. The fitted bleaching rate constant k2, association rate constant k<sup>1</sup> and dissociation rate constant k−<sup>1</sup> are:

$$k\_2 = 0.22 \pm 0.01 \text{ s}^{-1} \tag{4}$$

$$k\_1 = 1.5 \pm 0.2 \,\upmu\text{M}^{-1}\text{s}^{-1} \tag{5}$$

$$k\_{-1} = 6.9 \pm 0.7 \,\text{s}^{-1} \tag{6}$$

With the fitted k<sup>1</sup> and k−1, we can further acquire the dissociation constant K<sup>D</sup> = k−1 k1 = 4.6 ± 1.1 µM. For in vitro experiments, a portion of the protein specimens may lose their activity during purification and observation. Since our method only report active events, those inactive proteins are naturally screened out as a result. This leads to a smaller dissociation constant compared with bulk methods, such as nuclear magnetic resonance (Reinstein et al., 1990).

#### Discussion

#### Advantages of Single-Molecule Techniques

Single-molecule fluorescence methods have emerged as powerful tools for studying protein dynamics and reaction kinetics (Joo et al., 2008; Ryu, 2013). By detecting individual biological molecules, single-molecule fluorescence methods not only provide ensemble-averaged values of observables but also reveal information on a molecule's internal distribution. This unique advantage allows single-molecule fluorescence methods to play a crucial role in the investigation of protein kinetics and dynamics. Furthermore, these methods have obvious advantages over traditional biology methods for nonequilibrium systems. Elf et al. (2007) used one such method to directly observe the specific binding of a lac repressor to a chromosomal lac operator in living cells and characterize the kinetics of association and dissociation of the repressor in response to signaling molecules. In combination with reflected light-sheet microscopy, a singlemolecule fluorescence method was employed by Gebhardt et al. (2013) to observe the binding of fluorescent-protein-labeled transcription factors on DNA in living mammalian cells. Knight et al. (2015) directly visualized DNA interrogation by Cas9 and observed its off-target binding events in mammalian cells through single-particle tracking. These results demonstrated the advantages of single-molecule methods in life science research.

#### Potential Use and Challenges of in Vivo Single-Molecule Assays

In vitro single-molecule methods can be an alternative method of identifying the different features of an entire signaling pathway in living cells. However, some differences exist between in vitro and in vivo environments. For example, pH and ion concentration are generally heterogeneous in vivo, and the cytoplasm has various proteins relevant or irrelevant to the research targets. Therefore, in vivo single-molecule methods must be explored.

Thick plant cell walls limit the wide-field imaging of membrane receptors on the cells. Single-molecule assays can be employed through the biochemical removal of the cell wall to obtain protoplasts from the plant cells for observation (Zhang et al., 2011). A new technology, referred to as variable-angle TIRFM (VA-TIRFM), was developed in 2011 (Wan et al., 2011). An evanescent wave is produced inside the cytosol above a cell wall/cytosol interface when the excitation laser has an appropriate incidence angle. Using this constant-depth field, the TIRF image of targets on the plasma membrane can be captured

decay, we obtain the bleaching rate constant <sup>k</sup><sup>2</sup> <sup>=</sup> 0.22 <sup>±</sup>0.01 s−<sup>1</sup> . (C,D) Histograms of toff and ton event durations. By fitting with distribution shown in Equation (1) and (2), we obtain the association rate constant <sup>k</sup><sup>1</sup> <sup>=</sup> 1.5 <sup>±</sup> 0.2 <sup>µ</sup>M−<sup>1</sup> s <sup>−</sup><sup>1</sup> and the dissociation rate constant <sup>k</sup>−<sup>1</sup> <sup>=</sup> 6.9 <sup>±</sup> 0.7 s−<sup>1</sup> .

easily. Wang et al. (2015) explored the spatiotemporal dynamics of the BR receptor of Arabidopsis BRI1 in live cells using VA-TIRFM.

VA-TIRFM is not suitable for studying downstream signal transduction pathways with a single-molecule resolution, however, because its penetration depth is only 100–200 nm, which is substantially less than cell sizes, which are several micrometers. Epifluorescence microscopy or confocal microscopy can be used instead of VA-TIRFM (Moerner and Fromm, 2003). Nevertheless, autofluorescence from different cellular components—the chloroplasts in plant cells, for example—dramatically increases the background noise and leads to a low SNR and poor image quality (Kodama, 2016). Therefore, fluorophores with extremely high brightness are urgently required for the further development of in vivo single-molecule assays.

#### Perspectives of Applying Single-Molecule Assays to Study Plant Hormone Signal Transduction Systems

Although single-molecule methods are usually adopted in studies on protein dynamics and reaction kinetics, they are not commonly used to investigate plant hormone signal transduction pathways. Genetic and biochemical techniques have demonstrated remarkable advantages in identifying the key components of such pathways and illustrating their biological functions. However, single-molecule methods are more suitable than traditional methods for obtaining temporal and detailed molecular information about the components. In the ABA cascade, although a fully connected signal pathway has been elucidated using genetic and biochemical approaches, the rate constants between ABA receptor, phosphatase, and their substrates remain unknown. Moreover, the process through which proteins with several substrates, such as MAPK and BIN2, select their substrate in different cell environments remains unclear. Furthermore, it is unknown if the phosphorylation ability of such proteins changes depending on the air or ion concentration and then regulates different hormone signal transduction pathways. These questions are not easily addressed using conventional techniques. Moreover, for transmembrane receptors, such as BRI1 in the BR signaling pathway, the exact conformational changes occurring after their association with BR and that lead to signal transfer into cells are yet to be elucidated. All these dynamic events and detailed reaction mechanisms are difficult to decipher using conventional techniques, necessitating the development of alternative modern technologies.

With advantages such as a well-controlled interaction environment and molecule-by-molecule analysis, singlemolecule fluorescence methods can address the aforementioned questions and be powerful tools for studying plant hormone signal transduction pathways. Their potential applications can be increased if they are employed in combination with other modern technologies. For example, by using lipid vesicles to mimic a membrane, the dynamics or conformational changes in membrane receptors before and after ligand binding can be investigated (Boukobza et al., 2001). By combining a singlemolecule fluorescence method with optical tweezers, short DNA fragments can be fixed between two beads (Ashkin et al., 1986; Wang et al., 1997). Furthermore, the mechanism through which transcription factors distinguish target sequences or cooperate with each other can be observed directly. By labeling two sites on a receptor or a receptor and its ligand, nanoscale conformational changes as well as ligand binding events can be identified through smFRET. Therefore, the dynamics of macromolecular machinery can be revealed. In summary, single-molecule fluorescent methods open a new avenue for plant signal transduction pathway explorations.

#### CONCLUSION

We have introduced novel TIRFM-based single-molecule fluorescence methods for studying plant hormone signal transduction pathways. Unlike traditional biology methods, these methods focus on the detailed molecular mechanisms of the signaling pathway under investigation. During the preparation step involving protein immobilization on a slide, labeling methods and labeling fluorophores should be carefully determined, particularly while the dye-labeling sites of the protein are being selected. Before microscopic observation, slides should be thoroughly cleaned and passivated. This paper has introduced and validated single-molecule methods combined with monomer regulation, air-controlled flow cell construction, and CoSMoS to study different protein– protein interaction systems. For monomer regulation, divalent streptavidins were used to identify accurate stoichiometry

#### REFERENCES


during interactions. For studying signal transduction modulated by gas molecules, a flow cell was constructed to provide airtight conditions and maintain dissolved gas concentration consistency or even change the concentration with time. CoSMoS can be used to study interactions among two or more proteins. By combining CoSMoS with a flow cell model, the protein interaction order, stoichiometry, and molecular mechanisms of two or more fluorophore-labeled proteins can be directly observed simultaneously. On the basis of the observed system, an appropriate model can be constructed to identify the key time constants and conduct the quantitative assessment and dynamic analysis of signal transduction pathways.

#### AUTHOR CONTRIBUTIONS

SS and YT designed the experiment. SS, JC, and CM carried out the experiments. SS, CM, JC, and YT wrote the paper.

#### FUNDING

This work was supported by National Nature Science Foundation of China (No. 11274076 and 21773039).

#### ACKNOWLEDGMENTS

The authors thank Xuelu Wang and Haijiao Wang for providing BR related plasmids and discussion. This manuscript was edited by Wallace Academic Editing.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2017. 01888/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.

The reviewer YF and handling Editor declared their shared affiliation.

Copyright © 2017 Song, Chang, Ma and Tan. 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.

