# INTERACTIONS OF PLANTS WITH BACTERIA AND FUNGI: MOLECULAR AND EPIGENETIC PLASTICITY OF THE HOST

EDITED BY : Valentina Fiorilli, Marco Catoni, Luisa Lanfranco and Nicolae Radu Zabet PUBLISHED IN : Frontiers in Plant Science and Frontiers in Microbiology

#### Frontiers eBook Copyright Statement

The copyright in the text of individual articles in this eBook is the property of their respective authors or their respective institutions or funders. The copyright in graphics and images within each article may be subject to copyright of other parties. In both cases this is subject to a license granted to Frontiers. The compilation of articles constituting this eBook is the property of Frontiers.

Each article within this eBook, and the eBook itself, are published under the most recent version of the Creative Commons CC-BY licence. The version current at the date of publication of this eBook is CC-BY 4.0. If the CC-BY licence is updated, the licence granted by Frontiers is automatically updated to the new version.

When exercising any right under the CC-BY licence, Frontiers must be attributed as the original publisher of the article or eBook, as applicable.

Authors have the responsibility of ensuring that any graphics or other materials which are the property of others may be included in the CC-BY licence, but this should be checked before relying on the CC-BY licence to reproduce those materials. Any copyright notices relating to those materials must be complied with.

Copyright and source acknowledgement notices may not be removed and must be displayed in any copy, derivative work or partial copy which includes the elements in question.

All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For further information please read Frontiers' Conditions for Website Use and Copyright Statement, and the applicable CC-BY licence.

ISSN 1664-8714 ISBN 978-2-88963-654-9 DOI 10.3389/978-2-88963-654-9

#### About Frontiers

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

#### Frontiers Journal Series

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

#### Dedication to Quality

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

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

#### What are Frontiers Research Topics?

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

## INTERACTIONS OF PLANTS WITH BACTERIA AND FUNGI: MOLECULAR AND EPIGENETIC PLASTICITY OF THE HOST

Topic Editors: Valentina Fiorilli, University of Turin, Italy Marco Catoni, University of Birmingham, United Kingdom Luisa Lanfranco, University of Turin, Italy Nicolae Radu Zabet, University of Essex, United Kingdom

Citation: Fiorilli, V., Catoni, M., Lanfranco, L., Zabet, N. R., eds. (2020). Interactions of Plants With Bacteria and Fungi: Molecular and Epigenetic Plasticity of the Host. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-654-9

# Table of Contents

*05 Editorial: Interactions of Plants With Bacteria and Fungi: Molecular and Epigenetic Plasticity of the Host*

Valentina Fiorilli, Marco Catoni, Luisa Lanfranco and Nicolae Radu Zabet


Qiang Yan, Jierui Si, Xiaoxia Cui, Hao Peng, Maofeng Jing, Xin Chen, Han Xing and Daolong Dou

*47 Differences in Ear Rot Resistance and* Fusarium verticillioides*-Produced Fumonisin Contamination Between Polish Currently and Historically Used Maize Inbred Lines*

Elżbieta Czembor, Agnieszka Waśkiewicz, Urszula Piechota, Marta Puchta, Jerzy H. Czembor and Łukasz St¸epień

*61 Cytoprotective Co-chaperone BcBAG1 is a Component for Fungal Development, Virulence, and Unfolded Protein Response (UPR) of*  Botrytis cinerea

Honghong Zhang, Yurong Li, Martin B. Dickman and Zonghua Wang

*76 Spatial Effects and GWA Mapping of Root Colonization Assessed in the Interaction Between the Rice Diversity Panel 1 and an Arbuscular Mycorrhizal Fungus*

Hazel Davidson, Roshi Shrestha, Thomas Cornulier, Alex Douglas, Tony Travis, David Johnson† and Adam H. Price

*90 Plant Aquaporins in Infection by and Immunity Against Pathogens – A Critical Review*

Liyuan Zhang, Lei Chen and Hansong Dong

*106 Comparative Transcriptomics and Proteomics of* Atractylodes lancea *in Response to Endophytic Fungus* Gilmaniella *sp. AL12 Reveals Regulation in Plant Metabolism*

Jie Yuan, Wei Zhang, Kai Sun, Meng-Jun Tang, Piao-Xue Chen, Xia Li and Chuan-Chao Dai

*124 Diazotrophic* Paenibacillus beijingensis *BJ-18 Provides Nitrogen for Plant and Promotes Plant Growth, Nitrogen Uptake and Metabolism*

Yongbin Li, Yunlong Li, Haowei Zhang, Minyang Wang and Sanfeng Chen

*142 Arbuscular Mycorrhizal Symbiosis: Plant Friend or Foe in the Fight Against Viruses?*

Laura Miozzi, Anna Maria Vaira, Marco Catoni, Valentina Fiorilli, Gian Paolo Accotto and Luisa Lanfranco


# Editorial: Interactions of Plants With Bacteria and Fungi: Molecular and Epigenetic Plasticity of the Host

Valentina Fiorilli <sup>1</sup> \*, Marco Catoni <sup>2</sup> \*, Luisa Lanfranco<sup>1</sup> \* and Nicolae Radu Zabet <sup>3</sup> \*

<sup>1</sup> Department of Life Sciences and Systems Biology, University of Turin, Turin, Italy, <sup>2</sup> School of Biosciences, University of Birmingham, Birmingham, United Kingdom, <sup>3</sup> School of Life Sciences, University of Essex, Colchester, United Kingdom

Keywords: plant-bacterial interactions, plant-fungal interactions, molecular mechanisms, omics tools, genotypes

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

#### **Interactions of Plants with Bacteria and Fungi: Molecular and Epigenetic Plasticity of the Host**

#### Edited by:

Víctor Flors, University of Jaume I, Spain

#### Reviewed by:

Juan Antonio Lopez Raez, Experimental Station of Zaidín (EEZ), Spain

#### \*Correspondence:

Valentina Fiorilli valentina.fiorilli@unito.it Marco Catoni m.catoni@bham.ac.uk Luisa Lanfranco luisa.lanfranco@unito.it Nicolae Radu Zabet nzabet@essex.ac.uk

#### Specialty section:

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

Received: 14 January 2020 Accepted: 21 February 2020 Published: 04 March 2020

#### Citation:

Fiorilli V, Catoni M, Lanfranco L and Zabet NR (2020) Editorial: Interactions of Plants With Bacteria and Fungi: Molecular and Epigenetic Plasticity of the Host. Front. Plant Sci. 11:274. doi: 10.3389/fpls.2020.00274 In both natural and agricultural environments plants live in association with a multitude of microorganisms belonging to different microbial types, mainly bacteria and fungi. Some of these microbes regulate positively plant growth and productivity, while others can damage the host with important ecological and economic consequences. During the last decades, many studies contributed to unravel the multifaceted process at the basis of plant microbe-interactions (Cheng et al., 2019). However, many issues remain still unsolved, as for example how plants could discriminate between beneficial and pathogenic microbes or between different pathogen attackers, or which gene regulatory networks are responsible for host-microbe interactions and their degree of conservation among species.

When challenged by pathogens, plants trigger highly complex defense system, in order to recognize invader organisms and translate this signal into defense such as the expression of defense response genes. This plant immune system relies on a wide variety of different strategies, showing high plasticity in the response depending on the attacker lifestyle. These processes involve the regulation of genes, small RNAs, signal molecules, plant hormones, which can act locally or systemically through plant organs.

On the other hand, plants are strictly associated with microbial symbionts in natural environments, often poor in nutrients (van der Heijden et al., 2008). Beneficial microbes in the soil could help the host to overcome the nutritional and abiotic stresses, boost plant growth and fitness, and sustain plant productivity. In addition, beneficial associations can enhance the defensive ability of plants, resulting in faster and stronger defense activation upon pathogens attacks. While these phenomena have been widely described, the underling molecular mechanisms remain elusive. Changes in transcription, protein regulation and phytohormones accumulation have been reported during plant-beneficial microbe interactions. Moreover, epigenetic modifications triggering stable changes in plant's transcriptional capacity are emerging as relevant modulator of plant's responses to microbes, with potential role in memory and priming (Alonso et al., 2019).

With 13 original contributions, this Research Topic provides an overview of the current state of the art on the field of plant microbe-interactions. This Topic includes a combination of Reviews, Mini Reviews and Original Research Articles, focused on the role of the molecular infrastructures evolved by plants to manage different microbe-interactions, revealing that a complex plant–microorganism genotype and environment combinations could determine the outcome of the interaction.

Recently, genome wide RNA sequencing has become a popular approach to study transcriptional changes also in non-model organisms. Yuan et al. investigated the increased growth and sesquiterpenoids accumulation induced by the endophyte Gilmaniella in the medicinal herb Atractylodes lancea combining transcriptomic (RNAseq) and proteomic approaches. Authors observed that the presence of the endophyte induced in the host suppression of genes involved in plant immunity and signaling, while genes involve in both primary and secondary metabolism such as phenylpropanoid and zeatin biosynthesis were upregulated.

An emerging theme is the significance of genetic variation in differentiating the biological response to harmful and beneficial microorganisms. In this Research Topic a number of contributions addressed the role of different host genotypes in plant-microbe interactions. Czembor et al. studied 98 maize inbreed lines, historically used in Poland, in relation to resistance to Fusarium, across a 2 year in-field experiment. They coupled HPLC analysis for the detection of fumolisin content to NGS approach (ddRADseq) to infer genetic distances across the lines; they then correlate genetic distance with resistance to Fusarium. Authors observed large differences in resistance and fumolisin accumulation across the lines and concluded that old lines represent a valuable source of resistant traits against Fusarium.

Another emerging and powerful approach to identify novel genes involved in plant-microbe interactions is represented by Genome-Wide Association (GWA) studies. The contribution of plant genetic variability to the positive effects of arbuscular mycorrhizal (AM) symbiosis have been investigated in different crops (Diedhiou et al., 2016; Lehnert et al., 2017; Watts-Williams et al., 2019), and, in this line, Davidson et al. tested the responsiveness to AM colonization of 334 rice cultivars inoculated with the AM fungus Rhizophagus irregularis. GWA mapping for hyphal colonization revealed 23 quantitative trait loci (QTLs) with putative impact on AM fungal colonization and identified candidate genes associated to three QTLs.

More specific topics on plant-pathogen interactions are addressed by Cao et al. who investigated the hypersensitive response (HR) against Xanthomonas oryzae pv. oryzae (Xoo) in rice mediated by genes ascribed to the major disease resistance pathway. By observing the physiological response to infection in the xylem parenchyma, authors concluded that dominant resistant genes mediate HR by prevalent autophagy-like cell death, while recessive genes induce HR by vacuole-mediated cell death.

Legumes and vegetables pathogens affect field-grown and greenhouse-grown crops worldwide. Two works investigated, in different host plants, the role of a key component of the plant disease-associated signal transduction pathway (Guo et al.) and a negative regulator of programmed cell death (PCD) process (Yan et al.). Both reports highlighted how manipulating the expression of a specific gene of interest, using a genetic approach, could determine the biological function of genes in the plant resistance process. These results pave the way for the identification of molecular targets potentially useful for breeding programs to control pathogen resistance in crops.

Two studies, Zhang H. et al. and Yang et al., focused on the study of the microbial partner, characterizing two genes which contribute to Botrytis cinerea virulence, one of the most notorious pathogenic species, in different host plants.

An emerging theme in the field of plant-microbe interactions is the relevance of mineral nutrients availability. Mineral nutrients are not only important for the growth and development of plants and microorganisms but they also play a key role in the dynamics and the outcome of plant -pathogenic/beneficial microbes interactions. Li et al. studied the association of the diazotrophic bacteria Paenibacillus beijingensis with wheat, maize and cucumber in conditions of high and low nitrogen in the soil, and they observed beneficial effects on plant growth related to improved nitrogen uptake and assimilation, in line to previous findings from other bacterial/host combinations (Xie et al., 2016; Hao and Chen, 2017). Zhang L. et al. reviewed current knowledge on the roles of plant aquaporins (AQPs) of the plasma membrane intrinsic protein family. The authors highlighted that AQPs are not only involved in maintaining the plant water or mineral nutrition status, but they also contribute to control plant immune system and pathogen susceptibility.

In nature, plants are very likely subjected to multiple complex interactions involving several organisms, rather than single bipartite relationships. Although less investigated, multiple interactions represent a system more similar to the natural environment, and their study can assist to select or design optimal bio-fertilization or bio-control procedures. In this Research Topic, different contributions addressed this issue. Abbasi et al. compared the ability of different plant growth promoting rhizobacteria (PGPR) and the application of a chemical fungicide to antagonize Fusarium oxysporum f. sp. lycopersici race 3 (FOL) in tomato plants. Interestingly, all bacterial treatments mitigated FOL disease symptoms at the same level or better than chemical treatments. In a minireview, Miozzi et al. provided an overview of the impact of the arbuscular mycorrhizal symbiosis on plant viral diseases. The authors proposed the term "Mycorrhizal-Induced Susceptibility" (MIS) to describe the enhanced viral infection reported in many tripartite interactions (plant-mycorrhiza-virus), in opposition to the Mycorrhizal-Induced Resistance (MIR) that is often observed in response to infection by bacterial and fungal pathogens. Finally, Jones et al. summarized recent research that expands upon the role of keystone microbial species, phytohormones, and abiotic stress and how they relate to plant driven dynamic microbial structuring.

Plants are extremely plastic in their interaction with microorganisms, and the heterogeneous composition of the works published in this Research Topic well-represents the variety of responses, model and non-model organisms and experimental approaches used to investigate this subject. The knowledge of the determinants and the mechanisms that regulate plant-microbe interactions with different level of complexity can be instrumental for the development of new agrobiotechnological strategies of crop protection, with the aim to improve food security and environmental sustainability.

#### AUTHOR CONTRIBUTIONS

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

## FUNDING

Research in the laboratories of the Topic Editors was supported by the European Union's Horizon 2020 research and innovation

#### REFERENCES


programme under grant agreement no. 727929 (TOMRES) and from Competitive Research Grant CRG2017 given to LL from King Abdullah University of Science and Technology and 60% Project (University of Turin) given to VF.


**Conflict of Interest:** 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 © 2020 Fiorilli, Catoni, Lanfranco and Zabet. 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.

# The Botrytis cinerea Xylanase BcXyl1 Modulates Plant Immunity

Yuankun Yang, Xiufen Yang, Yijie Dong\* and Dewen Qiu\*

The State Key Laboratory for Biology of Plant Diseases and Insect Pests, Institute of Plant Protection, Chinese Academy of Agricultural Sciences, Beijing, China

Botrytis cinerea is one of the most notorious pathogenic species that causes serious plant diseases and substantial losses in agriculture throughout the world. We identified BcXyl1 from B. cinerea that exhibited xylanase activity. Expression of the BcXyl1 gene was strongly induced in B. cinerea infecting Nicotiana benthamiana and tomato plants, and BcXyl1 deletion strains severely compromised the virulence of B. cinerea. BcXyl1 induced strong cell death in several plants, and cell death activity of BcXyl1 was independent of its xylanase activity. Purified BcXyl1 triggered typically PAMP-triggered immunity (PTI) responses and conferred resistance to B. cinerea and TMV in tobacco and tomato plants. A 26-amino acid peptide of BcXyl1 was sufficient for elicitor function. Furthermore, the BcXyl1 death-inducing signal was mediated by the plant LRR receptorlike kinases (RLKs) BAK1 and SOBIR1. Our data suggested that BcXyl1 contributed to B. cinerea virulence and induced plant defense responses.

#### Edited by:

Luisa Lanfranco, Università degli Studi di Torino, Italy

#### Reviewed by:

Franco Faoro, Università degli Studi di Milano, Italy Mario Serrano, Universidad Nacional Autónoma de México, Mexico

#### \*Correspondence:

Yijie Dong nkdongyijie@163.com Dewen Qiu qiudewen@caas.cn

#### Specialty section:

This article was submitted to Plant Microbe Interactions, a section of the journal Frontiers in Microbiology

Received: 01 August 2018 Accepted: 04 October 2018 Published: 23 October 2018

#### Citation:

Yang Y, Yang X, Dong Y and Qiu D (2018) The Botrytis cinerea Xylanase BcXyl1 Modulates Plant Immunity. Front. Microbiol. 9:2535. doi: 10.3389/fmicb.2018.02535 Keywords: Botrytis cinerea, xylanase, BcXyl1, plant immunity, virulence

#### INTRODUCTIONS

Botrytis cinerea is a necrotrophic pathogen, causing widespread plant diseases and enormous economic losses in a large number of important crops throughout the world (Prins et al., 2000). B. cinerea can infect various organs in plants, including leaves, bulb, flowers, fruits, and root tubers. The infection process of B. cinerea mainly includes two typical stages: local lesions at an early stage and a late stage of fast-spreading lesions.

The plant cell wall is a natural barrier, which provides mechanical strength and rigidity to prevent pathogen infection. To establish successful colonization, B. cinerea, like other fungal pathogen, secretes a large number of cell wall-degrading enzymes (CWDEs) to degrade the plant defensive barriers during the infection process, thereby to permit pathogens to invade plant tissue and supply pathogens with nutrients (Cantarel et al., 2009; Kubicek et al., 2014). These CWDEs, including pectinases, cellulases, cutinases, and xylanases, are generally regarded as important virulence factors through the maceration of host tissues and the degradation of host macromolecules (Prins et al., 2000). The effects of targeted deletion of some genes encoding CWDEs support their direct involvement in the infection process. For example, deletion of the pectate lyase gene CcpelA and the pectate lyase gene PelB in Colletotrichum coccodes, resulted in a substantial loss of virulence on green tomato fruit and reduced virulence on avocado, respectively (Yakoby et al., 2001; Ben-Daniel et al., 2011). Targeted deletion of VdCUT11, a cutinase in V. dahliae, significantly compromised virulence on cotton plants (Gui Y. et al., 2017). However, the specific roles of the majority of CWDEs in pathogen virulence remain largely unknown, especially in B. cinerea.

**8**

To ward off microorganisms infection, plants have evolved elaborate systems to provide better immunity against pathogens (Zipfel, 2008). Recognition of conserved pathogen-associated molecular patterns (PAMPs) via pattern recognition receptors (PRRs) located on the cell surface constitutes the first layer of plant innate immunity and is termed as PAMP-triggered immunity (PTI). Intracellular responses associated with PTI include Ca2<sup>+</sup> influx, the burst of reactive oxygen species (ROS), the accumulation of defense hormone, the expression of defenserelated genes and callose deposition (Boller and Felix, 2009; Couto and Zipfel, 2016). In turn, during the coevolution of hosts and microbes, pathogens also employ numerous effectors to interfere with PTI and establish successful infection, which is regarded as effector-triggered susceptibility (ETS) (Chisholm et al., 2006; Jones and Dangl, 2006; Saijo et al., 2017). As a countermeasure, some plants recruit R proteins to recognize these effectors directly or indirectly termed effector-triggered immunity (ETI) (Houterman et al., 2008; Stergiopoulos and de Wit, 2009). Generally, ETI is often accompanied with stronger immune responses, such as hypersensitive response (HR).

Apart from the role of virulence factor, some CWDEs also function as PAMPs to activate the plant immune responses independent of their enzymatic activity. For instance, VdEG1, VdEG3 and VdVUT11 from Verticillium dahliae, XEG1 from Phytophthora sojae and BcXYG1, a secreted xyloglucanase from B. cinerea contributed to virulence and triggered plant immunity as PAMPs simultaneously (Ma et al., 2015; Gui Y. et al., 2017; Gui Y.-J. et al., 2017; Zhu et al., 2017).

Plants recognizes characteristic microbial molecules classically known as PAMPs by employing a multitier surveillance system, including PRRs (Couto and Zipfel, 2016). Plant PRRs include RLKs and receptor-like proteins (RLPs) (Boutrot and Zipfel, 2017). Currently, a handful of PRRs have been identified as receptors to participate in the recognition of PAMPs. The brassinosteroid insensitive 1 (BRI1)-associated receptor kinase 1 (BAK1) and the LRR receptor-like kinase (LRR-RLK) SUPPRESSOR OF BIR1-1 (SOBIR1) are involved in multiple PRR pathways and signal activation (Liebrand et al., 2014). For example, BcSpl1, XEG1, and VdCUT11 could trigger cell death in the plants, and the resulting immunity signal was mediated by the plant LRR RLKs BAK1 and SOBIR1 (Frías et al., 2011; Ma et al., 2015; Gui Y. et al., 2017).

Xylan is the major component of hemicellulose of the plant cell wall (Collins et al., 2005). Due to the complexity, the degradation of xylan requires several hydrolytic enzymes, of which xylanase is a crucial component for hydrolyzing the 1,4-β-d-xylosidic linkages in xylan. Xylanase has received more attention because of the special role in fungi pathogenicity. For example, a mutation in the xynB endoxylanase gene from Xanthomonas oryzae pv. oryzae resulted in attenuated virulence in rice (Pandey and Sonti, 2010). Moreover, the deletion of xylanases Xyn11A gene had a marked effect on the ability of B. cinerea to infect tomato leaves and grape (Brito et al., 2006). In addition to their roles in virulence, xylanases are regarded as elicitors to induce defense responses in plants. For example, ethylene-inducing xylanase (EIX) is a potent elicitor in tobacco and tomato. However, the function of the majority of xylanases in B. cinerea remains mostly undiscovered. Here, we reported on the identification and characterization of BcXyl1, a xylanase from B. cinerea. BcXyl1 contributes to B. cinerea virulence and triggers PTI responses in plants. Furthermore, a small peptide of BcXyl1 is sufficient for elicitor function. We found that the cell death signal is mediated by BAK1 and SOBIR1, and the xylanases activity is not necessary for the induction of necrosis.

#### MATERIALS AND METHODS

#### Fungal Cultures, Plants Grown

Botrytis cinerea B05.10 was used as wild-type strain and control strain in this study. All B. cinerea strains, including two independent BcXyl1 knockout mutants and two complementary transformants, were routinely maintained in 15% glycerol at −80◦C and grown on PDA at 22◦C, respectively. Agrobacterium tumefaciens AGL-1 were grown on LB (Kan and Rif) medium at 28◦C. To obtain conidia, B. cinerea grown on tomato-PDA plates (39 g of potato dextrose agar plus 250 g of homogenized tomato fruits per liter) as explained previously (Benito et al., 1998). N. benthamiana and tomato (Solanum lycopersicum) plants were grown at 27 ◦C in a greenhouse with a day/night period of 14/10 h and 60% relative humidity (RH).

#### Expression and Purification of Recombinant Protein

The open reading frame of BcXyl1 (amplified with primers BcXyl1 F/BcXyl1 R; **Supplementary Table S2**) and C130−<sup>155</sup> were amplified by PCR from cDNA of the wild-type strain B05.10 and the fragment fused with a myc tag and a 6xHis tag at the C terminus was cloned into the pPICZαA vector at the BamHI and EcoRI sites. The recombinant plasmid pPICZαA-BcXyl1 and pPICZαA-C130−<sup>155</sup> were linearized with PmeI and transformed into Pichia pastoris KM71H for expression. The transformed yeasts were grown and induced in BMGY (buffered glycerol complex medium) and BMMY (buffered methanol complex medium), respectively (Easy Select Pichia expression kit; Invitrogen). Then, the supernatant was collected (3000 g for 10 min at 4◦C) and purified using nickel affinity chromatography. The purified C130−155, BcXyl1rec, or BcXyl1 were kept in protein buffer (20 mM Tris, pH 8.0) and further detected via SDS-PAGE and Western blotting. The concentration of the purified protein was measured using Easy II Protein Quantitative Kit (BCA) and the protein was then stored at −80◦C.

#### Truncated Mutant Construction and Agroinfiltration Assay

To transiently express truncated mutants of the BcXyl1 protein in leaves, DNA sequences encoding different fragments (BcXyl1, N80, N130, N<sup>155</sup> C <sup>80</sup>, C130, C155, and C130−155) were amplified by PCR from cDNA of the wild-type strain B05.10 and inserted into pYBA1132 vector at the XbaI and BamHI sites and then transformed into the A. tumefaciens strain GV3101. Agroinfiltration assays were performed on N. benthamiana plants. Agrobacterium-mediated transient expression was performed as described (Ma et al., 2015). Leaves were scored and photographed 6 days after initial inoculation. Each assay was performed on six leaves from three individual plants, and repeated at least three times.

#### Site-Directed Mutagenesis

fmicb-09-02535 October 19, 2018 Time: 18:22 # 3

To determine the relationship between the enzymatic activity and cell death-inducing activity of BcXyl1, we constructed BcXyl1rec mutant, which abolished the enzymatic activity. According to multiple sequence alignment, two potentially highly conserved catalytic residues (E104 and E157) were the critical catalytic sites of BcXyl1. Next, two glutamic acid residues were substituted by Gln using the Quick ChangeTM Site-Directed Mutagenesis Kit (Stratagene, United States). BcXyl1rec was expressed in P. pastoris and carried out the enzyme assay.

#### Xylanase Assay

The xylanase activity was assayed via the method as described previously (Biely et al., 1988). The purified BcXyl1rec or BcXyl1 (500 ng) and substrate (1% beechwood xylan) were co-incubated in citrate phosphate McIlvaine buffer, pH 5, at 35◦C for 10 min (total volume: 125 µl). All samples were incubated at 100◦C for 10 min to end the assays reactions. The amount of reducing sugars released from xylan was quantified using a standard calibration curve obtained with the dinitrosalicylic acid procedure. The experiment was replicated three times.

#### Immunoblot Analysis

To confirm whether BcXyl1 was secreted into the apoplast and the relationship between enzymatic activity of BcXyl1 and cell death-inducing activity, transient expression in N. benthamiana was performed. Three sequences (BcXyl1, BcXyl1−1SP, and BcXyl1rec) were cloned into the pYBA1132 vector which contained a C-terminal GFP tag at the XbaI and BamHI sites, and then transformed into the A. tumefaciens strain GV3101. All primers are listed in **Supplementary Table S2**. Plant total protein extractions and immunoblots were assessed as previously described (Yu et al., 2012). All proteins were analyzed by immunoblots using anti-GFP-tag primary monoclonal antibody. The blots were visualized using the Odyssey <sup>R</sup> LI-COR Imaging System. Rubisco was used to confirm the equal protein loading.

#### Protein Infiltration Assays and Induction of PTI by BcXyl1

To test the induction of cell death or PTI responses, BcXyl1 and C130−<sup>155</sup> were dissolved in PBS and infiltrated into the leaves of N. benthamiana and tomato plants using a syringe. Plants were grown in a greenhouse with a day/night period of 14/10 h. The cell-death response was investigated after 48 h treated with BcXyl1, C130−155, or PEVC (P. pastoris culture supernatant from an empty vector control strain, purified in the same way as BcXyl1). To further investigate cell death, trypan blue staining was performed by boiling leaf tissues in a mixture of phenol, lactic acid, glycerol, and distilled water containing 1 mg/ml trypan blue (1:1:1:1) for 1 min. The samples were then soaked in 2.5 mg/ml chloral hydrate overnight. The accumulation of ROS in plant leaves was stained by 3<sup>0</sup> 3 diaminobenzidine (DAB) and Nitroblue Tetrazolium (NBT) solution as described previously (Bindschedler et al., 2006). To visualize callose deposition, 4-week-old N. benthamiana leaves were infiltrated with 1 µM recombinant proteins and stained with aniline blue at 24 h post-treatment, as described previously (Chen et al., 2012). To assay electrolyte leakage, the N. benthamiana leaves treated with proteins were harvested at different time points and submerged in sterile water at 4 ◦C. Ion conductivity was measured using a conductivity meter. To test whether BcXyl1 could confer plants disease resistance, the purified BcXyl1, C130−155, or PEVC was individually syringe-infiltrated into 4-week-old N. benthamiana and tomato leaves. Five microliters of 2 × 10<sup>6</sup> conidia/ml B. cinerea and TMV-GFP were placed on the systemic leaves, respectively. The inoculated plants were placed in a greenhouse at 25◦C with a day/night period of 14/10 h. Lesion diameter of B. cinerea and the number of TMV-GFP lesions on N. benthamiana leaves were evaluated at 2 and 4 days postinoculation, respectively. All experiments were performed three times.

#### Pathogenicity Assays

To test whether BcXyl1 functioned as a virulence factor of B. cinerea, the wild-type strain and derived mutants, including the BcXyl1 deletion (1BcXyl1-1 and 1BcXyl1-2) and complementary mutants (1BcXyl1-1-C and 1BcXyl1-2- C) were used in this study. Four-week-old N. benthamiana leaves were inoculated with 5 µL of 2 × 10<sup>6</sup> conidia/ml B. cinerea. The inoculated plants were placed in a greenhouse with a day/night period of 14/10 h. The lesion development of B. cinerea on the N. benthamiana leaves was evaluated at 2 days post-inoculation by determining the average lesion diameter. Tomato, grape, and apple fruits (commercially obtained) were washed under running tap water and surface sterilized by immersion for 5 min in ethanol. After air drying, fruits were inoculated with 5 µL of 2 × 10<sup>6</sup> conidia/ml B. cinerea. Fruits were incubated at 25◦C under conditions of high humidity on water-soaked filter paper in closed containers. The lesion development of B. cinerea on the fruits was evaluated at 3 days post-inoculation by determining the average lesion diameter. All the experiments were performed three times.

#### VIGS in N. benthamiana

To determine whether BAK1 or SOBIR1 participate in induction of cell death by BcXyl1, VIGS was performed. NbBAK1 or NbSOBIR1 gene was silenced using VIGS, as described previously (Kettles et al., 2016). A. tumefaciens strain harboring constructs (pTRV1, pTR::BAK1 or pTRV1, pTRV2::SOBIR1) were infiltrated into the N. benthamiana leaves. pTRV2::GFP was used as the control and the expression levels of BAK1 and SOBIR1 were determined by qRT-PCR. Agroinfiltration assays were performed on N. benthamiana plants using Bcl-2-associated X protein (BAX) as positive controls. Phenotypes were photographed 6 days after infiltration. All the experiments were performed three times.

#### RNA Extraction and qRT-PCR

fmicb-09-02535 October 19, 2018 Time: 18:22 # 4

To measure the expression of BcXyl1 during infection, 4 week-old N. benthamiana or 4-week-old tomato plants were inoculated with B. cinerea 2 × 10<sup>6</sup> conidia/ml. We selected 10 indicated time points during different stages of post-inoculation to determine expression patterns of BcXyl1 by qPCR. All samples were stored at −80◦C. Total RNA of B. cinerea was extracted with the E.Z.N.A. <sup>R</sup> Total RNA Kit I according to the manufacturer's instructions and stored at −80◦C. For the measurement of defense-related genes expression, leaves of 4-week-old N. benthamiana plants were treated with 1 µM purified BcXyl1, C130−155, or PEVC. The leaves were obtained at the indicated time points, immediately frozen in liquid nitrogen, and stored at −80◦C. The EasyPure Plant RNA Kit (TransGen Biotech) was used to extract total RNA. After isolation of total RNA, qPCR was performed using a TransStart Green qPCR SuperMix UDG according to the manufacturer's instructions (TransGen Biotech). qRT-PCR was performed under the following conditions: an initial 95◦C denaturation step for 10 min followed by 40 cycles of 95◦C for 15 s and 60◦C for 1 min. N. benthamiana EF-1a (P43643.1) and B. cinerea Bcgpdh gene (BC1G\_05277) were used as endogenous plant controls and used to quantify fungal colonization, respectively. qPCR assays were repeated at least twice, each repetition with three independent replicates (Livak and Schmittgen, 2001). All primers are listed in **Supplementary Table S2**. The relative transcript levels among various samples were determined using the 2−11CT method with three independent determinations (Livak and Schmittgen, 2001).

#### Generation of BcXyl1 Deletion and Complementary Mutants

BcXyl1 gene and 500 bp flanking sequences of the target gene were amplified from B. cinerea B05.10 wild-type strain genomic DNA. Two flanking sequences of the target gene and hygromycin resistance cassette were constructed into a fusion fragment using a nested PCR reaction, which is subsequently introduced into the binary vector pGKO2 gateway. To generated complementary transformants, the donor vector pCT-HN containing BcXyl1 gene was integrated into the mutant transformants using a previously described Agrobacterium-mediated transformation method (Liu et al., 2013). All mutants were identified using PCR with the corresponding primers. All primers are listed in **Supplementary Table S2**.

#### Statistical Analysis

All the experiments and data presented here were performed at least three repeats. The data are presented as the means and standard deviations. Statistical Analysis System (SAS) software was used to perform the statistical analysis via Student's t-test.

## RESULTS

#### Amino Acid Sequence Analysis of VdCP1 BcXyl1

BcXyl1 was identified by searching the B. cinerea genome sequence. The open reading frame of BcXyl1 (GenBank: ATZ53308.1) is 987 bp encoding a 329 aa protein with a predicted N-terminal signal peptide (1–20 aa), and no transmembrane helices of BcXyl1 were found, suggesting that it may be secreted into extracellular space. The bioinformatics analysis suggested that BcXyl1 belongs to SGNH hydrolase subfamily and has a highly strong similarity to fungal endo-β-1,4-xylanases.

#### BcXyl1 Contributes to B. cinerea Virulence

Previous studies showed that xylanases in pathogenic microorganisms were implicated in the pathogenicity. In

order to assess the role of BcXyl1 to B. cinerea virulence, we first analyzed the expression patterns of BcXyl1 during different stages of post-inoculation. qRT-PCR results suggested that when the spore suspension of B. cinerea was inoculated onto leaves of N. benthamiana and tomato, transcript level of BcXyl1 increased rapidly and reached a maximum of about 26-fold to 28-fold at 2 days post-inoculation, and then rapidly declined and maintained a level that was slightly higher than the initial level during later stages (**Figure 1**).

To more directly explore the biological roles of BcXyl1 during infection, we constructed two BcXyl1 deletion mutants in B. cinerea (1BcXyl1-1 and 1BcXyl1-2) and two rescued strains (1BcXyl1-1-C and 1BcXyl1-2-C), and the ability of the resulting mutants to infect various plant organisms was evaluated. All mutants showed no significant differences with the wild-type stain in growth rate and colony morphology on PDA plates (**Supplementary Figure S2**). N. benthamiana leaves were inoculated with spore suspension of the wild type and mutants, and lesion size was measured 48 h after inoculation. Interestingly, the deletion of BcXyl1 displayed significantly reduced virulence and produced much smaller lesions on leaves of N. benthamiana than the WT strain at 48 hpi (**Figures 2A,B**). The rescued strains recovered the high virulence phenotypes. And two BcXyl1 deletion mutants displayed much weaker disease symptoms and lesion diameter than the wild-type strain and the complement strains (1BcXyl1-1-C and 1BcXyl1-2-C) on grape, tomato, and apple fruits 72 h post-inoculation (**Figures 2A,B**). These results indicated that BcXyl1 functioned as a virulence factor that contributes to B. cinerea virulence on host plants.

#### BcXyl1 Is a Secreted Protein to Induce Cell Death in Several Plant Species

To further confirm whether BcXyl1 could induce cell death in N. benthamiana, we expressed BcXyl1 in the yeast P. pastoris using the pPICZαA vector (pPICZαA: BcXyl1). Moreover, the recombinant protein BcXyl1, with a size of 35 kDa, was infiltrated into the mesophyll of N. benthamiana leaves with different concentrations (**Supplementary Figure S3**). The area of necrosis occurred and increased with increasing concentrations of BcXyl1 from 800 nM to 2 µM after infiltration 3 days, whereas no cell death activity was detected in the leaves treated with PEVC (P. pastoris culture supernatant from an empty vector control strain, purified in the same way as BcXyl1) (**Figures 3A,B**).

To examined the cell death-inducing activity of BcXyl1 in plants other than N. benthamiana, we infiltrated BcXyl1 (1 µM) into the leaves of several plants, including tomato, soybean, and cotton. BcXyl1 could induce significant cell death in these plants while PEVC did not (**Figure 3C**). So, BcXyl1 has ability to induce cell death in several plant species.