# Dynamic Regulation of Auxin Response during Rice Development Revealed by Newly Established Hormone Biosensor Markers

Jing Yang1, 2, Zheng Yuan<sup>2</sup> , Qingcai Meng<sup>2</sup> , Guoqiang Huang<sup>2</sup> , Christophe Périn<sup>3</sup> , Charlotte Bureau<sup>3</sup> , Anne-Cécile Meunier <sup>3</sup> , Mathieu Ingouff <sup>3</sup> , Malcolm J. Bennett <sup>4</sup> , Wanqi Liang<sup>2</sup> \* and Dabing Zhang2, 5 \*

<sup>1</sup> Key Laboratory of Systems Biomedicine, Ministry of Education, Shanghai Center for Systems Biomedicine, Shanghai Jiao Tong University, Shanghai, China, <sup>2</sup> Joint International Research Laboratory of Metabolic & Developmental Sciences, Shanghai Jiao Tong University–University of Adelaide Joint Centre for Agriculture and Health, School of Life Sciences and Biotechnology, Shanghai Jiao Tong University, Shanghai, China, <sup>3</sup> CIRAD, UMR AGAP, Montpellier, France, <sup>4</sup> Centre for Plant Integrative Biology, School of Biosciences, University of Nottingham, Sutton Bonington, UK, <sup>5</sup> School of Agriculture, Food and Wine, University of Adelaide, Urrbrae, SA, Australia

#### Edited by:

Yunde Zhao, University of California, San Diego, USA

#### Reviewed by:

Clay Carter, University of Minnesota, USA Jong-Seong Jeon, Kyung Hee University, Korea

#### \*Correspondence:

Wanqi Liang wqliang@sjtu.edu.cn Dabing Zhang zhangdb@sjtu.edu.cn

#### Specialty section:

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

Received: 14 November 2016 Accepted: 10 February 2017 Published: 07 March 2017

#### Citation:

Yang J, Yuan Z, Meng Q, Huang G, Périn C, Bureau C, Meunier A-C, Ingouff M, Bennett MJ, Liang W and Zhang D (2017) Dynamic Regulation of Auxin Response during Rice Development Revealed by Newly Established Hormone Biosensor Markers. Front. Plant Sci. 8:256. doi: 10.3389/fpls.2017.00256 The hormone auxin is critical for many plant developmental processes. Unlike the model eudicot plant Arabidopsis (Arabidopsis thaliana), auxin distribution and signaling in rice tissues has not been systematically investigated due to the absence of suitable auxin response reporters. In this study we observed the conservation of auxin signaling components between Arabidopsis and model monocot crop rice (Oryza sativa), and generated complementary types of auxin biosensor constructs, one derived from the Aux/IAA-based biosensor DII-VENUS but constitutively driven by maize ubiquitin-1 promoter, and the other termed DR5-VENUS in which a synthetic auxin-responsive promoter (DR5rev) was used to drive expression of the yellow fluorescent protein (YFP). Using the obtained transgenic lines, we observed that during the vegetative development, accumulation of DR5-VENUS signal was at young and mature leaves, tiller buds and stem base. Notably, abundant DR5-VENUS signals were observed in the cytoplasm of cortex cells surrounding lateral root primordia (LRP) in rice. In addition, auxin maxima and dynamic re-localization were seen at the initiation sites of inflorescence and spikelet primordia including branch meristems (BMs), female and male organs. The comparison of these observations among Arabidopsis, rice and maize suggests the unique role of auxin in regulating rice lateral root emergence and reproduction. Moreover, protein localization of auxin transporters PIN1 homologs and GFP tagged OsAUX1 overlapped with DR5-VENUS during spikelet development, helping validate these auxin response reporters are reliable markers in rice. This work firstly reveals the direct correspondence between auxin distribution and rice reproductive and root development at tissue and cellular level, and provides high-resolution auxin tools to probe fundamental developmental processes in rice and to establish links between auxin, development and agronomical traits like yield or root architecture.

Keywords: rice, auxin, reporter, lateral root formation, inflorescence, spikelet, meristem

# INTRODUCTION

The phytohormone auxin (indole-3-acetic acid, IAA) regulates many critical growth and developmental processes in plants. IAA is synthesized in subsets of plant cells and then actively transported from cell to cell through polar transport. Development of effective hormone biosensors to visualize auxin distribution in vivo is needed to dissect the functions of this key hormone during plant development. In Arabidopsis, the most widely applied tool is DR5-GFP which uses a synthetic auxin-responsive promoter (DR5rev) to drive the expression of green fluorescent protein (Heisler et al., 2005). Auxin can be detected using DR5-GFP transgenic lines, despite of its indirect connection with auxin abundance in vivo, and the slow timescale of its auxin induced response (taking several hours from induction) which is not optimized to study fast biological processes such as tropic responses (Zhao et al., 2014). Brunoud et al. (2012) developed an alternative reporter system employing the CaMV35S promoter to constitutively drive expression of the DII-VENUS sequence in which the YFP VENUS reporter was fused to the auxin degron sequence called DII, present in Aux/IAA repressor proteins. The presence of auxin triggers the degradation of the DII-VENUS fusion protein, where the reduction in reporter fluorescence intensity is proportional to lAA levels in cells. Hence, subtle differences in auxin abundance can be visualized through changes in fluorescence, allowing high-resolution spatio-temporal changes in auxin distribution and response during plant growth and development (Brunoud et al., 2012). These two systems have been extensively used to characterize functions of genes associated with auxin signaling (Steenackers et al., 2016), gravitropic response (Band et al., 2012; Zou et al., 2016) and stomatal patterning (Le et al., 2014). New generations of DII-VENUS and DR5-GFP have also been recently developed. R2D2 integrates an auxin sensitive DII-VENUS and insensitive mDII-ntd TOMATO into one reporter to rapidly quantify changes in auxin using fluorescence ratio. DR5v2 is composed of the DR5 promoter and a novel binding site for ARF transcription factors designed to increase sensitivity and precision of auxin response visualization in Arabidopsis (Liao et al., 2015).

Rice exhibits divergent morphologies in root, shoot, inflorescence and flower tissue organization compared to dicotyledons. For instance, in Arabidopsis, a single primary root emerges from the embryo, later forming numerous lateral roots employing auxin-dependent initiation, patterning and emergence mechanisms (Lavenus et al., 2015). In contrast, rice develops a fibrous root system, composed of >100 crown roots bearing several lateral root types (Coudert et al., 2010). Auxin also regulates crown root (Inukai et al., 2005; Liu et al., 2005) formation and emergence in rice and also impacts lateral root formation (Liu et al., 2009). Similarly, in Arabidopsis floral meristems (FMs) initiate directly at the flank of IMs, and their formation is dependent on local auxin accumulation at the periphery of IMs (Yamaguchi et al., 2013). In contrast, rice exhibits a specialized inflorescence shape with primary and secondary branches, and spikelets attached on the branches (Zhang et al., 2013; Zhang and Yuan, 2014). To clarify the role of auxin during rice development, the DR5-GUS reporter was transformed into rice (Scarpella, 2003), to infer auxin distribution by analyzing GUS (β-glucuronidase) activities. However, the GUS reporter has low spatio-temporal resolution because of the longer protein turnover time of GUS protein and experimental variation in temperature, incubation time and pH, which frequently causes imprecise results of auxin location (Rahman et al., 2014).

In this study, we established and validated two auxin response reporter systems in rice: DR5-VENUS and DII-VENUS. Using these two reporters, we followed dynamic changes of auxin during rice development. This work describes new molecular tools for future auxin research in rice, but also provides the first insight in comparative auxin distribution and role in plant between monocot and dicot models.

# MATERIALS AND METHODS

## Plasmid Construction and Transformation

The DII-VENUS fragment containing the coding sequence for the degradation motif of the domain II of Arabidopsis AUX/IAA28 (AtIAA28) protein subcloned from 35s:: DII-VENUS plasmid (Brunoud et al., 2012) was inserted into the binary vector pUBI::CAMBIA1301 (CAMBIA) using Kpn I and Bam HI restriction sites, under the control of maize ubiquitin-1 promoter. The DR5rev::VENUS construct in pMLBART was composed of a generic synthetic promoter with nine repeats of core sequence (TGTCTC) reversely linked with CaMV minimal 35S promoter (Ulmasov et al., 1997; Friml et al., 2003), the triple VENUS sequence and the nuclear localization signal N7 (Cutler et al., 2000), which was harvested from Heisler et al. (2005). The two resultant vectors were transformed separately into rice japonica cultivar 9522 calluses with Agrobacterium tumefaciens EHA105 using Agrobacterium-mediated method (Hiei and Komari, 2008). We got 30 positive independent T0 transformants containing DR5-VENUS, Among these lines, 4 lines were identified as the homozygous plants showing similar and stable expression patterns during propagation. Among the nine positive T0 DII-VENUS lines, one line having the strongest and stable signals was selected for further analyses.

#### Multiple Sequence Alignment and Prediction of Putative ARF Binding Sites

Amino acid sequences of 31 OsAUX/IAA members and AtAUX/IAA28 protein from Rice Genome Annotation Project (http://rice.plantbiology.msu.edu/) and TAIR (http://www. arabidopsis.org/), respectively, were aligned using MUSCLE 3.6, and then adjusted manually in GeneDoc 2.6. ARF binding sites among the 3000-bp promoter region of each OsGH3 family was analyzed using PLANTPAN 2.0 (http://PlantPAN2.itps.ncku. edu.tw) (Chow et al., 2016).