BcXyl1 has a signal peptide with 20 amino acids and no transmembrane helices, implying that BcXyl1 might be a secreted protein. In order to check if, as previously hypothesized, BcXyl1 was secreted into the apoplast to induce cell death response, we transiently expressed the full length BcXyl1 and BcXyl1−1SP (lacking the signal peptide) in N. benthamiana by agroinfiltration. The results showed that BcXyl1 containing signal peptide induced cell death in N. benthamiana, whereas BcXyl1−1SP lacking signal peptide abolished the ability to trigger cell death at 5 days after agroinfiltration (**Supplementary Figure S4A**). The protein expression level of BcXyl1 and BcXyl1−1SP in N. benthamiana were detected by immunoblot (**Supplementary Figure S4B**). So, all results showed BcXyl1 was delivered into the apoplast to induce cell death in several plant species.

#### The Cell Death-Inducing Activity Is Independent of the Xylanase Activity of BcXyl1

Previous reports showed that xylanases from fungi had ability to degrade xylan (Brutus et al., 2005). Interestingly, purified BcXyl1 had a xylanase activity using low viscosity xylane (LVX) as substrate (**Supplementary Table S1**). The sequence alignment results showed that BcXyl1 included two potentially highly

conserved catalytic residues (E104 and E157), which are essential for the xylanase activity (**Supplementary Figure S1**). In addition, the enzymatic activity of CWDEs was required for cell death activity, and in a few cases, the cell death-inducing activity was found to be independent of the enzymatic activity. To determine the relationship between the enzymatic activity and cell death-inducing activity of BcXyl1, we generated a sitedirected mutant (BcXyl1rec) that two glutamic acid residues were substituted by Gln using site-directed mutagenesis and expressed the mutant protein in P. pastoris (**Figure 4A**). Enzymatic assays with purified BcXyl1rec showed the xylan-degrading xylanase activity was abolished (**Supplementary Table S1**). Surprisingly, although BcXyl1rec absent the ability of xylanase activity, retained the same cell death-inducing activity as the wild-type (BcXyl1) (**Figure 4B**). Further, A. tumefaciens infiltration assays showed that BcXyl1rec and BcXyl1 induced similar visible cell death symptoms in N. benthamiana leaves 4 days post-inoculation (**Figure 4B**). Western blot assays showed that the accumulation of BcXyl1 and BcXyl1rec was similar (**Figure 4C**). These results confirmed that BcXyl1 did not need to xylanase activity to induce cell death in N. benthamiana.

#### BcXyl1 Triggers the PTI Responses

Some cell death-inducing proteins are recognized by plant immune system and activate host PTI responses, bring a series of typical characteristics such as accumulation of ROS, leakage

FIGURE 4 | Cell death activity of BcXyl1 is independent of its xylanase activity in N. benthamiana. (A) Schematic presentation of the examined constructs. BcXyl1 (the native protein) and BcXyl1rec (replaced E104 and E157 with Gln). (B) Upper pictures: treatment of tomato leaves with 1 µM purified BcXyl1 or BcXyl1rec. Lower pictures: analysis of cell death produced by A. tumefaciens strains transiently expressing BcXyl1 or BcXyl1rec . (C) Immunoblot analysis of proteins from N. benthamiana leaves transiently expressing the examined proteins using a pYBA1132 vector. All the experiments were replicated three times.

of ion electrolytes, expression of defense genes, and callose deposition (Frías et al., 2012; Zhang et al., 2014, 2015). To examine whether BcXyl1 could induce typical PTI responses, the leaves of N. benthamiana and tomato plants were infiltrated with 1 µM BcXyl1. The ability of BcXyl1 to induce the accumulation of ROS in the infiltrated leaves was studied. The hydrogen peroxide (H2O2) and superoxide anion (O2−) production levels were assayed using DAB and NBT, respectively. A clear brown and blue precipitate was observed in leaves treated with BcXyl1, whereas the leaves treated with PEVC showed opposite patterns of DAB and NBT signal (**Figure 5A**). Meanwhile, BcXyl1 also induced electrolyte leakage and displayed an increase in conductivity, while PEVC exhibited barely change at the same concentration (**Figure 5B**). BcXyl1 was shown to cause significantly upregulation of seven genes associated with PTI and defense response in N. benthamiana leaves 12 h after treatment with BcXyl1; these genes included PR-1a and PR-5, which are involved in the SA-dependent defense pathway, PAL (phenylalanine ammonia lyase), NPR1 (the non-expressor of pathogenesis related 1), HSR203J and HIN1, which are two HR marker genes in tobacco, and COI1 (CORONATINE INSENSITIVE 1), which is JA responsive (**Figure 5C**). We finally examined callose deposition in leaves treated with BcXyl1, PEVC, or flg22. Furthermore, N. benthamiana leaves infiltrated with BcXyl1 or flg22 exhibited strong callose deposition compared with those infiltrated with PEVC, which exhibited undetectable levels of callose deposition (**Figure 5D**). These data indicated that BcXyl1 could induce typical PTI responses.

#### BcXyl1 Confers Plants Disease Resistance

Recent reports showed that fungi CWDEs could confer plants disease resistance (Gui Y. et al., 2017; Gui Y.-J. et al., 2017; Zhu et al., 2017). To further confirm the role of BcXyl1 in conferring resistance to plant diseases, the N. benthamiana leaves were treated with 1 µM BcXyl1or PEVC, and after 2 days, the systemic leaves were inoculated with TMV-GFP and B. cinerea spore suspension. BcXyl1-treated tobacco plants enhanced disease resistance against TMV, and the number of TMV-GFP lesions of BcXyl1-treated leaves was significantly decreased than that of the leaves treated with PEVC (**Figure 6A**). Meanwhile, BcXyl1 led to more resistance to the B. cinerea infection in N. benthamiana, as significantly lower lesions size on leaves compared with the leaves treated with PEVC controls (**Figure 6B**). Furthermore, in tomato plants that were pre-infiltrated with BcXyl1, lesion size on the B. cinerea-infected leaves was significant smaller compared with lesions size on leaves in plants that were pre-infiltrated with PEVC (**Figure 6C**). Together, these results strongly suggested that BcXyl1 conferred plants disease resistance.

FIGURE 5 | BcXyl1 induces PTI responses in plants. N. benthamiana or tomato leaves were infiltrated with 1 µM purified BcXyl1, C130−155, or PEVC. (A) ROS accumulation was detected in tobacco and tomato leaves 12 h after infiltration. The treated leaves were stained with DAB and NBT. All the experiments were replicated three times. (B) The conductivity was measured at the indicated time points. Error bars represent standard errors. All the experiments were replicated three times. (D) The expression of defense response genes was measured in N. benthamiana leaves by qPCR. Error bars represent standard deviation of three independent replicates. Student's t-test was performed to determine the significant differences between BcXyl1, C130−155, and PEVC. Asterisks "∗∗" indicate statistically significant differences at a p-value <0.01. (C) Callose deposition in N. benthamiana leaves were detected 2 days after infiltration; the treated leaves were stained with aniline blue. All the experiments were replicated three times.

#### A Small Peptide of BcXyl1 Is Sufficient for Elicitor Function

The plant receptors often recognize specific small protein epitopes of PAMP to induce plant immunity (Rotblat et al., 2002). To delineate the elicitor active peptide of BcXyl1, we generated N-terminal and C-terminal truncated mutants and detected the ability to induce cell death by agroinfiltration in N. benthamiana leaves (**Figure 7A**). We found that the N-terminal truncated mutant (N155) maintained the ability of cell death-inducing, whereas expression of N<sup>80</sup> and N<sup>130</sup> did not trigger cell death. The C-terminal truncated mutants (C<sup>80</sup> and C130) induced cell death, but C<sup>155</sup> resulted in the loss of cell death-inducing activity in N. benthamiana. Further, C130−<sup>155</sup> induced the same cell death symptom compared with full-length BcXyl1 in N. benthamiana (**Figure 7B**). Hence, C130−<sup>155</sup> was identified as the functional peptide of BcXyl1 to induce cell death in N. benthamiana. To probe whether C130−<sup>155</sup> induced plant immune responses, purified C130−<sup>155</sup> was used to infiltrate plants leaves. We found that like BcXyl1, C130−<sup>155</sup> could induce typical PTI responses, including accumulation of ROS, leakage of ion electrolytes, expression of defense genes, and callose deposition (**Figure 5**). Meanwhile, C130−<sup>155</sup> could also enhance resistance to B. cinerea and TMV in plants (**Figure 6**). These results suggested that a small peptide of BcXyl1 was sufficient for elicitor function.

#### BAK1 and SOBIR1 Mediates BcXyl1-Triggered Cell Death in N. benthamiana

The plant PRRs, such as the LRR RLKs BAK1 and SOBIR1, were employed to participate in multiple PRR pathways, including cell death induction (Monaghan and Zipfel, 2012; Liebrand et al., 2014; Gravino et al., 2016). For example, BAK1 was required for cell death inducing of GH12 members (Ma et al., 2015;

Zhu et al., 2017). As demonstrated above, BcXyl1 was secreted into the apoplast to induce cell death. To determine whether BAK1 and SOBIR1 participated in induction of cell death by BcXyl1, we used virus-induced gene silencing (VIGS) to induce the gene silencing of BAK1 or SOBIR1 in N. benthamiana leaves. Three weeks after viral inoculation to silence BAK1, transient expression of BcXyl1 in N. benthamiana did not result in cell death after agroinfiltration with BcXyl1 expression constructs. Treatment of BAK1-silenced plants with Bcl-2 associated protein X (BAX) was used as a control, which resulted in cell death induction (**Figure 8A**). The results of SOBIR1 silenced plants were in accordance with BAK1-silenced plants, BcXyl1 did not trigger cell death, while BAX was still capable of inducing cell death (**Figure 8A**). Immunoblotting confirmed that BcXyl1 were successfully expressed at the expected size in N. benthamiana plants inoculated with TRV::BAK, TRV::SOBIR1, or TRV::GFP (**Figure 8B**). qPCR analysis confirmed that the expression of BAK1 or SOBIR1 expression was markedly reduced upon inoculation with the TRV::BAK or TRV::SOBIR1, with an expression level about 20% in comparison with inoculation with TRV::GFP (**Figure 8C**). From these results, we inferred that BAK1 and SOBIR1 (a LRR-RLP/SOBIR1/BAK1 complex) were required for BcXyl1-triggered cell death in N. benthamiana.

#### DISCUSSION

Botrytis cinerea, a necrotrophic plant pathogen, attacks the plant organs, including leaves, flowers, fruits, bulb, and root tubers, and causes serious plant diseases and substantial losses in agriculture throughout the world (Williamson et al., 2007; Ky et al., 2012). Like other phytopathogenic fungi, B. cinerea secretes vast array of proteins during infection process (Fillinger and Elad, 2016). Cell wall-degrading enzymes (CWDEs) are the largest class of B. cinerea-secreted proteins (Kubicek et al., 2014). Recent studies have revealed that several CWDEs functioned as virulence factors in plant pathogens and were also recognized as PAMPs by plant PRRs to trigger the PTI responses, during plant–pathogen interactions (Ma et al., 2015). In this study, we described the identification and analysis of BcXyl1, a secreted xylanase from B. cinerea. BcXyl1 had the ability to induce cell death and plant PTI responses independent of its enzymatic activity. Furthermore, our study also found that a small peptide from BcXyl1 was sufficient for elicitor activity. VIGS assays showed that a LRR-RLP/SOBIR1/BAK1 complex modulates BcXyl1-triggered cell death in N. benthamiana. We also found that BcXyl1 functions as a virulence factor that contributes to B. cinerea virulence on host plants.

Increasing evidence demonstrated that xylanases are responsible for the pathogenesis of necrotrophic phytopathogens, including B. cinerea (Schouten et al., 2007; Frías et al., 2011). For instance, xyn11A was an endo-β-1,4-xylanase belonging to family 11 of glycoside hydrolase and required for virulence in B. cinerea, and the deletion of the xynB gene encoding an endoxylanase distinctly reduced the virulence of Xanthomonas oryzae pv. oryzae (Brito et al., 2006; Pandey and Sonti, 2010). In this study, we found BcXyl1 appeared to be a major virulence factor. Strikingly, BcXyl1 was strongly induced and accumulated during the early stage of infection, and the mutation of BcXyl1 had a severe effect on pathogenicity (**Figures 1**, **2**). It is noteworthy that not all fungal xylanases have been conclusively involved in pathogenicity and virulence. So far, gene deletion experiments in Fusarium oxysporum, Fusarium graminearum, Magnaporthe grisea, and Cochliobolus carbonum did not support an essential role for xylanases in fungal pathogenesis (Apel, 1993; Wu et al., 1997; Gómez-Gómez et al., 2002; Santhanam et al., 2013; Sella et al., 2013). In addition, previously study showed that a xylanase from B. cinerea could contribute to virulence by promoting the necrosis of the plant tissue surrounding the infection (Noda et al., 2010). Interestingly, a few nanograms of purified BcXyl1

resulted in a rapid leaf tissue necrosis in soybean, tomato, cotton, and N. benthamiana (**Figure 3C**). The range of plant species responding to BcXyl1 may be larger than we detected.

Previous studies showed that the enzymatic activity of many fungal CWDEs was required for cell death-inducing activity (Gui Y. et al., 2017). However, in certain cases, the cell death inducing activity was found to be unrelated to the enzymatic activity (Ma et al., 2014, 2015). For instance, Xyn11A, a xylanase from B. cinerea, and the Trichoderma viride EIX could induce cell death in plants independent of the xylanase activity (Furman-Matarasso et al., 1999; Noda et al., 2010). Although BcXyl1 is a xylanase, induction of cell death did not require the enzymatic activity (**Figure 4**).

Our results showed that BcXyl1 was localized to the plant apoplast by a signal peptide experiment, suggesting that the cell death-inducing activity may be mediated by surfacelocalized PRRs (**Supplementary Figure S5**). Plant surfacelocalized PRRs such as RLKs and RLPs were involved in the recognition of PAMPs (Boutrot and Zipfel, 2017). In addition, BAK1, as a co-receptor, plays a regulatory role in receptor complexes that mediate PTI (Schulze et al., 2010; Liu et al., 2013; Gravino et al., 2016; Yamada et al., 2016). And SOBIR1 is also specifically required for the function of receptor complexes (Liebrand et al., 2014). VIGS assays confirmed that tobacco BAK1 was required for BcXyl1-induced cell death, and the cell-death response also disappeared in VIGS-SOBIR1 plants (**Figure 8**). Hence, the RLP–SOBIR1– BAK1 complex mediated the cell death-inducing activity of BcXyl1.

The detection of PAMPs by plant PRRs to trigger PTI is a major component of plant defense responses. We confirmed that BcXyl1 triggered typical defense responses, including accumulation of ROS, leakage of ion electrolytes, deposition of callose, and expression of defense genes (**Figure 5**). We also found that the recombinant BcXyl1 proteins conferred systemic resistance in N. benthamiana, which offered protection against TMV and B. cinerea (**Figure 6**).

Generally, PAMPs are perceived by PRRs via specific epitopes, and the small peptides located on the surface of the proteins are sufficient to stimulate immune responses. For example, a 35 amino acid peptide of BcIEB1 could trigger necrosis and the PTI responses (González et al., 2017). Similarly, a 30-amino acid peptide of Xyn11A mediated the induction of cell death (Noda et al., 2010). The small peptide of VdEG3 from the GH12 domain was sufficient to induce cell death in N. benthamiana (Gui Y.-J. et al., 2017). In this study, progressive truncation of BcXyl1

confirmed that a region with 26 amino acids was sufficient for elicitor function (**Figures 5–7**).

Previous studies showed that many fungal xylanases involved in inducing plant defense responses immunity. For instance, the xylanase EIX from T. viride was an elicitor to induce defense responses in tomato, pepper and tomato plants (Rotblat et al., 2002). A xylanase from F. graminearum could induce cell death and hydrogen peroxide accumulation in wheat leaves (Sella et al., 2013; Moscetti et al., 2014). We have also determined that BcXyl1 induced plant defense responses and conferred tobacco and tomato plants disease resistance. Therefore, we speculated that fungal xylanases have the ability to trigger immunity in dicot and monocot plants.

Successful pathogens deliver effectors to interference the host PTI response and establish infection (Jones and Dangl, 2006; Gimenez-Ibanez et al., 2009). For example, a RXLR effector and CBM1 effector suppressed XEG1-triggered immunity in oomycetes and suppressed the GH12 protein and BcXyl1 triggered immunity in V. dahlia, respectively. Whether effectors mediate the suppression of BcXyl1, needs further investigation.

#### AUTHOR CONTRIBUTIONS

YD and DQ designed the experiments. YY performed most of the experiments and wrote the paper. XY participated in some part of the study and the Graduate Student Innovation Scientific Research Subject of Hainan Province (Hyb2017-17).

#### FUNDING

This study was supported by the National Natural Science Foundation of China (no. 31701782 and no. 31371984).

#### ACKNOWLEDGMENTS

We are grateful to X. F. Dai from the Institute of Food Science and Technology, Chinese Academy of Agricultural Sciences, for the generous gift of the vectors for gene knockout and complementation. We also thank L. Yao from Beijing Academy of Agriculture and Forestry for providing the pYBA1132 plasmid and pPICZαA plasmid. And YY thanks the meticulous care

#### REFERENCES


and selfless support of Ms. Nianhua Jiang over the past 4 years.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.02535/full#supplementary-material

FIGURE S1 | Sequence alignment of BcXyl1 and xylanases from other fungi. Two red triangles indicated possible catalytic residues of BcXyl1 (E104 and E157). Sequence data of all proteins can be found in the GenBank/EMBL data libraries under accession numbers: XynBc1 (ACF16413.1), BcXyl1 (ATZ53308.1), XynG1 (XP\_001258363.1), BcXyl2 (XP\_001546507.1), BcXyl3 (ATZ58346.1), BcXyl4 ( ATZ51455.1), NpGH11 (EOD46026.1), and SsGH11 (XP\_001588545).

FIGURE S2 | BcXyl1 deletion strains do not show developmental defects. BcXyl1 deletion mutants (1BcXyl1-1 and 1BcXyl1-2), rescued strains (1BcXyl1-1-C and 1BcXyl1-2-C). (A) The radial growth and colony morphology were observed after 8 days of incubation on PDA medium at 25◦C. (B) Conidial germination rate of each strain was determined after cultivated on Water-Agar media at 25◦C for 15 h. (C) Fungi were grown on PDA plates at 25◦C. Radial growth was measured every day, and the growth rate was calculated. All the experiments were replicated three times.

FIGURE S3 | SDS-PAGE analysis of BcXyl1 and BcXyl1rec recombinant proteins. BcXyl1 is the native protein; BcXyl1rec is the site-directed mutagenized protein, which E104 and E157 were substituted with Gln. Two recombinant proteins were stained with Coomassie blue.

FIGURE S4 | BcXyl1 is secreted into the apoplast to induce cell death. (A) BcXyl1 (the native protein) and BcXyl1−1SP (deleted the signal peptide). Cell death induction was detected in N. benthamiana leaves 5 days after infiltration with the examined various A. tumefaciens strains. (B) Immunoblot analysis of proteins from N. benthamiana leaves transiently expressing the examined proteins using a pYBA1132 vector. All the experiments were replicated three times.

FIGURE S5 | BcXyl1 confers disease resistance in plants. N. benthamiana or tomato leaves were infiltrated with 0.5 µM purified BcXyl1, C130−155, or PEVC. (A) The local leaves were inoculated with TMV-GFP, and the number of TMV-GFP lesions were measured. (B,C) The N. benthamiana or tomato local leaves were inoculated with 5 µL of 2 × 10<sup>6</sup> conidia/ml Botrytis cinerea. Lesions symptoms and diameter were observed and measured at 2 days post-inoculation, respectively. Error bars represent standard deviation of three independent replicates. Student's t-test was performed to determine the significant differences between BcXyl1, C130−<sup>155</sup> and PEVC. Asterisks "∗∗" indicate statistically significant differences at a p-value <0.01.

TABLE S1 | Hydrolysis activity test.

TABLE S2 | Primers used in this study.

leaves by Botrytis cinerea. Eur. J. Plant Pathol. 104, 207–220. doi: 10.1023/A: 1008698116106



is a necrotizing factor but is not essential for virulence. Plant Physiol. Biochem. 64, 1–10. doi: 10.1016/j.plaphy.2012.12.008


responses in plants. Plant Mol. Biol. 86, 495–511. doi: 10.1007/s11103-014- 0244-3


**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 © 2018 Yang, Yang, Dong and Qiu. 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.

# Dominant and Recessive Major R Genes Lead to Different Types of Host Cell Death During Resistance to Xanthomonas oryzae in Rice

Jianbo Cao1,2, Meng Zhang<sup>1</sup> , Jinghua Xiao<sup>1</sup> , Xianghua Li<sup>1</sup> , Meng Yuan<sup>1</sup> and Shiping Wang<sup>1</sup> \*

<sup>1</sup> National Key Laboratory of Crop Genetic Improvement, National Center of Plant Gene Research (Wuhan), Huazhong Agricultural University, Wuhan, China, <sup>2</sup> Public Laboratory of Electron Microscopy, Huazhong Agricultural University, Wuhan, China

#### Edited by:

Marco Catoni, University of Birmingham, United Kingdom

#### Reviewed by:

Philip Carella, University of Cambridge, United Kingdom Davide Pacifico, Istituto di Bioscienze e Biorisorse (IBBR), Italy Angela Feechan, University College Dublin, Ireland

\*Correspondence:

Shiping Wang swang@mail.hzau.edu.cn

#### Specialty section:

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

Received: 09 August 2018 Accepted: 02 November 2018 Published: 21 November 2018

#### Citation:

Cao J, Zhang M, Xiao J, Li X, Yuan M and Wang S (2018) Dominant and Recessive Major R Genes Lead to Different Types of Host Cell Death During Resistance to Xanthomonas oryzae in Rice. Front. Plant Sci. 9:1711. doi: 10.3389/fpls.2018.01711 The bacterial blight caused by Xanthomonas oryzae pv. oryzae (Xoo) is the most devastating bacterial disease of rice worldwide. A number of dominant major disease resistance (MR) genes and recessive MR genes against Xoo have been cloned and molecularly characterized in the last two decades. However, how these MR genes mediated-resistances occur at the cytological level is largely unknown. Here, by ultrastructural examination of xylem parenchyma cells, we show that resistances to Xoo conferred by dominant MR genes and recessive MR genes resulted in different types of programmed cell death (PCD). Three dominant MR genes Xa1, Xa4, and Xa21 and two recessive MR genes xa5 and xa13 that encode very different proteins were used in this study. We observed that Xa1-, Xa4-, and Xa21-mediated resistances to Xoo were associated mainly with autophagy-like cell death featured by the formation of autophagosome-like bodies in the xylem parenchyma cells. In contrast, the xa5- and xa13-mediated resistances to Xoo were associated mainly with vacuolar-mediated cell death characterized by tonoplast disruption of the xylem parenchyma cells. Application of autophagy inhibitor 3-methyladenine partially compromised Xa1-, Xa4-, and Xa21 mediated resistances, as did Na2HPO<sup>4</sup> alkaline solution to xa5- and xa13-mediated resistances. These results suggest that autophagy-like cell death is a feature of the dominant MR gene-mediated resistance to Xoo and vacuolar-mediated cell death is a characteristic of the recessive MR gene-mediated resistance.

Keywords: major disease resistance gene, bacteria blight, autophagy-like cell death, vacuolar-mediated cell death, ultrastructure

#### INTRODUCTION

Plant resistance against pathogens can be genetically classified into two classes based on the strength of resistance: qualitative or complete resistance conferred by major disease resistance (MR) genes and quantitative or partial resistance mediated by multiple genes or quantitative trait loci (Kou and Wang, 2010; Zhang and Wang, 2013). The molecular mechanisms of plant disease resistance are explained, in general, with a two-tiered innate immune system: pathogen-associated molecular

**21**

pattern-triggered immunity (PTI) or plant-derived damageassociated molecular PTI or basal resistance, and effectortriggered immunity (ETI) or gene-for-gene resistance (Jones and Dangl, 2006; Thomma et al., 2011; Monaghan and Zipfel, 2012). PTI is initiated by plasma membrane-localized plant pattern recognition receptors (PRRs), which are receptor-kinase proteins or receptor-like proteins, and ETI is initiated by cytoplasmic nucleotide-binding (NB)–leucine-rich repeat (LRR) type resistance proteins (Jones and Dangl, 2006; Macho and Zipfel, 2014). Thus, in general, PTI is quantitative resistance and ETI is qualitative resistance in many plant–pathogen pathosystems (Zhang and Wang, 2013). However, rice and biotrophic Xanthomonas oryzae pv. oryzae (Xoo), which causes the most devastating bacterial disease of rice worldwide, are a unique pathosystem for rice qualitative resistance against Xoo (Zhang and Wang, 2013). The rice MR genes resistant to Xoo can be an ETI or a PTI or other mechanisms that cannot be explained by ETI and PTI (Zhang and Wang, 2013; Ke et al., 2017).

Except for the rice-Xoo pathosystem, earlier studies have shown that plant qualitative resistance to biotrophic pathogens frequently featured a rapid hypersensitive response (HR), which is characterized by rapid and localized cell death to restrict pathogen replication during the early stage of the plant–pathogen interaction (Pontier et al., 1998; Mur et al., 2008). Further studies have revealed that HR is often, but not always, a part of ETI initiated by NB-LRR proteins (Coll et al., 2011). HR-associated cell death is a kind of programmed cell death (PCD). Evolutionarily conserved autophagy, which is intracellular self-digestion of cytoplasmic components characterized by the formation of membrane-bound autophagosomes carrying a portion of the cytoplasm to be degraded or organelle permeabilization, has been observed to be involved in plant PCD (Dickman and Fluhr, 2013; Kabbage et al., 2017). The autophagosomes have different ultrastructures: (1) the vacuolar membrane (tonoplast)-bound body (microautophagy) formed by a portion of the cytoplasm, cytoplasmic vesicles or organelles in the vacuole; (2) the double-membrane body (macroautophagy) formed by a portion of cytoplasm bound by an endoplasmic reticulum-like tubule derived double-membrane in the cytoplasm; (3) the multilamellar body formed by many membranes bound by a single membrane in the cytoplasm (van Doorn and Papini, 2013). In addition, vacuolar-mediated PCD, in which the tonoplast integrity is compromised or the tonoplast is fused with the plasma membrane resulting in the release of vacuolar components into the cytoplasm or the extracellular space leading to cell death, also occurs in HR (Hara-Nishimura and Hatsugai, 2011; Dickman and Fluhr, 2013). In contrast, plant necrosis is characterized by shrinkage of the protoplast and rupture of the plasma membrane (van Doorn et al., 2011).

One of the features that makes the qualitative resistance of rice to Xoo unique from other pathosystems is that one third of the 41 MR genes identified thus far are genetically recessive (Zhang and Wang, 2013; Ke et al., 2017). Eleven (Xa1, Xa3/Xa26, Xa4, xa5, Xa10, xa13, Xa21, Xa23, xa25, Xa27, and xa41) of these 41 genes, which have been cloned and molecularly characterized at present, are shown to encode diverse types of proteins. Among the dominant genes, Xa1 encodes a classic NB-LRR-type protein (Yoshimura et al., 1998), Xa3/Xa26 and Xa21 encode plasma membrane-localized LRR receptor kinase-type proteins (Song et al., 1995; Sun et al., 2004), Xa4 encodes a cell wall-associated protein kinase (Hu et al., 2017). The recessive gene xa5 encodes a mutated basal transcriptional factor IIA gamma (TFIIAγ) subunit 5 (TFIIAγ5 V39E) (Iyer and McCouch, 2004; Yuan et al., 2016), and xa13, xa25, and xa41 encode MtN3/saliva/SWEETtype membrane proteins with XA13 and XA25 localized in the plasma membrane (Chu et al., 2006; Liu et al., 2011; Hutin et al., 2015; Cheng et al., 2017). The dominant MR genes Xa1, Xa4, Xa21 and the recessive MR genes xa5, xa13 are race-specifically resistant to Xoo, while the recessive genes xa1, xa4, xa21 and the dominant Xa5, Xa13 are susceptible to Xoo (Zhang and Wang, 2013).

Xa27, xa13, and xa5 can trigger a HR (although the HR of xa5 is weak) leading to the brown symptoms on infected rice leaves after infiltrating inoculation with avirulent Xoo strains (Gu et al., 2005; Yang et al., 2006; Iyer-Pascuzzi et al., 2008). The ectopically expressed Xa10 and Xa23 can only induce HR in Nicotiana benthamiana (Tian et al., 2014; Wang et al., 2015). Nonetheless, the formation of HR in rice xylem vessel tissue against Xoo bacteria, which multiply in xylem vessels of rice leaves under natural infection conditions (Kou and Wang, 2010), needs to be investigated.

To address the HR of rice MR gene-mediated resistance to Xoo, we compared tissue phenotypes and ultrastructural morphologies of rice leaves from lines containing the dominant Xa1, Xa4, and Xa21 MR genes and the recessive xa5 and xa13 MR genes. We found that autophagy-like cell death is the major characteristic in dominant MR gene-mediated resistance and that vacuolar-mediated cell death is the main feature in recessive MR gene-mediated resistance. These findings suggest that the different types of HR-PCD contribute to the resistance of rice against Xoo by different MR genes.

#### MATERIALS AND METHODS

#### Rice Materials

IRBB1, IRBB4, IRBB21, IRBB5, and IRBB13 are near-isogenic rice lines (NILs) carrying dominant MR genes Xa1, Xa4, and Xa21 and recessive MR genes xa5 and xa13, respectively, in the genetic background of the rice variety IR24. Each of these five lines confers race-specific resistance to Xoo bacteria with different resistance spectra (Zhang and Wang, 2013; Hu et al., 2017).

#### Pathogen Inoculation

Rice plants were inoculated with 10<sup>9</sup> cells ml−<sup>1</sup> of Xoo strains T7174 (Japanese race (1), PXO61, PXO86, PXO112, PXO99, or PXO341 (Philippine race 1, 2, 5, 6 or 10) suspension at either the 4-leaf stage (rice lines IRBB4, IRBB5, IRBB13, and IR24) or 7-leaf stage (rice lines IRBB1, IRBB21, and IR24) by the leaf-clipping method (Chen et al., 2002). For mock treatment, water without Xoo was used by clipping rice leaves. Disease was scored by measuring the lesion length at 2 weeks after inoculation. To study the cell responses to Xoo and the effects of 3-methyladenine (3-MA) and Na2HPO<sup>4</sup> on resistance, rice plants

were inoculated by infiltrating leaves with a bacterial suspension of Xoo using a needleless syringe method (Schaad et al., 1996). The bacterial suspension with 10<sup>9</sup> cells ml−<sup>1</sup> contained 5 mM 3-MA (Sigma, SIGMA-ALDRICH, St. Louis, MO, United States) or 2 mM Na2HPO<sup>4</sup> (Sinopharm, Signopharm Chemical Reagent Co., Ltd., Shanghai, China). For mock treatment, leaves were infiltrated only by 5 mM 3-MA or 2 mM Na2HPO<sup>4</sup> solution. Disease was scored by counting the number of infiltrating sites with water-soaked symptom at 3 days after inoculation. The inoculated leaves were photographed using scientific scanner (Image Scanner III, GE, Sweden). All the inoculation of plants with Xoo was biologically repeated at least twice with similar results, and one replicate was shown.

#### Transmission Electron Microscopy

The ultrastructure of rice leaf cells was studied by transmission electron microscopy. The leaves were sampled at 0 day after inoculation (sampling at 1 h after inoculation) (0 DAI), 3 DAI, 5 DAI, and 14 DAI. The leaf tissues at the inoculation sites were cut into 1 × 3 mm pieces and fixed in 2.5% (w/v) glutaraldehyde in 0.1 M phosphate buffer solution (PBS) (pH 7.2) at 4◦C overnight. The fixed tissues were washed in PBS three times for 30 min each at room temperature (20–25◦C), postfixed for 2 h in 1% osmium tetroxide, dehydrated in a graded series of acetone, infiltrated with Spurr resin (SPI, SPI Chem, West Chester, PA, United States), and polymerized at 65◦C for 48 h. The samples were cut into ultrathin sections (60–70 nm thick), stained with 2% uranyl acetate, and examined with a Hitachi transmission electron microscope (H-7650; Hitachi, Japan) at 80 kv. Each sample had 3 biological replicates with each replicate having at least 3 ultrathin sections observed under the electron microscope. To quantify the cells containing autophagosomelike body, tonoplast disruption and protoplast shrinkage, 208– 520 xylem parenchyma cells were observed from at least six or nine leaf xylem veins of six or nine plants in two or three independent inoculations. In each xylem vein, the total xylem parenchyma cells of one vein were observed, then the frequencies (%) of the cells with the above three structures in the total xylem parenchyma cells were calculated. 57–80 mesophyll cells (approximately 10 cells from each plant) were observed and calculated for the frequencies (%) of cells with the above three structures from six leaves of six plants in two independent inoculations.

#### Expression Analysis of Autophagy-Related Genes and Vacuolar Processing Enzyme Gene

The 3-cm leaf fragments next to the inoculation sites were used for RNA isolation. Quantitative reverse transcription-PCR (qRT-PCR) was conducted as described previously using genespecific primers (**Supplementary Table S1**; Qiu et al., 2007). The expression level of actin gene was used to standardize the RNA sample for each qRT-PCR, and then the expression level relative to the control was calculated. The assays were biologically repeated twice with similar results, and only one replicate was presented.

## Statistical Analysis

The significant differences of lesion length, gene expression level and the number of cells with autophagosome-like body, tonoplast disruption or protoplast shrinkage ultrastructures between resistant and susceptible plants, were assessed using pairwise Student's t-test in Excel (Microsoft<sup>1</sup> ).

#### RESULTS

#### The Leaf Tissue Morphology at Rice–Xoo Interaction Sites

Leaf tissue is a major infection site of Xoo (Niño-Liu et al., 2006). To study whether different types of rice MR gene-mediated resistance against Xoo have different types of cell death, we first analyzed the leaf morphology of rice–Xoo interaction sites in resistant IRBB1, IRBB4, IRBB5, IRBB13 and IRBB21 lines each carrying a different MR gene and susceptible line IR24 against Xoo strains T7174, PXO61, or PXO99 (**Figure 1**). At 5 days after infection (DAI), the brown HR-like lesions appeared on all the inoculation sites of rice leaves of the NILs carrying MR genes, while the inoculation sites of susceptible IR24 leaves formed approximately 0.5-cm-long chlorotic water-soaked symptoms (**Figure 1A**). At 14 DAI, all the infection sites of leaf tissue turned into yellow lesions in the resistant rice lines and the susceptible rice line (**Figure 1B**). The average lesion length of IR24 was about 6-fold longer than that of IRBB1, IRBB4, IRBB5, IRBB13, and IRBB21 at 14 DAI (**Figure 1C**).

#### Ultrastructural Morphotypes of Xylem Parenchyma Cells in Different MR Gene-Mediated Resistances to Xoo

Xoo multiply and spread in the xylem vessels of rice leaves (Kou and Wang, 2010). The leaf xylem parenchyma cells surrounding xylem vessels are in direct contact with Xoo (Niño-Liu et al., 2006). Thus, we examined the ultrastructure of xylem parenchyma cells in the process of cell death after Xoo infection in NILs by transmission electron microscopy. We observed three types of abnormal ultrastructures in xylem parenchyma cells of infection sites at 3 DAI (**Figures 2A–C**). The first type of abnormal ultrastructure was observed mostly in resistant IRBB4 and IRBB21 lines, which showed autophagosome-like bodies formed by autophagy processes including double-membranelike bodies in the cytoplasm, single-membrane-bound bodies containing multiple small vesicles in the cytoplasm, and tonoplast-bound bodies in vacuoles (**Figure 2A**). The second type of abnormal ultrastructure was tonoplast disruption that was commonly observed in the resistant IRBB13 line (**Figure 2B**). The third type of abnormal structure was protoplast shrinkage and rupture of the plasma membrane observed in the susceptible IR24 line (**Figure 2C**).