#### Plant Growth and Vibratome Sectioning

Rice seedlings were grown vertically in sterile square petri dishes (Corning, 431301; 20 cm × 20 cm) under controlled conditions (day/night temperature of 28/25◦C, a 12 h photoperiod, and a light intensity of 500 µEm-2s-1) for 3 days. Tissue parts of rice root, stem base, leaves and shoot apices were dissected and embedded in 3% agarose blocks (Lartaud et al., 2014). After solidification and reshaping, materials were cut into 70 µm slices in thickness with Thermo Vibratome 750. Agar parts of slices were carefully removed in water, and samples were quickly transferred on slides and immersed in a drop of 10% glycerol for imaging.

#### Chemical Treatments

For live imaging, 3-days old DR5-VENUS seedlings were treated for 1 day in 100 nM auxin transport inhibitor N-1- Naphthylphthalamic acid (NPA), and 3 days separately in 500 nM 1-Naphthaleneacetic acid (NAA) and 500 nM trans-zeatin (TZ) water solutions. For mRNA analysis, 6-days old wild-type seedlings were treated for 1.5 h in 1 µMNPA, 5 µMNAA, and 5 µM TZ water solutions, respectively.

#### Root Gravitropism Assay

Firstly, rice seeds were sterilized using 50% bleach for 10 min with gentle shaking, and then washed for 6∼7 times with sterile double distilled water. Seeds were dried for 3 min, then laid on half Murashige and Skoog (MS) solid medium and grown them vertically for 5 days. Following plate rotation to 90 degrees, timeserial pictures were taken at 30 min intervals automatically. Root tip angles were measured in ImageJ software.

#### Sample Preparation and Microscope Observation

Fluorescence images were taken on Zeiss LSM510 SP5 confocal, or TLSM 7MP/OPO two photon microscopy. For tissue organization observation, root tips were stained using 10 µg/ml Propidium Iodide (PI) solution for 10 min in dark and rinsed in double distilled water for 3 times, then included in low melting 0.5% agarose, mounted between a slide and a cover slip of 170 ± 1µm for TLSM observation. Cell organization of rice vegetative tissues was visualized using chlorophyll autofluorescence. Fresh sections or intact tissues were immersed in a drop of 10% glycerol for LSM510 live imaging.

Under the SP5 microscope, Z-stacks were scanned every 1.5 µm in thickness and maximum projections were generated. For the TLSM, VENUS and PI emissions were collected in separate channels with excitation at 950 nm (Chameleon Ultra II) and 1,096 nm (Chameleon Compact OPO) with a gain set at 600 nm using 2PMT NDD and 2 PMT BiG detectors.

#### Gene Expression Analysis

Root samples of 6-days old plants after drug or water treatment were collected instantly. After fixation in liquid nitrogen, samples were ground and then transferred into tubes filled with Trizol (Sigma). Total RNA was extracted using the traditional chloroform method, DNA was removed with DNA eraser reagent at 42◦C for 2 min and cDNA was reverse transcribed from 1 µg total RNA by using Takara PrimeScriptTM RT reagent Kit. Real-time qRT-PCR was performed on Bio-Rad CFX96 machine by the three-step method. Expression levels of those genes were normalized using those of tublin β-4 and ubiquitin 2 as the reference. Specific primers were in Supplementary Table S1.

#### Immunostaining

Flower materials were fixed, wax-embedded and sectioned following the whole mount protocol (Paciorek et al., 2006). After clearing sections using Histoclear solution with increasing proportions of ethanol (100% Histoclear, 2:1 solution of Histoclear and absolute ethanol, 1:2 solution of Histoclear: 2ethanol, 100% ethanol), samples were rehydrated gradually, with ethanol 95, 70, 50, 30, and TBS buffer (100 mM Tris-HCl, 150 mM Nacl, pH: 7.5), 3∼5 min for each step. The crosslink formed by paraformaldehyde was destroyed by treating slides for 30 min with target retrieval solution (DakoCytomation) at 33◦C. After BSA solution (0.5% BSA, 0.02% Tween-20 in TBS) blocking slides for 1 h at room temperature, PIN1 proteins were detected by applying primary mouse monoantibody (1:1,000) obtained from Professor Klaus Plame (Pasternak et al., 2015) at 4 ◦C overnight, and Alexa Fluor 488-conjugated goat anti-mouse secondary antibody (1:800) at RT for 2 h. Specific fluorescent signals were then captured through Leiss LSM510 confocal system.

# RESULTS

#### Rice Genome Has Conserved Auxin-responsive Elements and Auxin-interacting Domain Sequences

To reveal whether the auxin responsive element AuxRE or ARF transcription binding sites located at promoter regions of primary auxin responsive gene families in Arabidopsis, such as GH3, AUX/IAA, and SAUR (Abel and Theologis, 1996; Ulmasov et al., 1999; Chen et al., 2014) genes are conserved in rice, we searched for multiple AuxRE sites by scanning the 3,000 bp promoter regions upstream of translation start sites of 11 OsGH3 genes. We observed that auxin responsive sequences (ARS, TGTCTC) were highly enriched in rice promoter regions of OsGH3.3, OsGH3.5, OsGH3.12 (Supplementary Table S2), while no ARS was present within the OsGH3.10 promoter, which are well in line with the responses of increased expression of OsGH3.3, OsGH3.5, OsGH3.12, and no detectable change in transcriptional level of OsGH3.10 induced by auxin treatment (Jain et al., 2006b; Terol et al., 2006). Therefore, we decided to directly use the synthetic DR5rev promoter containing ARS sequences to monitor auxin responsive expression in rice tissues.

The DNA fragment encoding the DII degradation domain of AtIAA28 was used in Arabidopsis auxin sensor DII-VENUS owing to its relatively long half-life (Brunoud et al., 2012). In rice, there are 31 AUX/IAA proteins (Jain et al., 2006a), and through the alignment of Arabidopsis IAA28 protein, we observed that rice AUX/IAA members share the consensus degron sequence GWPPV, and the conserved dipeptide KR between the first two domains (Supplementary Figure S1). Because little is known about the stability of rice AUX/IAAs in vivo, we generated the UBI::DII-VENUS via inserting the cDNA sequence between KQ (KR for other AtIAAs) and DII of AtIAA28 together with VENUS and nuclear localization signal N7 under the control of the maize ubiquitin-1 promoter (Supplementary Figure S2), which has been proved to have a relative stronger transcriptional ability in reproductive tissues than CaMV35S promoter (McElroy and Brettell, 1994).

## DR5-VENUS Is Applicable for the Detection of Auxin Relocation and Cellular Level in Rice

#### DR5-VENUS Is Sensitive to Exogenous NAA Treatment

To test the sensitivity of rice DR5-VENUS line and the authenticity of these auxin response, we treated the transgenic plants using active synthetic auxin NAA. Consistent with previous observation (Rahman et al., 2007), NAA inhibits rice primary root elongation in a dose-dependent manner (Supplementary Figure S3A). At the rice root tip under NAA treatment, the auxin reporter signal is visible in the root cap zone, outermost epidermal layer, as well as the root hair zone (**Figure 1A**, bottom panel), compared to the untreated control (**Figure 1A**, top panel). Moreover, with higher NAA concentrations, the signal at the root tip was gradually increased (Supplementary Figures S3B,C) confirming the dose-dependent response of DR5-VENUS to auxin levels. Consistent with the increased DR5-VENUS signal, exogenous auxin treatment enhanced the transcription of the auxin inducible gene OsGH3.2 and auxin transporter genes OsPIN1a, 1b, 1c, 1d, and OsAUX1 (**Figure 1B**). These results suggest that signal distribution of DR5-VENUS measures auxin presence in situ in rice. As application of exogenous auxin also induced quantitative changes in DR5-VENUS signal, these results suggest the reporter provides a reliable means to measure auxin levels.

#### The Auxin Transport Inhibitor NPA Disrupts DR5-VENUS Pattern

To further assess authenticity of the auxin response in the DR5-VENUS marker line, we blocked polar auxin transport of the marker line using the auxin transport inhibitor NPA. This treatment caused the auxin gradient between epidermis and inner tissues to disappear, the intensity of the auxin maxima located in root cap and vasculature to weaken, and the VENUS fluorescent signals in the QC also became broadly diffuse (**Figure 1C** and