To study whether the dominant or recessive MR genemediated resistance and the susceptible reaction are associated with different abnormal structures in rice-Xoo interactions,

<sup>1</sup>http://www.microsoftstore.com

we counted the xylem parenchyma cells containing the three abnormal ultrastructures in all the rice lines at 0 and 5 DAI. There were integrated protoplasts and intact organelles in xylem parenchyma cells of all the rice lines and no significant difference in the numbers of xylem parenchyma cells with the three abnormal ultrastructures at 0 DAI among different lines (**Figures 3A,C,E**, **4A,C**, **5A–C** and **Supplementary Figures S1**, **S2**). However, at 5 DAI, autophagosome-like bodies, tonoplast disruption, and protoplast shrinkage were observed in xylem parenchyma cells of all the rice lines, but the frequencies of cells with the three ultrastructures were very different among these rice–Xoo interactions (**Figures 3**–**5** and **Supplementary Figure S2**). In IRBB1, IRBB4, and IRBB21 lines resistant to Xoo strains T7174, PXO61 or PXO112, PXO99, or PXO61, autophagosome-like bodies were the major feature in xylem parenchyma cells (**Figures 3A,C,E** and **Supplementary Figures S2A–B**). The number of rice cells containing autophagosome-like bodies was 4- and 3-fold higher than the number of cells containing tonoplast disruption and plasmolysis, respectively, in IRBB1; was 5- and 4-fold higher than the number of cells containing tonoplast disruption and plasmolysis, respectively, in IRBB4; and was 5- and 3 fold higher than the number of cells containing tonoplast disruption and plasmolysis, respectively, in IRBB21 (**Figure 3G** and **Supplementary Figure S2D**). However, protoplast shrinkage was the major features in xylem parenchyma cells in IRBB1, IRBB4 and IRBB21 lines susceptible to compatible Xoo strain PXO99, PXO341 (**Figures 3B,D,F**). The number of cells with protoplast shrinkage was 4- and 6-fold higher than the number of cells containing autophagosome-like bodies and tonoplast disruption (**Figure 3G**).

In contrast, tonoplast disruption was observed to be the major feature in xylem parenchyma cells during the IRBB5, IRBB13 lines resistant to Xoo strains PXO61 or PXO86, PXO99

arrows) in the susceptible reaction.

(**Figures 4A,C** and **Supplementary Figure S2C**). The number of rice cells containing tonoplast disruption was 3-fold higher than the number of cells containing autophagosome-like bodies or cells showing protoplast shrinkage in IRBB5, and was 3-fold higher than the number of cells containing autophagosome-like bodies and 6-fold higher than cells showing protoplast shrinkage in IRBB13 (**Figure 4E** and **Supplementary Figure S2D**). However, protoplast shrinkage was also the major features in xylem parenchyma cells in IRBB5, IRBB13 lines susceptible to compatible Xoo strains PXO99, PXO341 (**Figures 4B,D**). The number of cells with protoplast shrinkage was 4- and 6-fold higher than the number of cells containing autophagosomelike bodies and tonoplast disruption (**Figure 4E**). Meanwhile, the average lesion length of rice lines containing MR genes inoculated with compatible strains all exceeded 6 cm at 14 DAI (**Supplementary Figure S3**).

Furthermore, protoplast shrinkage was observed to be the major ultrastructure feature in xylem parenchyma cells in rice susceptible reactions to Xoo (**Figure 5**). Many xylem parenchyma cells showed protoplast shrinkage in IR24 that was susceptible to Xoo strains T7174, PXO61, and PXO99 (**Figures 5A–C**). The number of cells with protoplast shrinkage was 4- and 7-fold higher than the number of cells containing autophagosome-like bodies and tonoplast disruption, respectively, in all IR24–Xoo interactions (**Figure 5D**).

#### Expression of Autophagy-Related Genes and Vacuolar Processing Enzyme Gene in Different Resistances to Xoo

Autophagy-related genes (ATG) control the formation of autophagosome-like body in plants (Liu et al., 2005; Xia et al., 2011). Vacuolar processing enzyme genes (VPE) regulate vacuolar mediated cell death in plant-pathogens interactions and H2O2-induced PCD (Hatsugai et al., 2004; Deng et al., 2011). We analyzed transcription level of autophagy-related genes OsATG5

bacterium; V, vacuole; XV, xylem vessel; dark arrow, autophagosome-like body; dark arrowhead, tonoplast disruption; white arrowhead, protoplast shrinkage; white arrow, rupture of plasma membrane. (A–F) Many autophagosome-like bodies in xylem parenchyma cells of IRBB1, IRBB4, and IRBB21 plants at 5 days after inoculation (DAI) with Xoo strains T7174, PXO61, or PXO99 comparison with plants at 0 DAI and plants susceptible reaction to compatible strains PXO99, PXO341. (G) The percentage of cells with autophagosome-like bodies, tonoplast disruption, and protoplast shrinkage in micrographs of xylem parenchyma cell in rice leaves at 5 DAI. Data represent mean (at least nine leaf xylem parenchyma cells were observed from nine different plants in two or three independent inoculations) ± SD. The double asterisk (∗∗) stands for a significant difference between frequency of cells with autophagosome-like body and frequency of cells with tonoplast disruption or protoplast shrinkage at P < 0.01. n, the total number of observed cells.

FIGURE 4 | Tonoplast disruption ultrastructure of xylem parenchyma cells in recessive MR genes xa5 and xa13 mediated-resistance. V, vacuole; XV, xylem vessel; and B, Xoo bacterium. Dark arrow, autophagosome-like body; dark arrowhead, tonoplast disruption; white arrowhead, protoplast shrinkage; and white arrow, rupture of plasma membrane. (A–D) Many xylem parenchyma cells with tonoplast disruption in IRBB5 and IRBB13 plants at 5 days after inoculation (DAI) with Xoo strains PXO61 and PXO99 comparison with plants at 0 DAI and plants susceptible reaction to compatible strains PXO99, PXO341. (E) The percentage of cells with autophagosome-like bodies, tonoplast disruption and protoplast shrinkage in micrographs of xylem parenchyma cell in rice leaves at 5 DAI. Data represents mean (at least nine leaf xylem parenchyma cells were observed from nine different plants in two or three independent inoculations) ± SD. The double asterisk (∗∗) stands for a significant difference between frequency of cells with tonoplast disruption and frequency of cells with autophagosome-like body or protoplast shrinkage at P < 0.01. n, the total number of observed cells.

and OsATG7, VPE gene OsVPE2 in IRBB1, IRBB4, IRBB21, and IRBB13 lines resistant to Xoo strains T7174, PXO61, and PXO99 (**Supplementary Figure S4**). On 8 and 24 h after inoculation, OsATG5 and OsATG7 were markedly induced to higher levels in IRBB1, IRBB4, and IRBB21 lines than that in susceptible IR24 control lines; OsVPE2 was not induced in IRBB1, IRBB21 lines but was markedly induced to higher levels in IRBB13 plants compared to IR24 (**Supplementary Figure S4**).

#### Ultrastructural Morphotypes of Xylem Parenchyma Cells at Late Stage of Rice–Xoo Interaction and at Mock Treatment

To study the ultrastructure of xylem parenchyma cells at the late stage of the rice–Xoo interaction and the ultrastucture of xylem parenchyma cells of rice lines at mock treatment

(clipping leaf only with water), we examined the three abnormal structures, autophagosome-like bodies, tonoplast disruption, and protoplast shrinkage, in the infection sites of the rice lines at 14 DAI and the mock inoculation sites at 3, 5, 14 DAI (**Figure 6** and **Supplementary Figure S5**). Protoplast shrinkage was observed to be the major ultrastructural feature in xylem parenchyma cells at 14 DAI of all the rice–Xoo interactions and all the mock treatments (**Figures 6A–H** and **Supplementary Figures S5A–F**). The number of rice cells containing protoplast shrinkage was 3- to 6-fold higher than the number of cells with autophagosome-like bodies or tonoplast disruption, respectively, in all rice lines (**Figure 6I** and **Supplementary Figures S5G,H**). However, the number of rice cells containing autophagosomelike bodies was still 2- to 3-fold higher than the number of cells containing tonoplast disruption in dominant genes Xa1-, Xa4-, and Xa21-mediated resistance (**Figure 6I**). The number of rice cells containing tonoplast disruption was still 2- and 3-fold higher than the number of cells containing autophagosomelike bodies in recessive genes xa5- and xa13-mediated resistance (**Figure 6I**). Furthermore, at 3 and 5 DAI, there were integrated protoplasts and intact organelles in xylem parenchyma cells and no significant difference among the numbers of xylem parenchyma cell with the three abnormal ultrastructures at mock treatment (**Supplementary Figure S5**).

#### Effects of Autophagy Inhibitor 3-Methyladenine and Na2HPO<sup>4</sup> Alkaline Solution on Different MR Gene-Mediated Resistances to Xoo

The infiltrated inoculation sites on rice leaves with deep inkcolored water-soaked symptoms defined the rice susceptibility to Xoo (Sugio et al., 2005; Yang et al., 2005). Na2HPO<sup>4</sup> alkaline solution could neutralize the low pH (5.2–6.0) liquid released from disrupted vacuole (Martiniere et al., 2013). To determine if the autophagy inhibitor 3-methyladenine (3-MA) and Na2HPO<sup>4</sup> alkaline solution could influence the resistance of rice lines with different MR genes, we observed whether there were watersoaked symptoms on the infiltrated inoculation sites (susceptible reaction) when the rice lines with MR genes were inoculated with Xoo bacteria in 3-MA solution or Na2HPO<sup>4</sup> alkaline solution at 3 DAI (**Figures 7A–E**). The infiltration sites appeared water-soaked (deep inky color) in IRBB1, IRBB4, and IRBB21 inoculated with Xoo in 3-MA solution and IRBB5 and IRBB13 inoculated with Xoo in Na2HPO<sup>4</sup> alkaline solution (**Figures 7A–E**). However, all the corresponding IRBB lines with MR genes inoculated with only the incompatible Xoo strain bacteria (resistant reaction) did not have water-soaked symptoms (**Figures 7A–E**). Furthermore, all the numbers of infiltration site with water-soaked symptom (susceptible reaction) in the inoculation of IRBB1, IRBB4 and IRBB21 with Xoo in 3-MA solution and the inoculation of IRBB5 and IRBB13 with Xoo in Na2HPO<sup>4</sup> alkaline solution were significantly (P < 0.01) more than that in each rice lines inoculated only with Xoo (**Supplementary Table S2**). In contrast, the susceptible line IR24, when inoculated with Xoo, Xoo in 3-MA solution and Xoo in Na2HPO<sup>4</sup> alkaline solution, had water-soaked symptoms on all infiltration sites (**Figures 7F–H**). Whereas there was no water-soaked symptom on infiltration sites in the rice lines only injected with 3-MA solution or Na2HPO<sup>4</sup> alkaline solution (**Figure 7**).

To investigate whether the 3-MA and Na2HPO<sup>4</sup> alkaline solution affected the ultrastructure of mesophyll cell in rice– Xoo interaction, we analyzed the mesophyll cell with the three

abnormal structures in rice lines with MR genes inoculated with Xoo bacteria in 3-MA solution or Na2HPO<sup>4</sup> alkaline solution at 3 DAI (**Supplementary Figure S6**). In rice lines with MR genes infiltrated by only 3-MA or Na2HPO<sup>4</sup> alkaline solution, the mesophyll cells represented intact protoplast and the numbers of mesophyll cell with three abnormal ultrastructures did not have difference (**Supplementary Figures S6A–E,I–J**). However, in IRBB1, IRBB4, IRBB21 lines inoculated with Xoo strains, the auphagosome-like bodies were the major ultrastructural features in mesophyll cells and the numbers of mesophyll cells with autophagosome-like bodies were 3- and 4-fold higher than the number of cells containing tonoplast disruption or protoplast shrinkage (**Supplementary Figures S6A–C,I**). In the IRBB1, IRBB4, and IRBB21 lines inoculated with Xoo strains in 3-MA solution, the IRBB5, IRBB13 lines inoculated with Xoo strains in Na2HPO<sup>4</sup> alkaline solution and the IR24 lines only inoculated with Xoo strains, there were significantly more mesophyll cells with protoplast shrinkage than the mesophyll cells with autophagosome-like body or tonoplast disruption (**Supplementary Figure S6**). More mesophyll cells with tonoplast

disruption were observed in IRBB5 and IRBB13 lines inoculated with only Xoo strains, even in IRBB5 and IRBB13 lines inoculated with Xoo strains in Na2HPO<sup>4</sup> alkaline solution (**Supplementary Figures S6D–E,J**).

These results indicated that the mixing Xoo bacteria in 3- MA and Na2HPO<sup>4</sup> alkaline solution significantly reduced the dominant genes Xa1-, Xa4-, and Xa21- mediated resistance and the recessive genes xa5- and xa13- mediated resistance, respectively.

#### DISCUSSION

A wide variety of pathogens can lead to lesion formation on infected plant tissue and trigger hypersensitive responseprogrammed cell death (HR-PCD) during plant resistance against pathogens (Mur et al., 2008; Kabbage et al., 2017). Most research on HR-PCD focused on the pathogens which grow and spread in the intercellular spaces of plant cells (Kabbage et al., 2017). However, the Xoo bacteria multiply and spread in the xylem vessels, a vascular structure surrounded by xylem parenchyma cells in rice leaves (Niño-Liu et al., 2006; Kou and Wang, 2010). In comparison with the chlorotic watersoaked symptoms of inoculated sites of susceptible rice line, the brown HR-like lesions on the Xoo-infected leaves of resistant rice lines (**Figure 1**) indicate that Xoo triggered the HR-PCD of resistant rice leaf cells. The xylem parenchyma cells, which directly interact with the Xoo bacterium (Niño-Liu et al., 2006), and the mesophyll cells appeared to have many classic autophagosome-like bodies and tonoplast rupture structures in resistant rice lines–Xoo interactions (**Figures 2**–**6** and **Supplementary Figures S2**, **S6**). OsATG5 and OsATG7 were only markedly induced expression in Xa1-, Xa4-, and Xa21-mediated resistance, as did OsVPE2 only in xa13-mediated resistances (**Supplementary Figure S4**). Autophagy inhibitor 3-MA partially impaired the Xa1-, Xa4-, and Xa21-mediated resistances through reducing the number of mesophyll cells with autophagosomelike bodies (**Figure 7** and **Supplementary Figure S6**). Meanwhile, in all the susceptible reactions, the xylem parenchyma cells showed protoplast shrinkage and plasma membrane disruption (**Figures 3**–**6**). In all the resistant rice lines, there were little cells with autophagosome-like bodies and tonoplast rupture structures in control treatment (0 DAI) and mock treatment (3, 5, 14 DAI) (**Figures 3**–**5** and **Supplementary Figures S1**, **S5**), Therefore, the HR-PCD of xylem parenchyma cells belongs to autophagy-like cell death in Xa1-, Xa4-, and Xa21-mediated resistance and vacuolar-mediated cell death in xa5- and xa13 mediated resistance. However, there were approximately 70%

of xylem parenchyma cells that had autophagy-like cell death and vacuolar-mediated cell death at 5 DAI (**Figures 3**, **4** and **Supplementary Figure S2**). Thus, these findings suggest that autophagy-like cell death and vacuolar-mediated cell death are the major types partly mixed with other types in resistances against Xoo.

The dominant Xa1, Xa4, and Xa21 encode a NB-LRR-type protein, a cell wall-associated kinase and a plasma membranelocalized LRR receptor kinase, respectively (Yoshimura et al., 1998; Chen et al., 2010; Hu et al., 2017). The Arabidopsis thaliana MR genes Cf9, Pto, PRS2, and RPS4 encode a membraneanchored glycoprotein, a cytoplasmic serine-threonine protein kinase, a CC-NB-LRR-type protein and a TIR-NB-LRR-type protein, respectively (Pedley and Martin, 2003; Hofius et al., 2009; Chakrabarti et al., 2016). The Cf9, Pto, PRS2 and RPS4 proteins as receptors can trigger autophagy cell death to mediate A. thaliana resistance to Pseudomonas syringae pathovar (pv) tomato, Pst (Liu et al., 2005; Hofius et al., 2009; Gururani et al., 2012). Although XA1, XA4, and XA21 belong to different types of receptor proteins, they can accept and transfer resistance signals into rice cells during the resistance response (Yoshimura et al., 1998; Wang et al., 2006; Hu et al., 2017). As receptors, XA1, XA4, and XA21 proteins presumably have triggered autophagylike cell death to partially mediate rice resistance against Xoo. The xylem parenchyma cells with autophagy-like cell death still had intact morphologies in Xa1-, Xa4-, and Xa21-mediated resistance (**Figure 3** and **Supplementary Figure S2**). The intact xylem parenchyma cells possibly provide a vital cell environment to facilitate rice resistance against Xoo.

The low expression level of susceptible genes such as SWEET11/Xa13, SWEET13, and SWEET14/Xa41 limits the growth of Xoo bacteria in xa5-, xa13-, and xa41- mediated resistances (Chen et al., 2010; Hutin et al., 2015; Huang et al., 2016; Yuan et al., 2016). The above SWEET genes encode glucose transporters which localize on plasma membrane and take part in pumping glucose to extracellular space (Chen et al., 2010; Hutin et al., 2015; Huang et al., 2016). A lot of other glucose transporters on tonoplast play important roles in uptake glucose into vacuole in A. thaliana and rice plant cells (Cho et al., 2010; Hedrich et al., 2015). The SWEET genes such as wheat leaf rust R genes Lr67 and Lr34 can lead to the intracellular glucose accumulation and the leaf senescence in resistance reactions (Krattinger et al., 2009; Chen et al., 2010; Moore et al., 2015). Therefore, in xa5 and xa13-mediated resistances, glucose may accumulate in xylem parenchyma cell where it is pumped into vacuole. Consequently, the tonoplast is disrupted by the high concentration of glucose in vacuole. The vacuole of plants has a low pH (5.2–6.0) to maintain the activity of acid hydrolytic enzymes and defense proteins (Neuhaus et al., 1991; Martiniere et al., 2013). These hydrolytic enzymes can inhibit Pst bacteria growth after being released into the extracellular matrix by tonoplast fusion with the plasma membrane in A. thaliana (Hatsugai et al., 2009). Meanwhile, the destructive vacuolar-mediated cell death mediated by VPE, which is characterized by vacuole collapse by tonoplast disruption and vacuole collapse leading to cytoplasmic content degradation and rapid cell death, is involved in N. benthamiana resistance to tobacco mosaic virus and A. thaliana resistance to Pst or Botrytis cinerea (Hatsugai et al., 2004; Rojo et al., 2004). We also found the tonoplast disintegration of most xylem parenchyma cells in xa5- and xa13-mediated resistances and the higher expression level of OsVPE2 in xa13-mediated resistance (**Figures 4**, **6** and **Supplementary Figure S4**). Based on our results, vacuolar-mediated cell death is a form of destructive cell death. The Na2HPO<sup>4</sup> alkaline solution (pH 9.0) dramatically increased the susceptibility but did not change the number of mesophyll cells with tonoplast disruption in xa5- and xa13 mediated resistances (**Figure 7**, **Supplementary Figure S6** and **Supplementary Table S2**), which suggests that the alkaline solution partially impairs xa5- and xa13-mediated resistance against Xoo. Destructive vacuolar-mediated cell death may depend on the low pH created by vacuole collapse to inhibit Xoo growth. Alternatively, hydrolytic enzymes and defense proteins released into the cytoplasm possibly participate in xa5- and xa13-mediated resistance.

The plant cells with shrinkage of protoplast and rupture of plasma membrane is regarded as necrosis cell death (van Doorn et al., 2011). The necrosis cell death typically happens under abiotic stress (van Doorn et al., 2011; Gao et al., 2015). However, the protoplast shrinkage also occurs in fungal toxin victorin-induced cell death (van Doorn et al., 2011). Meanwhile, the cell wall deformation, protoplast shrinkage and swelling of chloroplasts are observed in susceptible reaction of a black-rotsusceptible cultivar (Golden Acre) inoculated with Xanthomonas campestris pv.campestri (Bretschneider et al., 1989). At 3 and 5 DAI, the xylem parenchyma cells appeared protoplast shrinkage and plasma membrane rupture only in rice susceptible line IR24– Xoo interactions and the rice lines with MR genes but susceptible to compatible Xoo strains (**Figures 3**–**5**). Therefore, the cell death of xylem parenchyma cells in rice lines susceptible to Xoo can be categorized into necrosis at early stage of infection. At 14 DAI of resistant reaction, susceptible reaction or mock treatment, most parenchyma cells of rice line all represented protoplast shrinkage (**Figure 6** and **Supplementary Figure S5**), which indicates that the necrosis may be the general features of cell death in rice during late stages of pathogen infection and wound stress. In rice susceptible to Xoo, there are more bacteria in infection site than that in resistant reaction (Song et al., 1995; Yoshimura et al., 1998; Iyer and McCouch, 2004; Chu et al., 2006; Hu et al., 2017). The high bacterial population in xylem vessel of susceptible rice line may damage the xylem parenchyma cells to form necrosis. The necrosis of xylem parenchyma cells with plasma membrane rupture can possibly release nutrients into xylem vessels where Xoo bacteria multiply.

Plant cell often represents protoplast shrinkage and cell corpse at late stage of PCD (van Doorn et al., 2011). At 14 DAI, the leaves of infection site becoming yellow lesion (**Figure 1B**) indicated that the cell death of xylem parenchyma cells reached to late stage in resistant rice plants. Meanwhile, most xylem parenchyma cells appeared protoplast shrinkage and cell corpse in resistant rice plants at 14 DAI (**Figure 6**). But, most xylem parenchyma cells also represented autophagosome-like bodies and tonoplast disruption ultrastructures in resistant reactions (**Figures 6A–E,I**). Based on our results, the cell death progression in resistant rice lines happens as the following steps. Firstly, the xylem parenchyma cells always keep autophagy-like cell death or vacuolar-mediated cell death from 3 DAI to 14 DAI. Secondly, the long-time autophagosome-like body formation or tonoplast disruption processes possibly lead to the rupture of plasma membrane. Lastly, the protoplast shrinkage leads to cell corpse at late stage. However, the cell death progression of xylem parenchyma cells in susceptible rice line is always keeping necrosis i.e., protoplast shrinkage from 3 DAI to 14 DAI (**Figures 2**, **5**, **6F–H**).

#### AUTHOR CONTRIBUTIONS

fpls-09-01711 November 19, 2018 Time: 14:42 # 12

JC designed and performed most of the experiments, analyzed the data, and drafted the manuscript. MZ performed the qRT-PCR assays. JX and XL provided biochemical analysis support and field management. MY revised part of the manuscript. SW supervised the project, interpreted the data, and revised the manuscript. All authors read and approved the manuscript.

#### FUNDING

This work was supported by grants from the National Key Research and Development Program of China (2016YFD0100903) and the National Natural Science Foundations of China (31500977 and 31821005).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018.01711/ full#supplementary-material

FIGURE S1 | Percentage of cells with autophagosome-like bodies, tonoplast disruption, and protoplast shrinkage in micrographs of xylem parenchyma cells in rice leaves at 0 DAI. Data represent mean (at least nine leaf xylem parenchyma cells were observed from nine different plants in two or three independent inoculations) ± SD. No difference between frequency of cells with protoplast shrinkage and frequency of cells with tonoplast disruption or autophagosome-like body at P < 0.05. n, the total number of observed cells.

FIGURE S2 | Autophagosome-like body and tonoplast disruption ultrastructural features of xylem parenchyma cells in rice lines with MR genes resistant to another incompatible Xoo strains. V, vacuole; XV, xylem vessel; and B, Xoo bacterium. Dark arrow, autophagosome-like body; dark arrowhead, tonoplast disruption; white arrowhead, protoplast shrinkage; and white arrow, rupture of plasma membrane. (A–C) Many autophagosome-like bodies and tonoplast disruption in xylem parenchyma cells of IRBB4, IRBB21, and IRBB5 plants at 5 days after inoculation (DAI) with Xoo strains PXO112, PXO61, and PXO86 comparison with plants at 0 DAI. (D) The percentage of cells with autophagosome-like bodies, tonoplast disruption, and protoplast shrinkage in micrographs of xylem parenchyma cell in rice leaves at 0 and 5 DAI. Data represent mean (at least six

#### REFERENCES

Bretschneider, K. E., Gonella, M. P., and Robeson, D. J. (1989). A comparative light and electron microscopical study of compatible and incompatible interactions between Xanthomonas campestris pv. campestris and cabbage leaf xylem parenchyma cells were observed from six different plants in two independent inoculations) ± SD. The double asterisk (∗∗) stands for a significant difference between frequency of cells with autophagosome-like body and frequency of cells with tonoplast disruption or protoplast shrinkage in at P < 0.01. n, the total number of observed cells.

FIGURE S3 | Lesion length of rice lines with MR genes Xa1, Xa4, xa5, xa13, and Xa21 infected by compatible Xoo strains at 14 DAI. Bars represent mean (10 to 15 leaves from four plants) ± standard deviation (SD).

FIGURE S4 | Expression pattern of OsATG5, OsATG7, and OsVPE2 in resistant (IRBB1, IRBB21, IRBB4, and IRBB13) and susceptible (IR24) rice lines–Xoo interactions. The expression of autophagy-related genes (OsATG5, OsATG7) and a vacuolar processing enzyme (OsVPE2) were analyzed by qRT-PCR in IRBB1/IR24 rice plants inoculated with Xoo strain T7174, IRBB21/IR24, and IRBB13/IR24 rice plants inoculated with Xoo strain PXO99, IRBB4/IR24 rice plants inoculated with Xoo strain PXO61. The rice plants were sampled on ck, 8 and 24 h after inoculation. Data are means (three replicates) ± standard deviations. ck, before inoculation. The letters "a" indicates statistically significant differences between ck and inoculated plants of the same rice plant at P < 0.01. Double astericks (∗∗P < 0.01) indicate statistically significant differences resistant rice plant and susceptible rice plant inoculated at same time.

FIGURE S5 | Ultrastructural features of xylem parenchyma cell in rice lines with mock treatment. V, vacuole; XV, xylem vessel; and B, Xoo bacterium. White arrowhead, protoplast shrinkage; and white arrow, rupture of plasma membrane. (A–F) Many xylem parenchyma cells with protoplast shrinkage at 14 day after inoculation (DAI) comparison with the normal xylem parenchyma cells without three abnormal ultrastructures at 3 DAI and 5 DAI in IRBB1, IRBB4, IRBB21, IRBB5, IRBB13, and IR24 plants. (G,H) Percentage of cells with autophagosome-like bodies, tonoplast disruption, and protoplast shrinkage in micrographs of xylem parenchyma cell in rice leaves at 3, 5, and 14 DAI. Data represent mean (at least six leaf xylem parenchyma cells were observed from six different plants in two independent inoculations) ± SD. The double asterisk (∗∗) stands for a significant difference between frequency of cells with protoplast shrinkage and frequency of cells with tonoplast disruption or autophagosome-like body at P < 0.01. n, the total number of observed cells.

FIGURE S6 | Effects of 3-methyladenine and Na2HPO<sup>4</sup> on the ultrastructural features of mesophyll cell in rice lines infiltrated with Xoo strains in different solutions at the 3rd day. B, Xoo bacterium; V, vacuole; IS, intercellular space; Ch, chloroplast; N, nucleus; dark arrow, autophagosome-like body; dark arrowhead, tonoplast disruption; white arrowhead, protoplast shrinkage; and white arrow, rupture of plasma membrane. (A–H) the ultrastructural features of mesophyll cell in IRBB1, IRBB4, IRBB21, IRBB5, IRBB13, and IR24 leaf infiltration sites with Xoo strains T7174, PXO61, and PXO99 in H2O solution (T7174/PXO61/PXO99), in 5 mM 3-methyladenine (3-MA) solution (T7174/PXO61/PXO99 + 3-MA), in 2 mM Na2HPO<sup>4</sup> solution (T7174/PXO61/PXO99 + Na2HPO4) and with only 5 mM 3-MA or 2 mM Na2HPO<sup>4</sup> solution. (I,J) percentage of cells with autophagosome-like bodies, tonoplast disruption, and protoplast shrinkage in micrographs of xylem parenchyma cell in rice leaves at 14 DAI with Xoo. Data represent mean (at least six leaf mesohyll cells were observed from six different plants in two independent inoculations) ± SD. The double asterisk (∗∗) stands for a significant difference between frequency of cells with protoplast shrinkage and frequency of cells with autophagosome-like body or tonoplast disruption, between frequency of cells with autophagosome-like body and frequency of cells with tonoplast disruption in resistant plants, at P < 0.01. n, the total number of observed cells.

TABLE S1 | PCR primers used for quantitative RT-PCR assays.

TABLE S2 | Effect of 3-methyladenine and Na2HPO4 on the percentage of infiltrating inoculation site with water-soaked symptoms.

(Brassica oleracea). Mol. Plant Pathol. 34, 285–297. doi: 10.1016/0885-5765(89) 90026-X

Chakrabarti, A., Velusamy, T., Tee, C. Y., and Jones, D. (2016). A mutational analysis of the cytosolic domain of the tomato Cf-9 disease-resistance protein shows that membrane-proximal residues are important for Avr9-dependent necrosis. Mol. Plant Pathol. 17, 565–576. doi: 10.1111/mpp. 12315


resistance gene xa5. Phytopathology 98, 289–295. doi: 10.1094/PHYTO-98-3- 0289


fpls-09-01711 November 19, 2018 Time: 14:42 # 13


**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 Cao, Zhang, Xiao, Li, Yuan 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) 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.

# GmDAD1, a Conserved Defender Against Cell Death 1 (DAD1) From Soybean, Positively Regulates Plant Resistance Against Phytophthora Pathogens

Qiang Yan1,2, Jierui Si<sup>2</sup> , Xiaoxia Cui<sup>1</sup> , Hao Peng<sup>3</sup> , Maofeng Jing<sup>2</sup> , Xin Chen<sup>1</sup> , Han Xing<sup>4</sup> and Daolong Dou<sup>2</sup> \*

1 Institute of Industrial Crops, Jiangsu Academy of Agricultural Sciences/Jiangsu Key Laboratory for Horticultural Crop Genetic Improvement, Nanjing, China, <sup>2</sup> Department of Plant Pathology, Nanjing Agricultural University, Nanjing, China, <sup>3</sup> Department of Crop and Soil Sciences, Washington State University, Pullman, WA, United States, <sup>4</sup> National Center for Soybean Improvement, Nanjing Agricultural University, Nanjing, China

#### Edited by:

Valentina Fiorilli, University of Turin, Italy

#### Reviewed by:

Fabiano Sillo, University of Turin, Italy Marino Moretti, Max-Planck-Institut für Terrestrische Mikrobiologie, Germany Ivan Fernandez Lopez, Helmholtz-Zentrum für Umweltforschung (UFZ), Germany

> \*Correspondence: Daolong Dou ddou@njau.edu.cn

#### Specialty section:

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

Received: 10 September 2018 Accepted: 23 January 2019 Published: 08 February 2019

#### Citation:

Yan Q, Si J, Cui X, Peng H, Jing M, Chen X, Xing H and Dou D (2019) GmDAD1, a Conserved Defender Against Cell Death 1 (DAD1) From Soybean, Positively Regulates Plant Resistance Against Phytophthora Pathogens. Front. Plant Sci. 10:107. doi: 10.3389/fpls.2019.00107 Initially identified as a mammalian apoptosis suppressor, defender against apoptotic death 1 (DAD1) protein has conserved plant orthologs acting as negative regulators of cell death. The potential roles and action mechanisms of plant DADs in resistance against Phytophthora pathogens are still unknown. Here, we cloned GmDAD1 from soybean and performed functional dissection. GmDAD1 expression can be induced by Phytophthora sojae infection in both compatible and incompatible soybean varieties. By manipulating GmDAD1 expression in soybean hairy roots, we showed that GmDAD1 transcript accumulations are positively correlated with plant resistance levels against P. sojae. Heterologous expression of GmDAD1 in Nicotiana benthamiana enhanced its resistance to Phytophthora parasitica. NbDAD1 from N. benthamiana was shown to have similar role in conferring Phytophthora resistance. As an endoplasmic reticulum (ER)-localized protein, GmDAD1 was demonstrated to be involved in ER stress signaling and to affect the expression of multiple defense-related genes. Taken together, our findings reveal that GmDAD1 plays a critical role in defense against Phytophthora pathogens and might participate in the ER stress signaling pathway. The defenseassociated characteristic of GmDAD1 makes it a valuable working target for breeding Phytophthora resistant soybean varieties.

Keywords: Glycine max, Phytophthora resistant, defender against apoptotic death 1 (DAD1), programmed cell death (PCD), ER stress

#### INTRODUCTION

As sessile organisms, plants are continually exposed to various biotic and abiotic stresses. Therefore, complex stress perception, signal transduction and adaptation strategies have evolved in plants to cope with adverse environmental conditions. In particular, the programmed cell death (PCD) pathway has been demonstrated to play key roles in plant responses to both abiotic and biotic stresses (Dickman et al., 2001; Lam et al., 2001; Williams et al., 2010). In plant defense against

pathogens, PCD restricts microbe growth and spreading in host tissue by eliminating excessive damaged cells (Kimchi, 2007).

Several PCD repressors have been identified in plants, including Bax inhibitor 1 (BI-1), B-cell lymphoma2 (Bcl-2)-associated athanogene (BAG), ER-luminal binding immunoglobulin protein (BiP), and defender against apoptotic death 1 (DAD1) (Gallois et al., 1997; Matsumura et al., 2003; Doukhanina et al., 2006; Williams et al., 2010; Jing et al., 2016; Li et al., 2016a,b). These repressors may increase or decrease plant resistance to different pathogens (Kawai-Yamada et al., 2004, 2009; Babaeizad et al., 2009; Watanabe and Lam, 2009; Eichmann et al., 2010; Ishikawa et al., 2011).

Among these PCD repressors, DAD1 is unique as it is conserved from yeast to mammals (Nakashima et al., 1993). Initially identified in a temperature-sensitive mutant hamster tsBN7 cell line, DAD1 is a subunit in the oligosaccharyltransferase (OST) complex, which is a core component for catalyzing N-glycosylation in ER (Yan et al., 2005; Peristera and Stephen, 2012). N-glycosylation is the attachment of oligosaccharides to certain asparagine residues of specific nascent proteins, which ensures their successful folding and export from ER. In Drosophila melanogaster, DmDAD1 is essential for efficient N-glycosylation in developing tissues (Zhang et al., 2016). Disruption of DmDAD1 increases accumulation of unfolded or misfolded proteins, which triggers stress signaling in ER and initiates PCD. In contrast, its overexpression stabilizes or increases N-glycosylation (Zhang et al., 2016).

Different hypotheses have been proposed for the roles of DAD1 in maintaining cell viability. DAD1 may facilitate the targeting of OST complex to proteins directly responsible for cell viability. On the other hand, since DAD1 interacts with Mcl1, a Bcl2-family protein acting as an apoptosis inhibitor (Makishima et al., 2000), DAD1 may also affect cell viability in an OST-independent manner.

Plant DAD1 orthologs from Arabidopsis thaliana and rice can rescue hamster tsBN7 cells from apoptosis (Gallois et al., 1997; Tanaka et al., 1997), which indicates they may also function as cell death repressors. Subsequent studies demonstrate that AtDAD1 protects Arabidopsis protoplast cells against ultraviolet-C-induced PCD (Danon et al., 2004) and DAD1 expression in Gladiolus decreases drastically during petal senescence (Yamada et al., 2004). Regarding the roles of DAD1 proteins in plant defense, Wang X. J. et al. (2011) reported that TaDAD2 silenced wheat leaves have attenuated resistance to Puccinia striiformis with down-regulated expression of several defenserelated genes. However, how this protein modulates plantpathogen interactions has not been well characterized overall.

In this study, a DAD1 orthologous gene was identified from soybean (Glycine max). Spatial and temporal expression of GmDAD1 upon P. sojae infection, as well as its protein subcellular localization, were investigated. The function of GmDAD1 in conferring Phytophthora resistance was dissected in soybean hairy roots with GmDAD1 specifically silenced by RNAi, and Nicotiana benthamiana transgenic lines overexpressing GmDAD1 or suppressing native NbDAD1. Our findings demonstrate that GmDAD1 plays a critical role in Phytophthora resistance probably via regulating ER stress signaling.

## MATERIALS AND METHODS

## Plant Materials and Growth Conditions

Two soybean varieties were used in this research: Williams 82 carrying the gene Rps1k, which confers resistance to P. sojae race 2 (Bernard and Cremeens, 1988) and Williams which does not carry any known Rps resistance gene (Bernard and Lindahl, 1972). Seeds of Williams 82 and Williams were sown in small plastic pots containing disinfected soil and maintained in greenhouse at 25◦C and 16h:8h light/dark photoperiod. N. benthamiana plants were grown under identical conditions as described above.

## Culture of Phytophthora Pathogens

Phytophthora sojae isolates P6497 and P6497-RFP, which is a P. sojae strain constitutively expressing red fluorescence protein (RFP) (Xiong et al., 2014) were routinely cultured on 10% V8 juice agar plates at 25◦C in the dark. Phytophthora parasitica was grown under the same conditions.

#### P. sojae Inoculation and Soybean Samples Collection

Root, stem and leaf samples of the soybean varieties Williams 82 and Williams were collected at seedling and pod-filling stages. Hypocotyl inoculation of P. sojae was performed on Williams 82 and Williams plants as described previously (Sun et al., 2014). Agar disks containing hyphae were cut from fresh cultures and inoculated onto hypocotyl incision. After inoculation, the seedlings were placed in growth chamber to keep moisture. Inoculated stems were collected at 0, 6, 12, 24, and 48 h post inoculation (hpi). All samples were frozen immediately in liquid nitrogen and stored at −70◦C. Three biological replicates were performed for each time point.

## DNA and RNA Extraction and RT-qPCR

Following supplier instructions, all DNA and RNA samples were extracted using the Hi-DNAsecure plant kit and the RNA simple Total RNA kit (Tiangen, China), respectively. For RNA samples, elimination of genomic DNA contamination and reverse transcription were performed using the HiScript II Q RT SuperMix reagent Kit (Vazyme, China).

qPCR reactions were performed on an ABI PRISM 7500 real-time PCR system (Applied Biosystems, United States) using the ChamQTM SYBR qPCR Master Mix reagent (Vazyme, China). Relative gene expression levels were calculated using the comparative 2−11CT method (Livak and Schmittgen, 2001). Statistical analysis was conducted using the Student's t-test with Excel 2010 software and the data were considered statistically significant for P < 0.05. qPCR primers for GmDAD1 were designed from its conserved region. PsTEF (GenBank ID EU079791) was selected for determining P. sojae biomass (Yan et al., 2014). GmCons4 (GenBank ID BU578186.1) was selected as

endogenous reference in soybean (Libault et al., 2008). NbEF1a (GenBank ID AY206004) was used as N. benthamiana reference in the VIGS (virus-induced gene silencing) assay.

Defense-related genes analyzed in this research include five pathogenesis-related (PR) genes: PR1a, PR2, PR3, PR4 and PR5 (Bertini et al., 2003; Chen et al., 2007; Mazarei et al., 2007; Maldonado et al., 2014); the JA-regulated defense gene plant defensin 1.2 (PDF1.2) (Lorenzo and Solano, 2005); the ethylene (ET) signaling marker gene ethylene response factor 1 (ERF1) (Lorenzo et al., 2003); the reactive oxygen species (ROS) biosynthetic gene NADP oxidase (NADPHOX) and two ROS scavenging genes: catalase (CAT) and ascorbate peroxidase (APX) (Perez and Brown, 2014). We employed the sequences of G. max if the genes have been reported already, or obtained them by searching in the soybean EST and genome databases<sup>1</sup> using orthologous sequences from A. thaliana as queries. All primers were designed using the Primer Premier 5 software. Primer specificity was evaluated by sequence similarity comparison and melting curve results of RT-qPCR. The primers of ER related genes were designed used the same strategy. The analyzed ERstress related genes were the binding immunoglobulin protein (Bip), the protein disulphide isomerase (PDI), the calnexin1 (CNX1), the ER lumen-localized Dnaj protein3a (ERdj3A), the luminal binding domain/glucose-regulated protein 94 (GRP94), the basic region/leucine zipper motif 17 (bZIP17) and the downstream gene vacuolar processing enzyme (VPE) (Rojo et al., 2004; Cai et al., 2014; Tiziana and Roberto, 2014). All primers used in this study and detailed information were listed in **Supplementary Table S1**.

#### Subcellular Localization of the GmDAD1 Protein

For subcellular localization, the full-length coding sequence (CDS) of GmDAD1 was amplified from cDNAs of the Williams variety using primer pair pBIN-G-DAD-F/R (**Supplementary Table S1**). The 351-bp GmDAD1 CDS was then translationally fused with GFP after cloning into pBIN-GFP (Zhang et al., 2014) using KpnI and XbaI sites. After sequencing validation, GmDAD1-GFP and mCherry-HDEL constructs were introduced into Agrobacterium tumefaciens stain GV3101. The two Agrobacterium liquid cultures were mixed and co-infiltrated into N. benthamiana leaves using a blunt syringe. After maintained for 48 h in greenhouse, agroinfiltrated leaves were detached and visualized with a laser scanning confocal microscope (Zeiss, GERMANY) at 488 and 591 nm for GFP and mCherry detection, respectively.

#### Plasmid Construction for Soybean Cotyledon Transformation

The pBIN-GFP-GmDAD1 construct which was used to determine GmDAD1 subcellular localization was also used to overexpress GmDAD1 in soybean hairy roots, and the pBIN-GFP empty vector was used as control which allows expression of the GFP only. To make the GmDAD1-RNAi construct, partial GmDAD1 gene was amplified (using primers p12-DAD-F and p12-DAD-R) and cloned into pDONR221 (Invitrogen, United States) and then entered in pHellsGate12:GFP via Gateway LR reaction. Modified from pHellsGate12 (Wesley et al., 2001), pHellsGate12:GFP harbors a 35S:GFP:nos expression cassette (Yan et al., 2014). After sequence validation, the pBIN-GFP-GmDAD1, GmDAD1-RNAi, the empty pBIN-GFP and pHellsGate12:GFP vectors were introduced into Agrobacterium rhizogenes strain K599 by electroporation.

#### Plasmid Construction for N. benthamiana Transformation

To overexpress GmDAD1 in N. benthamiana, the full length of GmDAD1 CDS was obtained from cDNAs of the Williams variety using primer pair pDONR-DAD-F/R (**Supplementary Table S1**) and then cloned into the entry vector pDONR221 via Gateway BP reaction. After sequencing validation, the fragment was then entered in pEarlyGate202 via LR recombination reaction between the entry clone and the destination vector (Invitrogen, United States) (Earley et al., 2006). To make Tobacco Rattle Virus (TRV)-based VIGS construct targeting NbDAD1, partial fragment of NbDAD1 was amplified using primer pair TRV:NbDAD-F/R and cloned into pTRV2 (Liu et al., 2002) using KpnI and EcoRI sites. All constructs were validated by sequencing and transformed into A. tumefaciens strain EHA105 for N. benthamiana transformation and GV3101 for VIGS experiment.

#### Soybean Cotyledon Transformation

Surface-sterilized soybean seeds were soaked in sterilized water overnight and then germinated on medium containing 0.5% sucrose and 1.2% agar in growth chamber with 16h:8h light/dark photoperiod. About 5 days after germination, unblemished cotyledons were harvested for A. rhizogenesmediated transformation. Transformation was performed as described previously (Yan et al., 2014). After about 3 weeks of cultivation, transformed hairy roots became abundant at inoculated cotyledons. Positive transformants were selected by detecting GFP signal under fluorescence microscopy, cut off from cotyledons, and cultivated on White medium (**Supplementary Table S2**) for further verification and resistance level test.

#### N. benthamiana Transformation and Virus-Induced Gene Silencing (VIGS)

Nicotiana benthamiana plants overexpressing GmDAD1 were generated via A. tumefaciens mediated leaf disk transformation (Horsch et al., 1985). The T1 seeds harvested from selfpollinated T0 plants were surface-sterilized with 70% ethanol for 30 s, and 10% sodium hypochlorite solution for 5 min, then washed by sterilized water for five times. The sterilized seeds were germinated on MS medium with 100 mg/L glufosinate ammonium (Sigma, United States). T2 seeds were collected and sown in small plastic pots. After 2 weeks, the seedlings were sprayed with 100 mg/L glufosinate ammonium solution. Resistant were transplanted to new pots and confirmed by both genomic DNA and cDNA PCR using gene-specific

<sup>1</sup>https://www.soybase.org/GlycineBlastPages/

primers (DAD-Test-F/R). The T2 plants were used for functional characterization.

For TRV-VIGS assay, Agrobacterium cultures harboring pTRV1 and pTRV2-VIGS (TRV2-NbDAD1, TRV2 empty vector or TRV2-NbPDS used as positive control of silencing ) were mixed and infiltrated into N. benthamiana leaves using a blunt syringe (Fu et al., 2002). Inoculated plants were maintained at 20◦C in greenhouse for effective virus infection and spread.

#### Resistance Assay of N. benthamiana Against Phytophthora parasitica

Leaves from 5 to 6-week-old N. benthamiana plants were detached and inoculated with 20 µl P. parasitica zoospores (10<sup>4</sup> ml−<sup>1</sup> ) per leaf. Inoculated leaves were then kept in a moist chamber and lesion diameters were measured at 36 and 60 hpi. Representative infected leaves were photographed at 60 hpi under a UV lamp and then stained with trypan blue to visualize the infected area. The experiment was repeated three times with similar results and at least 20 leaves were inoculated for each biological replicate. Two weeks after infiltration, leaves from TRV and NbDAD1-VIGS plants were inoculated with P. parasitica using the same strategy. Lesion diameters were measured at 36 and 48 hpi due to the semi-dwarf phenotype of NbDAD1- VIGS plants. At least 10 lesions per construct were measured with three biological repeats. Student's t-test was used to analyze the significance of differences. Difference were considered as significant when P < 0.05.

#### Root Infection and Observation

After verification by detection of GFP fluorescence and qPCR, transgenic hairy roots of similar length (approximately 3 cm) were excised and dipped in the zoospore suspension (10<sup>4</sup> zoospores per ml) of P. sojae race P6497-RFR for 5 min as described previously (Xiong et al., 2014). Inoculated roots were placed in Petri dishes containing 0.6% agar in the dark at room temperature. At 12, 24, and 36 hpi, the infection progression was monitored under an OLYMPUS MVX10 (OLYMPUS, Japan) fluorescence microscope via RFP fluorescence detection at 535 nm. The P. sojae-specific gene PsTEF was used for qPCR quantification of the relative biomass of P. sojae. For each sample, about 10 infected hairy roots were collected and pooled for DNA/RNA extraction which helps to reduce bias and increase statistical accuracy (Graham, 1991; Subramanian et al., 2005; Graham et al., 2007).

#### Western Blotting Assay

About 10 transgenic roots with GFP fluorescence were collected and ground in liquid nitrogen. Total proteins were extracted with the extraction buffer (50 mM Tris-HCl, pH 7.5, 5 mM EDTA, 2 mM DTT, 1% triton, 2% polyvinylpolypyrrolidone and Roche complete protein inhibitor tablets). The samples were boiled for 10 min in 6× sodium dodecyl sulfate (SDS) loading buffer. SDS-PAGE and immunoblotting were performed in a mini-gel apparatus and submarine gel transfer systems (Bio-Rad, United States), respectively. Proteins were then transferred onto polyvinylidene fluoride (PVDF) membranes and then membranes were blocked with 5% non-fat dry milk in 0.01 M PBST for 1 h and then incubated with anti-GFP (1:1,000) (Sigma, United States) for 2 h at room temperature. After washing by TBST three times, the membrane was incubated with IRDye <sup>R</sup> 800CW Goat anti-rabbit IgG (LI-COR, United States) secondary antibody at room temperature for 1 h. Protein bands were detecting using the Odyssey <sup>R</sup> CLx quantitative fluorescence imaging system (LI-COR, United States).

#### Sequence Analysis and Alignment

The conserved and transmembrane domains of GmDAD1 were analyzed with InterProScan and TMPRED respectively (Hofmann and Stoel, 1993; Jones et al., 2014). Multiple sequence alignment was performed using the BioEdit software (Hall, 1999).

## RESULTS

#### ER-Located GmDAD1 Shares Conserved Regions With Other Plant DAD1 Orthologs

GmDAD1 (Gma.7542.2.S1\_at) was identified from an Affymetrix Genechip microarray data analysis on soybean and P. sojae interaction (Zhou et al., 2009). GmDAD1 was up-regulated in soybean varieties with different degrees of resistance to P. sojae (Zhou et al., 2009). Sequence analysis of GmDAD1 (cloned from the Williams variety) revealed that its open reading frame (ORF) encodes a protein of 117 amino acid residues. GmDAD1 shares 91, 54, and 36% identities with DAD1 orthologs in Arabidopsis thaliana, Homo sapiens, and Saccharomyces cerevisiae, respectively. Similar to other plant DAD1 orthologs, GmDAD1 contains three transmembrane regions (residues 27–52, 61–81, and 95–115) and a subunit of OST (residues 13–116) (**Figure 1A**). To investigate the subcellular localization of GmDAD1, a GmDAD1-GFP fusion construct driven by the CaMV 35S promoter was expressed in N. benthamiana leaves. GmDAD1-GFP co-localized in the cytoplasm with mCherry-HDEL, an endoplasmic reticulum (ER) marker, demonstrating the ER localization of GmDAD1 (**Figure 1B**).

#### GmDAD1 Expression Is Induced Upon P. sojae Infection

GmDAD1 transcript can be detected ubiquitously in roots, stems and leaves during plant development in cv Williams, with root being the organ exhibiting highest expression (**Figure 2A**). Interestingly, leaves showed much higher GmDAD1 transcript accumulation at pod filling stage than seedling stage (**Figure 2A**). Similar GmDAD1 expression pattern was detected in Williams 82 variety in the seedling stage (**Supplementary Figure S1**). On the contrary, the expression of GmDAD1 is higher in roots at the pod filling stage in Williams 82 than in Williams.

After inoculation with P6497, a P. sojae isolate of race 2, the compatible variety Williams showed

FIGURE 1 | Molecular characterization and subcellular localization of GmDAD1 protein. (A) Sequence alignment of GmDAD1 and other defender against cell death (DAD) proteins. The darkblue (100%), pink (75%), and cyan (50%) boxes represent levels of amino acid identity or similarity. TM, transmembrane domain; OT, oligosaccharyltransferase domain. At, Arabidopsis thaliana; Hv, Hordeum vulgare; Zm, Zea mays; Os, Oryza sativa; Ta, Triticum aestivum. The asterisk indicates the stop codon. (B) Subcellular localization of GmDAD1 was performed via transient expression system in Nicotiana benthamiana. Green and red fluorescence represent the signal of GFP fusion protein and ER marker mCherry-HDEL, respectively. The reticulate fluorescence pattern of GmDAD1-GFP and its co-localization with mCherry-HDEL indicate accumulation in the ER.

elevated GmDAD1 expression which peaked at 24 hpi and subsequently decreased (**Figure 2B**). In the incompatible variety Williams 82, GmDAD1 was also significantly induced by P. sojae infection at 24 hpi (**Supplementary Figure S1**).

#### GmDAD1 Enhances Resistance to P. sojae in Soybean Hairy Roots

RT-qPCR analysis of ten mixed hairy roots displaying GFP fluorescence indicated that expression of GmDAD1 in GmDAD1- GFP overexpression (OE) plants was nearly 14-fold higher that in the control (GFP) (**Figure 3A**). Western blotting also showed the accumulation of the GmDAD1-GFP fusion protein (**Figure 3B**). When OE and GFP hairy roots were inoculated with P. sojae P6497-RFP (Xiong et al., 2014), the biomass of P. sojae was significantly and consistently less in OE hairy roots than in GFP samples at 12, 24, and 36 hpi (**Figure 3C**). In the GFP control, the invasion hyphae emerged at 12 hpi, rapidly extended at 24 hpi, and almost filled the entire tissue at 36 hpi (**Figure 3D**). In contrast, hyphal growth was limited and the invasion hyphae were much sparser in GmDAD1-GFP overexpression roots (**Figure 3D**), which is consistent with the lower accumulation of P. sojae biomass (**Figure 3C**).

#### Silencing of GmDAD1 Reduces Resistance to P. sojae in Soybean Hairy Roots

RNAi-directed silencing of GmDAD1 in soybean hairy roots (**Figure 4A**) was performed as described previously

(Yan et al., 2014). Both GmDAD1-RNAi (RNAi) and EV control (EV) roots were inoculated with P. sojae P6497- RFP. Compared with control, GmDAD1-RNAi roots showed gradually increased P. sojae biomass accumulation over time (**Figure 4B**). Furthermore, a greater hyphal growth and higher oospore germination can be observed in GmDAD1-RNAi roots (**Figure 4C**). Our results indicated that GmDAD1 is important for soybean resistance against P. sojae.

#### GmDAD1 Affects the Expression of Multiple Defense-Related Genes

To further determine whether the expression of defense-related genes was affected by GmDAD1 silencing, we assessed the expression of several genes in hairy roots inoculated with P. sojae, including the marker genes of SA, and JA/ET signaling pathways, ROS generation and scavenging. The expression of PR1a, PR2, PR3, PR5 and ERF1 were decreased in GmDAD1-RNAi roots after P. sojae inoculation. It is to note that the expression of PR1a was also dramatically suppressed without inoculation (**Figure 5**). In contrast, the expression of PDF1.2, PR4, and two ROS scavenging genes, CAT and APX, were induced in the GmDAD1 silencing roots infected with P. sojae (**Figure 5**). No significant change of NADPHOX expression was observed when GmDAD1 was silenced (**Figure 5**).

## GmDAD1 Is Involved in P. sojae-Activated ER Stress Signaling

Since DAD1 catalyzes the first step of protein N-linked glycosylation, disruption of GmDAD1 is expected to trigger unfolded protein response (UPR), which facilitates proper protein folding in ER via inducing the expression of a series of relevant genes (Li et al., 2011). After P. sojae inoculation, the transcript accumulations of six UPR marker genes were examined in soybean hairy roots, including Bip, PDI, CNX1, ERdj3A, GRP94, and bZIP17. All these genes are induced at the onset of ER stress and mark the activation of adaptive UPR. Expression changes of VPE were also monitored since its protein product possesses caspase-1-like activity and acts downstream of UPR and is part of the ER-PCD pathway. Compared to EV control, GmDAD1-RNAi roots showed significantly higher transcript accumulations of all seven UPR/ER stress marker genes at both 24 and 36 hpi (**Figure 6**). VPE was upregulated at 12 hpi and its expression decreased at 24 and 36 hpi in EV hairy roots, On the contrary, different trend was observed in GmDAD1 silencing hairy roots. The expression increased continuously through the selected time course, and was significantly higher at 24 and 36 hpi (**Figure 6**).

FIGURE 4 | Silencing of GmDAD1 reduces resistance to P. sojae in soybean hairy roots. (A) Relative expression of GmDAD1 was determined by RT-qPCR in inoculated hairy roots in which GmDAD1 was silenced via RNAi (GmDAD1-RNAi) or empty vector (EV) at 0, 12, 24, and 36 hpi. (B) Relative biomass of P. sojae was determined in inoculated hairy roots GmDAD1-RNAi or EV at 12, 24, and 36 hpi. Values represent the means of three replicates ± SD. Data were analyzed by using Student's t-tests (∗P < 0.05, ∗∗P < 0.01 compared with the control). (C) Microscopic analysis of P. sojae colonization in soybean hairy roots. The control EV and GmDAD1-RNAi hairy roots were inoculated with zoospore suspension (10<sup>4</sup> zoospore/ml) of the P. sojae P6497-RFP. Photos were taken at 12, 24, 36 hpi. The white arrows indicate germinating oospores.

FIGURE 5 | GmDAD1 affects the expression of multiple defense-related genes. RT-qPCR analysis of the expression patterns of defense-related genes in the EV and GmDAD1-RNAi transgenic hairy roots after inoculation with P. sojae. Values represent the means of three replicates ± SD. Data were analyzed by using Student's t-tests (∗P < 0.05, ∗∗P < 0.01 compared with the control).

#### GmDAD1 Enhances Resistance to P. parasitica in N. benthamiana

To test whether GmDAD1 confers resistance against other Phytophthora pathogens, transgenic N. benthamiana plants overexpressing GmDAD1 were generated and verified (**Supplementary Figure S2**). Compared to wild-type (WT) and empty vector controls (EV) both GmDAD1 overexpression lines tested (4-1 and 8-4) showed reduced disease symptoms (**Figures 7A,B**) and significantly smaller lesion diameters on leaves (**Figure 7C**) when infected with P. parasitica zoospores. The results suggest that GmDAD1 overexpression enhances N. benthamiana resistance against P. parasitica.

#### Silencing of NbDAD1 in N. benthamiana Reduces Resistance to P. parasitica

Since plant DADs are highly conserved, the native NbDAD1 in N. benthamiana was silenced via TRV-based VIGS system for functional analysis. Compared to TRV-infected controls, plants infiltrated with TRV-NbDAD1 displayed a semi-dwarf phenotype with increased branching (**Figures 8A,B**), which implies a possible role of NbDAD1 in modulating growth and development. Three verified NbDAD1 knock-down lines and TRV-infected controls were challenged with P. parasitica zoospores on detached leaves (**Figure 8C**). Silencing of NbDAD1 led to significantly larger lesion diameters at both 36 and 48 hpi (**Figures 8D–F**), which indicates that NbDAD1 is similar as GmDAD1 in the function of conferring resistance against P. parasitica.

#### DISCUSSION

Being one of the most important crops worldwide, soybean can be infected by several major diseases, including the Phytophthora

(A) Detached leaves from wild type (WT), empty vector control (EV) and GmDAD1 overexpression plants (4-1 and 8-4) were inoculated with P. parasitica zoospores. Photographs were taken at 60 hpi under a UV lamp. (B) Trypan blue staining of the P. parasitica inoculated N. benthamiana leaves. (C) Lesion diameter of inoculated leaves measured at 36 and 60 hpi. The lesion size was calculated from 20 leaves ± SD with three biological repeats. Data were analyzed by using Student's t-tests (∗P < 0.05, ∗∗P < 0.01 compared with the control).

stem and root rot caused by P. sojae (Tyler, 2007). Continual efforts have been made to characterize novel defense genes against Phytophthora pathogens (Sugimoto et al., 2012). Here we identified GmDAD1, an ER-membrane protein from soybean, and dissected its function in plant–Phytophthora interactions.

Being evolutionary conserved across plant and animal species, DAD1 is a subunit of the OST complex, which catalyzes the first step of protein N-linked glycosylation in ER (Kelleher and Gilmore, 1997; Sanjay et al., 1998). In both animals and plants, the expression of DAD1 orthologs responds to a wide range of adverse environmental stimuli, including injury (Zhu et al., 2008), temperature (Lee et al., 2003), and pathogen infection (Wang X. J. et al., 2011). DAD1 inhibits undesired cell death triggered by host defense.

N-glycosylation has been reported to play a critical role in plant–pathogen interactions. For example, site-mutation on the N-glycosylation motif of A. thaliana receptor kinase EFR bleaches its ligand binding and results in oxidative burst elicitation capacity resulting in higher susceptibility of the plant to bacterial pathogens (Haweker et al., 2010). Several reports on the role of DAD proteins in plant defense have been published so far. The Arabidopsis dad1 mutant shows reduced secretion of PR proteins and resistance against pathogens (Wang et al., 2005). In wheat, knock-down of TaDAD2 suppresses the expression of PR1, PR2, and PR5 in response to the infection of Puccinia striiformis f. sp. tritici (Wang X. J. et al., 2011). We hence propose that GmDAD1 may also play a role in soybean disease resistance.

In soybean, GmDAD1 expression can be induced by P. sojae infection in both compatible and incompatible varieties, which indicates that GmDAD1 serves as a nonspecific defense gene to some extent. However, GmDAD1 has consistently higher expression after P. sojae inoculation in the incompatible variety Williams 82, and its expression does not drop dramatically afterward at 48 hpi, as it happens in the compatible variety Williams. Therefore, GmDAD1 may be subjected to distinct transcriptional regulations in P. sojae compatible and incompatible soybean varieties.

Since GmDAD1 has highest transcript accumulation in roots, we adopted the soybean hairy root infection system for P. sojae resistance test. GmDAD1 gain- and loss-of-function mutants exhibit opposite P. sojae resistance phenotypes, which indicates that GmDAD1 contributes to the resistance of soybean against P. sojae. Similarly, knock-down of NbDAD1, the native DAD1 ortholog in N. benthamiana, reduces plant resistance to another Phytophthora pathogen, P. parasitica. Heterologous expression of GmDAD1 in N. benthamiana enhances resistance to P. parasitica. Our results reveal that DAD1 is a potential valuable defense gene against Phytophthora pathogens and this disease resistance function is conserved across plant species.

Phytohormone signaling, which is mediated by SA during biotrophic and hemibiotrophic plant–pathogen interactions and JA and ET for necrotrophic plant pathogens, plays important roles in plant resistance (Glazebrook, 2005). Previously studies demonstrated that the resistance to P. sojae is mediated by the SA and ET signaling pathways (Moy et al., 2004; Sugano et al., 2014). Therefore, we assessed the expression of several key defense related genes by RT-qPCR. When GmDAD1 silencing hairy roots were inoculated with P. sojae, the transcription of PR1a, PR2, PR3, PR5, and ERF1 were significantly reduced. Since the PR genes are generally regarded as early markers of resistance response, the suppressed expression of these genes may be responsible for the compromised resistance at the begin of the infection process (from 0 to 24 hpi). Moreover, the two JA-dependent signal marker genes PDF1.2 and PR4 were upregulated after P. sojae infection in the silenced hairy roots (later than 24 hpi). We inferred that this JA resistance signaling activation might be lately induced, and the up-regulation might be caused by the antagonistic effect of JA and SA pathways.

Reactive oxygen species are important messenger molecules in defense signal regulation. The expression of ROS-generating gene NADPHOX showed no difference between EV and GmDAD1- RNAi hairy roots, however, the ROS-scavenging genes CAT and APX were statistically significant up-regulated after P. sojae infection in the silencing roots, this means that the ROS signaling was not completely affected by GmDAD1 silencing.

AS a core subunit of OST complex, DAD1 plays an important role in protein N-glycosylation (Peristera and Stephen, 2012), the defeat of protein N-glycosylation cause accumulation of misfolded proteins in ER and subsequently ER stress (Li et al., 2011; Cai et al., 2014). In soybean hairy roots infected by P. sojae,

we found that GmDAD1 acts as a repressor for multiple UPR marker genes. In detail, all tested genes become up-regulated at later stages of the infection when GmDAD1 is silenced, indicating severe ER stress. We believe that this situation is caused by a less efficient or delayed defense signaling transduction. However, whether the suppression of defense-related genes was directly caused by the ER stress due to GmDAD1 silencing need to be further investigated.

Under extreme condition such as pathogen infection, a prolonged ER stress is known to eventually activate the ER-PCD pathway. Phytophthora pathogens are hemibiotrophic. They initially establish a biotrophic relationship with their hosts, and switch to necrotrophic phase later than 15 hpi (Enkerli et al., 1997). In EV hairy roots, a sharp increase of VPE, a cystein proteinase mediating PCD via the maturation and activation of vacuolar proteins, was observed at 12 hpi most likely to limit and overcome the biotrophic phase of P. sojae infection. In GmDAD1- RNAi roots, VPE expression was relatively suppressed at the same infection stage, suggesting the failure of PCD induction. However, elevated expression of VPE was detected at 24 and 36 hpi indicating a later activation of ER-PCD pathway. This late apoptosis overlaps with the necrotrophic phase of P. sojae, which may be one of the reasons of the increased P. sojae accumulation in GmDAD1 silencing hairy roots.

Disruption of DAD1 causes growth defect or even embryonic lethality in animal systems (Brewster et al., 2000; Zhang et al., 2016). In this study, we have observed significantly reduced transformation rate when silencing GmDAD1 in soybean hairy roots (**Supplementary Figure S3**). Moreover, knockdown of NbDAD1 by VIGS caused a semi-dwarf phenotype in N. benthamiana. These results suggest that DAD1 may play a similar role of regulating growth in plants most likely by acting on the N-glycosylation pathway of key proteins involved in plant development.

#### REFERENCES


#### CONCLUSION

We observed that GmDAD1, a conserved component of the OST complex, via participating in the ER-PCD and UPR pathways and affecting the expression of multiple defense-related genes, confers resistance to Phytophthora pathogens. Moreover, GmDAD1 regulates plant growth and development likely by the effect on the N-glycosylation pathway. Taken together, GmDAD1 can be considered as a promising target for the molecular breeding of Phytophthora-resistant soybean varieties.

#### AUTHOR CONTRIBUTIONS

DD and QY designed the project. QY, JS, and XaC performed the experiments and analyzed the data. XnC, HX, and DD guided the experimental work. DD, QY, HP, and MJ wrote the manuscript. All authors read and approved the final manuscript.

## FUNDING

This work was supported by grants from the National Natural Science Foundation of China (31625023, 31721004, and 31672008) and Special Fund for Agro-scientific Research in the Public Interest (201503112).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.00107/ full#supplementary-material



gladiolus petals is down-regulated prior to the onset of programmed cell death. J. Plant Physiol. 161, 1281–1283. doi: 10.1016/j.jplph.2004.06.005


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

Copyright © 2019 Yan, Si, Cui, Peng, Jing, Chen, Xing and Dou. 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.

# Differences in Ear Rot Resistance and Fusarium verticillioides-Produced Fumonisin Contamination Between Polish Currently and Historically Used Maize Inbred Lines

#### Edited by:

*Marco Catoni, University of Birmingham, United Kingdom*

#### Reviewed by:

*Javier Plasencia, National Autonomous University of Mexico, Mexico John Laurie, Agriculture and Agri-Food Canada, Canada Alessandra Lanubile, Catholic University of Sacred Heart, Italy*

> \*Correspondence: *Elzbieta Czembor ˙ e.czembor@ihar.edu.pl*

#### Specialty section:

*This article was submitted to Plant Microbe Interactions, a section of the journal Frontiers in Microbiology*

Received: *17 October 2018* Accepted: *20 February 2019* Published: *18 March 2019*

#### Citation:

*Czembor E, Waskiewicz A, ´ Piechota U, Puchta M, Czembor JH and St ¸epien Ł (2019) Differences in ´ Ear Rot Resistance and Fusarium verticillioides-Produced Fumonisin Contamination Between Polish Currently and Historically Used Maize Inbred Lines. Front. Microbiol. 10:449. doi: 10.3389/fmicb.2019.00449* Elzbieta Czembor ˙ 1 \*, Agnieszka Waskiewicz ´ 2 , Urszula Piechota<sup>3</sup> , Marta Puchta<sup>3</sup> , Jerzy H. Czembor <sup>3</sup> and Łukasz St ¸epien´ 4

*<sup>1</sup> Department of Grasses, Legumes and Energy Plants, Plant Breeding and Acclimatization Institute – NRI, Radzikow, Blonie, Poland, <sup>2</sup> Department of Chemistry, Poznan University of Life Sciences, Pozna ´ n, Poland, ´ <sup>3</sup> National Centre for Genetic Resources, Plant Breeding and Acclimatization Institute – NRI, Radzikow, Blonie, Poland, <sup>4</sup> Department of Pathogen Genetics and Plant Resistance, Institute of Plant Genetics, Polish Academy of Sciences, Poznan, Poland ´*

Poland is the fifth largest European country, in terms of maize production. Ear rots caused by *Fusarium* spp. are significant diseases affecting yield and causing grain mycotoxin contamination. Inbred lines, which are commonly used in Polish breeding programs, belong, mostly, to two distinct genetic categories: flint and dent. However, historically used lines belonging to the heterotic Lancaster, IDT and SSS groups were also present in previous Polish breeding programs. In the current study, 98 inbred lines were evaluated across a 2-year-long experiment, after inoculation with *F. verticillioides* and under natural infection conditions. Lancaster, IDT, SSS and SSS/IDT groups were characterized as the most susceptible ones and flint as the more resistant. Based on the results obtained, the moderately resistant and most susceptible genotypes were defined to determine the content of fumonisins (FBs) in kernel and cob fractions using the HPLC method. Fumonisin's content was higher in the grain samples collected from inoculated plants than in cobs. The association of visible *Fusarium* symptoms with fumonisin concentration in grain samples was significant. Conversely, the cobs contained more FB<sup>1</sup> under natural infection, which may be related to a pathogen's type of growth, infection time or presence of competitive species. Using ddRADseq genome sampling method it was possible to distinguish a basal relationship between moderately resistant and susceptible genotypes. Genetic distance between maize genotypes was high. Moderately resistant inbreed lines, which belong to IDT and IDT/SSS belong to one haplotype. Genotypes which belong to the flint, dent or Lancaster group, and were characterized as moderately resistant were classified separately as the same susceptible one. This research has demonstrated that currently grown Polish inbred lines, as well the ones used in the past are a valid source of resistance to *Fusarium* ear rot. A strong association was observed between visible *Fusarium* symptoms with fumonisin concentration in grain samples, suggesting that selection in maize for reduced visible molds should reduce the risk of mycotoxin contamination. NGS techniques provide new tools for overcoming the long selection process and increase the breeding efficiency.

Keywords: maize inbred lines, Fusarium ear rot, fumonisin accumulation, genetic distance assessment, ddRADseq

#### INTRODUCTION

Maize has become one of the most important crops for food and feed production worldwide–both as silage and crop residue. It is also used industrially for starch and oil extraction. The production area in Poland has reached 1200 thousand ha. Fungal diseases are among the most important factors limiting the yield and grain quality of maize, with Fusarium ear rots (FER) caused by Fusarium at the top of the list. Fusaria has the ability to produce toxins which are harmful to both humans and animals alike. The environment has a huge impact on these processes and this is also being studied. In Europe the most common ear rotting fungal species are: F. verticillioides (Sacc.) Nirenberg, causing pink ear rot, and F. graminearium Schwabe, causing red ear rot. However, a number of minor species are also present: F. temperatum, F. poae, F. proliferatum, F. avenaceum, F. culmorum F. subglutinans, and F. sporotrichioides. Maize grain is mostly contaminated with fumonisins produced by F. verticillioides and F. proliferatum, and/or by deoxynivalenol and other trichothecenes along with zearalenone produced by F. graminearum, which affect the health of human and animals (Desjardins and Plattner, 2000; Logrieco et al., 2002; Bennett and Klich, 2003; Munkvold, 2003b; Oldenburg and Ellner, 2005; Voss et al., 2006; Dorn et al., 2009; Czembor et al., 2014, 2015; Gallo et al., 2015; Stoycho, 2015; Miedaner et al., 2017). The maximum acceptable levels of mycotoxin content have been established for both maize and maize products. The guidance has been placed for foodstuffs (EC, 2007) and for animal feed in the European Union (EC, 2006). Contamination of maize grain or feed product is influenced by environmental conditions, agricultural practices, genotype resistance and the interaction between all of these factors (Maiorano et al., 2009a,b; Vasileiadis et al., 2011; Zijlstra et al., 2011; Mesterhazy et al., 2012; Cao et al., 2014b; Miedaner et al., 2017). Fusarium verticillioides can cause disease, at all developmental stages of the plant, in some cases without displaying any symptoms and, consequently, fumonisins are present in symptomless infected kernels (Desjardins and Plattner, 2000; Desjardins et al., 2002; Miedaner et al., 2010).