FIGURE 1 | Auxin responses to outside stimuli at DR5-VENUS root tip. (A) 500 nM NAA treated 3-days old DR5-VENUS rice seedlings for 3 days; images were taken at root tip and root hair zone. NAA raises auxin response at the epidermal (arrowheads), and columella cells (arrows). Scale bar: 50 µm. (B) Quantification of relative mRNA levels of auxin transporter genes after NAA treatment. Auxin responsive gene OsGH3.2 is used as positive control. All the data were expressed as mean ± SEM from three biological and three technical repeats. Statistically significant differences (Student's t-test, P < 0.01) between mock control (CK) and NAA treatment are indicated by two asterisks. Expression levels of all auxin transporters were increased after NAA treatment, including OsAUX1. (C) 100 nM NPA inhibition effect on DR5-VENUS rice primary root for 1 day. Auxin profile becomes much broader after NPA treatment and expands into cortex layers inside meristem. Scale bar: 50 µm. (D) Quantification of relative mRNA levels of auxin transporter genes after NPA treatment. OsGH3.2 is used as reference gene induced by auxin. All the data were expressed as mean ± SEM from three biological and three technical repeats. Statistically significant differences (Student's t-test, P < 0.05) are indicated by one asterisk. NPA has no effect on OsPIN1 genes except for OsPIN1c slightly induced by NPA. (E) Root gravitropism assay of rice wild type 9522 cultivar (CK). Root tip angle reaches up to 45 degrees at about 2.5 h after rotation in Rice. Root tip angles were calculated using image J software. All the data were expressed as mean ±SEM (Root Number = 8). (F) DR5 signal redistribution at epidermis cell layer and root cap region under a gravity stimulus of 2.5 h. "U" represents the epidermis layer at upper side. DR5-VENUS signals at epidermis upper side are weaker under gravistimulation. Scale bar: 25 µm. White dotted lines highlight the lower region of lateral root cap where auxin accumulates. (G) DR5-VENUS signals at the root tip after 500 nM Trans-zeatin (cytokinin) application on 3-days old rice seedlings. Scale bar: 50 µm. Trans-zeatin induces relocation of auxin response from root cap to epidermis and cell initials inside meristem. (H) Quantification of relative mRNA levels of auxin transporter genes after cytokinin treatment. Type-A cytokinin responsive genes OsRR3 and OsRR6 are selected as positive control genes. All the data were expressed as mean ± SEM from three biological and three technical repeats. Statistically significant differences (Student's t-test, \*\*P < 0.01; \*P < 0.05) are marked out. OsPIN1c and OsPIN1d are significantly reduced while OsPIN1a slightly increased under CK treatment. (A,C,F,G) Red channel, propidium iodide; Green channel, VENUS.

FIGURE 2 | Auxin expression sites in rice vegetative tissues. In (A,B), auxin highly accumulates in leaf primordia but is lacking in meristem center. Medial longitudinal view of SAM in (A) and top view of SAM in (B). The median longitudinal view of 9,522 SAM is used as a control (C) "P" marks leaf primodium, and numerals denote first, third, fourth, fifth leaf primordia, respectively. "M" indicates meristem center. (D) DR5-VENUS expression at tiller buds and their connected vasculature (V, arrows). (E) DR5-VENUS expression in veins (arrows) of mature leaf surface. (F) DR5-VENUS expression at root cap and root cap initials (RC) of a crown root primodium (CRP) at rice stem base. Scale bar: 25 µm. (A–F) are images under the fluorescent field. Red channel, chloroplast autofluorescence; Green channel, VENUS.

signals at both axils and thus the lowest auxin level. (B) Auxin reaches the maximum at the tip of tillering bud where DII signals are undetectable. (C) Highest auxin expression level is detected at the tip of root cap region of an emerging lateral root. Asterisk indicates the root cap region with no detectable DII signal. The level of DII-VENUS is shown in pseudocolor from blue to red (Spectrum LUT bar in the top right). Blue, no signal; Red, strong saturated signal intensity. Scale bar: 25µm.

Supplementary Figure S4). As to the auxin transporters, the expression level of OsPIN1c gene was statistically up-regulated, most probably due to local auxin accumulation in response to NPA treatment (**Figure 1D**). Our results suggest that the DR5- VENUS reporter can be altered indirectly by disrupting polar auxin transport.

#### Dynamic Changes in DR5-VENUS at the Root Tip Following a Gravity Stimulus

Additionally, we used DR5-VENUS to monitor dynamic changes of auxin gradients during root gravitropism. After placing the rice root horizontally, the root tip took approximately 4.5 h to return to its vertical position (**Figure 1E**). After 2.5 h, the

FIGURE 4 | Auxin distribution in rice roots. Auxin distributions in radicles of DR5-VENUS (A) and DII-VENUS (B) transgenic lines. Expression profiles of DR5-VENUS and DII-VENUS are fully complementary. Auxin maxima is visible in QC, columella cells, initials cells, X, and Epi while virtually absent from cortex, ground tissues and lateral root cap. In developing lateral roots (C,D), auxin accumulates inside QC, root cap and in flanking zone of lateral (Continued)

#### FIGURE 4 | Continued

root. During early LRP formation (E,F), auxin is visible inside developing lateral root meristem. Transverse sections show auxin localization at median planes (G,I) and nearby layers (H,J) of lateral root primodium ready to emerge from primary root. Auxin is visible in phloem (p) and again in pQC, prc and lacking inside lateral root apical meristem. Epi, epidermis; X, central metaxylem; QC, quiescent center; pQC, putative quiescent center; prc, putative root cap; p, phloem. Asterisk indicates no signal inside lateral root primodium. Red channel, propidium iodide; Green channel, VENUS. Scale bar: 25 µm.

FIGURE 5 | Morphological features of cortex cells in differentiation zone of rice roots. (A) Longitudinal section above the median plane inside the primary root of Lti6a:CFP;H2B:mCherry transgenic seedlings. Chromatin in cell nucleus was marked in red, and cell membrane was marked in green. (B1) and (B2) are the sagittal and radial views at the intersection point (C) obtained with the ortho view function of Zeiss Zen software. Arrows indicate the membrane retractation of cortex cells, and arrowheads points out the cortex cells with unrecognizable cell outlines. Images were taken and analyzed using ZEN (Zeiss) and Fiji software. Red channel, mCherry; Green channel, CFP.

most notable asymmetric pattern of DR5-VENUS activation was first seen when the root angle reached 45◦ . Weak DR5- VENUS signals in the lower half of lateral root cap cells adjacent to columella cells (highlighted with dotted lines) appeared, which were dramatically increased compared with their upper counterparts. Besides, in the meristematic zone, the fluorescent intensity at the upward side of epidermis was largely weakened, while the signals underneath remained stable (**Figure 1F** and Supplementary Figure S5), which was complementary to that of DII-VENUS (Supplementary Figure S6), but differs from that of Arabidopsis in which DR5-VENUS expression was pronouncedly elevated in lower epidermal cells after gravistimulation (Band et al., 2012; Brunoud et al., 2012). These results suggest the existence of a complex pattern of auxin distribution within root cap and epidermal tissues in rice following gravistimulation. Moreover, lateral root cap and epidermis formation results from distinct initials in monocotyledons, compare to dicotyledons where the latter differentiates from a common ones (Clowes, 1994). The existence of a root cap junction clearly separating root cap and meristem in monocotyledons may be responsible for the divergent pattern of auxin relocation in rice roots compared to Arabidopsis (Rebouillat et al., 2008; Wang L. et al., 2014).

#### Cytokinin Indirectly Induces Changes in DR5-VENUS Spatial Expression

In agreement with observations in Arabidopsis (Ruzicka et al., 2009; Shimizu-Sato et al., 2009; Shen et al., 2014), cytokinin application caused a significant increase in OsRR3, OsRR6, and OsPIN1a transcript abundance and an up regulation of auxin response in the epidermal, stele and quiescent center (QC) cells. This treatment also decreased auxin content at the root cap zone through down-regulating expression of two auxin carriers OsPIN1c and OsPIN1d (**Figures 1G,H**), confirming the existence of crosstalk between auxin and cytokinin in rice roots. This result is consistent with the antagonistic effect of gene expression profiles related to these two phytohormones in the root apex (Takehisa et al., 2012).

# Auxin Distribution during Rice Leaf Development

In the shoot apical meristem (SAM), DR5-VENUS signals were only detected at the adjacent leaf primordia, while DII-VENUS was found at the apical meristem (**Figures 2A–C**, **3A**), especially enriched at both leaf axils, suggesting that the rice SAM represents a zone of an auxin limitation, at least at a certain period of vegetative development, instead of being an auxin sink, contrasting that reported in Arabidopsis and suggesting an intricate auxin mechanism in regulating rice SAM function. Given the presence of DR5 expression in adjacent leaf primordia, auxin may be locally synthesized and contribute to leaf growth (Qin, 2005; Cheng et al., 2007; Li et al., 2008). Consistently, auxin depletion at leaf axils of Arabidopsis and tomato has been shown to be essential for axillary meristem formation (Wang Q. et al., 2014).

At the rice stem base, we observed that the DR5-VENUS signal accumulated in vascular tissues and apices of nearby tillering buds, leading us to speculate that auxin is transported from newly formed leaves basipetally down through the vasculature system for suppressing their outgrowth (**Figures 2D**, **3B**). Consistently external IAA application inhibited the growth of tiller buds by decreasing the endogenous level of cytokinin in rice (Liu et al., 2011).

Our reporter analysis also suggested that auxin may be also involved in leaf vein development and root cap formation of emerging crown root at the stem base (**Figures 2E,F**). Supportively, previous research using rice mutants showed that the impaired polar auxin transport induced defects in leaf vascular patterning (Qi et al., 2008). Moreover, auxin can affect crown root formation in rice by regulating the expression of CRL1 gene through transcription factor ARF (Inukai et al., 2005).

# Auxin Plays a Key Role in Lateral Root Development and Emergence

Rice develops a much larger and ramified root architecture compared to Arabidopsis (Chu et al., 2013; Wang L. et al., 2014; Kochian, 2016). DII-VENUS and DR5-VENUS reporters revealed that auxin forms maxima at the root cap, putative QC, stem cells and vasculature. Moreover, we found a relative higher signal in the epidermal layer of the meristematic and elongation zone in the primary roots, compared with Arabidopsis (**Figures 4A,B**), which was also confirmed by the auxin response at root hair and root surface at the differentiation zone (**Figure 1A** and Supplementary Figure S3C). Strong DR5-VENUS signals in central metaxylem, protoxylem and companion cells of the phloem were also clearly visible (Supplementary Figure S7).

Strong DR5-VENUS signals were also observed in rice lateral roots at the root tip (**Figures 4C,D**, **3C**), in LRP (**Figures 4G–J**) as well as in cortex cells overlying LRP (**Figures 4E,F**). Surprisingly, in these cells, the DR5-VENUS signal was cytoplasmic instead of the always-observed nuclear localization of the VENUS signal. We then used Lti6a:CFP;H2B:mCherry transgenic lines to follow cortex cell differentiation in mature root parts (C. Périn and M. Ingouff unpublished) (Zhang et al., 2011; Howe et al., 2012). In these lines, plasma membrane and chromatin are marked by CFP and mCherry, respectively. Cortex cells in the differentiation zone of rice roots were undergoing programmed cell death, with indistinct cell borders (arrowheads), membrane retraction (arrows) and abnormal disaggregating nuclei (**Figure 5**), paving the way for the LRP to later emerge. This result suggests there is an increase in auxin level in cortex cells surrounding LRP that may be responsible for the collapse of cortex cells during root organ emergence in rice, to be compared with the cell wall breakdown triggered in endodermal cells during LRP emergence in Arabidopsis (Peret et al., 2009).

### Auxin Distribution Is Associated with Rice Inflorescence Branching

Local auxin accumulation is required for reproductive organ initiation in Arabidopsis (Reinhardt et al., 2000; Benkova et al., 2003; Heisler et al., 2005; Yamaguchi et al., 2013), however the role of auxin distribution during rice flower development remains unclear. Specifically during the rice inflorescence formation, highly branched architecture is mainly produced from the inflorescence meristem (IM) (Zhang et al., 2013). In **Figure 6**, three new potential sites for the coming primary BMs had the obvious DR5-VENUS signals at IM (**Figure 6A**). While, as the elongation of the primary BMs, auxin response was shifted to the first several layers of BMs, and the locations where several secondary BMs were going to be formed (**Figures 6B,C**, **7A**). Consistently, auxin response were also observed in the first layer of the BMs of maize tassel IM (Gallavotti et al., 2008), although the inflorescence morphology of rice differs from that of maize (**Figures 8B,C**). Notably, no obvious accumulation of DR5-VENUS signal was documented in the first layer of Arabidopsis BMs, and maize ear IMs which produce floral or spikelet pair meristem directly without generating branching meristem (Gallavotti et al., 2008; Gallia

FIGURE 6 | Auxin distribution during rice inflorescence branch formation using DR5:VENUS. (A) Primary branch meristems (pBMs) are marked out at the rachis. Scale bar: 25 µm. (B) Late pBM, suggesting emerging sites for secondary branch meristem (sBM). Relative weak signals at the first layer were also observed (arrowheads). Inserted image is the magnified version of the pbm in the process of generating two incipient sbms in the dotted region. Scale bar: 25 µm. (C) Developing sBMs. Signals at the first layer were observed (arrowheads). Inserted image is the magnified region in the dotted box. Scale bar: 50 µm. (D) Fluorescence in spikelet meristem (SM) shows strong auxin response at glume primordia (including rgp and egp) and the presumptive vascular strands (V, arrow). rgp, rudimentary glume primodium; egp, empty glume primodium. Scale bar: 50 µm. Yellow channel, VENUS.

et al., 2015; **Figures 8A,D**), suggesting that auxin maximum at the first layer of BMs represents a sign for inflorescence branching. After secondary lateral branches are generated, the SM at the terminus of primary branch and others at secondary branches are initiated in succession, where auxin was traced at the developing vasculature of inflorescence, and also in primitive and maturing glume primordia (**Figure 6D**). These results suggest that auxin accumulation is a key determinant of rice inflorescence morphogenesis, particularly the formation of the characteristic branches.

#### Auxin Distribution in Rice Spikelet

Spikelet is a unique and fundamental structure within grass inflorescences, which bears glume instead of petal structures enclosing the floret (Zhang and Wilson, 2009; Zhang and Yuan, 2014). Unlike Arabidopsis (**Figure 8E**), SMs of rice produce a pair of glume primordia at the very onset (Itoh, 2005), during which auxin is limitedly expressed at cells of the top joint zone where the rudimentary glume attaches to the meristem (arrowhead), the incipient site for the sterile lemma (arrow), and the first cell layer of the SM (**Figure 9A**). With the growth of a pair of sterile lemmas, auxin response was seen at the transition zone of the rudimentary glume (arrowhead), the floral meristem, and sterile lemma primordia (arrow) (**Figures 9B**, **7B**), which are totally absent in maize flower.

In contrast to those in maize spikelet pairs, rice florets are bisexual since initiation (**Figures 8F–H**). Among rice spikelet organs, the lemma is the first one appeared showing a strong DR5-VENUS signals at the apex, which was verified by complete exclusion of DII-VENUS expression at this region, and relatively low auxin at the first several layers of the meristem (**Figures 9C**, **7C**), then the palea emerged out at the location where DII-VENUS signal was invisible

FIGURE 7 | Auxin distribution during flower formation using DII-VENUS sensor. (A) Secondary branch primordia formation at the primary branch. Auxin reaches maxima at these potential sites (Asterisks). (B) Strong auxin levels observed in glume primordia (Asterisks), including rudimentary and empty glume primordia. (C) Lemma primordium marked out with strong auxin (Asterisk). (D) High auxin levels at palea, lemma and floral meristem (Asterisks). (E) Three stamen primordia indicated by asterisks. Lodicule primordium (arrowhead) besides the left stamen primordium has lower auxin level. (F) Relatively strong auxin signals are visible in young stamens and apices of glume primordia (Asterisks). Green channel, VENUS. Scale bar: 15 µm.

(**Figure 7D**). Comprehensive analysis of lodicules showed that DR5-VENUS expression at lodicule primordia was detectable but relatively weak (**Figures 9D**, **7E**). At the earlier stage, an auxin response was observed located at the first layer of the stamen primordia from the top view (**Figures 9E**, **7E**), and at the inner vascular tissues of stamens seen from the longitudinal direction (**Figures 7F**, **9F**), suggesting auxin may participate in rice anther development (Qu et al., 2014). In addition, DR5-VENUS signals were detectable at the stamen and pistil primordia (**Figures 9E,F**, **10E,F**), which were also reported in maize unisexual floret (Gallavotti et al., 2008), although the development of gynoecia in maize tassel flowers, and stamen in ear flowers became a complete abortion, suggesting that auxin is essential for floral organ initiation instead of growth. Taken together, our observations suggest that auxin signaling may be essential for rice spikelet organ development.

#### Auxin Transport in Rice Spikelet Organs

Auxin flow is achieved via specific transport proteins, including influx carriers AUX1/LAXs that determine in which tissues the hormone accumulates (Band et al., 2014), and polar efflux carriers PINs whose orientations can infer the intercellular direction of auxin movement (Wisniewska et al., 2006). To probe the relationship between rice flower development and auxin transport, sub-cellular localization of PIN1 homologs in rice (Supplementary Figure S8) were determined using immunostaining of AtPIN1 antibody applicable in rice (Pasternak et al., 2015; **Figures 10A,C,E**). After the formation of two empty glumes, the incipient sites of lemma and palea primordia were specified by PIN1 polar localization which may direct auxin movement through the basal vasculature (**Figures 10A,B**). Moreover, PIN1 localization overlapped with the spatial distributions of DR5-VENUS in spikelet primordia at stage 4 when the palea primodium formed, denoting that the

auxin maxima at the spikelet primordia may be generated by the PIN1 action (**Figures 10C,D**). At the final stage of spikelet development, PIN1 exhibited strong expression in inner vascular bundles of anthers, suggesting a large amount of auxin possibly being delivered to young pollen grains (Feng et al., 2006). The PIN1 and auxin signals also remained in anther filaments during vascular tissue differentiation (**Figures 10E,F**). Besides, PIN1 signal was also visible at pistil primordium (**Figure 10E**), demonstrating that auxin may have function in affecting ovule development (Wu et al., 2015).