Ear rot caused by F. verticillioides favored warm and dry conditions, however, warm and wet conditions following silking have been reported to be conducive for disease development (Munkvold, 2003b). Weather conditions during flowering are critical for primary infection, as well as for toxin accumulation during flowering and then before harvesting (De La Campa et al., 2005; Maiorano et al., 2009a,b; Cao et al., 2014b). Low rainfall and a high number of days with maximum temperatures around 30-35◦C during flowering favor disease development. Additionally, precipitation stimulates mycotoxin accumulation before the maturity stage because of the extended harvest period.

Appropriate agronomic practice is one of the most effective approaches to reduce mycotoxin contamination of maize grain (Munkvold, 2003a; Ariño et al., 2009; Mesterhazy et al., 2012; de Galarreta et al., 2015) along with the use of less susceptible hybrids, which can be developed either by traditional breeding methods or by transgenic technology (Munkvold, 2003a; Smith et al., 2004; Presello et al., 2005, 2011; Butrón et al., 2006; Smith, 2007; Toldi et al., 2008; Eller et al., 2010; Lanubile et al., 2011; Mesterhazy et al., 2012; Czembor et al., 2015). Genetic resistance is needed in currently used cultivars and it can be deployed from available intra-specific variability. The understanding of the mechanisms underlying maize resistance to ear rot is still limited. Their nature is polygenic and mapped resistance quantitative trait loci (QTL) have relatively small effects and are not consistent between populations (Pérez-Brito et al., 2001; Robertson-Hoyt et al., 2006; Ding et al., 2008; Xiang et al., 2010; Ali and Yan, 2012; Chen et al., 2012; Yuan et al., 2013; Zila et al., 2013, 2014; Mideros et al., 2014; Butrón et al., 2015; Lanubile et al., 2017). The accumulation of mycotoxins can also be affected by the plant genotype. Pedigree breeding has caused maize inbreeds to become not only more-elite but also genetically more uniform. Older generations of inbred lines are still used in inbred line development and genetic studies or as testers in many breeding programs (Presello et al., 2005, 2011; Mesterhazy et al., 2012). The most utilized inbreds belong into the heterotic groups such as Reid Yellow Dent, Iowa Stiff Stalk Synthetic (SSS), Lancaster, Iodent (IDT) and have not been subjected to phenotypic selection for FER resistance. Flint and popcorn hybrids tend to exhibit less FER severity and grain mycotoxin contamination than dent ones, though resistant dent hybrids are also available (Presello et al., 2007). The stability of the earlier germplasms from Argentina was evaluated in Argentinian and Canadian environments (Presello et al., 2004, 2005, 2011). German conditions were evaluated by Bolduan et al. (2009) and Löffler et al. (2010) who concluded that the sources of resistance are effective in different locations. South African conditions were evaluated by Small et al. (2012), Italian by Balconi et al. (2014) Central European (Poland) by Czembor and Frasinski (2018) and Czembor et al. (2013a,b, 2015).

The breeding programs intensified the development of more resistant genotypes and the identification of new resistance sources, the plant selection procedures and QTL mapping. The resistance mechanism is also being studied. Recently, next-generation sequencing (NGS) technology has emerged as a cutting-edge approach for high-throughput sequence determination. It has improved the efficiency and speed of gene discovery, reducing the time, labor, and cost (Robertson et al., 2006; Elshire et al., 2011; Mesterhazy et al., 2012; Peterson et al., 2012; Zila et al., 2014).

In summary the objectives of this study were: (i) to evaluate the variation of FER resistance and mycotoxin contamination caused by F. verticilliides among a broad base of early, midearly and late groups of maize elite inbred lines belonging to currently used flint and dent groups as well as historical heterotic groups such as Lancaster, IDT, SSS, and (ii) to estimate the level of genetic diversity among and within these groups using NGS technology. Inbreds possessing resistance to FER and fumonisin accumulation would be valuable materials for future breeding programs.

#### MATERIALS AND METHODS

#### Plant Materials

Ninety-eight inbred lines were evaluated, belonging to currently used flint (n = 23) and dent (n = 39) inbred groups, as well as historically used heterotic groups such as Lancaster (n = 4), Iodent Reid (IDT, n = 9), Stiff Stalk Synthetic (SSS, n = 8) and SSS/IDT (S/I, n = 5).A heterotic group of 10 lines was unknown. Based on phenotypic FER assessment, genotypes were divided into 4 groups: highly resistant, moderately resistant, moderately susceptible and very susceptible. Moderately resistant and very susceptible genotypes, representing each heterotic group (30 inbreds: 10 dents, 8 flints, 3 IDT, 2 Lancaster, 3 SSS, 2 SSS/IDT, 2 of the unknown heterotic group), were selected for further research aimed at determining the content of fumonisins in grain and cob fractions (**Figure 1**).

FIGURE 1 | Ears morphology of inbred lines belonging to different heterotic groups: flint, dent, Lancaster, Iodent (IDT), Stiff Stalk Synthetic (SSS), SSS/IDT and unknown groups and symptoms of the *Fusarium* ear rot (FER) disease after kernel inoculation with *F. verticillioides*.

#### Field Experiment and Phenotypic Ear Rot Resistance Assessment

Two-year field experiments were conducted in Radzików, Central Poland (N: 52.21929, 20.63137, sea level: 87 m), across 2011-2012 seasons, as was suggested by Bolduan et al. (2009). They recommended a two-stage selection: an online personal contribution by artificial infection at one place and the next between test hybrids in several (2–3) locations. An RCBD (randomized complete block design) model was used for the field experiment. About 25 plants of maize inbred lines were grown in one row, in three replications (0.75 m between rows and 0.25 m between plants in the row).

To produce inoculum, a well-characterized fumonisinproducing F. verticillioides strain was used. This isolate was selected from the collection carried out by Plant Breeding and Acclimatization Institute-NRI based on the results of a 2-year experiment as the most aggressive in relation to two hybrids of maize. A PDA (Potato dextrose agar) plug with 7 day-old culture was transferred to a 10-mL vial containing 12 autoclaved toothpicks and 8-mL SNA (Synthetic low-Nutrient Agar) and incubated for 2 weeks at 25◦C. Prior to autoclaving, the toothpicks were washed and boiled in excess water for 1 h to remove any toxic substances that might inhibit fungal growth (Jardine and Leslie, 1992). After incubation the toothpicks were removed from the vials and air-dried on a sterile bench overnight.

Inoculation of individual ears was conducted 10–12 days after silking time. At least 20 plants were inoculated for each genotype (6–7 plants in three replicates). Control plants were inoculated using toothpicks without the pathogen. At maturity, ears from each plot were dehusked and harvested manually, dried to approximately 15% grain moisture content and individually rated for FER symptoms using a seven-point scale: 1 = no visible disease symptoms, 2 = 1–3%, 3 = 4-10%, 4 = 11– 25%, 5 = 26–50%, 6 = 51–75%, and 7 = 76–100% of kernels exhibiting visual symptoms of infection, such as brown, pink, or reddish discoloration of kernels and pinkish or white mycelial growth (Clements et al., 2003a,b; **Supplementary Figure 1**). Ear rot severity scores were converted to percentages of the ear exhibiting symptoms replacing each score with the mid-point of the interval.

#### Fumonisins Analysis

Fumonisin B1, B2, B<sup>3</sup> standards were purchased with a standard grade certificate from Sigma-Aldrich (Steinheim, Germany). Sodium dihydrophosphate, potassium chloride, acetic acid and o-phosphoric acid were purchased from POCh (Gliwice, Poland). Organic solvents (HPLC grade), disodium tetraborate, 2 mercaptoethanol and all the other chemicals were also purchased from Sigma-Aldrich (Steinheim, Germany). Water for the HPLC mobile phase was purified using a Milli-Q system (Millipore, Bedford, MA, USA).

Extraction and purification procedure: 10 g of homogenized ground samples of maize kernels were prepared for analysis. Mycotoxins (FBs) were extracted and purified according to the detailed procedure described earlier (Wa´skiewicz et al., 2013a,b). The eluates were evaporated to dryness at 40◦C under a stream of nitrogen. The dry residue was stored at −20◦C until the HPLC analyses. The chromatographic system consisted of Waters 2695 high-performance liquid chromatography (Waters, Milford, USA) with Waters 2475 Multi λ Fluorescence Detector (λex = 335 nm, λem = 440 nm) with an XBridge column (3.0 × 100 mm). Quantification of mycotoxins was performed by measuring the peak areas at the retention time according to a relevant calibration curve. The limit of detection was 0.01 µg g−<sup>1</sup> .

#### Genetic Distance Assessment

The level of genetic diversity among 17 distinct maize genotypes was analyzed using NGS technology. They were selected as moderately resistant and susceptible to ear rot, based on the evaluation under field conditions after inoculation and fumonisin content.

For different combinations of restriction enzymes, it is possible to perform sequencing from 0.01 to 1.99% of maize genome size due to the ddRAD in silico pipeline (Yang et al., 2016). That fraction selected by the ddRADseq pipeline is enough for a broad spectrum of research. The ddRADseq method is used in phylogeny analysis and establishing a genetic relationship among different genotypes. Different restriction enzymes and broad or narrow size selection make the library proper to MiSeq capacity. DdRADseq does not need a reference sequence (Peterson et al., 2012).

Genomic DNA was extracted from 200 mg of 14-day-old single plant leaf tissue by CTAB method (Wa´skiewicz et al., 2013b) with RNase A (Qiagen) treatment before precipitation (30 min at 37◦C). DNA quality was evaluated on 1% agarose gel and quantity was measured by NanoDrop 1000 (NanoDrop Technologies Inc., USA).

Double-digest restriction-associated DNA sequencing (ddRADseq) protocol described previously by Peterson et al. (2012) was used to detected molecular data. The library was prepared using 200 ng genomic DNA of each sample. DNA was digested at 37◦C for 3 h using the 8U HindIII and 8U FspBI restriction enzymes. In the next step, digested DNA was ligated to adapters using T4 Ligase for 2 h at 16◦C. Adapters included Illumina primer sequences and unique barcodes. Purification of ligation products was prepared by AMPure (Beckman Coulter, USA) with beads: sample ratio 1:1. Cleaned DNA was re-suspended in 150 µl of water. Fifteen microlite of ligation products were amplified with NeBNext Ultra II Q5 Master Mix (New England Biolabs, UK) using primers with Illumina index sequences due to the single-indexing method. PCR programme: preliminary denaturation 98◦C/30 s., 16 cycles of denaturation at 98◦C/10 s. and combined annealing and elongation 65◦C/75 s., final elongation 65◦C/5 min., hold at 4◦C. All PCR products were pooled and cleaned-up with AMPure (Beckman Coulter, USA) with bead: sample ratio 1:1 and next 0.6: 1 to size selection of 400-600 bp. A library was quantified on fluorimeter Qubit (ThermoFisher Scientific, USA) by Quant-iTTM High-Sensitivity dsDNA Assay Kit. A library was diluted to 4 nM and denatured with 0.2N NaOH for 5 min. A denatured library was diluted to 20 pM. 600 µl of a prepared library was sequenced using Illumina MiSeq with Miseq Reagent Kit v3 and final PhiX concentration of 12.5 pM and pair reading protocol for 151 cycles.

Quality control of reads was checked in FastQC software (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/; FastQC A Quality Control tool for High Throughput Sequence Data by: S. Andrews). Reads-analysis were prepared with Geneious 11.0 software (https://www.geneious.com, Kearse et al., 2012). The parameters of the substitution model define a rate matrix that can be used to calculate the probability of evolving from one base to another in a given period of time. Most models are variations of two sets of parameters–the equilibrium frequencies and relative substitution rates. Genetic Distance Model: Tamura-Nei. This model also assumes different equilibrium base frequencies. In addition to distinguishing between transitions and transversions, it also allows the two types of transitions (A ↔ G and C ↔ T) to have different rates (Tamura and Nei, 1993). Before analyzing, the reads were split by barcode data. Then, the ends were trimmed, paired reads merged and duplicate reads removed (**Supplementary Figure 2**). The reads were de novo assembled using original data by the Geneious de novo overlap assembler to a high accuracy level (**Supplementary Figure 3**). A genotype tree was made using the neighbor-joining method with a genetic distance model. The tree was constructed with standard software settings. In this method, neighbors are defined as a pair of leaves with one node connecting them, minimizing the total branch length at each stage of clustering, starting with a star-like tree. The branch lengths and an unrooted tree topology can quickly be obtained by using this method without assuming a molecular clock (Saitou and Nei, 1987).

#### Meteorological Data

Field trials were conducted in Radzików, Central Poland. The generative stages of hybrids belonging to the middle-early group were as follows: silking (R1)–I decade of July, blister (R2)–II decade of July, milk (R3)–III decade of July, dough (R4)–I decade of August, dent (R5)–from the II decade of August until the end of I decade of September. Physiological maturity (R6) stage and harvesting time were in the II decade of September. The generative stages of hybrids belonging to the middle-late group started 7 days later (II decade of July) and lasted until the half of the III decade of September. The average temperatures and rainfalls during this time are presented in **Figure 2**.

#### Statistical Analyses

Comparison between inbred lines was done using Fisher's least significant difference test using InfoStat software. In addition to the mold-covered surface proportion of the maize grain, fumonisin content in the infected maize grain meal and the content of the infected maize grain meal, the so-called Toxin-Mold-Index (TMI), was also calculated according to the method described by Arpad et al. (1997), by multiplying the sum of mycotoxin contents, expressed in mg per kg, by the proportion of mold-covered surfaces of maize ear expressed as percentage of the total surface where: TMI = (FB1+FB2+FB3) <sup>∗</sup>A; FB1, FB2, FB<sup>3</sup> content in maize meal obtained from grain and corn cobs; A - proportion of the mold-covered area of ears in % of the total areas of ears.

Differences between inbred lines within years were determined with Fisher's protected least significant difference (LSD) test. Linear correlation coefficients were determined for the relationship between FER severity and fumonisin concentration in grain and cobs sampled from non-inoculated and inoculated plants in 2012 using InfoStat software.

#### RESULTS

#### Fusarium Ear Rot Severity and Fumonisin Contamination

All inoculations resulted in visible disease symptoms (DI = 100%). A summary of statistical data such as mean, SD, minimum and maximum of ear rot ratings of flint (n = 23), dent (n = 39), IDT (n = 9), Lancaster (n = 4), SSS (n = 8), SSS/IDT (n = 5), and unknown (n = 10) heterotic groups under natural infection (non-inoculated) and inoculated with F. verticillioides in 2011 and 2012 are shown in **Table 1**. The Variability of ear rot incidence was observed between, and within, heterotic groups. **Figure 1** shows ears representing individual origin groups and symptoms of the disease after inoculation.

In July, when the primary infection usually takes place, maximum temperatures were higher in 2012 than in 2011 by ca. 5◦C and favored the development of F. verticillioides, used for artificial infection (**Figure 2**). In 2011, high precipitation in July (112 mm, 55 mm and 127 mm in the first, second and third decades, respectively) and August (62 mm and 74 mm in the first and second decades, respectively) did not stimulate disease development. The effects of maize heterotic groups and inbred lines on disease severity differed between years.

Significant differences in FER severity were observed among inbred lines and heterotic groups evaluated under natural infection and after inoculation in both years. Fusarium ear rot severity was significantly higher in 2012 than in 2011 with means of 4.9 and 1.8% of the ear exhibiting disease symptoms under natural conditions, respectively, and 15.2 and 8.07% after inoculation with F. verticillioides. Under natural infection, FER severity for all 98 inbred lines ranged from 1.6 to 45.5% of the ear exhibiting disease symptoms in 2012 and from 0.0 to 7.0% of the ear in 2011. After inoculation, FER severity ranged from 3.0 to 63.0% of the ear exhibiting disease symptoms in 2012 and from 1.4 to 50.5% of the ear in 2011. In 2012 the means of ear rot severity of evaluated groups ranged from 10.9 to 30.99% of ears covered by disease symptoms. The most susceptible group was SSS/IDT (**Table 1**). The most resistant inbred lines belonged, mostly, to the unknown heterotic group. The SD in all heterotic groups ranged from 1.13 to 2.2 and from 2.10 to 6.76 under natural infection in 2011 and 2012, respectively. They were much higher after artificial inoculation; ranging from 2.32 to 17.89 in 2011 and 2012, respectively.

On average, for more than 91% of all evaluated inbred lines, the symptoms of the disease did not exceed 10% of the total ear area under natural infection (**Figure 3**). In 2011, for 73% of all evaluated inbred lines, the symptoms of the disease covered <10% of the ear area, and in 2012 the same applied to 43.9% of lines tested. In 2012, for 10.2% of evaluated inbred lines, the disease symptoms were observed within a range of 26–50% of ear area.

A simple correlation analysis and a principal component analysis (PCA, **Figure 4**) were performed to decipher the relationships between FER severity under natural infection and after inoculation. Using Pearson's correlation coefficients it was shown that the FER severity under natural infection in 2012 was significantly correlated with FER after the F. verticillioides kernel inoculation (r = 0.48<sup>∗</sup> , p < 0.005). Furthermore, the results of FER scored in 2011 after inoculation was positively correlated with the results obtained in 2012 (0.46<sup>∗</sup> , p < 0.005). The purpose of the PCA was to group individual parameters in a more comprehensive manner. Two principal components (PC) explained 69.2% of the variability. The length of the vectors represents the PCA loadings of the variables on the first two principal components, PC1 (44.5%) had strong positive loading from FER after inoculation in 2011 and 2012 and FER under natural infection in 2012. PC2 (24.7%) had positive loading from FER under natural infection in 2011. The highest diversity among inbreds was



*<sup>a</sup>Mean ear rot severity is reported as the average of the entry least square means (back-transformed to the original 0-100% disease severity scale).*

associated with FER, evaluated after inoculation in 2012. The most susceptible and moderately resistant inbred lines from each group of origin were selected for fumonisin content analysis.

Ear rot severity of these genotypes in 2011 and 2012, fumonisin contamination in grain and cobs sampled from inoculated and non-inoculated plants in 2012 and TMI index are shown in **Table 2**. Under natural infection, cobs were more contaminated with fumonisin B<sup>1</sup> than grain samples. The frequency of cob samples contaminated with FB<sup>1</sup> was 66.7% and for grain samples contaminated with FB<sup>1</sup> it was 29.6%. Maize inbred lines accumulated different levels of fumonisins in grain and in cobs. In grains sampled from inbred lines of dent, flint, SSS, IDT, SSS/IDT and Lancaster heterotic groups after kernel inoculation with F. verticillioides, the FB<sup>1</sup> contamination ranged from 0.99 to 99.04 mg/kg, FB<sup>2</sup> contamination ranged from 0.05 to 21.44 mg/kg and FB<sup>3</sup> from 0.0 to 2.21 mg/kg. In cob samples FB<sup>1</sup> contamination ranged from 0.02 to 23.05 mg/kg, FB<sup>2</sup> from 0.0 to 4.5 mg/kg and FB<sup>3</sup> from 0.0 to 1.82 mg/kg. Based on the TMI index it was possible to identify highly resistant and susceptible genotypes. A hierarchical clustering of the inbred lines representing resistant and very susceptible genotypes belonging to each heterotic group was created based on the TMI index (**Figure 5**). Inbreds with low fumonisin content belonged to two separate clusters with low distance between them. Susceptible inbreds belonged to the very inconsistent group with large distances from each other.

In 2012, ear rot severity positively correlated with fumonisin content in grain samples (FB1, FB2, and FB3), with r factor ranging from r = 0.82∗∗∗ to 0.91 ∗∗∗ . The correlations between

individual fumonisins were also highly significant (in grain samples they ranged from r = 0.90∗∗∗ to r = 0.93∗∗∗ p < 0.005 and for cobs they ranged from r = 0.92∗∗∗ to 0.97∗∗∗ p < 0.005). No correlations were found between the ear rot severity and toxin content in maize cobs.

#### Molecular Markers and Population Genotyping

The level of genetic heterogeneity among inbreds with contrasting ear rot resistance was studied using the ddRADseq genome sampling method. They belonged to the following groups: flint, dent, Lancaster, IDT, SSS and SSS/IDT. The range of quality values across all bases at each position for our results in the FastQ file was high and it was conceivable to observe a difference for a number of reads for genotypes, even if they belonged to the same origin group (**Figure 6**; **Supplementary Figures 2, 3**). The genotype tree was made using the neighbor-joining method with a genetic distance model. Branch length expressed nucleotide differences. The relative genetic distances between haplotypes were high. One haplotype contained moderately resistant inbred lines belonging to the IDT historical heterotic group (lines 11, 21, 30), SSS / IDT (line 22) and SSS (line 32). Susceptible inbreds belonged to the SSS (line 45) and SSS/IDT groups (line 44) and were classified separately. Likewise, lines from the Lancaster group, described as resistant (line 49) and susceptible (line 39) did not belong to the same haplotype. Furthermore, in the dent group, resistant and susceptible lines were separate. Interestingly, resistant lines from the flint heterotic group belonged to the same haplotype and susceptible flint line was classified separately. Haplotypes of the genotype tree corresponded to the data of the co-ancestry heat map (**Figure 7**).

#### DISCUSSION

The rapid increase of maize cultivation area, the use of inappropriate crop rotation and global warming have resulted in a rise in frequency of diseases including ear rot fusariosis (Fusarium spp.). Using highly resistant hybrids is one of the most important methods of integrated plant protection (Vasileiadis et al., 2011; Zijlstra et al., 2011).

Artificial inoculations are necessary for the evaluation of maize FER under field conditions because the disease symptoms under natural infection are not repeatable and largely depend on local environmental factors (Santiago et al., 2015). Different methods of inoculation were developed and described (Drepper and Renfro, 1990; Schaafsma et al., 1993; Munkvold and Carlton, 1997; Munkvold et al., 1997; Reid et al., 2002). These methods correspond with natural disease development processes. Flowering and kernel drying are critical periods for disease development and kernel contamination with fumonisins (Munkvold, 2003b; Bush et al., 2004; De La Campa et al., 2005;


**54**


Maiorano et al., 2009a,b; Venturini et al., 2011; Cao et al., 2014a,b; Czembor and Matusiak, 2014). Weather conditions affect all of these processes. We could not confirm that high rainfall or high temperature just before harvest are conducive for fumonisin contamination, particularly concerning rainfall data for 2011 and 2012.

In our study, the inoculation technique was very effective, both for FER assessment and for the assessment of the fumonisin contamination of the grain. Kernel samples collected from the most resistant inbred line were also not heavily contaminated with FBs after inoculation. The correlation between FB1, FB<sup>2</sup> and FB<sup>3</sup> contamination of grain samples and FER severity were highly significant. No significant correlations were observed between the content of fumonisins in samples collected from cobs, both under natural infection and after inoculation. Also in other studies the correlation between the concentration of fumonisin and the severity of FER after inoculation is usually found. Desjardins et al. (2002) and Desjardins and Plattner (2000) reported a very high correlation between fumonisin concentration and visible symptoms, with fumonisin levels being higher in symptomatic kernels and lower in symptomless kernels, respectively. However, in some studies the correlations were not present (Clements et al., 2003a,b).

NGS techniques provide new tools for overcoming the long selection process and increase the breeding efficiency. The knowledge of the QTL genomic location is essential to investigate their potential, as well as their effectiveness toward FER and environmental conditions examined locally (Campos-Bermudez et al., 2013). Here, ddRADseq genome sampling method was deployed to distinguish genetic differences between genotypes with contrasting ear rot resistances (selected based on the field evaluation after inoculation and fumonisin quantification) but also belonging to different heterotic groups: flint and dent groups, as well as to the historical heterotic groups, such as Lancaster, IDT, SSS. The genetic distances between haplotypes were large, indicating an ancient gene pool and possible interspecific hybridization events in maize ancestry. However, among the among the lines of the flint group, as well as in the historical groups, it was possible to find relationship between moderately resistant and susceptible genotypes. Lanubile et al. (2012) summarized reports on mapping disease resistance genes in maize. Using SNPs it was found that QTL located on chromosomes 1, 5, and 9 play significant roles in FER resistance (Zila et al., 2013, 2014). Seven SNPs in six genes associated with FER resistance were identified on chromosomes 4, 5, and 9 based on evaluation of the collection of 1,687 maize inbred lines. Maschietto et al. (2017) found eight QTL associated with FER resistance and FB<sup>1</sup> contamination and Lanubile et al. (2014) mapped 24 candidate genes for FER resistance on the same chromosomal regions. Based on these findings, it is possible to conclude that the QTL with the highest effect for FER resistance was located distally on Chromosome 1. Wisser et al. (2006) postulated that known disease resistance QTL covered 89% of the maize genome. This high degree of coverage is unlikely and possibly related to the relatively

FIGURE 5 | Hierarchical clustering of the inbred lines representing moderately resistant and very susceptible genotypes selected from the collection of 98 inbred based on the disease assessment of non-inoculated and inoculated ears with *F. verticillioides* in 2011 and 2012 years using the Euclidean distance of the elements of relationship matrix A (based on TMI index genotypes in a red cluster represent a very susceptible group). Thirty inbred lines belonging to different heterotic groups 10 dent (D), 8 flint (F), 3 Iodent (IDT), 2 Lancaster (L), 3 Stiff Stalk Synthetic (SSS), 2 SSS/IDT, and 2 unknown (UN). Red color circle: susceptible group, blue color circle: moderate resistant group.

low precision and accuracy of QTL mapping and not taking into account other relevant factors. Our study stands in line with this hypothesis. It is because of the strong influence of environmental factors, plant phenotype and/or interaction of factors, on the spread of the disease in different locations, where FER resistance studies are conducted. That's why QTL mapping results have often been contradictory and their verification in different genetic backgrounds is necessary (Pérez-Brito et al., 2001; Robertson-Hoyt et al., 2006; Ding et al., 2008; Santiago et al., 2013). A new approach to identify candidate genes and QTL for resistance is represented by plant metabolome investigation after pathogen infection (Lanubile et al., 2017) and differential transcriptomic analyses showing genes potentially involved in resistance processes.

Maize breeding has undergone significant changes through the last four decades and so have the materials used. Facing


FIGURE 7 | Co-ancestry heat map (the color code corresponds to the degrees of similarity-from green means the most similar to red as the least similar). Inbred lines belong to heterotic group: dent (D), flint (F), Iodent (IDT), Lancaster, Stiff Stalk Synthetic (SSS), SSS/IDT and unknown pedigree (un).

the dramatic climate changes it becomes obvious that pathogen populations are being changed as well, along with the maize lines grown. Fusarium ear rot resistance remains one of the most important traits in constant selection of materials and sources of this resistance have never been compared until now. This research has demonstrated that currently grown Polish inbreds and also the ones grown in the past may serve as a valid source of resistance to FER. A strong association was observed between visible Fusarium symptoms after inoculation with fumonisin concentration in grain samples, suggesting that selection in maize for reduced visible molds should reduce the risk of mycotoxin contamination. Under natural infection, the cobs contained more fumonisin B<sup>1</sup> than grain samples. This finding suggests that F. verticillioides contaminates grain more intensively when applied artificially. When found naturally, it tends to overgrow preferably the cobs and not the grain. Moreover, under natural infection one has to take into account the presence of other species like F. graminearum, F. poae, F. temperatum or F. proliferatum, which may compete for nutrients and it has been shown that mycotoxin biosynthesis is one of the strategies used by the pathogen to get the advantage over a competitor. To validate this hypothesis a different experimental approach will be established in the future research. Resistance to FER and to fumonisin accumulation is determined by many factors, such as: environmental conditions, infection timing, presence of other competitive species, ear morphology, the kernel's structure and nutrient components, as well as by the interaction between all these factors, which makes drawing conclusions from a single experimental approach challenging. The use of NGS technique in combination with disease symptom screening and mycotoxin

#### REFERENCES

Ali, F., and Yan, J. (2012). Disease resistance in maize and the role of molecular breeding in defending against global threat. J. Integr. Plant Biol. 54, 134–151. doi: 10.1111/j.1744-7909.2012.0 1105.x

contamination allows for fast analysis which genotypes will be most valuable in incorporating the resistance into current breeding programs.

#### DATA AVAILABILITY

This manuscript contains previously unpublished data. The name of the repository and accession number are not available.

#### AUTHOR CONTRIBUTIONS

EC designed and conducted field experiments. AW conducted toxins analyses. MP, UP, and JC performed the molecular studies. EC, AW, MP, and UP interpreted the data. EC, AW, MP, UP, JC, and LS drafted the manuscript, designed figures and tables. All authors read and approved the final manuscript.

#### FUNDING

This work was funded by Polish Ministry of Agriculture and Rural Development.

#### ACKNOWLEDGMENTS

This work was funded by The Polish Ministry of Agriculture and Rural Development.

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.00449/full#supplementary-material

Supplementary Figure 1 | Seven-point scale used to score FER: 1 = no visible disease symptoms, 2 = 1–3%, 3 = 4–10%, 4 = 11–25%, 5 = 26–50%, 6 = 51–75%, and 7 = 76–100% of kernels exhibiting visual symptoms of infection, such as brown, pink, or reddish discolouration of kernels and pinkish or white mycelial growth (Clements et al., 2003a,b).

Supplementary Figure 2 | Number of reads before (blue) and after (red) removing duplicates from data to reduce the impact of a heterogenous PCR step prior to sequencing.

Supplementary Figure 3 | The range of quality values across all bases at each position in the FastQ file. The y-axis shows the quality scores. The higher the score, the better the base call. The background of the graph divides the y axis into very good quality calls (green), calls of reasonable quality (orange), and calls of poor quality (red). The elements of the plot are as follows: the yellow box represents the inter-quartile range (25–75%), the upper and lower whiskers represent the 10 and 90% points. The blue line represents the mean quality.

Supplementary Table 1 | 1. Table of relevant maize accession numbers with indexes and barcodes used, which allow to de-multiplexing NGS resultant sequences. 2. Link to the cloud with raw resultant sequencing data, which is available for everyone after logging to BaseSpace Cloud.


– the Toxin –Mould-Index (TMI). Period. Politech. Ser. Chem. Eng. 41,3–10.


grown in Poland. PLoS ONE 10:e0133644. doi: 10.1371/journal.pone.01 33644


disease in the U.S.A. maize inbred line collection. BMC Plant Biol. 14:372. doi: 10.1186/s12870-014-0372-6

Zila, C. T., Samayoa, L. F., Santiago, R., Butrón, A., and Holland, J. B. (2013). A genome-wide association study reveals genes associated with Fusarium ear rot resistance in a maize core diversity panel. G3 Genes Genomes Genet. 3, 2095–2104. doi: 10.1534/g3.113.007328

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

Copyright © 2019 Czembor, Wa´skiewicz, Piechota, Puchta, Czembor and Ste¸pien. ´ 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.

# Cytoprotective Co-chaperone BcBAG1 Is a Component for Fungal Development, Virulence, and Unfolded Protein Response (UPR) of Botrytis cinerea

#### Edited by:

Marco Catoni, University of Birmingham, United Kingdom

#### Reviewed by:

Mehdi Kabbage, University of Wisconsin-Madison, United States Chang Ho Kang, Gyeongsang National University, South Korea

> \*Correspondence: Zonghua Wang

wangzh@fafu.edu.cn

†Present address: Yurong Li, Corteva AgriscienceTM, Johnston, IA, United States

#### Specialty section:

This article was submitted to Plant Microbe Interactions, a section of the journal Frontiers in Microbiology

Received: 07 December 2018 Accepted: 19 March 2019 Published: 09 April 2019

#### Citation:

Zhang H, Li Y, Dickman MB and Wang Z (2019) Cytoprotective Co-chaperone BcBAG1 Is a Component for Fungal Development, Virulence, and Unfolded Protein Response (UPR) of Botrytis cinerea. Front. Microbiol. 10:685. doi: 10.3389/fmicb.2019.00685

#### Honghong Zhang1,2,3, Yurong Li2,3† , Martin B. Dickman2,3 and Zonghua Wang1,4 \*

<sup>1</sup> Fujian University Key Laboratory for Plant-Microbe Interaction, College of Plant Protection, Fujian Agriculture and Forestry University, Fuzhou, China, <sup>2</sup> Institute for Plant Genomics and Biotechnology, College of Agriculture and Life Sciences, Texas A&M University, College Station, TX, United States, <sup>3</sup> Department of Plant Pathology and Microbiology, College of Agriculture and Life Sciences, Texas A&M University, College Station, TX, United States, <sup>4</sup> Institute of Oceanography, Minjiang University, Fuzhou, China

The Bcl-2 associated athanogene (BAG) family is an evolutionarily conserved group of co-chaperones that confers stress protection against a variety of cellular insults extending from yeasts, plants to humans. Little is known, however, regarding the biological role of BAG proteins in phytopathogenic fungi. Here, we identified the unique BAG gene (BcBAG1) from the necrotrophic fungal pathogen, Botrytis cinerea. BcBAG1 is the homolog of Arabidopsis thaliana AtBAG4, and ectopic expression of BcBAG1 in atbag4 knock-out mutants restores salt tolerance. BcBAG1 deletion mutants (1Bcbag1) exhibited decreased conidiation, enhanced melanin accumulation and lost the ability to develop sclerotia. Also, BcBAG1 disruption blocked fungal conidial germination and successful penetration, leading to a reduced virulence in host plants. BcBAG1 contains BAG (BD) domain at C-terminus and ubiquitin-like (UBL) domain at N-terminus. Complementation assays indicated that BD can largely restored pathogenicity of 1Bcbag1. Abiotic stress assays showed 1Bcbag1 was more sensitive than the wildtype strain to NaCl, calcofluor white, SDS, tunicamycin, dithiothreitol (DTT), heat and cold stress, suggesting BcBAG1 plays a cytoprotective role during salt stress, cell wall stress, and ER stress. BcBAG1 negatively regulated the expression of BcBIP1, BcIRE1 and the splicing of BcHAC1 mRNA, which are core regulators of unfolded protein response (UPR) during ER stress. Moreover, BcBAG1 interacted with HSP70-type chaperones, BcBIP1 and BcSKS2. In summary, this work demonstrates that BcBAG1 is pleiotropic and not only essential for fungal development, hyphal melanization, and virulence, but also required for response to multiple abiotic stresses and UPR pathway of B. cinerea.

Keywords: BAG protein, co-chaperone, Botrytis cinerea, pathogenicity/virulence, unfolded protein response

## INTRODUCTION

fmicb-10-00685 April 8, 2019 Time: 7:54 # 2

Co-chaperones are proteins that assist chaperones in protein folding, oligomeric assembly, and protein transportation and degradation (Hartl, 1996). The Bcl-2-associated athanogene (BAG) family is a group of broadly conserved co-chaperones of 70-kilodalton heat shock protein (HSP70) (Bracher and Verghese, 2015). In mammals, the BAG family was initially identified by screening mouse cDNA library for Bcl-2 interaction proteins (Takayama et al., 1995). Using the ATPase domain of HSC70/HSP70 as molecular bait in yeast two-hybrid screening, additional BAG family members were identified from human, Caenorhabditis elegans and the fission yeast Schizosaccharomyces pombe (Takayama et al., 1999). The human BAGs contain six members, BAG1 to BAG6, and share a conserved BAG domain (BD) of approximately 45 amino acids, located near the C terminus (Takayama and Reed, 2001). As nucleotide exchange factors (NEFs) of HSP70, BAG proteins play a major role in both positively and negatively modulating HSP70 ATP activity via the BAG domain (Gassler et al., 2001). Moreover, BAGs act as scaffolds between HSP70 and target transcription factors or proteins, thus affecting diverse physiological events (Townsend et al., 2003; Kabbage and Dickman, 2008).