Auxin uptake depends on OsAUX1 (LOC\_Os01g63770) permease that modulates root initiation and elongation in rice (Yu et al., 2015). Using ProOsAUX1:OsAUX1-sGFP transgenic lines, we observed OsAUX1 specific expression in rice floral tissues (**Figure 11**). Strong OsAUX1-sGFP accumulation was visible in floral primordium after the emergence of lemma primordium (**Figures 11A,B**), and the subsequent palea primordium (**Figures 11C,D**). The OsAUX1-sGFP signals were also seen in cells at the first several outer layers of stamen primordia (**Figures 11E,F**), and carpel primodium

FIGURE 9 | Auxin distribution during spikelet development using DR5-VENUS. (A) Developing SM, during the formation of glume primordia. Signals are firstly observed at the first cell layer of SM and glume primordia, including rudimentary glume primodium (arrowheads) and empty primodium (arrows). (B) Strong auxin level in empty glume primordia (arrows and arrowheads) and SM. (C) At stage 3 of spikelet development, lemma primordium (arrows) is formed. Arrowhead indicates elongated empty glume. (D) Two lodicule primordia (arrows) with weak auxin response. (E) Strong auxin response observed at the first layer of stamen primodium (arrowheads), and vasculature of palea and lemma. (F) Longitudinal observation shows presence of strong auxin levels (arrows) inside stamen. Yellow channel, VENUS. Scale bar: 25 µm.

(**Figures 11G,H**), which overlapped with DR5-VENUS signals previously observed. Taken together, uniformly sub-localized OsAUX1-sGFP signals at cell membranes (see the magnified zones in **Figure 11**) coincided well with DR5-VENUS signals (**Figures 10**, **11**), implying that PIN1 and OsAUX1 work together to convey and redistribute auxin during rice floral organ formation.

#### DISCUSSION

We demonstrated that two traditional auxin reporters DR5- VENUS and DII-VENUS, proven useful in Arabidopsis and Maize (Gallavotti et al., 2008), are also capable of revealing auxin distribution in rice. Similar hormone distribution maps at high spatial and temporal resolution were developed in all three experimental model plants, confirming the importance of polar auxin transport in regulating plant morphogenesis both in dicot and monocot species.

The response efficiency of synthetic DR5rev promoter in vivo is higher in rice, which has broader expression profile compared to that in Arabidopsis. At rice primary root apex, besides marked signals at QC, columella cells and xylem, additional ones were also found at the epidermis layers of meristematic and elongation zones, and also within phloem cells at maturation region, which were invisible in Arabidopsis DR5-GFP or GUS lines (**Figure 2A** and Supplementary Figure S7), while partially of them were supplemented by using another auxin responsive element DR5v2 (Liao et al., 2015). Through identifying AuxRE sites in 3000-bp promoter regions of auxin early responsive gene family GH3 in rice and Arabidopsis, strikingly, we found out that the occurrence of canonical sequence TGTCTC in rice is dramatically increased (Supplementary Figure S9). We hypothesize that TGTCTC containing sequences may have a greater contribution ability to the auxin response through increase of ARF binding level in vivo in rice. Thus, screening the natural promoter conditions of auxin responsive gene families is probably a much more advisable strategy before transforming it into other plant species.

For the first time, this study provides a clear evidence that auxin plays a crucial role in rice flower formation. The overlapping localization patterns suggest that, DR5-specified auxin and its transporters PIN1, OsAUX1 signals are capable of providing positional information for flower primodium initiation (**Figures 9**, **10**). Two weeks of NPA (30 µM) treatment during

rice transition phase from vegetative to reproductive growth brought out yellow sterile inflorescence without any spikelet, at the same time, longer time and higher level of NPA (50 µM) adoption was lethal to rice plants, with deformed inflorescence arrested inside (data not shown), while IAA or NAA treatment prompted shoot apex differentiation into flower initials, further advanced rice flowering (Sircar and Kundu, 1955). However, rice Osaux1 T-DNA mutants present out inconspicuous defects in spikelet structure or fertility, which may be explained by the genetic redundancy of AUX1-like gene family in rice. Therefore,

FIGURE 11 | OsAUX1 protein localization overlaps with auxin maxima during floral development. (A,B) Spikelet meristems of ProOsAUX1:OsAUX1-sGFP transgenic lines are observed under confocal microscope. At stage 3, OsAUX1 is expressed in the floral meristem, including lemma and glume primordia, where DR5-VENUS signals are also present. Scale bar: 25 µm. (C,D) At stage 4 of spikelet development, both OsAUX1 and auxin are strongly expressed in palea primodium. Scale bar: 25 µm. (E,F) OsAUX1 is observed at lemma and first several layers of stamen primordia, overlapping with DR5 expression sites. Scale bar: 25 µm. (G,H) OsAUX1 also has a relative high expression level at carpel primodium at stage 7. Scale bar: 25 µm. The closeup view in (A,C,E,G) are magnified pictures of signals visible inside dotted squares. Green channel in (A,C,E), GFP; Green channel in (B,D,F), VENUS.

local auxin gradient formed by auxin polar transport is required for rice flower organogenesis.

In this work, combining with DR5-VENUS, we use DII-VENUS patterns as negative controls for well defining auxin distributions in rice. DII-VENUS labels out strong auxin signals with minimum of florescence, profiles of which are quite complementary with DR5 signals in almost all conditions and most notably during rice spikelet development (**Figures 5**–**7**). However, in few conditions at specific tissues, DII-VENUS reporter doesn't work well, for example in root (**Figure 4**) and shoot apical meristem (**Figure 3A**), a feature possibly caused by limited expression abilities of maize ubiquitin-1 promoter in these tissues. We hypothesize that it will be better if we replace domain II fragment of IAA28 Arabidopsis, which might possess differentiated stability and half-life characteristics in rice, with rice-specific ones, although the key residues for auxin interaction are considered generic and could be transformed into any plant species (Dreher et al., 2006; Zhang et al., 2016).

In conclusion, comparison of auxin localization and dynamic relocation between Arabidopsis and rice could help shed light on the auxin functions in angiosperms; these two biosensors represent important tools to understand the auxin signaling pathway in diverse rice developmental processes by transformation or genetic crossing method, but also to further reveal the strong link between auxin flow and agronomical traits of interest like aerial and root architecture or yield.

#### AUTHOR CONTRIBUTIONS

WL, ZY, DZ, and JY designed the experiments. JY performed the experiments. QM and GH assisted in

#### REFERENCES


immunostaining technique. CP, CB, AM, and MI generated and analyzed the Lti6a:CFP;H2B:mCherry data. JY, WL, CP, and MB analyzed the data. JY, WL, and DZ wrote the article.

#### ACKNOWLEDGMENTS

We thank Professor Klaus Plame (University of Freiburg, Germany) and Heisler MG (California Institute of Technology, USA) for providing PIN1 antibody and pDR5rev:: 3XVENUS-N7 plasmids, respectively; Yanhua Qi (State Key Laboratory of Plant Physiology and Biochemistry, Zhejiang University, China) for kindly providing ProOsAUX1:OsAUX1-sGFP seeds.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2017. 00256/full#supplementary-material


Aux/IAA gene family in rice (Oryza sativa). Funct. Integr. Genomics 6, 47–59. doi: 10.1007/s10142-005-0005-0


and polar auxin transport in root gravitropism. J. Exp. Bot. 67, 5325–5337. doi: 10.1093/jxb/erw294

**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, Yuan, Meng, Huang, Périn, Bureau, Meunier, Ingouff, Bennett, Liang and Zhang. 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.

# GSHR, a Web-Based Platform Provides Gene Set-Level Analyses of Hormone Responses in Arabidopsis

Xiaojuan Ran1,2, Jian Liu1,2, Meifang Qi1,2, Yuejun Wang1,2, Jingfei Cheng1,2 and Yijing Zhang1,2 \*

<sup>1</sup> National Key Laboratory of Plant Molecular Genetics, CAS Center for Excellence in Molecular Plant Sciences, Institute of Plant Physiology and Ecology, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai, China, <sup>2</sup> University of Chinese Academy of Sciences, Beijing, China

Phytohormones regulate diverse aspects of plant growth and environmental responses. Recent high-throughput technologies have promoted a more comprehensive profiling of genes regulated by different hormones. However, these omics data generally result in large gene lists that make it challenging to interpret the data and extract insights into biological significance. With the rapid accumulation of theses large-scale experiments, especially the transcriptomic data available in public databases, a means of using this information to explore the transcriptional networks is needed. Different platforms have different architectures and designs, and even similar studies using the same platform may obtain data with large variances because of the highly dynamic and flexible effects of plant hormones; this makes it difficult to make comparisons across different studies and platforms. Here, we present a web server providing gene set-level analyses of Arabidopsis thaliana hormone responses. GSHR collected 333 RNA-seq and 1,205 microarray datasets from the Gene Expression Omnibus, characterizing transcriptomic changes in Arabidopsis in response to phytohormones including abscisic acid, auxin, brassinosteroids, cytokinins, ethylene, gibberellins, jasmonic acid, salicylic acid, and strigolactones. These data were further processed and organized into 1,368 gene sets regulated by different hormones or hormone-related factors. By comparing input gene lists to these gene sets, GSHR helped to identify gene sets from the input gene list regulated by different phytohormones or related factors. Together, GSHR links prior information regarding transcriptomic changes induced by hormones and related factors to newly generated data and facilities cross-study and cross-platform comparisons; this helps facilitate the mining of biologically significant information from large-scale datasets. The GSHR is freely available at http://bioinfo.sibs.ac.cn/GSHR/.