The identification and preliminary characterization of plant BAG proteins is still underway. Using a combination of bioinformatics and structural algorithms, seven BAGs have been identified from Arabidopsis thaliana (AtBAG1–AtBAG7) (Doukhanina et al., 2006), several of which have been selected for functional characterization. The structural and biochemical data of AtBAGs demonstrate that the AtBAGs function as NEFs for HSP70/HSC70 and the regulation mechanism of HSP70/HSC7 is conserved in plants (Fang et al., 2013). Our previous work has uncovered that AtBAG4, 6, and 7 exhibit different cytoprotective specificities. Briefly, AtBAG4 appears to protect plants from various abiotic stress stimuli, e.g., salt and drought (Doukhanina et al., 2006). AtBAG6 is activated via proteolytic cleavage by a specific plant aspartyl protease that is required for autophagic cell death in planta and subsequent resistance to the necrotrophic fungus Botrytis cinerea (Kabbage et al., 2016; Li and Dickman, 2016; Li et al., 2016). Consistently, overexpression of AtBAG6 induced programmed cell death (PCD) in both yeast and plants (Kang et al., 2006). The ER-localized AtBAG7 is an essential component of the unfolded protein response (UPR) and directly interacts with UPR regulator AtBIP2 (Williams et al., 2010). Under heat stress, AtBAG7 is also proteolytically processed in the ER lumen and translocate from the ER to the nucleus, where it interacts with the transcriptional factor WRKY29 for heat tolerance (Li et al., 2017). A number of BAGs have also been reported in other plant species. For example, OsBAG4 from rice interacts with an E3 ubiquitin ligase EBR1 and regulates PCD, which controls plant immunity and broad-spectrum disease resistance (You et al., 2016). Taken together, plant BAGs are multifunctional and modulate numerous physiological and biological processes.

Although the functions of human and plants BAG family members are extensively studied, there is limited knowledge on the roles of the BAGs in fungi. Previous studies showed that SNL1, a mammalian homolog of BAG1 in Saccharomyces cerevisiae, is functionally linked to the nuclear pore complex and plays a role in promoting both protein biogenesis and translation by recruiting ribosomes and HSP70 to the ER membrane (Ho et al., 1998; Verghese and Morano, 2012). Two other BAG1 homologs from Schizosaccharomyces pombe, BAG101 and BAG102, are co-factors of 26S proteasomes, and play as HSP70 chaperones (Kriegenburg et al., 2014). Overexpression of BAG101 and BAG102 inhibit cell growth by triggering HSP70 to release and activate HSF1 (heat shock factor 1) (Poulsen et al., 2017). To date, the only example in filamentous fungi is BAGA from Aspergillus nidulans, which impacts fungal sexual development and modulates secondary metabolism (Jain et al., 2018).

The endoplasmic reticulum (ER) is the central intracellular organelle for protein translocation, protein folding, and protein post-translational modifications, allowing further transport of proteins to the Golgi apparatus and ultimately to vesicles for secretion or display on the plasma surface. Perturbations in ER function, named "ER-stress," unfolded or misfolded proteins accumulate within the ER and disrupt ER homeostasis to activate an intracellular signaling pathway, known as the UPR eventually culminating in cell death (Malhotra and Kaufman, 2007). As a conserved survival pathway to counteract the lethal effects caused by ER stress, UPR can mitigate accumulation of unfolded proteins and restore ER homeostasis by reducing protein translation and while increasing and misfolded proteins degradation aided by molecular chaperones (e.g., binding immunoglobulin proteins, BiPs) (Sitia and Braakman, 2003; Ron and Walter, 2007). BiPs work as HSP70 chaperones and carry aberrant proteins from the ER to the cytoplasm for degradation by the proteasome (Gething, 1999). The IRE1-HAC1/XBP-1 pathway (HAC1 mRNA in yeast and XBP1 mRNA in metazoans) is a major branch of UPR that is remarkably conserved from yeast to human (Back et al., 2005). UPR is initiated by the activation of the ER stress sensor IRE1, which transmits the signal by removing a non-conventional intron from a transcription factor HAC1/XBP-1 mRNA to produce potent transcriptional activator of UPR targets (Ron and Walter, 2007). Comparing to the extensive studies of UPR in human and plant systems, UPR has only been delineated in small number of fungal pathogens, including Aspergillus fumigatus (Feng et al., 2011), Alternaria brassicicola (Joubert et al., 2011), Ustilago maydis (Heimel et al., 2013; Lo Presti et al., 2016), and all of which demonstrate that UPR regulation is correlated with fungal pathogenicity.

In this study, we identified a unique BAG gene in the necrotrophic fungal pathogen Botrytis cinerea, the causal agent of gray mold diseases to over 1,400 species of cultivated plants worldwide (Elad, 2016). Target gene replacement of BcBAG1 resulted to defect in vegetative growth, conidiation, sclerotial formation, penetration and attenuated virulence in B. cinerea, 1Bcbag1 mutants were more sensitive to various stress conditions indicating that BcBAG1 regulates stress tolerance in B. cinerea. In particular, BcBAG1 deletion mutants significantly increased susceptibility to diverse ER stress-inducer including heat, cold, tunicamycin (Tm), and dithiothreitol (DTT). We demonstrated that BcBAG1 binds to the ER chaperone BcBIP1 and negatively regulate UPR components, including

the expression of BcBIP1, BcIRE1 and the splicing of BcHAC1 mRNA, suggesting BcBAG1 is necessary for the maintenance of the UPR in B. cinerea. Collectively, we present the evidence of identification and functional characterization for BcBAG1 a member of BAG family in B. cinerea, which is vital for fungal virulence on host plants and is required for ER stress response with regards to maintenance of UPR.

#### MATERIALS AND METHODS

#### Strains and Culture Conditions

BcBAG1 deletion mutant, 1Bcbag1, was generated from the B. cinerea WT strain B05.10 (Quidde et al., 1999). All strains (**Supplementary Table S1**) were maintained on potato dextrose agar (PDA) medium at 22◦C. Mycelia used for protoplast preparation, genomic DNA and total RNA extraction were grown in YEPD (1% peptone, 0.3% yeast extract, 2% glucose, pH 6.7) at 150 rpm, 22◦C for 36 h. B. cinerea protoplast was recovered on SH medium (20% sucrose, 0.5 mM hepes, 1 mM NH4H2PO4, pH 7.0) at 22◦C for 12–16 h. The selectable marker, 100 µg/ml hygromycin B (VWR) or 100 µg/ml nourseothricin sulfate (Research Products International) was supplemented to PDA containing 1% agar.

#### Bioinformatic Analysis

Preliminary BAG protein search and DNA sequence downloading were conducted in B. cinerea B05.10 genome database<sup>1</sup> . The phylogenetic tree was generated through MEGA v7.0 based on the neighbor-joining method (Kumar et al., 2016). Domain is predicted by Pfam<sup>2</sup> and InterPro programs<sup>3</sup> . The multiple alignment of BAG domain sequence were constructed using Clustal X (Larkin et al., 2007).

#### BcBAG1 Gene Deletion and Complementation

Primers used in this study are listed in **Supplementary Table S2**. The replacement constructs for BcBAG1 were generated through the split-marker approach as described before (Goswami, 2012). Briefly, the 800-bp upstream and 855-bp downstream fragments of BcBAG1 were amplified with primer pairs AF/AR and BF/BR (**Supplementary Table S2**), respectively. The resulting amplicons ligated with the hygromycin phosphotransferase (hph) fragments by using Splicing Overlap Extension (SOE)- PCR. The resulting PCR products (20 µg) were transformed into protoplasts of B. cinerea. Protoplast preparation and PEG-mediated transformation of B. cinerea were performed as the established protocol (Gronover et al., 2001). After transformation, hygromycin-resistant transformants were picked individually and PCR analyses with designated primer pairs OF/OR, UF/UR, and AF/BR were performed to identify transformants that carried the insertion of hph at the BcBAG1 locus. Then all positive transformants were confirmed by subsequent RT-PCR and Southern blotting.

The full length of BcBAG1 was amplified from B05.10 genomic DNA, then ligated into pNAH-OGG with NcoI to create BcBAG1- GFP. GFP-fusion constructs were transformed into B05.10 for subcellular localization analysis. For complementation, the CDS of BcBAG1 with native promoter sequence were amplified with relative primer pairs (CF/CR) and cloned into pNAH-OMG harboring nourseothricin acetyltransferase gene (NAT1) with SpeI/NcoI to make BcBAG1-Com, conferring resistance to antibiotics nourseothricin sulfate. Similar method was used to create the truncated BcBAG1A1−<sup>141</sup> and BcBAG1B142−<sup>298</sup> constructs. These constructs were transformed into 1Bcbag1 mutants. Transformants with resistance to both nourseothricin and hygromycin were selected and confirmed by PCR and RT-PCR.

#### Southern Blotting and Real-Time PCR

Fungal genomic DNA was extracted as described (Raeder and Broda, 1985). For Southern blotting, the genomic DNAs were digested with PvuI (NEB) for 24 h at 37◦C. Probe labeling, hybridization and detection were performed in accordance with the manufacturer's instructions for the Digoxigenin High Prime DNA Labeling and Detection Starter Kit I (Roche Applied Science).

Total RNA was isolated using EastepTM Total RNA extraction Kit, following the manufacturer recommendation (Corp). The first-strand cDNA was synthesized with the M-MLV (Moloney Murine Leukemia Virus) reverse transcriptase (Life Technologies). Quantitative RT-PCR (qRT-PCR) was performed with SYBR Green PCR master mix (Applied Biosystems). The fungal actin gene was used as an internal reference. The relative expression levels were calculated by the 2−11Ct method (Livak and Schmittgen, 2001).

#### Mycelial Growth, Conidiation, and Sclerotial Formation Tests

For vegetative growth assays, 5 mm diameter mycelial plugs were cultured on fresh PDA in the dark at 22◦C. Radial growth was measured by colony diameters after 3 days. Determination of the sensitivity of 1Bcbag1 to environmental stresses were performed on modified PDA plates with: 1 M NaCl, 1.2 M D-sorbitol, 2 µg/ml Tm, 2.5 mM DTT, 20 mM H2O2, 0.5 mM tert-butyl hydroperoxide (TBHP), 15 µM bortezomib (Bort), 200 µM MG132, 600 µg/ml calcofluor white (CFW), 0.02% SDS, 1 µg/ml iprodione (Ipro), 0.1 µg/ml fludioxonil (Flud). The percentage of mycelial radial growth inhibition (RGI) was calculated using the formula RGI% = [(C − N)/C ∗ 100], where C is colony diameter of the control, and N is colony diameter of the experimental treatment.

For conidiation assays, conidia of WT and mutants were collected from a 10-day-old PDA plate with 5 ml sterile water and spores were counted microscopically with a hemocytometer. For sclerotial formation, 5 mm diameter mycelial plugs were inoculated to PDA and incubated in the dark at 22◦C, the number of mature (melanized) sclerotia were counted after 4 weeks.

<sup>1</sup>http://fungi.ensembl.org/Botrytis\_cinerea/Info/Index

<sup>2</sup>https://pfam.xfam.org/

<sup>3</sup>https://www.ebi.ac.uk/interpro/

#### Conidial Germination and Fungal Penetration Assays

Conidial germination was conducted as described by Doehlemann et al. (2006). Briefly, conidial suspensions were adjusted to 1.0–1.5 × 10<sup>5</sup> spores/ml in 10 mM KH2PO<sup>4</sup> and 10 mM glucose solution and placed in the center of a glass slide. Incubation was kept in a moist chamber at 22◦C for 12 h, 24 h, and 48 h. For the onion infection assay, 20 µl droplets (5.0 × 10<sup>4</sup> spores/ml) were deposited on the hydrophobic epidermis layers of onion and incubated for 48 h in the dark and moist chamber at 22◦C (Viefhues et al., 2014).

#### Pathogenicity Assay

Three-day-old mycelial plugs with 5 mm diameter or 10-day-old conidial suspensions (1.0–1.5 × 10<sup>5</sup> spores/ml) were inoculated on 4-week-old detached tomato leaves and grape. Inoculated plant materials were incubated in 16 h daylight humid chamber at 22◦C. Results were recorded after 4 days and 7 days. The experiment was repeated at least three times.

#### Yeast Two-Hybrid (Y2H) Assay

The Y2H assay was conducted according to the manufacturer's standard instructions (Clontech). The cDNA of BcBAG1 was cloned into pGBKT7 as the bait vector and the cDNAs of HSP70-type chaperones were cloned into pGADT7 as the prey constructs, respectively. The pGBKT7-BcBAG1 and each prey vector were co-transformed into the AH109 yeast strain to evaluate interactions. The positive and negative controls were from the Kit (Clontech).

#### Arabidopsis Complementation Assays

Arabidopsis thaliana Col-0 and atbag4 T-DNA knock out mutants were obtained from Arabidopsis Stock Center<sup>4</sup> . Atbag4 homozygous mutants (SALK\_027577C) were confirmed by PCR. The BcBAG1 full-length cDNA was cloned into pCB302ES containing the 35S promoter and the HA-epitope tag (Hwang and Sheen, 2002). This construct was transferred into the atbag4 knockout mutants by the floral dipping method (Zhang et al., 2006).

For salt stress assays, seeds were surface sterilized in 70% ethanol for 10 min and in 5% bleach solution for 5 min, and germinated on 1/2 Murashige and Skoog (MS) medium (Invitrogen) at 23◦C for 5 days. The seedlings were transferred to fresh 1/2 MS medium containing 100 mM NaCl and grown at 23◦C for 2 weeks to 5 weeks.

#### RESULTS

#### Identification and Characterization of BcBAG1

To identify BAG proteins in B. cinerea, we searched B. cinerea B05.10 genome database with "BAG" as a query, and obtained one hit (Bcin10g01250.1/BC1G\_05107). Additionally, based

<sup>4</sup>https://www.arabidopsis.org/

on Pfam and SMART programs, we search for all gene with BAG domain (BD domain) in genome database. Results indicated that there is only one gene with a single copy (Bcin10g01250.1/BC1G\_05107) containing BAG domain, designated BcBAG1 hereafter. Phylogenetic analysis revealed that BcBAG1 shares low similarity with BAGs in yeast, plants and animals while it is closely related to BAG homologs from other filamentous fungi, e.g., Sclerotinia sclerotiorum (86.53%), Magnaporthe oryzae (57.04%), Fusarium oxysporum (52.01%), and Aspergillus nidulans (45.30%) (**Supplementary Figure S1A** and **Supplementary Table S3**). BcBAG1 contains a ubiquitin-like domain (UBL) at the N-terminus and a BAG domain at the C-terminus, encoding a 35 kD protein with 298 amino acids (**Supplementary Figure S1B**). The alignment also showed that most of the key interaction residues for BAG-HSC70/HSP70 binding are conserved in BAG proteins across filamentous fungi, yeast, Arabidopsis, and human (**Supplementary Figure S1C**).

To address the function of BcBAG1, we generated knockout mutants of BcBAG1 (1Bcbag1), in the wild-type (WT) strain B05.10 (**Supplementary Figure S2A**). Two individual 1Bcbag1 lines, K3-7 and K8-6, were validated and selected for later use (**Supplementary Figures S2B,C**). We obtained three complemented strains by transforming the full-length BcBAG1 under its native promoter into mutants and all strains equally restored the defects of 1Bcbag1. Thus one complemented strain (BcBAG1-Com) was used in the following studies.

#### BcBAG1 Is Required for Vegetative Growth, Conidiation, and Sclerotial Formation

To evaluate the role of BcBAG1 during vegetative growth of B. cinerea, we examined hyphal growth on PDA. As shown in **Figures 1A,C**, the diameters of the 1Bcbag1 colonies were similar to WT strain B05.10 and the complementation strain BcBAG1-Com. However, 1Bcbag1 formed a thick hyphal layer on the surface of plates and the amount of aerial hypha drastically increased in comparison with B05.10 and BcBAG1- Com (**Figure 1B**). These results indicate that BcBAG1 influences on vegetative growth of B. cinerea.

Given that wind-dispersal of conidia determines the severity of gray mold disease in the field (Leroch et al., 2013), we assessed the role of BcBAG1 in conidial production. Conidiation of B05.10, 1Bcbag1 and BcBAG1-Com from 10-day-old PDA culture was measured using microscopic examination. Although 1Bcbag1 still produced aerial mycelia (**Figure 1B**), conidiogenesis of 1Bcbag1 is significantly reduced (**Figure 2A**), in detail, the mutants produced approximately 2.1 × 10<sup>7</sup> conidia/PDA plate, while the WT produced approximately 2.0 × 10<sup>8</sup> conidia/PDA plate (P < 0.01) (**Figure 2D**). However, the conidia exhibited normal morphology between 1Bcbag1 and WT (**Figure 2B**). These results indicate that BcBAG1 involved in conidial production but do not affect conidial morphology in B. cinerea.

Melanization of sclerotia is considered as a survival strategy of various fungi when encountering harsh environments like over-wintering (Williamson et al., 2007). Here, we examined sclerotial formation following BcBAG1 deletion in B. cinerea.

After 4-week incubation on PDA in the dark, we observed that 1Bcbag1 mutants were unable to produce sclerotia, while B05.10 and BcBAG1-Com produced abundant sclerotia on PDA (**Figures 2C,E**), suggesting BcBAG1 is essential for sclerotial formation. Taken together, we reasoned that BcBAG1 plays a crucial role in regulating vegetative growth, conidiation and sclerotial formation in B. cinerea.

#### BcBAG1 Is Involved in the Regulation of Hyphal Melanization

After incubating on PDA for 7 days, we noticed that 1Bcbag1 displayed increased generation of black pigment when compared to WT (**Figure 3A**). It has been reported that the dark pigmentation in fungi is due to the accumulation of 1,8 dihydroxynaphthalene (DHN) melanin (Henson et al., 1999). Melanin is a dark durable pigment that protects fungi against diverse environmental stresses, such as UV irradiation and temperature extremes (Bell and Wheeler, 1986; Butler and Day, 1998). We therefore examined whether BcBAG1 participates in melanin biosynthesis. The 1Bcbag1, BcBAG1-Com and the WT strains were cultured on PDA supplemented with 50 µg/ml tricyclazole, a fungicide that specifically inhibits DHNmelanin biosynthesis (Thompson et al., 2000). The result showed tricyclazole was able to repress the massive melanin synthesis in 1Bcbag1 mutants caused by BcBAG1 deletion (**Figure 3A**). Additionally, we instituted RT-PCR assays to monitor the expression level of THR1 (1,3,8-trihydroxynaphthalene reductase gene), a key component in melanin biosynthesis pathway (Perpetua et al., 1996). Corresponding results obtained from RT-PCR analysis revealed a significantly up-regulation (about 10 folds) in the expression pattern of THR1 in 1Bcbag1 compared to WT (**Figure 3B**). These data indicate that BcBAG1 negatively regulates melanin biosynthesis pathway to suppress melanin production in B. cinerea.

## BcBAG1 Is Required for Virulence of B. cinerea

To determine the role of BcBAG1 in pathogenicity and virulence of B. cinerea, we conducted infection assays by inoculating mycelial plugs containing the WT, 1Bcbag1 and BcBAG1-Com, on detached tomato leaves and grapes, respectively. Four days post-inoculation (dpi), 1Bcbag1 only initiated a small localized lesions, whereas the WT and the complemented strains have produced fully expanded lesions that were already at the soft rot stage (**Figure 4A**). We also performed infection assays on tomato leaves, with conidial suspensions (1.0 × 10<sup>5</sup> spores/ml). Consistently, 1Bcbag1 showed apparently attenuated virulence compared to the WT and BcBAG1-Com (**Figures 4A,B**). These results indicate that BcBAG1 functions as a virulence factor by enhancing colonization on the hosts.

During infection, B. cinerea produces three types of penetration structures, including germ tube apices (GA), appressoria (HA) and infection cushions (IC) (Vandenheuvel and Waterreus, 1983). To investigate whether the weak virulence of 1Bcbag1 resulted from penetration defects, we evaluated conidial germination using hydrophobic coverslips and the penetration on onion epidermis cells for microscopic observation (**Figure 5A**). Although all strains initiated germination 12 h post-incubation (hpi), the germination rate of 1Bcbag1 was significantly reduced (by 50%) compared to the germination recorded for the WT and BcBAG1-Com (**Figure 5C**). Moreover, deletion of BcBAG1 delayed the onset of appressoria formation at 24 hpi and the development of infection cushions (IC) at 48 hpi (**Figure 5A**). Cotton blue was used to stain the appressorium-like structure, the infection cushion and the hyphae of B. cinerea on onion epidermal cells. As shown in **Figure 5B**, the conidia of WT and BcBAG1-Com germinated, and the infection cushions surrounded by abundant invasive hyphae were typically developed for penetration on the onion epidermis 48 hpi. In contrast, 1Bcbag1 attenuated the ability to form infection cushions and only few germ tubes and invasive hyphae appeared on onion cells. These results suggest that the reduced virulence of 1Bcbag1 is correlated with the developmental defects of infection and penetration structures.

#### Both UBL and BAG Domains Contribute to BcBAG1 Function

BcBAG1 possesses an N-terminal ubiquitin-like (UBL) domain and a conserved C-terminal BAG domain (BD) (**Supplementary Figure S1B**). Human BAG-1 interacts with HSP70 via its BAG domain and utilizes the UBL domain in targeting the chaperone cofactor to the 26S proteasome for degradation (Demand et al., 2001). The UBL/BAG domain proteins in S. pombe, SpBAG101 and SpBAG102, display similar interaction pattern to human BAG-1 (Kriegenburg et al., 2014; Poulsen et al., 2017). To validate the functionality of the UBL and the BAG domain of BcBAG1, we generated two truncated forms of BcBAG1; BcBAG1A1−<sup>141</sup> and BcBAG1B142−<sup>298</sup> (**Supplementary Figure S1B**), containing UBL and BAG domain, respectively. The BcBAG1A1−<sup>141</sup> and

BcBAG1B142−<sup>298</sup> constructs were subsequently transformed into 1Bcbag1 for complementation, designated as N3-11 and C2-8, respectively. Both N3-11 and C2-8 could not effectively rescue defects in conidia production and sclerotia formation (**Figure 2**). Interestingly, C2-8 could partially restored conidial germination, infection structure formation and pathogenicity to WT levels (**Figures 4**, **5**). These data suggest that both UBL and BAG domains are necessary for integral BcBAG1 function and we further inferred that the BAG domain plays an indispensable role for pathogenicity of B. cinerea.

#### BcBAG1 Modulates Multiple Stress Responses in B. cinerea

BAG family members are involved in cell protection during variable biotic and abiotic stress responses (Doukhanina et al., 2006; Behl, 2016). To investigate the function of BcBAG1 in response to environmental stress, we examined the sensitivity of 1Bcbag1 to various abiotic stress stimuli. As shown in **Figure 6**, 1Bcbag1 was more sensitive to salt stress (1 M NaCl) than WT, but no different with WT to another osmotic stress inducer 1.2 M D-sorbitol. Mycelial growth in response to cell wall stress inducers (0.6 mg/ml CFW; and 0.02% SDS) was measured. 1Bcbag1 showed dramatically increased sensitivity to both CFW and SDS compared to WT (**Figure 6**). In contrast, when exposing to oxidative stimuli, 20 mM H2O<sup>2</sup> and 0.5 mM TBHP, 1Bcbag1 led to an average ∼7% (20 mM H2O2) and ∼8% (0.5 mM TBHP) lower growth inhibition rate than the WT strain (**Figure 6**), indicating that disruption of BcBAG1 is more resistance to oxidative stress.

Human BAG-1 is a coupling factor between HSP70 and 26S proteasome (Luders et al., 2000). BAG3, as a co-chaperone, forms a complex with HSP70 to facilitate the degradation of ubiquitinated proteins via the proteasome or autophagy pathways (Gamerdinger et al., 2011; Minoia et al., 2014). Thus, we examined the role of BcBAG1 in proteasome degradation by testing the sensitivity of 1Bcbag1 to proteasome inhibitors, MG132 and Bort (Huang and Chen, 2009). Unexpectedly, results showed that the growth rate of 1Bcbag1 is not statistically different from WT and BcBAG1-Com (**Figure 6**), implying that BcBAG1 is not involved in response to proteasome inhibitors, MG132 and Bort. Although human BAG3 mediate the responses to Bort and MG132 (Judge et al., 2017), BcBAG1 does not share the same role in this perspective as the human counterpart.

#### BcBAG1 Negatively Regulates Unfolded Protein Response (UPR)

Previous work revealed that Arabidopsis BAG7 (AtBAG7) functions as an ER stress co-chaperone to maintain the UPR and protect plants from ER stress-induced cell death (Williams et al., 2010). Except AtBAG4, BcBAG1 shares relative higher identity to AtBAG7 in comparison to other Arabidopsis BAGs (**Supplementary Table S3**), we therefore examined whether BcBAG1 plays a role in ER stress signaling pathway. ER stress can be induced by chemical compounds, e.g., Tm or DTT (Oslowski and Urano, 2011). Besides, environmental/abiotic stress including excessive heat and cold also trigger ER stress (Williams et al., 2010). Accordingly, we cultured the WT, 1Bcbag1 and BcBAG1-Com on media supplemented with 2 µg/ml Tm and 2.5 mM DTT for 3 days at 22◦C. The results showed that the growth of 1Bcbag1 was strongly inhibited by Tm and DTT with a much higher inhibition rate comparing to WT and BcBAG1-Com (**Figure 6**). In addition, after incubating on PDA under heat (30◦C) and cold (4◦C) conditions for 7 days, 1Bcbag1 was more sensitive than the WT and BcBAG1-Com to both heat and cold treatments (**Figures 7A–C**). Quantitative realtime (qRT-PCR) analysis demonstrated that the transcription of BcBAG1 was highly induced upon above ER stress conditions, including heat treatment (50◦C for 30 min), Tm (2 µg/ml for 1 h), or DTT (20 mM for 1 h) (**Figure 7C**), indicating that BcBAG1 is responsible for ER stress tolerance.

To cope with ER stress, eukaryotes utilize UPR to alleviate the detrimental effects (Schroder and Kaufman, 2005). In S. cerevisiae, UPR is sensed by transmembrane protein kinase and ribonuclease (RNase) IRE1 and initiated with IRE1-mediated splicing of an unconventional intron (252-nucleotide) from the HAC1 transcript (Cox and Walter, 1996; Ruegsegger et al., 2001). HAC1 encodes a basic leucine zipper-type (bZIP) transcription factor, and the splicing of HAC1 regulates the expression of UPR target genes thus to mitigate ER stress (Mori et al., 1996). To explore the underlying mechanism within the sensitivity of 1Bcbag1 to ER stress, we detected the presence of the spliced and unspliced form of BcHAC1 mRNA by RT-PCR. Spliced form of BcHAC1 (BcHAC1<sup>S</sup> ) in B05.10 was significantly increased upon ER stress, suggesting B. cinerea shares the similar HAC1 splicing process with other organisms during UPR (**Figure 8A**). Notably,

under normal conditions, the amount of BcHAC1<sup>S</sup> was drastically more abundant in 1Bcbag1 than the WT (**Figure 8A**, upper panel). More interestingly, BcHAC1 was constitutively spliced in 1Bcbag1 under both conditions that are with/with-out stress (**Figure 8A**). DNA sequencing confirmed that a 20 nucleotide of the fragment was absent in the spliced form compared to the unspliced form (**Figure 8A**, lower panel). Meanwhile, we examined the expression of UPR-related genes in B. cinerea, BcBIP1 and BcIRE1, homologs of ER chaperone KAR2/BIP1 and ER stress sensor/transducer IRE1 in S. cerevisiae, respectively (Pincus et al., 2010). Both BcBIP1 and BcIRE1 were induced following ER stress in the WT (**Figure 8B**). Without stress, the expression levels of BcIRE1 and BcBIP1 were increased by threefold and fivefold, respectively, in 1Bcbag1 (**Figure 8B**). All these results indicates that, BcBAG1 deletion causes constitutive activation of UPR through negatively regulating the expression of BcBIP1, BcIRE1 and the splicing of BcHAC1 mRNA. We speculate that BcBAG1 effectively repress the harmful and excessive constitutive activation of UPR, thus to maintain the proper UPR level during ER stress signaling pathway.

#### BcBAG1 Binds HSP70-Type Chaperones

The heat shock proteins HSP70 family play crucial roles in assisting a variety of protein folding processes (Mayer and Bukau, 2005). Human BAG-1 binds the ATPase domain of Hsc70 to stimulates Hsc70 ATP hydrolysis which results in the

release of ADP from Hsc70, thereby regulates specific protein folding and maturation pathways (Hohfeld and Jentsch, 1997). To ascertain whether BcBAG1 bind HSP70-type chaperone in B. cinerea, we performed the yeast two-hybrid assay to establish possible interaction between BcBAG1 and the HSP70 family members, including BcBIP1, BcSSC1, BcSKS2, BcSSA1, BcPSS1, and BcLHS1, which are homologs of HSP70 family in the fission yeast. Results showed that BcBAG1 only interacted with BcBIP1 and BcSKS2 (**Figure 8C**), demonstrating BcBAG1 does function as a co-chaperone of HSP70 proteins, in accordance with that has been reported in other systems (Gassler et al., 2001; Sondermann et al., 2002; Poulsen et al., 2017).

#### BcBAG1 Restores Salt Stress Tolerance to atbag4

Using NCBI blastp and blastn programs, we found BcBAG1 is predicted to be the closest homolog of Arabidopsis BAG protein AtBAG4, in light of the amino acid sequences of the BAG domains between BcBAG1 and AtBAG4 share 33% identity and 50% similarity (**Supplementary Figures S1A,C** and **Supplementary Table S3**). To determine whether BcBAG1 is able to functionally complement the AtBAG4 T-DNA mutants,

full-length BcBAG1 was overexpressed using the cauliflower mosaic virus 35S promoter in the atbag4 T-DNA mutants (atbag4::BcBAG1). Expression levels of BcBAG1 in atbag4 was confirmed by RT-PCR (**Figure 9A**). Previous studies indicated that atbag4 mutants were more susceptible to salt stress (100 mM NaCl) compared to the WT Col-0 (Doukhanina et al., 2006). Here, we performed the salt tolerance assay for atbag4::BcBAG1, taking Col-0 and atbag4 as the positive and negative controls, respectively. Arabidopsis seedlings were cultured on 1/2 MS medium supplemented with 100 mM NaCl. After 2 weeks treatment, no difference was observed from seedlings. However, 5 weeks later, atbag4 mutants displayed massive chlorosis and bleaching of leaves, while Col-0 and atbag4::BcBAG1 plants grew well and exhibited nearly the same growth tendency under salt conditions (**Figure 9B**). These observations demonstrate that BcBAG1 can be ectopically expressed in Arabidopsis and fully restore salt tolerance in atbag4.

#### DISCUSSION

As an evolutionarily conserved group, BAG proteins from yeast to plants and mammals have been associated with

regulation of PCD and cell protection. Recently, they have also been found to play an important role in autophagy, UPR and ubiquitin-proteasome system (Kabbage et al., 2017). However, the identification and characterization of BAG proteins in phytopathogenic fungi is rare. In this study, we explored the B. cinerea genome and identified a unique BAG gene, BcBAG1. Targeted deletion of BcBAG1 exerted strong adverse effects on vegetative growth, conidiation, sclerotia formation, hyphal melanization, stress response, conidial germination, penetration and virulence, suggesting that the pleiotropic function of BAGs delineated in mammals and plants appears to be maintained in B. cinerea. The BAG domain of BcBAG1 shares highest similarity with AtBAG4 in Arabidopsis, also an ectopic expression of BcBAG1 fully restored salt tolerance of atbag4 mutants (**Figure 9**). In addition, deletion of BcBAG1 resulted in increased sensitivity to salt, cell wall stressors, ER stress inducers, and heat or cold treatments. These results parallel plant AtBAG4 studies in which AtBAG4-overexpressing transgenic tobacco plants confer tolerance to a wide range of stresses such as UV light, cold, salt treatments (Doukhanina et al., 2006). From this perspective, the cytoprotective function of BAGs in response to diverse stresses is relatively conserved between the fungi and plants.

Our previous studies have addressed the importance of AtBAG7 in the maintenance of the UPR and the mechanisms of ER-localized co-chaperone AtBAG7 in stress protection (Williams et al., 2010; Li et al., 2017). Interestingly, BcBAG1 showed a close evolutionary relationship with AtBAG7 among plant BAGs (**Supplementary Figure S1A** and **Supplementary Table S3**). Moreover, 1Bcbag1 were more sensitive to ER stress stimuli (**Figure 6**) and expression of BcBAG1 was induced under ER stress (**Figure 7B**), suggesting that both BcBAG1 and AtBAG7 are functionally associated with the ER stress. Of note, we found that the deletion of BcBAG1 gives rise to the constitutive activation of UPR with high levels even without stress (**Figures 8A,B**), indicating that BcBAG1 is necessary for the inhibition of excess UPR under normal condition. Given the defects in fungal development and differentiation caused by deletion of BcBAG1, we speculate that under normal conditions, the abnormal activated UPR is actually harmful to the fungus. It has been reported that

moderate activation of UPR is necessary for ER recovery when responding to stress in S. cerevisiae (Chawla et al., 2011). Thus, we suggest that constitutive activation of UPR in 1BcBag1 results in loss of normal ER proteinfolding capacity, which impinges upon the ability to sustain resistance to ER stress. Thus, 1Bcbag1 displayed growth defects during ER stress. In addition, as a multifaceted HSP70 molecular chaperone, BIP ensures an appropriate response to restore protein folding homeostasis to the ER by providing a buffer for inactive IRE1 (Pincus et al., 2010). Our result revealed that disruption of BcBAG1 increases the expression of BcBIP1 and BcIRE1 (**Figure 8B**) and BcBAG1 interacts with BcBIP1 (**Figure 8C**). We conclude that the regulation of UPR by BcBAG1 correlates to BcBIP1. The relationship between BcBAG1/BcBIP1/BcIRE1 and how BcBAG1 modulates the ER machinery require further studies.

Previous studies of human BAG-1/HSC70 complex revealed that BAG-1 exploits Glu212, Asp222, Arg237, and Gln<sup>245</sup> residues in the BAG domain to bind with Hsc70 ATPase domain (Sondermann et al., 2001). Multiple alignment showed that these residues, with the exception of Glu212, are highly conserved in BcBAG1 BD (**Supplementary Figure S1C**), implying that BcBAG1 interacts with HSP70 in a similar manner as with human BAG-1. We did examine the interaction between BcBAG1 and HSP70 proteins by Y2H and found that BcBAG1 exclusively interacts with BcBIP1 and BcSKS2 in the HSP70 family, but fails to bind to BcSSC1, BcSSA1, BcPSS1, and BcLHS1 (**Figure 8C**). Previous studies demonstrated that the S. cerevisiae BAG protein SNL1 interacts with HSP70

family members including SSA1, SSB1, SSB2, SEE1, and SEE2 (Sondermann et al., 2002; Verghese and Morano, 2012; Abrams et al., 2014). While the S. pombe BAG proteins SpBAG101 and SpBAG102 exclusively interact with SSA1, SSA2, and SKS2 (Poulsen et al., 2017). These data suggest BAGs from different origins show different affinity and specificity to HSP70s members. Notably, the interaction between BcBAG1 and BcSKS2 is conserved from that in yeast studies. BcSKS2 is the homolog of fission yeast ribosome-associated chaperone SKS2, and the S. cerevisiae orthologs of fission yeast SKS2, called SSB1 and SSB2. SSB1/2 chaperones play a dual role in de novo protein folding and ribosome biogenesis (Mudholkar et al., 2017). We conjectured that the binding to primary HSP70 chaperone might be responsible for the functional pleiotropy of BcBAG1.

Apart from the BAG domain, BcBAG1 also contains a UBL domain. Human BAG-1 function as a link between HSC/HSP70 and 26S proteasome degradation system via its UBL domain, chaperone-bound substrates are released and degraded (Luders et al., 2000; Alberti et al., 2002). Co-precipitation experiments in fission yeast provide direct evidence that both SpBAG101 and SpBAG102 interact with 26S proteasomes depend on the UBL domain (Poulsen et al., 2017). However, we found that response to proteasome inhibitors, e.g., Bort and MG132, by BcBAG1 deletion was unaffected in fungus (**Figure 6**). Besides, the result of ubiquitination assay confirmed that deletion of BcBAG1 does not alter the levels of ubiquitination (**Supplementary Figure S3**). Therefore, we inferred that BcBAG1 is not a key player in the proteasome degradation. However, we cannot exclude the possibility that other proteasome inhibitors may work on BcBAG1 or BcBAG1 can be targeting some proteasome substrate for degradation.

Botrytis cinerea, however, integrates a number of hurdles that must be traversed for successful colonization and defense against the plethora of plant hosts that are encountered (Elad et al., 2007). Therefore, it is relatively difficult to present a precise mechanism for the alteration of pathogenicity and virulence in the host. Based on our data, we supposed that the attenuated pathogenicity of 1Bcbag1 is directly or indirectly related to the following reasons. First, 1BcBag1 showed a decrease of conidial germination, and delayed the formation of appressoria and infection cushions on an artificial surface and onion epidermis (**Figure 6**). At the same time, we observed that BcBAG1 resides in the cytoplasm throughout growth stages, but the localization is altered and most likely concentrated at the infection spots during invasion (**Supplementary Figure S4**), suggesting the expression of BcBAG1 contributes to the formation of penetration structures. Therefore, the impairment of penetration structures in 1BcBag1 weakens the ability of breaching the host tissues to effectively cause disease. Second, it is reported that cell wall integrity is crucial for B. cinerea virulence and pathogenicity (Soulie et al., 2003, 2006; Cui et al., 2013). 1BcBag1 increased sensitivity to cell wall stressors, CFW and SDS (**Figure 6**), indicating BcBAG1 involves in cell wall integrity pathway. Therefore, defect in cell wall integrity is one key to the reduction of virulence. Third, melanin is a factor affecting the virulence of B. cinerea. BcPKS13 and BcBRN1, encoding polyketide synthase and tetrahydroxynaphthalene (THN) reductases, respectively, both are involving in fungal DHN melanin biosynthesis (Zhang et al., 2015). Loss of BcPKS13 and BcBRN1 blocks melanization resulting in enhanced virulence. Conversely, overexpression of BcBRN1 enhances melanization, decreases secretion for virulence factors such as several hydrolytic enzymes and oxalic acid, and attenuated virulence (Zhang et al., 2015). From this perspective, an increment of melanin biosynthesis negatively affects the pathogenesis of B. cinerea. Consistently, our results showed that BcBAG1 negatively regulates melanin biosynthesis (**Figure 3**), thus the increased mycelial melanin biosynthesis suppress fungal virulence for 1BcBag1. Fourth, previous studies indicated UPR plays as a central regulator of fungal pathogenesis (Heimel et al., 2013; Guillemette et al., 2014; Krishnan and Askew, 2014a,b). Our study found that BcBAG1 is responsible for proper maintenance of UPR, as a result, abnormal activation of UPR in 1Bcbag1 could cause the attenuated virulence.

In summary, this paper details the biological functions of BcBAG1. We have shown that BcBAG1 exhibits functional versatility and is involved in fungal development, differentiation, stress response, and pathogenicity. BcBAG1 acts as a cochaperone of HSP70 and is a key regulator required for maintenance of the UPR. In light of these findings, future studies involving translocation of BcBAG1 during infection and identification of other targets are of interest. Taken together, these results demonstrate the importance of the BAG family in filamentous fungus cell death pathways and cytoprotection.

#### AUTHOR CONTRIBUTIONS

fmicb-10-00685 April 8, 2019 Time: 7:54 # 13

HZ and MD designed the research. HZ performed the experiments. HZ, YL, and MD analyzed the data. HZ, YL, MD, and ZW wrote the article. MD and ZW revised and approved the manuscript.

#### FUNDING

This work was partially supported by NSFC (Grant No. 31471739).

#### ACKNOWLEDGMENTS

We thank Professor Matthias Hahn (Kaiserslautern University, Germany) for providing vector pNAH-OGG and pNAH-OMG. We also thank the Chinese CSC Scholarship Program for the grant of HZ.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.00685/full#supplementary-material

FIGURE S1 | Phylogenetic and sequence analysis of BAG1 in Botrytis cinerea (BcBAG1). (A) The phylogenetic tree of BAG proteins. Evolutionary analyses were conducted in MEGA7.0. The evolutionary history was inferred by using a neighbor-joining method based on the amino acid sequences. The numbers at nodes inferred the percentage of their occurrence in 10,000 bootstrap replicates. Species names and GenBank accession numbers of each sequence are represented as follows: SsBAG1 (Sclerotinia sclerotiorum, XP\_001591798.1); MoBAG1 (Magnaporthe oryzae, XP\_003710309.1); FoBAG1 (Fusarium oxysporum f. sp. Lycopersici, XP\_018239181.1); NcBAG1 (Neurospora crassa, XP\_961586.1); AnBAG1 (Aspergillus nidulans, XP\_661815.1); UmBAG1 (Ustilago maydis, KIS67500.1); SpBAG101 (Schizosaccharomyces pombe, NP\_596760.1);

#### REFERENCES


SpBAG102 (Schizosaccharomyces pombe, NP\_595316.1); ScSNL1 (Saccharomyces cerevisiae, KZV10602.1); HsBAG-1M (Homo sapiens, NP\_001336215.1); HsBAG2 (Homo sapiens, NP\_004273.1); HsBAG3 (Homo sapiens, NP\_004272.2); HsBAG4 (Homo sapiens, NP\_004865.1); HsBAG5 (Homo sapiens, NP\_001015049.1); HsBAG6 (Homo sapiens, P46379.2); AtBAG1 (Arabidopsis thaliana, NP\_200019.2); AtBAG2 (Arabidopsis thaliana, NP\_568950.2); AtBAG3 (Arabidopsis thaliana, NP\_196339.1); AtBAG4 (Arabidopsis thaliana, NP\_190746.2); AtBAG5 (Arabidopsis thaliana, NP\_172670.2); AtBAG6 (Arabidopsis thaliana, AEC10664.1); AtBAG7 (Arabidopsis thaliana, NP\_201045.1). (B) Schematic diagram of BcBAG1. Purple red and green boxes indicate the UBL and BAG domain of BcBAG1, respectively. (C) Sequence alignment of the conserved BAG domain from different organisms. Three predicted helixes labeled on the top. Conserved residues which involving in binding of BAG protein to Hsc70 ATPase domain in human HsBAG-1M are indicated by red arrow, and residues critical to packing interactions are highlighted by blue arrow.

FIGURE S2 | Sketch of gene deletion and identification of deletion mutants and complemented transformants. (A) Diagram of targeted gene replacement. (B) RT-PCR confirmation. Lane 1:100 bp DNA Ladder (NEB); lanes 2 and 3: K3-7 and K8-6, respectively; lane 4: B05.10; lanes 5 and 6: complemented transformants; lane 7: PCR negative control. (C) Southern blotting confirmation. Total genomic DNAs were digested by PvuI, and a DNA fragment in the upstream of 5<sup>0</sup> terminus of BcBAG1 was selected and labeled as the probe shown in panel (A). M: 1 kb DNA Ladder (NEB).

FIGURE S3 | Disruption of BcBAG1 does not affect protein ubiquitination. Total protein extracts from corresponding strains was western blotted with (upper panel) an anti-Ub antibody (P4D1) and stained with Coomassie brilliant blue (lower panel) as the loading control. B05.10: the wild-type strain; K8-6: 1Bcbag1 mutant lines; C2-8 (C-terminus of BcBAG1) and N3-11 (N-terminus of BcBAG1).

FIGURE S4 | Subcellular localization of BcBAG1. GFP-BcBAG1 was overexpressed in the 1Bcbag1 mutant (OG2-7) and the fluorescence for BcBAG1 localization at different stages was visualized by confocal microscopy. (A) Vegetative hyphae; (B) conidia; (C,D) conidia on hydrophobic glass slides for 12 and 24 h, respectively. (E,F) Both are conidia on onion epidermal cells for 24 h and 48 h, respectively. Scale bars: 20 µm.

TABLE S1 | Wild-type and mutant strains of Botrytis cinerea used in this study.

TABLE S2 | Oligonucleotide primers used in this study.

TABLE S3 | Similarity of full length and BAG domain amino acid sequence between BcBAG1 and other BAGs.



behavior of the unfolded protein response. PLoS Biol. 8:e1000415. doi: 10.1371/ journal.pbio.1000415


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

Copyright © 2019 Zhang, Li, Dickman 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) 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.

# Spatial Effects and GWA Mapping of Root Colonization Assessed in the Interaction Between the Rice Diversity Panel 1 and an Arbuscular Mycorrhizal Fungus

Hazel Davidson, Roshi Shrestha, Thomas Cornulier, Alex Douglas, Tony Travis, David Johnson† and Adam H. Price\*

Institute of Biological and Environmental Sciences, University of Aberdeen, Aberdeen, United Kingdom

If water saving methods of rice management are to be adopted, the interaction between rice plants and arbuscular mycorrhizal (AM) fungi will grow in agronomic significance. As yet there are very few studies on the interaction between rice and AM fungi and none on host genetics. A subset 334 cultivars from the Rice Diversity Panel 1 were grown in 250 L boxes filled with phosphorus (P) deficient aerobic soil without addition, with added rock phosphate and with rock phosphate and the AM fungus Rhizophagus irregularis. Statistical analysis of position of plants revealed a positive effect of their neighbors on their dry weight which was stronger in the presence of rock phosphate and even stronger with rock phosphate and AM fungi. A weak but significant difference in the response of cultivars to AM fungus treatment in terms of shoot dry weight (SDW) was revealed. Neighbor hyphal colonization was positively related to a plant's hyphal colonization, providing insights into the way a network of AM fungi interact with multiple hosts. Hyphal colonization ranged from 21 to 89%, and 42% of the variation was explained by rice genotype. Colonization was slightly lower in aus cultivars than other rice subgroups and high in cultivars from the Philippines. Genome wide association (GWA) mapping for hyphal colonization revealed 23 putative quantitative trait loci (QTLs) indicating there is an opportunity to investigate the impact of allelic variation in rice on AM fungal colonization. Using published transcriptomics data for AM response in rice, some promising candidate genes are revealed under these QTLs being a calcium/calmodulin serine/threonine protein kinase at 4.9 Mbp on chromosome 1, two ammonium transporters genes at 24.6 Mbp on chromosome 2 and a cluster of subtilisin genes at 1.2 Mbp on chromosome 4. Future studies should concentrate on the biological significance of genetic variation in rice for AM colonization.

Keywords: Oryza sativa, Rhizophagus irregularis, Glomus intraradices, common mycorrhizal network, GWA mapping, QTL

## Edited by:

Marco Catoni, University of Birmingham, United Kingdom

#### Reviewed by:

Katsuharu Saito, Shinshu University, Japan Vanesa Silvani, University of Buenos Aires, Argentina Cristiana Sbrana, Istituto di Biologia e Biotecnologia Agraria (IBBA), Italy

> \*Correspondence: Adam H. Price a.price@abdn.ac.uk

#### †Present address:

David Johnson, School of Earth and Environmental Sciences, The University of Manchester, Manchester, United Kingdom

#### Specialty section:

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

Received: 31 January 2019 Accepted: 26 April 2019 Published: 17 May 2019

#### Citation:

Davidson H, Shrestha R, Cornulier T, Douglas A, Travis T, Johnson D and Price AH (2019) Spatial Effects and GWA Mapping of Root Colonization Assessed in the Interaction Between the Rice Diversity Panel 1 and an Arbuscular Mycorrhizal Fungus. Front. Plant Sci. 10:633. doi: 10.3389/fpls.2019.00633

**76**

#### INTRODUCTION

fpls-10-00633 May 16, 2019 Time: 14:41 # 2

The symbiotic interaction between land plants and mycorrhizal fungi is ancient and is thought to be driven by the provision of nutrients (particularly P and N) to the plant in exchange for carbon (Smith and Read, 2008) in the form of sugar and lipids (Keymer and Gutjahr, 2018). Most grasses, including major crop plants, form symbioses with arbuscular mycorrhizal (AM) fungi that can affect agricultural plant productivity (Van Der Heijden et al., 2015), but research on the interaction between the grass crop rice and AM fungi has been very limited. This is probably because in flooded soil, in which most rice is grown, this association is traditionally considered unimportant. It has, for example, been shown that in flooded conditions AM fungi are very scarce on rice (Ilag et al., 1987) while it has been demonstrated that flooding of aerobically-grown rice roots decreases AM fungal colonization within 7 days- although it does not eliminate AM fungi completely (Vallino et al., 2014). Fieldwork has begun to provide evidence for an impact of AM fungi on rice productivity in upland (aerobic) soils (Maiti et al., 2011) but not for flooded rice. Yet as water availability and sustainability of crop production grow as drivers for crop management decisions [see for example page 8–9 in Bouman et al. (2007)], it can be predicted that increasing amounts of the world's rice production will adopt one of a number of water saving techniques including aerobic rice, alternate wetting and drying or the system of rice intensification (SRI) (Bouman et al., 2007, p. 19–28). Under these growing conditions, rice-AM fungi interactions are more likely to be important. Indeed it has recently been shown that AM fungi colonization and AM fungal species richness were substantially higher under SRI (which includes a more aerobic soil condition) than conventional flooding (Watanarojanaporn et al., 2013). So it is timely to investigate the interaction more closely in order to consider the possibility that future rice breeding might need to incorporate the knowledge gained.

Molecular characterization of the AM fungi-rice interactions has provided insight into the biology of the interaction (Güimil et al., 2005; Campos-Soriano et al., 2012; Gutjahr et al., 2015). However, there has been almost no studies investigating if there are differences between rice cultivars in their interaction with AM fungi and no studies on the host genetics of the interaction. In a comparison of six upland cultivars from China, one was found to have much lower colonization when infected by either of two AM fungi (Gao et al., 2007), while Li et al. (2016) found differences in colonization of six cultivars with Rhizophagus intraradices. In the field in Italy no differences in colonization between 12 cultivars was detected (Vallino et al., 2009). Suzuki et al. (2015) assessed the growth response of 64 cultivars of rice to Funneliformis mosseae and found wide differences (from −4 to +119%). The authors assessed colonization in 12 cultivars (names not specified) and although this did differ between cultivar it did not related to the degree of growth response. We are not aware of any other studies that assessed differences in colonization between rice cultivars.

While there have been no studies on genetic mapping AM fungal interaction in rice, there have been on other crops. Leiser et al. (2016) found very little evidence of QTLs for AM colonization in sorghum probably because of weak repeatability between replicates. In contrast, Lehnert et al. (2017) examining wheat found strong genetic variation but still rather weak evidence of quantitative trait loci (QTLs) suggesting AM colonization is controlled by small effect genes, meaning QTLs and potential candidate genes reported are speculative. Finally, De Vita et al. (2018) screened 108 durum wheat cultivars for colonization by R. irregularis and F. mosseae, detecting substantial variation and seven putative QTLs by genome wide association (GWA) mapping.

In this study, the well characterized global collection of rice cultivars, the Rice Diversity Panel 1 (Zhao et al., 2011) was screened for growth and colonization in response to inoculation with the 'model' AM fungus R. irregularis (formerly known as Glomus intraradices). This species has been found colonizing aerobically-grown rice roots in India (Bhattacharjee and Sharma, 2011) and dominating them in Italy (Vallino et al., 2009), and was the first AM fungus sequenced (Tisserant et al., 2013; Van Der Heijden et al., 2015).

The main aim of this study was to test the hypothesis that there is genetic variation within rice for the colonization and the impact of AM fungus R. irregularis, and that these are correlated. Further, we test the hypothesis that genetic variation for AM fungal colonization can be genetic mapped. This was done using 334 accessions of the Rice Diversity Panel 1, allowing GWA mapping since there are 5.2 million SNP markers available for this population (Wang et al., 2018). To provide phenotype data that are relevant, discriminating and achievable required that a time-efficient screening system be developed. The design of the boxes used, specifically the competitive nature of the arrangement of plants, allowed three further hypotheses to be tested; (1) that spatial relationships between the rice plants affect the growth of the rice plant, (2) that the degree of neighbor interactions is different if an AM fungus is present, and (3) that AM fungal colonization rates are affected by the size and AM fungal colonization of neighboring plants.

#### MATERIALS AND METHODS

#### Rice Genotypes

A random set of 334 accessions of the Rice Diversity Panel 1 (RDP1) (Zhao et al., 2011) was used for this study (see **Supplementary Table S1**). RDP1 represents 372 cultivars which come from the five subpopulations of rice (indica, aus, tropical japonica, temperate japonica, and aromatic). The cultivars have been genotyped using the High-Density Rice Array (HDRA) of 700,00 SNPs, with ∼1 informative SNP per kb (McCouch et al., 2016) and this data base has recently been extended to 5.2 million SNPs by imputation (Wang et al., 2018). In addition to the RDP1, cultivars Azucena and Bala which have been continuously grown in Aberdeen from seed originally obtained from the International Rice Research Institute were used as local check cultivars. These are distinguished from the Azucena and Bala in the RDP1 population by referring to them as "Own Azucena" and "Own Bala."

#### AM Inoculum

fpls-10-00633 May 16, 2019 Time: 14:41 # 3

Rhizophagus irregularis inoculum was purchased from INVAM, United States and amplified in maize by growing three plants per pot in four pots of 2.5 L capacity filled with fine sand, each pot receiving 60 g of inoculum. These were fed weekly with zero P Yoshida's nutrient solution (Yoshida et al., 1976) and on two occasions 1/4 P was included in the nutrient solution. After 3 months growth, when 100% hyphal colonization was confirmed in subsampled roots, the maize plants were dried by withholding water, the roots cut into fine pieces and the roots and sand used as inoculum.

#### Soil

Throughout, the soil used was a mixture of 1 part subsoil from Insch Field Farm [previously described in MacMillan et al., 2006 (available P 12.24 µg g−<sup>1</sup> , total P 814 µg g−<sup>1</sup> )] and 3 parts fine sand (Sibelco UK; available P 1.95 µg g−<sup>1</sup> , total P 12.2 µg g−<sup>1</sup> ). This was chosen as previous research had shown it suitable for experiments on P response (Al-Ogaidi, 2013).

#### Greenhouse Conditions

All experiments were conducted in the Cruickshank Greenhouse of the University of Aberdeen where the temperature is set to 25◦C and plants received ambient light plus 12 h a day of supplementary light of approximately 150 µM m−<sup>2</sup> s <sup>−</sup><sup>1</sup> PAR.

#### AM Colonization Measurement

The proportion of roots containing hyphae, arbuscules or vesicles of the AM fungus was assessed using the magnified intersections method of McGonigle et al. (1990) and slight modifications of the staining method of Vierheilig et al. (1998). Roots were washed to be free of soil and stored in 60% ethanol. Approximately 8–12 fine root sections of 3–4 cm in length were selected at random from the central portion of the root. After washing out the ethanol, they were bathed in 10 ml of 10% KOH at 90◦C for 10 min. After clearing, samples were rinsed in cold water and acidified by dipping in 1% HCl for 30 s and then transferred to test tubes containing 8 ml of 1% ink (Parker's Quink Black) in 1% HCl at 85◦C for 4 min. This was followed by degassing under vacuum (while still in the stain) for 2 min. Samples were then washed in water to remove excess stain. Finally, they were transferred into sample storage jars containing lactic acid and glycerol (1:1:14 glycerol:water:lactic acid) de-stain for at least 24 h. These were mounted on microscope slides in glycerol. The proportion of approximately 100 root-graticule intersections observed for each plant containing a visible fungal hyphae, or an arbuscle or a vesicle was recorded. Hyphal colonization was calculated as the sum of all three of those since an arbuscle and a vesicle must be accompanied by a hyphae.

#### Optimizing the Screening Method

Four preliminary experiments were conducted in order to optimize the screening methodology.

#### Dose Response

A dose experiment was conducted in the greenhouse using 1 L pots where 200 ml of soil/sand mix was placed at the bottom, 500 ml was placed in the middle and 100 ml on top. Different concentrations of inoculum were incorporated into the middle portion to give seven treatments being of 0, 1, 2, 5, 10, and 20% with an addition 10% autoclaved inoculum. Four replicates were used. Surface sterilized seeds of Azucena were placed in each pot and grown for 8 weeks with 200 ml of Yoshida's nutrient minus P being given twice a week. SDW and colonization were assessed. This revealed no impact of the autoclaved inoculum on plant growth, AM fungi substantially reduced growth, 1% inoculum had a smaller effect, while the other treatments were identical. Hyphal colonization were not different between 2, 5, and 20% inoculum. No more than 2% was therefore considered an appropriate rate of inoculation. Importantly, no hyphae were observed in the 0 and 10% autoclaved treatments (in 12–14 roots of 4 replicate plants of both treatments) indicating the subsoil used is not a source of AM inoculum.

#### Time Course

A time course was conducted in which 1 and 10% inoculation (using the same pots, soil, filling methods, rice genotype and growth conditions as described above) were harvested at 21, 36, and 51 days. Arbuscles, vesicles, and hyphal colonization was assessed (**Figure 1**). The colonization rate increased with time (from 19–65%) but while 1% inoculum consistently showed a lower colonization rate than 10% inoculum, there was no interaction between dose and time (based on two way ANOVA P > 0.05). It was considered that a dose of 1% or more and a 4 weeks growth period should be suitable for studying the genetics of AM fungus-rice interactions.

#### AM Fungi and Rock Phosphate

A third preliminary experiment was conducted to examine the interaction between AM fungus and rock phosphate using pots, soils, soil filling, rice genotype and plant growth conditions as above. Treatments with 0, 50, 100, 200, and 300 mg of rock phosphate per pot were applied with and without AM fungi at 2% with four replicates. At 4 weeks, SDW and leaf P uptake were assessed while the AM colonization was assessed on the AM fungus treatment. AM treatment reduced SDW by 25% but rock phosphate did not affect it. AM fungus reduced leaf P concentration (from 1.47 to 1.08 mg g−<sup>1</sup> : P < 0.001) while rock phosphate increased it (from 1.08 to 1.46 mg g−<sup>1</sup> : P = 0.022) and there was no interaction. Rock phosphate increased hyphal colonization in a linear fashion (0 = 27%, 50 = 36%, 100 = 42%, 200 = 49%, and 300 = 61%; P = 0.035). This experiment suggested adding rock phosphate did not affect growth of these plants at this young age but did increase AM colonization.

#### Genetic Variation in Two Cultivars

A final preliminary experiment determined if there were genetic differences in colonization rates by testing cultivars Azucena and Bala [parents of the mapping population described in Price et al. (2000)]. Zero and 300 mg rock phosphate and zero and 2% AM fungus inoculation were used in a fully factorial design with four replicates. Other conditions were as above. Root and shoot fresh weight (not dry weight) were recorded. The AM fungus reduced the root weight of both cultivars (P = 0.010) but it did not

affect shoot weight while rock phosphate increased shoot weight (P = 0.013) but did not affect root weight. There was a highly significant genotype by rock phosphate by AM fungus interaction for % root mass (root fresh weight/total fresh weight × 100) (P = 0.004) which was high in both cultivars in the absence of rock P (65–67%) but was lowest for Azucena with rock phosphate and AM fungus (61%) and lowest in Bala with rock phosphate but without AM fungus (58%). There was a highly significant difference in hyphal colonization rate between Bala and Azucena (P < 0.001) (Azucena 55.1%, Bala 30.4%), and a significant interaction between rock phosphate and cultivar (P = 0.010) where colonization was halved in Bala by rock phosphate (from 40.7 to 20.1%) but Azucena was not affected. It was concluded that treatment with rock phosphate may increase the ability to discriminate between cultivars.

#### Screening the Rice Diversity Panel 1

Boxes were built from plywood 3 m long, 0.85 m wide and approximately 0.35 m deep and lined with plastic sheet. They were then divided into three 1 m long sections orientated north to south using stiff plastic and each section filled with subsoil/sand mix according to three treatments to a depth of 0.3 m (approximately 250 L of subsoil/sand in each) (**Figure 2**). A control treatment had no additions, a rock phosphate treatment (RP) had 100 mg per plant placed in a band in the top 10 cm and a rock phosphate and AM fungus treatment (RP + AM) had 100 mg per plant rock phosphate and 1% AM fungal inoculum placed in a band in the top 10 cm. Two surfacesterilized seeds of 334 cultivars of the Rice Diversity Panel 1 were sown completely randomized in a grid pattern of 17 × 20 with a spacing of 5 cm between plants. The same randomization was used for each of the three treatments within a replicate run. After germination, plants were thinned to one. Plants were watered daily while every other day they received Yoshida's nutrient solution without P so that they received approximately 0.125 L each (equivalent to 5 mg of N which would produce a plant of approximately 0.25 g shoot dry weight (SDW) if not restricted by P). After 4 weeks growth, shoots were harvested for dry weight measurement while for the AM treatment alone the roots were harvested for colonization assessment. The experiment was repeated in four replicate runs starting on 15th May, 29th May, 9th July, and 23rd July 2013 each with a different randomization of the 334 genotypes. The order of the treatments was also randomized between runs and consisted of, from south to north run 1 Control:RP + AM:RP; run 2 AM + RP:RP:Control; run 3 RP + AM:RP:Control; run 4 RP:Control:RP + AM. Note AM hyphal colonization were assessed on 4–6 plants from the control and RP treatment in every run. On one occasion, 12% colonization was found, but the vast majority of samples contained no evidence of colonization.

#### Spatial Statistics

Spatial and genetic variation in SDW was modeled using generalized additive mixed models, assuming a Gaussian error after log transformation, in order to stabilize the variance. Hyphal colonization (%) was modeled using generalized additive mixed models, assuming a Gaussian error. The models included a combination of fixed, smooth and random effects listed in **Table 1**, as well as second-order interactions between the fixed effects. The models were fitted by REML with the gamm and gamm4 functions of the R packages 'mgcv' and 'gamm4,' respectively (Wood et al., 2013). For the SDW analysis, the 'Variety' random intercept was nested in 'Subgroup,' and a genotype by treatment interaction was evaluated by fitting the corresponding random effect (see **Table 1**). We tested if separate spatial trends (smoothing splines) were required for each block, box or run, or if all 4 runs could be suitably described by a common spatial trend, using the Bayesian Information Criterion as an indication of most parsimonious model. For the hyphal colonization analysis, we used a 'Variety' random intercept nested in 'Subgroup.' Spatial dependence of hyphal colonization of the focal plant on the colonization of its immediate neighbors was

FIGURE 2 | RDP1 screen design. Above, plan of the three treatments (randomly allocated) of each replicate run showing internal and external edges. Below, expanded plan of single central treatment showing individual plants (marked with X) highlighting plants on external and internal edges, plus a graphic explanation of the meaning of neighbors in relation to a focal plant.

modeled using a Markov random field (MRF) smoother within each block. For both analyses, fixed effects were selected or discarded based on the significance of the p-values of their coefficients (Null hypothesis: coefficient value is zero).

#### A Pot Experiment on 16 Cultivars

Using 16 cultivars selected to have a range of colonization in the screen of the RDP1, colonization rate was assessed in plants grown in 500 ml pots as described in the optimization steps above using a complete randomized block design with four replicate blocks. They were inoculated with 2% AM fungus inoculum, sown on the August 6th, 2015 and grown in the greenhouse as described earlier. After 32 days they were harvested and colonization assessed as described above.

#### Genome Wide Association Mapping

Genome wide association mapping was performed with the 5.2 m SNP database (Wang et al., 2018) using a mixed model approach on all the cultivars implemented using EMMA (Efficient Mixed Model Analysis) (Yu et al., 2006), as described in Norton et al. (2014) (a SNP was selected as worth reporting if the P-value was <0.0001 and if the minor allele frequency (MAF) was >5%). The false discovery rate (FDR) of detected associations was estimated using the R-language Bioconductor "multtest" library to calculate Benjamini–Hochberg adjusted probabilities (Benjamini and Hochberg, 1995). A significance threshold of 10% FDR is normally used to identify putative SNP associations (McCouch et al., 2016) but it is not uncommon for no SNPs to reach this conservative threshold.

Significant SNPs within 200 kb of each other (based on consideration of linkage disequilibrium decay reported in this population as suggested by Zhao et al., 2011) were considered to represent SNP clusters of the same QTL and singleton SNPs were not considered QTLs (there must be at least 1 other SNP with P < 0.001 within 200 kb). Gene annotation was examined 200 kb either side of the most significant SNP of a QTL.

TABLE 1 | Predictors used in the spatial statistical models.


Predictors used in shoot dry weight model are denoted by <sup>∗</sup> and those in the hyphal colonization model are denoted with §.

#### RESULTS

#### Shoot Dry Weight

The mean values for SDW and AM colonization rate are provided in **Supplementary Table S1**. SDW was affected by treatment such that the RP + AM plants were smaller while the other two treatments were only marginally different. Replicate runs, neighbor's size, position within the box and cultivar all had an effect on SDW at least as large as treatment. With replicate run, there was a gradual increase in SDW with number such that the averages were 78, 87, 102, and 125 mg per plant for runs 1–4, respectively. The spatial statistical analysis showed that plants on an external edge were slightly bigger than those that were not (103 mg vs. 98). More strikingly, plants on an internal edge were bigger than those that were not (110 mg vs. 97). The focal plant was larger when its neighbors were larger, and this positive effect of the neighbors' average size was greater in the AM + RP than in the RP treatments, and least in the control treatment (**Figure 3**).

The most parsimonious description of the spatial trends in plant SDW involved a significant spatial trend, common across all replicates (P = 1.28 × 10−12, Bayesian Information Criterion difference with models involving different spatial trends per box or block was greater than 15), which suggested an effect of the general orientation of the experimental setup in the greenhouse (see spatial trend in **Supplementary Figure S1**), independent of the position of the treatment in the box.

Once variation due to position and neighbors was removed from the data by subtracting their estimated effect from the observations, a two way analysis of variance (factors; treatment and genotype) on the SDW of the controls and rock phosphate treatment demonstrated a weakly significant effect of the rock phosphate treatment (P = 0.045) explaining less than 1% of the variation with rock phosphate stimulating growth (84.6 vs. 83.6 mg) and no cultivar by treatment interaction, while cultivar explained 68% of the non-spatial variation. This contrasts to ANOVA performed before the spatial analysis was conducted in which treatment and cultivar together only explained 30% of the total variation. When comparing rock phosphate to rock phosphate and AM fungus, the proportion of variation explained by treatment, cultivar and their interaction was 67% where it was only 49% before spatial correction. In this comparison, cultivar explained 50.1% of the variation, treatment 15.8% and the interaction 3.9%, the latter being small but highly significant (P < 0.001). Thus there is no evidence that cultivar relative performance differed between the control and rock phosphate treatment but it did when AM fungus was introduced. Ranking cultivars by their responsiveness to AM treatment (SDW ratio RP + AM:RP treatment) did not reveal any discernible pattern (between subpopulations or country/region of origin for example).

Scatter plots of SDW for control vs. rock phosphate and for rock phosphate vs. rock phosphate plus AM fungus are shown in **Figures 4A,B**, respectively. These show the regression line and the 1:1 line (dashed). It clearly demonstrates the lack of effect of rock phosphate (**Figure 4A**), and the negative effect of AM fungus on SDW (**Figure 4B**). The correlation is better for the upper graph indicating that cultivar performance is more divergent in the comparison between rock phosphate and rock phosphate plus AM fungus.

There was a difference in the SDW between subpopulations of the Rice Diversity Panel 1 which was consistent between treatments (P < 0.001, R <sup>2</sup> = 14.7–18.2%), where aus (n = 54) and indica (n = 63) were high, tropical japonica (n = 83) were lower and temperate japonica (n = 82) were lowest (**Figure 5**).

#### Hyphal Colonization

Hyphal colonization was roughly normally distributed in each run (**Supplementary Figure S1**) while both proportion of roots with arbuscles and vesicles were skewed toward low numbers and many zeros (**Supplementary Figure S1**). One way ANOVA on the raw data indicated heritability for proportion of roots with hyphae, arbuscles, and vesicles were 39, 29, and 35%, respectively. Mean hyphae colonization was highly correlated with mean arbuscles (r = 0.554, **Supplementary Figure S1**), and mean vesicles (r = 0.700, **Supplementary Figure S2C**). Based on these observations, it was decided that further analysis would be conducted on the hyphal colonization data alone.

Spatial analysis indicated hyphal colonization of a particular plant increased with its SDW (P = 0.026) and the mean SDW of its nearest (8) neighbors (P = 0.0012). The interaction between the effects of focal plant and neighbors' SDWs was close to significant (P = 0.066) (**Figure 6B**) so the data did not provide clear support for one model over the other. In other words, the difference in the predictions of the simple additive model and of the interactive model was small (**Figures 6A,B**). Both models agreed that large SDW of either the focal plant or its neighbors increased hyphal colonization of focal plant in similar ways.

The model indicated significant and idiosyncratic spatial trends in each of the four replicates (Markov Random Field component: P < 2 × 10−16; see **Supplementary Figure S3**). This spatial analysis allowed for a correction of estimation of the colonization of the plants such that the proportion of variation in colonization explained by cultivar (from one way ANOVA) increased from 35 to 42% with heritability increased from 39 to 54%. The mean colonization rates for each cultivar are presented in **Supplementary Table S1** while **Figure 7** shows a histogram of the distribution. The average coefficient of variation for the trait was 0.25 (it was 0.30 before spatial analysis). The range in colonization was considerable, from the Bangladesh aus cultivar DZ 193 at 22.9 ± 8.3% (±standard deviation) to the Chinese indica cultivar Kun Min Tsieh Hunan at 89.8 ± 9.3%. Colonization rates differed between subgroup (P < 0.001) where indica, temperate japonica and tropical japonica were similar (58.0, 54.5, and 55.0%, respectively) while aus were lower (50.0%) (**Figure 5**). Within each subgroup there were large variations but it is considered noteworthy that while there were several aus cultivars in the top 30 cultivars, there were no indicas in the lowest 30 cultivars. Differences were also detected between countries or regions (country groups when individual countries had less than 10 cultivars) (P < 0.001, R <sup>2</sup> = 12.7%) where the Philippines and China were high and Bangladesh and United States were low (**Supplementary Figure S4**). These differences are probably partly driven by different subgroup compositions for countries (e.g., Bangladesh is dominated by aus cultivars) but in the tropical japonica subgroup there is still a detectable country/region difference (P = 0.006, R <sup>2</sup> = 26.3%) where the six Philippine cultivars had higher colonization (71.9%) than 13 South American (55.7%), 15 West African (55.0%), 10 SE Asian (55.0%), and 15 United States (49.5%) cultivars.

#### Pot Experiment on 16 Cultivars

The colonization of 16 cultivars grown in pots differed significantly (P = 0.002) and cultivar explained 50% of the variation. Conducting a two way ANOVA on colonization data from both the full RDP1 screen and the pot experiment with factors cultivar and method (pot or RDP screen) found the model explained 62% of the variation and indicated all factors were significant at P < 0.001, including the cultivar by method interaction (F = 27.8, 5.6, and 3.1 for cultivar, method and interaction, respectively). Colonization in pots is plotted against the results obtained in the RDP1 screen in **Figure 8**.