Keywords: GSHR, Arabidopsis thaliana, web-based platform, hormone, transcriptome, gene set-level, cross-study, cross-platform

# INTRODUCTION

Phytohormones are typically small endogenous compounds regulating every aspect of plant life, from plant growth and development to responses to environmental changes (Santner and Estelle, 2009; Wolters and Jurgens, 2009; Scheres and van der Putten, 2017). Specific phytohormones are synthesized in response to distinct developmental or environmental cues, which are perceived

#### Edited by:

Diwakar Shukla, University of Illinois at Urbana–Champaign, United States

#### Reviewed by:

Manpreet Singh Katari, New York University, United States Balaji Selvam, University of Illinois at Urbana–Champaign, United States

> \*Correspondence: Yijing Zhang zhangyijing@sibs.ac.cn

#### Specialty section:

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

Received: 29 June 2017 Accepted: 08 January 2018 Published: 24 January 2018

#### Citation:

Ran X, Liu J, Qi M, Wang Y, Cheng J and Zhang Y (2018) GSHR, a Web-Based Platform Provides Gene Set-Level Analyses of Hormone Responses in Arabidopsis. Front. Plant Sci. 9:23. doi: 10.3389/fpls.2018.00023

**405**

by specific receptors and further activate cascades of signaling pathways, leading to remarkable biochemical and physiological changes (Santner and Estelle, 2009; Wolters and Jurgens, 2009; Scheres and van der Putten, 2017). The major hormones were traditionally classified into two categories, those mainly regulating developmental processes including auxin, brassinosteroid (BR), cytokinin (CK), and gibberellic acid (GA) and strigolactones, and those regulating stress responses, including abscisic acid (ABA), ethylene (ET), jasmonic acid (JA), and salicylic acid (SA). Over the past decades, wide-spread crosstalk was revealed among various phytohormone pathways in response to specific developmental or environmental changes (Depuydt and Hardtke, 2011; Robert-Seilaniantz et al., 2011; Murphy, 2015). Typical examples include the antagonistic role of GA and ABA in controlling seed dormancy (del Carmen Rodríguez-Gacio et al., 2009; Ye and Zhang, 2012; Shu et al., 2016), the interaction between auxin and CK in regulating meristem development (Kuderova and Hejatko, 2009; Su et al., 2011; Azizi et al., 2015). The crosstalk happens on multiple layers of regulation, among which transcriptional control is one major component.

Genetic studies identified various specific transcription factors mediating phytohormone signaling pathways; mutations in these factors resulted in plants that became more sensitive or insensitive to hormone treatments (Moore et al., 2011; Eckardt, 2015; Tsuda and Somssich, 2015). The interaction among factors regulating hormone signaling and factors controlling growth and responses mediates the crosstalk of different hormones in particular conditions. For example, the antagonistic interaction between auxin and CK in controlling root meristem development is achieved via the regulatory circuit between auxin signaling factors SHORT HYPOCOTYL 2 (SHY2/IAA3) and CK signaling regulator ARABIDOPSIS RESPONSE REGULATOR1 (ARR1) (Kuderova and Hejatko, 2009; Su et al., 2011). In another example, the central component regulating dark morphogenesis phytochrome interacting factor (PIF4) actively participated in auxin and BR-mediated pathways (Oh et al., 2012, 2014; Wang et al., 2014; Chaiwanon et al., 2016; Franciosini et al., 2017). To systematically characterize the crosstalk is non-trivial due to the complicated signaling network. The transcription of 100–1000 of genes display significant changes in response to various hormone treatments, and to dissect their common and unique targets is the pre-requisite for a comprehensive understanding of the crosstalk. During the past decades, high-throughput technologies have promoted systematic characterization of the signaling pathways triggered by various phytohormones and related factors. For example, microarray and RNA-seq experiments were performed to profile the transcriptomic changes triggered by phytohormones (Paponov et al., 2008; Garg et al., 2012) while chromatin immunoprecipitation followed by sequencing (ChIP-seq) studies have facilitated the identification of genes regulated by hormone-related factors (Oh et al., 2012, 2014; Song et al., 2016; Zhu et al., 2016). It was revealed that both phytohormones and related factors have wide-spread targets, and much research on phytohormones now involves investigating how plant hormones interact with each other and with related factors to orchestrate plant transcriptomic network (Depuydt and Hardtke, 2011; Eckardt, 2015). Accordingly, it is important to compare the high-throughput data obtained from different studies and from different platforms. However, such comparison is a challenging task for a number of reasons. Firstly, different platforms have different architectures and designs, and the results are not directly comparable (Barnes et al., 2005; Nookaew et al., 2012). For example, the probe intensity obtained from the Affymetrix platform cannot directly be compared to that from the Agilent platform due to different technical procedures such as the probe design, chip fabrication and data analysis (Sah et al., 2010; Del Vescovo et al., 2013); the results obtained from ChIP-seq experiments are not directly comparable to those from RNA-seq studies. Secondly, given the dynamic and flexible effects of phytohormones, similar experiments from different labs or even from different repeats in the same lab may generate data with large variances that are not readily comparable (Lee and Park, 2010; Rivas-San Vicente and Plasencia, 2011). To obtain more robust results, researchers proposed the idea of modulelevel analysis; this derives its power by focusing on gene sets sharing common biological functions or expression behavior. The basic procedure is to pre-define specific gene sets based on the comprehensive collection of information from dry-lab and wet-lab experiments. For any given gene list, the potential function of these genes could be deduced based on large amounts of prior information. Typical examples include Gene Set Enrichment Analysis (GSEA) and the Database for Annotation, Visualization and Integrated Discovery (DAVID), which are mostly for animals (Dennis et al., 2003; Subramanian et al., 2005), PlantGSEA for plants (Yi et al., 2013), and Comprehensive Annotation of Rice Omics-data (CARMO) for rice (Wang et al., 2015).

Given the vast amount of hormone-related transcriptomic data accumulated in public databases, the ability to make full use of available high-throughput data is of great significance. Here, we present gene set-level analyses of hormone responses in Arabidopsis (GSHR), a web server that provides analyses of phytohormone responses based on the integration of hormoneresponsive gene modules. GSHR collected 333 RNA-seq and 1,205 microarray samples from the Gene Expression Omnibus that were organized into 1,368 gene lists regulated by different hormones or hormone-related factors. This gene list-level analysis helps reach more reliable and robust conclusions about which genes under study were affected by what types of plant hormones or related factors; three examples were used to illustrate the power of GSHR in data mining of hormone related information.

#### MATERIALS AND METHODS

#### Data Collection

1,538 phytohormone-related transcriptomic datasets in Arabidopsis thaliana were collected from Gene Expression Omnibus (GEO<sup>1</sup> ), including 73 studies with 1,205 datasets

<sup>1</sup>http://www.ncbi.nlm.nih.gov/geo/

generated using microarray platforms (Affymetrix or Agilent), and 21 studies with 333 datasets generated by RNA-seq based on Illumina sequencing platform (Supplementary Table 1). All relevant comparisons between datasets of the same study were performed, then we defined the up-regulated and downregulated genes in one pair comparison respectively as a gene set, finally generating 1,368 gene sets. We extracted the hormones or the transcription factors, mutants, or genes involved in hormone signaling used in the comparison as hormone or hormone-related factors of the gene set.

#### Process of RNA-Seq Data

RNA-seq reads with MAPQ < 20 were filtered followed by adapter trimming using trim galore<sup>2</sup> with default settings. The cleaned reads were mapped to Col-0 genome (Araport11) using STAR (Dobin et al., 2013), and the number of reads in each gene was counted. Differentially expressed genes were calculated via DEseq (Anders and Huber, 2010) with the combined criteria: |log2(fold change)| >1, p value<0.05.

#### Process of Microarray Data

Microarray data from Affymetrix platform were pre-processed with Bioconductor package affy (Gautier et al., 2004; Gentleman et al., 2004), and gcrma function was used for background correction and normalization (Wu and Irizarry, 2004). Microarray data from Agilent platform were normalized using quantile normalization (Bolstad et al., 2003). In the case of genes with multiple probes, the probe with the largest expression intensity was kept. For both platforms, linear model in Bioconductor package limma (Smyth, 2005) was applied for identifying differentially expressed genes with combined criteria:|log<sup>2</sup> (fold change)| >1 and p value<0.01. Comparisons resulting in >10 differentially expressed genes were kept.

#### Enrichment Analysis

Fisher's exact test (Routledge, 2005) was used for the enrichment analysis, which is displayed as follows:

$$\mathbf{p} = \frac{\binom{\mathbf{n}}{\mathbf{k}} \binom{\mathbf{N} \cdot \mathbf{n}}{\mathbf{K} \cdot \mathbf{k}}}{\binom{\mathbf{N}}{\mathbf{K}}}$$

where N is the total number of genes as background. Different N is used for different analysis. For comparison of gene sets from transcriptomic data, there are 24,327 genes in total; and the numbers for KEGG, gene ontology and InterPro domain analysis are 4,797, 30,468 and 15,816, respectively. n is the number of input genes. K is the total number of genes in one gene set and k strands for the overlapped genes between input and the pre-defined gene sets. For multiple testing correction, the adjusted P values calculated by FDR, Bonferroni correction and Benjamini and Hochberg method were provided.

The enriched fold change is the fraction of input genes involved in the given gene set divided by the fraction of all annotated genes involved in the gene set. This is commonly used for functional term enrichment analysis, e.g., DAVID bioinformatics Resources: https://david.ncifcrf.gov/term2term. jsp. The calculation formula is:

$$\text{Fold Change} = \frac{\text{n/k}}{\text{N/K}}$$

The values of N, n, k and K are descripted as above and presented in the download file, corresponding to PopTotal, ListTotal, Count and PopHits columns, respectively.

#### Construction of Co-expression Network

We selected 22,090 genes present in both microarray and RNAseq datasets for construction of co-expression network. log2(fold change) of each gene from all pair-wise comparisons were collected, and the function rcorr in R package Hmisc (Harrell, 2017) was used to calculated the Pearson correlation coefficients and p-values between a pair of genes. The genes pairs with coefficients above 0.70 were used to construct the co-expression network.

#### Web Server Implementation

The GSHR was constructed on Apache HTTP Server based on Linux system, MySQL was used for storage and operation of the database, the web interface is supported by PHP and JavaScript scripts. We used Python for statistical analysis and data processing, cluster analysis was performed in R. The JavaScript Library d3.js was used for co-expression network visualization. All scripts are available upon request.

<sup>2</sup>http://www.bioinformatics.babraham.ac.uk/projects/trim\_galore/

# RESULTS

#### Structure of GSHR

fpls-09-00023 January 23, 2018 Time: 16:5 # 4

GSHR integrates thousands of hormone-related transcriptomic data sets and offers a user-friendly interface for gene list-level analyses of user input gene lists (**Figure 1**). GSHR accepts gene lists with Arabidopsis thaliana Genbank IDs (**Figure 2A**) and returns the hormones and hormone-related factors potentially regulating the input gene list, ranked by statistical significance. Detailed descriptive and statistical information about the comparisons is also presented (**Figures 2B–D**). For those hormones regulated input gene lists, further functional exploration could be directly performed on the resulting page, including hierarchical clustering of expression pattern

or factors associated with the comparisons are listed in the 1st and 2nd columns. (E) Optional functional analyses of selected genes from the summary page shown in (D) including clustering and visualization of the expression pattern, co-expression network, enrichment analysis of pathways (KEGG), functions (Gene Ontology, GO) and domains (InterPro).

and factors (A), the clustering and visualization of the expression pattern in response to abscisic acid (ABA) and auxin (IAA) and ethylene (ET) (B), the co-expression network (C) and the enriched GO terms (D).

#### in response to related hormones, co-expression network, enrichment analyses of pathways, gene ontology terms and InterPro domains. The purpose is to facilitate a more comprehensive understanding about the input gene list, and helping pinpointing essential genes of interest from the input (**Figure 2E** and Supplementary Figure 1). The manual page of the website provides detailed guidelines.

#### Examples of Use

#### Case 1: Hormones Involved in Plant Response to Phosphate Deficiency

To illustrate the usage of GSHR, we used genes up-regulated by phosphate starvation in Arabidopsis thaliana roots as input (Sun et al., 2016). The output given by GSHR showed that a significant proportion of the input genes were regulated by


FIGURE 4 | The resulting page summarizing relevant hormones and functional terms enriched for cadmium stress response. The most relevant hormones and factors (A), the enriched pathways (B), and GO terms (C).

ABA, auxin and ET (**Figure 3A**), indicating that these hormones were potentially involved in plant responses to phosphate deficiency. In support of the results, ABA-responsive genes were induced when cultured in low level of phosphate (Ciereszko and Kleczkowski, 2002; Chiou and Lin, 2011). In addition, low phosphorus enhanced the sensitivity of root response to

auxin and ET in Arabidopsis, resulting in root architecture changes (Lopez-Bucio et al., 2002; Ma et al., 2003). Following identification of the relevant hormones, the enriched gene sets could be selected for further functional analyses. Clustering and visualization were provided to characterize their expression responses to selected hormones. The heatmap suggested that the selected gene sets were preferentially up-regulated by IAA and ABA, and both up and down-regulated genes were observed by treatment of ET (**Figure 3B**). In addition, genes with high degree in the co-expression network are good candidates of key factors linking signaling pathways between phosphate signaling and hormone responses (**Figure 3C**). Enriched GO terms include ET biosynthesis and ABA signaling pathway as expected (**Figure 3D**).

#### Case 2: Hormones Participated in Plant Response to Cadmium Stress

Cadmium is a toxic non-biodegradable heavy metal which negatively affect plant growth and development (Bucker-Neto et al., 2017). We used a set of cadmium-induced gene from Arabidopsis root as input to search for hormones involved in plant response to cadmium stress (Lopez-Martin et al., 2008). The GSHR identified SA, ABA, GA and JA as top enriched plant hormones (**Figure 4A**). This result is consistent with previous studies. SA influenced the absorption and transport of cadmium and acted as a signaling molecule activating cadmiumtolerant genes (Liu et al., 2016); JA could alleviate negative impacts of cadmium stress by increasing activities of antioxidant enzymes in soybean (Keramat et al., 2009); GA reduced nitric oxide accumulation in roots and suppressed cadmium uptake in Arabidopsis to alleviate cadmium toxicity (Zhu et al., 2012); ABA concentration was increased when exposed to cadmium, which induced transient MAP kinase activity thereby increasing cadmium tolerance (Burnett et al., 2000; Bucker-Neto et al., 2017). Further KEGG and GO enrichment results of selected gene sets indicated that these genes mainly related to plant circadian rhythm, starch and sucrose metabolism, ABA and stress response (**Figures 4B,C**), which were also reported to be involved in cadmium responses (Devi et al., 2007; Maistri et al., 2011).

#### Case 3: PIF4 Targets Were Enriched for Auxin- and ABA-Responsive Genes

Besides differentially expressed genes from transcriptomic studies, GSHR accepts gene lists obtained from any source including different types of high-throughput data. Here, PIF4 target genes identified via ChIP-seq experiment (Pedmale et al., 2016) were used as input to illustrate this application of GSHR. AUXIN RESISTANT 2 (AXR2)- and BRASSINOSTEROID INSENSITIVE 1 (BRI1)-regulated genes were among the most enriched gene sets (**Figure 5A**). AXR2 belongs to the Auxin/Indole Acetic Acid protein family essential for mediating auxin signaling (Nagpal et al., 2000) that was also reported to be involved in the BR signaling network (Nemhauser et al., 2004; Nakamura et al., 2006). BRI1 is a leucine-rich repeat receptor kinase involved in BR signal transduction (Wang et al., 2001; Kinoshita et al., 2005). Consistent with the enrichment result, recent genetic and high-throughput studies revealed that PIF4 is closely associated with auxin- and BR-mediated signaling pathways. (Oh et al., 2012, 2014; Wang et al., 2014; Franciosini et al., 2017). PIF4 targets a significant proportion of the genes targeted by AUXIN RESPONSE FACTOR 6 (ARF6) and BRASSINAZOLE-RESISTANT 1 (BZR1), the core mediators of auxin and BR signaling, respectively (Chaiwanon et al., 2016). Input genes could be further clustered according to their expression changes in arx2 and bri1 mutants (**Figure 5B**).

#### DISCUSSION AND FUTURE PLAN

GSHR provides data-driven analysis to search for hormones and hormone-related factors affecting transcription of input genes. The purpose of this platform is to link previously generated transcriptomic information to newly generated data, thereby helping to underpin biological insights from omics data and provide clues for subsequent research. This is different from traditional plant hormone databases that have mostly focused

#### REFERENCES


on providing comprehensive hormone-related information for individual genes (Peng et al., 2009; Jiang et al., 2011).

Focusing on gene sets facilitates comparison across different platforms and different experiments. However, pair-wise comparison generates multiple similar gene sets resulted from comparisons from similar experiments. The resulting page could be more concise via further clustering or removing redundant results. We choose to provide all pair-wise comparison results because different growth conditions and different way of treatments may affect non-identical gene sets, and all these information may be informative and helpful to users. With the rapid accumulation of high-throughput data, annotations, and references, the gene sets in GSHR will be updated annually.

#### AUTHOR CONTRIBUTIONS

XR and JL constructed the web server. XR, MQ, YW, and JC collected and processed public data. XR and YZ wrote the manuscript.

### FUNDING

This work was supported by grants from National Natural Science Foundation of China (31570319 and 31770285), and sponsored by CAS Center for Excellence in Molecular Plant Sciences and Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018.00023/ full#supplementary-material



Zhu, X. F., Jiang, T., Wang, Z. W., Lei, G. J., Shi, Y. Z., Li, G. X., et al. (2012). Gibberellic acid alleviates cadmium toxicity by reducing nitric oxide accumulation and expression of IRT1 in Arabidopsis thaliana. J. Hazard. Mater. 239-240, 302–307. doi: 10.1016/j.jhazmat.2012.08.077

**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 BS and handling Editor declared their shared affiliation.

Copyright © 2018 Ran, Liu, Qi, Wang, Cheng and Zhang. 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.