There is no correlation between the colonization rates obtained in the two experiments. However, it is remarkable how close most of the cultivars are to the 1 to 1 line indicating most performed very similarly in both experiments. Three cultivars (Ta Mao Tsao, Lusitano and Agostano) appear to have very substantially different colonization rates in the pot. Indeed, if these three are removed, the remaining 13 cultivars correlate with r = 0.806.

#### GWA Mapping

The results of GWA mapping for the hyphal colonization with all subgroups is presented in **Supplementary Figure S5** and summarized in **Table 2** while the results of analysis with the All analysis (all cultivars) is graphed for more clarity in **Figure 9**. No SNPs were considered significant using the FDR of 10%. In total 23 putative QTLs were revealed in the analysis of all cultivars using the criteria that for at least 1 SNP P < 0.0001 and MAF > 5% and there is at least one other SNP with P < 0.001 within 200 kb. A further two putative QTLs were identified either in the aus subpopulation or both the aus and tropical subpopulation. No putative QTLs were detected in the indica or temperate japonica subpopulations. Genes within 200 kb of the most significant SNPs were listed and their responsiveness to AM colonization as reported by Fiorilli et al. (2015) and Gutjahr et al. (2015) is given in **Supplementary Table S2**. The number of genes in the lists ranges from 50 to 99 and they had an average of 8.8% of genes being differentially expressed in either Fiorilli et al. (2015) or Gutjahr et al. (2015) or both.

FIGURE 6 | 3D plots of the impact of neighbor and first order neighbor SDW (SDW) on hyphal colonization. The additive model (A) and the interactive model (B) are equally well supported by the data (AIC difference < 2). The black surface is the fitted model while the green and red surfaces are the 95% confidence intervals. Units are g for SWD and % colonization for hyphae.

#### DISCUSSION

#### Cultivars Differ in Colonization by R. irregularis and the Impact of the AM Fungus on Plant Growth

The main aim of this study was to test the hypothesis that there is genetic variation within rice for colonization and the impact of the AM fungus on rice growth. The study clearly showed there is. Firstly, there was highly significant differences between cultivars in colonization with a very wide range from 23–90% (**Figure 7**). Secondly, it shows that impact of AM fungus on shoot growth, all be it almost entirely negative, is different between cultivars (**Figure 4B**). There are few studies showing cultivar differences in rice. A meta-analysis of the impact of AM fungi on annual crop plants has been conducted on 39 papers containing 320 species by Lehmann et al. (2012) which concluded that cultivars differed in colonization rates, and interestingly old and new cultivars had less colonization that ancestral ones, and that colonization rate was quite strongly correlated to responsiveness (of host growth) to AM fungi. However, the analysis included only one study specific to rice (Gao et al., 2007). Gao et al. (2007) pot experiments found one of six aerobic rice cultivars had lower root colonization. Differences between an upland and a lowland cultivar were found to depend on the species of AM fungus used and the amount of arsenic applied as a treatment (Li et al., 2011). Suzuki et al. (2015) assess 64 rice cultivars for growth response to a different AM fungus and measured colonization in 12 of them. Only two cultivars are in common between that study and the current one (Nipponbare and Kasalath) and while they appeared to differ in growth response in Suzuki et al. (2015), here they do not. Considering examples of studies within other grass species, differences in five Canadian durum wheats have been reported including an interaction between soil fertility and cultivar (Singh et al., 2012). Amongst six winter wheats and maize accessions, Zhu et al. (2001) found cultivar differences,

while Chu et al. (2013) found differences in four maize cultivars but not at the lowest P addition. Recently three studies have examined a wide range of either sorghum (Leiser et al., 2016), or wheat (Lehnert et al., 2017) or durum wheat (De Vita et al., 2018). Leiser et al. (2016) tested 187 sorghum cultivars grown in pots 38 days and found only limited variation in AM fungal colonization rates in terms of range (38.5–72.9%) or heritability (15%). Lehnert et al. (2017) tested 94 wheat cultivars grown in pots until mature. They showed very strong genotypic differences but a lower range of colonization (24–56%). De Vita et al. (2018) found large differences in colonization of durum wheat roots with both AM fungi R. irregularis and F. mosseae, and discovered seven QTLs apparently common between the species suggestive of common genetic regulation within the host.

The differential cultivar response of shoot growth to AM fungal plus rock phosphate treatment relative to the rock phosphate treatment alone detected in this experiment was highly significant but small and did not fall into any discernible pattern (e.g., between subgroups or country of origin). This implies the host genetics of rice-AMF interaction are influential in determining plant growth, but much more research is needed to determine its biological significance. Despite the high statistical difference and big range of variation in colonization detected here, there is no particularly striking pattern that provides conclusive evidence of the biological implications. While aus cultivars are generally lower than other subpopulations (**Figure 5**), and there are differences related to the country of origin of the cultivars, these do not explain a large proportions of the variation or provide clear reasons for such differences. Aus cultivars, which are revealed to have a low colonization on average, are notable as the donors of a number of important abiotic resistance traits [e.g., submergence tolerance from FR 13A (Xu et al., 2006); phosphorus starvation tolerance from Kasalath (Gamuyao et al., 2012); drought resistance from N22 and Dular (Gowda et al., 2011)] and might be expected to be adapted to poor soils and might therefore be expected to have strong associations with AM fungi. It is perhaps notable then that Kasalath does have a high colonization rate (ranked 19th) while the other three cultivars mentioned above are not remarkable. Kasalath was identified as having high phosphorus uptake efficiency of 30 rice genotypes in a study which also identified the Brazilian tropical japonica IAC 25 as high for P uptake efficiency and the United States tropical japonica Lemont as low (Wissuwa and Ae, 2001). IAC 25 is ranked 9th here so is high for AM colonization, but Lemont ranked 33rd for colonization which is definitely not low.

The fact that cultivars from the Philippines and China have high root colonization and those from Bangladesh are low (**Supplementary Figure S2**) is not currently explicable. One hypothesis that could be forwarded would be that upland cultivars adapted to aerobic soils would have a greater need of, and therefore stronger association with AM fungi than flooded cultivars. Unfortunately information on the hydrological environment of the cultivars in the RDP1 is lacking.

The repeated test of 16 cultivars in a pot experiment produced an interesting result (**Figure 9**), being a remarkably strong agreement in colonization rates between the large box screen and pots for 13 of the cultivars, but for three cultivars their colonization in pots was much lower than in the box screen. It can be hypothesized that this reflects cultivar differences in the impact of soil properties on colonization. The preliminary experiments reported here demonstrate that colonization of Bala was sensitive to rock phosphate while in Azucena it was not. Since Bala was included amongst the 16 tested in pots, and its colonization was very similar to the box screen colonization rate, it seems unlikely the same edaphic factor is acting in the pot experiment reported in **Figure 8** and the preliminary experiments. The cultivar difference in the impact of edaphic factors on root colonization by AM fungi merits further study.

#### Impact of R. irregularis on the Rice Plants

The optimization experiments do not provide convincing evidence that P uptake efficiency and root colonization



by AM fungi are linked, which may occur because P uptake is more likely to be related to proliferation of extra-radical mycelium (Sawers et al., 2017), or that under these conditions, R. irregularis is behaving toward the parasitic end of the plant-mycorrhiza continuum (Johnson, 2010). It will be important to test if different root colonization rates detected here reflect colonization obtained with other species of AM fungi and colonization in the field environment.

In the preliminary experiments and the main screen described in this report, the AM treatment invariably reduced plant growth, which is not a unique observation (Van Der Heijden and Horton, 2009). We would suggest that this is a feature of the soil medium used here since other experiments in Aberdeen with

the same rice genotypes and the same AM fungal inoculum, but with different growing medium, do not always have a negative effect. What is highly remarkable from study is the strong positive neighbor effects detected on both plant growth and root colonization in the presence of R. irregularis. The positive nature of these effects of neighbors suggest R. irregularis has formed a common mycorrhizal network (CMN) inter-connecting roots of individual plants (Selosse et al., 2006). Recent work has shown that CMNs can affect growth and nutrition of neighbors grown in pairs (Walder et al., 2012). This should be tested on rice to determine the repeatability, extent and mechanism of this apparent facilitation. If proven more than just specific to the conductions used in this experiment, the result has major implications for consideration of mycorrhizal crop plants and their productivity since it demonstrates that when connected by a CMN, one plant can have a positive impact on its neighbors. This phenomenon merits further study, especially at the field level.

#### Putative QTLs for Root Colonization by AM Fungi by GWA Mapping

In this study, 23 putative QTLs for root colonization by AMF were detected, mostly in the analysis of the whole population. Two important observations need to be made. First, no association was above the 10% FDR; this does not mean they are not real, but they must be treated with caution, hence they are termed putative QTLs. Second, the number is high. If these QTLs are to be believed, the implication is that there are many genes involved in determining the degree of colonization in rice–AM fungal symbiosis and the trait is truly quantitative. In the study of sorghum (Leiser et al., 2016) the GWA mapping revealed no SNP associated with colonization with a P-value below P = 0.0001 which they considered reflected the low heritability (15%). In contrast, the wheat study of Lehnert et al. (2017), which had a heritability of 54% for root colonization (the same as reported here for rice), found six QTLs. More studies are needed before we can determine if root colonization in grasses is generally regulated by multiple, small effect genes as implied in the studies to date.

It is worth noting that none of the QTLs reported here match the 97 and 16 loci found associated with either controlled environment or field resistance to the fungal pathogen blast in this population [as reported by Kang et al. (2016) and Zhu et al. (2016), respectively]. This implies that the loci reported for AM colonization are not associated with genes affecting the basic plant/fungus interaction process.

#### Candidate Genes for Putative QTLs

Identifying the most promising functional candidate genes from the gene lists for the 23 putative QTLs reported here (**Table 2** and **Supplementary Table S2**) will depend on individual and community understanding of the molecular mechanisms involved in the AM fungus-rice interaction which will improve with time. However, some notable observations on examination of these lists are given below.

In QTL 1.1 is LOC\_Os01g09580, a calcium/calmodulin serine/threonine protein kinase which is massively downregulated in AMF in both the Fiorilli et al. (2015) and Gutjahr et al. (2015) study. Over expression in rice of Medicago DMI3 (a Ca+/calmodulin-dependent serine/threonine protein kinase that is known to be a component of the common symbiotic pathway between AM fungi and rhizobium) supported elevated AM fungal colonization (Ortiz-Berrocal et al., 2017).

In QTL 2.2 is a pair of excellent candidate genes. The genes LOC\_Os02g40710 and LOC\_Os02g40730 are annotated as ammonium transporters. These genes are OsAMT1.2 and OsAMT1.3, respectively. The transport of ammonia from AM fungi to plant host within the arbuscule is considered a central process in AM symbiosis (Guether et al., 2009). Transcriptomics studies on rice exposed to R. irregularis identified ammonium transporters (although not these two) as responsive (Güimil et al., 2005; Gutjahr et al., 2015), while the expression of ammonium transporter OsAMT3.1 has been used as a marker for rice-AM interaction (Vallino et al., 2014). Furthermore, AMT2:3 appears to be important in influencing the life span of Medicagoassociated AM fungi (Breuillin-Sessoms et al., 2015). Fiorilli et al. (2015) revealed that LOC\_Os02g40730 is strongly upregulated by R. irregularis inoculation (2.2-fold in large laterals, 9-fold in fine laterals). Both OsAMT1.2 and OsAMT1.3 have a highly root-specific expression pattern in rice (Sato et al., 2013).

In QTL 4.1 is a cluster of seven subtilisin or subtilisin-like proteases including LOC\_Os04g02980 annotated as OsSub33 which is upregulated in both root types in the Fiorilli et al. (2015) study. Subtilisins are a class of serine proteases also known as subtilases. A review on subtilases (Schaller et al., 2017) highlights studies that suggests some subtilases are linked to AM colonization, having a role in arbuscule development. Further, a subtilase was identified by Kistner et al. (2005) as one of seven host genes required for mycorrhiza and symbiotic bacteria to enter root epidermal or cortical cells based on the study of a mutant of Lotus japonicus that are impaired for nodulation. While this makes these genes good candidates for rice-AM fungal interaction, it must be noted that there are 71 genes in rice annotated as subtilisin-like or Subtilisin homologs.

Recently a 53 kbp deletion at 29 Mbp on chromosome 1 regulating strigolactone production and strongly associated with resistance to the parasitic plant Striga was reported (Cardoso et al., 2014). This deletion was tested by PCR on the RDP1 and used as a marker in the GWA mapping conducted here. It is surprising perhaps that there is no association detected at this locus since the production of strigolactone is considered to be an evolutionary adaptation critical to the signaling by plants to mycorrhiza (Akiyama et al., 2005).

#### CONCLUSION

These experiments clearly demonstrate that rice cultivars differ very widely for colonization by R. irregularis and reports putative QTLs which appear to contribute to that variation. While the pattern of variation in colonization in rice offers little insight into the evolution or biological significance, it lays a foundation for further research. This may be determining if variation detected here matches variation in the field and if it impacts nutrient uptake in low nutrient, aerobic environments.

Some of the genes under the QTLs shown here should be further investigated. A supplementary finding is strong evidence of positive interactions between neighboring rice plants facilitated by a common mycorrhizal network.

#### AUTHOR CONTRIBUTIONS

AP conceived and designed the experiments and largely wrote the manuscript. HD conducted preliminary experiments and the screen of RDP1. RS conducted the study of 16 cultivars in pots. TC did the spatial statistical analysis. AD and TT conducted the GWA mapping. DJ advised on the role of mycorrhiza in common mycorrhizal networks. All authors contributed to editing the manuscript.

#### FUNDING

The majority of the research reported here is an output of EU project "EURoot" (FP7-KBBE-2011-5 Grant Agreement No. 289300) project. RS's contribution was funded by FACCE-JPI NET project "GreenRice" (Sustainable and environmental friendly rice cultivation systems in Europe) and was funded by the BBSRC award BB/M018415/1.

#### REFERENCES


#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.00633/ full#supplementary-material

FIGURE S1 | Spatial trend in shoot dry weight (SDW), shared between all four replicate runs (showing deviations from predicted value, on the log-scale in grams). Values increase from red to white.

FIGURE S2 | Raw AM colonization data showing; (A) frequency distribution of hyphal colonization in each run; (B) scatter plot of mean hyphal colonization vs. mean arbuscle colonization and; (C) hyphal colonization vs. mean vesicle colonization.

FIGURE S3 | Spatial trends in hyphal colonization in each of the four replicate runs (deviation from predicted value). Units are % colonization.

FIGURE S4 | Boxplots of hyphal colonization in Rice Diversity Panel 1 according to country of origin (with n in brackets). Letters above bars represent results of Tukey's test of difference between groups.

FIGURE S5 | Manhattan and QQ plots for hyphal colonization for all cultivars, aus (AUS), indica (IND), tropical japonica (TRJ), and temperate japonica (TEJ) cultivars. Red dotted lines indicate putative QTLs from the all analysis, green in subgroups.

TABLE S1 | List of rice accession and phenotype data.

TABLE S2 | List of genes observed in each of the putative QTLs with information on available transcriptomics data for response to AM treatment [using data from Fiorilli et al. (2015) or Gutjahr et al. (2015)].



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

Copyright © 2019 Davidson, Shrestha, Cornulier, Douglas, Travis, Johnson and Price. 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.

# Plant Aquaporins in Infection by and Immunity Against Pathogens – A Critical Review

Liyuan Zhang<sup>1</sup> , Lei Chen<sup>1</sup> and Hansong Dong1,2 \*

<sup>1</sup> Plant Immunity Research Group, National Key Laboratory of Crop Science, Department of Plant Pathology, Shandong Agricultural University, Tai'an, China, <sup>2</sup> Plant Immunity Laboratory, Department of Plant Pathology, Nanjing Agricultural University, Nanjing, China

Plant aquaporins (AQPs) of the plasma membrane intrinsic protein (PIP) family face constant risk of hijack by pathogens aiming to infect plants. PIPs can also be involved in plant immunity against infection. This review will utilize two case studies to discuss biochemical and structural mechanisms that govern the functions of PIPs in the regulation of plant infection and immunity. The first example concerns the interaction between rice Oryza sativa and the bacterial blight pathogen Xanthomonas oryzae pv. oryzae (Xoo). To infect rice, Xoo uses the type III (T3) secretion system to secrete the proteic translocator Hpa1, and Hpa1 subsequently mediates the translocation of T3 effectors secreted by this system. Once shifted from bacteria into rice cells, effectors exert virulent or avirulent effects depending on the susceptibility of the rice varieties. The translocator function of Hpa1 requires cooperation with OsPIP1;3, the rice interactor of Hpa1. This role of OsPIP1;3 is related to regulatory models of effector translocation. The regulatory models have been proposed as, translocondependent delivery, translocon-independent pore formation, and effector endocytosis with membrane protein/lipid trafficking. The second case study includes the interaction of Hpa1 with the H2O<sup>2</sup> transport channel AtPIP1;4, and the associated consequence for H2O<sup>2</sup> signal transduction of immunity pathways in Arabidopsis thaliana, a non-host of Xoo. H2O<sup>2</sup> is generated in the apoplast upon induction by a pathogen or microbial pattern. H2O<sup>2</sup> from this source translocates quickly into Arabidopsis cells, where it interacts with pathways of intracellular immunity to confer plant resistance against diseases. To expedite H2O<sup>2</sup> transport, AtPIP1;4 must adopt a specific conformation in a number of ways, including channel width extension through amino acid interactions and selectivity for H2O<sup>2</sup> through amino acid protonation and tautomeric reactions. Both topics will reference relevant studies, conducted on other organisms and AQPs, to highlight possible mechanisms of T3 effector translocation currently under debate, and highlight the structural basis of AtPIP1;4 in H2O<sup>2</sup> transport facilitated by gating and trafficking regulation.

Keywords: aquaporin, plasma membrane intrinsic protein, H2O<sup>2</sup> transport, immunity signaling, translocon, type III effectors

Edited by:

Valentina Fiorilli, University of Turin, Italy

#### Reviewed by:

Raffaella Balestrini, Italian National Research Council (IPSP-CNR), Italy Kalyan K. Mondal, Indian Agricultural Research Institute (ICAR), India

> \*Correspondence: Hansong Dong hsdong@njau.edu.cn

#### Specialty section:

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

Received: 08 January 2019 Accepted: 26 April 2019 Published: 28 May 2019

#### Citation:

Zhang L, Chen L and Dong H (2019) Plant Aquaporins in Infection by and Immunity Against Pathogens – A Critical Review. Front. Plant Sci. 10:632. doi: 10.3389/fpls.2019.00632

## INTRODUCTION

fpls-10-00632 May 25, 2019 Time: 17:6 # 2

Aquaporins (AQPs) are membrane-intrinsic proteins initially defined as water (H2O) transporting channels in all organisms and subsequently found to have many other substrate specificities (de Groot and Grubmuller, 2001; Maurel et al., 2008, 2015; Sutka et al., 2017), such as hydrogen peroxide (H2O2; Tian et al., 2016). In plants, AQPs are classified into five major families (Chaumont et al., 2001; Maurel, 2007), including the plasma membrane intrinsic proteins (PIPs), tonoplast intrinsic proteins (TIPs), nodulin 26 like intrinsic proteins (NIPs), small basic intrinsic proteins (SIPs), and X intrinsic proteins (XIPs). The PIP family is further divided into the PIP1 subfamily made of PIP1;1 to PIP1;5 and the PIP2 subfamily consisting of PIP2;1 to PIP2;8 in most plant species (Maurel, 2007; Gomes et al., 2009; Laloux et al., 2018). While AQPs of the other four families function in substrate trafficking between organelles, PIPs are responsible for substrate transportation between the exterior and interior of cells (Maurel, 2007; Gomes et al., 2009; Kaldenhoff et al., 2014; Li et al., 2015; Bao, 2017).

Recently discovered functions of AQPs surpass the original "water channel" concept (Preston et al., 1992; Wudick et al., 2009; Heckwolf et al., 2011; Brown, 2017), and suggest implications in infection and immunity in both animals (Hara-Chikuma et al., 2015; Yang, 2017) and plants (Maurel et al., 2015; Wang F. et al., 2014; Zhang et al., 2018; Li et al., 2019). The functions of animal AQPs are no longer confined to substrate-transportbased processes such as urinary concentration and body fluid homeostasis (Brown, 2017), and are now known to include roles in various disease conditions and pathological states (Yang, 2017). Similarly, functional diversity – redundancy, overlapping, and extension beyond substrate transport – is a property of plant AQPs, especially PIPs (Ji and Dong, 2015b; Li et al., 2015; Zhang et al., 2018, 2019; Li et al., 2019). The functional scope of PIPs goes far beyond water relations or drought tolerance, extending to the subcellular transport of reactive oxygen species (ROS), including H2O<sup>2</sup> (Tian et al., 2016; Smirnoff and Arnaud, 2019). H2O<sup>2</sup> transport connects with signaling between the cell exterior and interior and between organelles, resulting in plant resistance to pathogen infection (Tian et al., 2016).

PIPs possess extracellular regions exposed to the outside environment (Maurel et al., 2015), and have potential to partake in plant responses to biotic and abiotic stresses. Here are several examples. Previous uses of induced resistance in crop protection (for example: Chen et al., 2008a,b; Fu et al., 2014; Wang F. et al., 2014) confirm the practical value of PIP-mediated immunity signal transduction (Tian et al., 2016). The correlation of PIP function in water transport with stress response results in promising strategies for improvement of plant tolerance to abiotic stresses, including drought (Balestrini et al., 2018). Drought tolerance in a variety of plant species is related to arbuscular mycorrhizal (AM) symbiosis, in which AM fungi (Rhizophagus spp.) show enhanced expression of AQP-encoding genes (Bárzana et al., 2014, 2015; Calvo-Polanco et al., 2016; Ruiz-Lozano et al., 2016; Sánchez-Romera et al., 2016; Ruiz-Lozano and Aroca, 2017). Surprisingly, the AM fungus R. clarus contributes its aquaglyceroporin (glycerol/water-transporting AQP) RcAQP3 to the mediation of long-distant polyphosphate translocation from the fungal vacuoles into cells of plant roots and leaves (Kikuchi et al., 2016). Genetic resources of plants, including the AQP transcriptome, can be used in responses to environmental cues, symbiotic microbes (AM fungi and rhizobia), and microbial pathogens (Desaki et al., 2018; Rey and Jacquet, 2018; Wang R. et al., 2018).

Due to their direct contact with the extracellular environment, PIPs risk being appropriated by plant pathogens to expedite infection (Zhang et al., 2018; Li et al., 2019). When infection is imminent, the real-time function of PIPs may switch from substrate transport to the regulation of plant responses to pathogens (Zhang et al., 2018; Li et al., 2019). This is either favorable or unfavorable to plant growth and development, depending on plant responses to pathogenicity determinants, called effectors, whose functions are subject to regulation of PIPs (Tian et al., 2016; Wang X. et al., 2018; Li et al., 2019; Zhang et al., 2019).

This review will summarize recent studies on the roles of PIPs in plant infection and immunity, and discuss the molecular, biochemical, and structural mechanisms involved. Discussion of infection will focus on type III (T3) effector translocation (T3ET) from Xanthomonas oryzae pv. oryzae (Xoo) into rice cells. Discussion of immunity will focus on the response of Arabidopsis to pathogens or pathogen-associated molecular patterns (PAMPs), also termed microbial patterns. This review will reference studies investigating AQPs in animals, microbes, and other plants to highlight the broad importance of PIP function, from substrate transport to infection and immunity in plants.

#### THE CIRCUMSTANTIAL FUNCTION OF A PIP IN T3ET

PIPs possess three extracellular regions that are exposed to the outside environment (Maurel et al., 2015). As a result, they are at a constant risk of being hijacked by pathogens attempting to infect plants, and inevitably partake in immunity against infection. Therefore, PIPs are required to extend their function from substrate transport to plant infection and immunity when the circumstances demand it. Emerging evidence suggests the implication of OsPIP1;3 in rice infection by Xoo (Zhang et al., 2018; Bian et al., 2019; Li et al., 2019). In this case, OsPIP1;3 functions with the bacterial hydrophilic protein Hpa1, which belongs to the harpin-group proteins secreted by the T3 secretion pathway of Gram-negative plant-pathogenic bacteria (Wei et al., 1992; Tejeda-Dominguez et al., 2017; Zhang et al., 2018; Li et al., 2019). Hpa1 produced by X. oryzae (Zhu et al., 2000; Chen et al., 2008a) is involved in the virulence of bacterial pathogens (Wang X. et al., 2018). Hpa1 modulates physiological and pathological processes in plants in association with PIPs (Sang et al., 2012; Li et al., 2013, 2014, 2015, 2019; Ji and Dong, 2015b; Zhang et al., 2018). The virulence role of Hpa1 is determined by its biochemical properties. Hpa1 is a one-domain harpin, which share a unitary hydrophilic "harpin" domain distinct from the enzymatic domain

present in two-domain harpins (Kvitko et al., 2007; Choi et al., 2013; Ji and Dong, 2015b). Two-domain harpins have potential to associate with the bacterial periplasm or plant cell walls to facilitate assembly of the T3 secretion machinery (Mushegian et al., 1996; Koraimann, 2003; Zhang et al., 2008; Dik et al., 2017; Hausner et al., 2017). One-domain harpins, including Hpa1, target plasma membranes (PMs), where they serve as T3 translocators to mediate T3ET (Kvitko et al., 2007; Bocsanczy et al., 2008; Wang X. et al., 2018; Bian et al., 2019; Li et al., 2019).

In Xoo-infected rice plants, secreted Hpa1 translocates at least two transcription activator-like (TAL) effectors – AvrXa10 and PthXo1, which are also produced via the pathway (Wang X. et al., 2018). Translocated effectors exert virulent or avirulent effects depending on the susceptibility of the plant variety (Yang et al., 2006; Büttner, 2016; Schreiber et al., 2016; Schwartz et al., 2017; Zhang et al., 2018). The rice variety Nipponbare is susceptible to the TAL effector PthXo1 secreted by PXO99A, a well-studied Xoo strain (Yang et al., 2006; Wang X. et al., 2018; Zhang et al., 2018). To infect Nipponbare plants, PXO99<sup>A</sup> secretes Hpa1 and delivers it to the cell surface, where Hpa1 interacts with OsPIP1;3 to facilitate the translocation of subsequently secreted PthXo1 into Nipponbare cells (Wang X. et al., 2018; Li et al., 2019). PthXo1 then induces virulence by activating its regulatory target – the host susceptibility gene OsSWEET11 (Yang et al., 2006) in an OsPIP1;3-dependent manner (Zhang et al., 2018; Li et al., 2019). If the OsPIP1;3 gene is silenced by hairpin or knocked out by TALEN14, both PthXo1 translocation and OsSWEET11 expression incur concomitant impairments up to 70%, highly alleviating virulence as a consequence (Zhang et al., 2018; Li et al., 2019). In contrast, both events acquire >2-fold enhancements if OsPIP1;3 is overexpressed, causing marked aggravations in virulence (Li et al., 2019).

AvrXa10 is an avirulent effector secreted by the Xoo strain PXO86, and induces immune responses in the resistant rice variety IRBB10 (Tian et al., 2014). The plant immunity is determined by the disease-resistant gene Xa10, which is the target of AvrXa10 (Tian et al., 2014). Xa10 has two homologs in the Nipponbare genome – Xa10-Ni and Xa23-Ni, both of which function similarly to confer immune responses in Nipponbare plants inoculated with recombinant PXO99<sup>A</sup> strains that deliver the matching artificially designed TAL effectors (Wang et al., 2017). When avrXa10 is transferred from PXO86 into the PXO99<sup>A</sup> genome, the resulting PXO99A/avrXa10 recombinant delivers AvrXa10 in IRBB10 cells (Wang X. et al., 2018). Thereafter, AvrXa10 activates the disease resistant gene Xa10- Ni to confer the plant resistance against the blight disease (Wang X. et al., 2018). The AvrXa10 translocation and Xa10-Ni activation incur concomitant impairments in plants inoculated with the hpa1-deleted mutant; the absence of hpa1 markedly reduces the quantity of AvrXa10 translocation, decreasing the expression level of Xa10-Ni (Wang X. et al., 2018). The AvrXa10 translocation and Xa10-Ni activation requires OsPIP1;3, and both events are enhanced by OsPIP1;3 overexpression but inhibited by OsPIP1;3 silencing (Bian et al., 2019).

These findings demonstrate the important role of OsPIP1;3 in the translocation of T3 effectors, at least the TAL effectors PthXo1 and AvrXa10, from bacterial cells into the cytosol of rice cells. OsPIP1;3 functions either as a disease-susceptibility or resistance factor, depending on a virulent or avirulent function of the translocated effector.

#### POTENTIAL MECHANISMS OF T3ET REGULATION

Passages of proteic T3 effectors are 1.2–5.0 nanometers in width (Ji and Dong, 2015b; Guignot and Tran Van Nhieu, 2016), in contrast to PIP/AQP channels with an aperture around 3 Å, which is permeable to small substrates (Heckwolf et al., 2011; Li et al., 2015; Tian et al., 2016) but impossible for proteins to pass (Li et al., 2019). Presumably, the role of OsPIP1;3 in T3ET complies with one of regulation models currently in debate (Domingues et al., 2016; Prasad et al., 2016; Santi-Rocca and Blanchard, 2017; Scheibner et al., 2017; Tejeda-Dominguez et al., 2017; Gaytán et al., 2018; Wagner et al., 2018; Shanmugam and Dalbey, 2019). Three models have been proposed as the canonical translocon-dependent delivery (Büttner, 2012; **Figure 1** middle purple route) and the translocon-independent pore formation (Rüter et al., 2010; **Figure 1** left black route) and endocytosis (Santi-Rocca and Blanchard, 2017; **Figure 1** right red route). To date, studies on the three models have obtained empirical genetic evidence (Finsel and Hilbi, 2015; Domingues et al., 2016; Dong et al., 2016; Chakravarthy et al., 2017; Scheibner et al., 2017), but the structural basis of each model remains to be analyzed.

The first model of T3ET (**Figure 1** middle purple route) was proposed to emphasize molecular interactions between T3 translocators and molecular interactions of T3 translocators with PM receptors (Büttner and Bonas, 2002; Büttner et al., 2008; Büttner, 2012; Ji and Dong, 2015b), either lipids (Haapalainen et al., 2011; Li et al., 2011), or proteins (Oh and Beer, 2007; Li et al., 2015; Adam et al., 2017). T3 translocators include one hydrophilic protein, such as Hpa1 from xanthomonads – bacteria in the Xanthomonas genus (Zhu et al., 2000; Chen et al., 2008a; Wang X. et al., 2018), and two hydrophobic proteins (Büttner et al., 2008; Ji and Dong, 2015b), such as HrpF from the same bacteria (Büttner et al., 2002; Li et al., 2011; Hausner et al., 2017). Recognition of the hydrophilic translocator by a component of the PM composition is the first step towards translocon assembly (Goure et al., 2004; Mueller et al., 2008; Sawa et al., 2014). Then, the translocon is finalized by the binding of lipids to hydrophobic translocators (Büttner et al., 2008; Büttner, 2012; Ji and Dong, 2015b). A completed translocon possesses an inner conduit that opens into a target cell and accommodates bacterial effector translocation (Büttner et al., 2008; Chatterjee et al., 2013; Ji and Dong, 2015b; Büttner, 2016).

Although there is no evidence so far to verify the T3 translocon assembly, many studies suggest the involvement of T3 translocators in effector translocation from animal- and plantpathogenic bacteria into cells of their corresponding eukaryotic hosts (summarized in Scheibner et al., 2017). Mounting evidence indicates the engagement of PM phospholipids in T3ET, especially phosphatidylinositol phosphates PI(n)Pn (Lee et al., 2001a,b; Büttner et al., 2002; Weber et al., 2006; Hubber and Roy, 2010; Li et al., 2011; Finsel and Hilbi, 2015; Dong et al., 2016). For T3ET

FIGURE 1 | Hypothetic routes of T3ET using Xoo as an example. Effector translocation may use the left black route (Rüter et al., 2010; Scheibner et al., 2017) or the right red pathway (Lu et al., 2007; Santi-Rocca and Blanchard, 2017) according to recently proposed models. In a previously proposed model, T3ET occurs via the translocon (the middle purple route) hypothetically assembled by interactions between translocators, and their receptors in eukaryotic PMs (Büttner, 2012; Ji and Dong, 2015b). Three translocators have been identified in animal-pathogenic bacteria, but the number of T3 translocator remains unknown in plant-pathogenic bacteria including Xoo. In Xoo, the hydrophilic protein Hpa1 (Wang X. et al., 2018) and the hydrophobic protein HrpF (Büttner et al., 2002; Li et al., 2011) were determined to function as T3 translocators, but whether the third translocator exists is unclear (question marks). Regarding molecular interactions during the translocon assembly, OsPIP1;3 has been verified to interact with Hpa1 at rice PMs to expedite the translocation of TAL effectors AvrXa10 and PthXo1 (Zhang et al., 2018; Li et al., 2019). In the cartoon, numbers 1 through 5 refer to the order of the translocator in self oligomerization to form the homogenous complex, which is assumed to be consisting of 5 or 8 monomers (Mueller et al., 2008). In vitro assays indicated HrpF binding to lipids (Büttner et al., 2002; Li et al., 2011), such as PI4P, but no evidence was available to demonstrate the lipid binding at plant PMs and the subsequent effect on T3 effector translocation.

from xanthomonads, lipids in the plant PM associates with the bacterial hydrophobic T3 translocator HrpF (Büttner et al., 2002; Li et al., 2011). HrpF was the first reported T3 translocator and is regarded as a marker of T3 translocon in xanthomonads (Büttner et al., 2002; Scheibner et al., 2017). HrpF is highly conserved in xanthomonads (Sugio et al., 2005) and has been shown to mediate the translocation of AvrBs3 from X. campestris pv. vesicatoria (Xcv) – the bacterial spot pathogen of pepper (Büttner et al., 2002; Noël et al., 2002), and from X. oryzae pv. oryzicola — the pathogen that causes bacterial leaf streak in rice (Li et al., 2011). Evidence is further provided by our demonstrations that the hydrophilic T3 translocator Hpa1 of Xoo interacts with OsPIP1;3 at rice PMs to expedite translocation of TAL effectors PthXo1 and AvrXa10 from Xoo cells into the cytosol of rice cells (Zhang et al., 2018; Bian et al., 2019; Li et al., 2019).

The second model of T3ET (**Figure 1** left black pathway) is the translocon-independent pore formation by bacterial effectors characteristic of cell-penetrating peptide (RPP; Scharnert et al., 2013; Rüter and Schmidt, 2017). Pore forming in eukaryotic PMs is momentary, occurs quickly upon recognition of bacterial effectors, and is regulated by membrane repair mechanisms (Scharnert et al., 2013). RPPs are either autonomously transported across the membrane or delivered by endocytosis (Wang F. et al., 2014). Autonomous translocation was found with the T3 effector YopM from Yersinia enterocolitica (Rüter et al., 2010). The YopM sequence contains two N-terminal α-helices, which determines the interaction with eukaryotic PMs (Li et al., 2019), and two putative nuclear localization signals at the C-terminus (Benabdillah et al., 2004). Therefore, YopM can be translocated directly into the cytosol of target cells and further transported into the nucleus via vesicle trafficking (Skrzypek et al., 1998).

Little is known about the translocon-independent translocation of T3 effectors from plant-pathogenic bacteria except for the TAL effector AvrBs3 from Xcv. Preliminary infection experiments with Xcv translocon mutants and endocytosis inhibitors deny a contribution of endocytosis to the delivery of AvrBs3 (Scheibner et al., 2017). A possible route for AvrBs3 translocation from the translocon mutants is a direct transportation through pore formation. The pore could be proteolipidic (Gilbert et al., 2014) and could be generated by means of proteic and lipidic constituents, which are required for the translocation of T3 effectors from xanthomonads (Büttner et al., 2002; Li et al., 2011, 2019). However, the efficiency of AvrBs3 translocation from the translocon mutants is much lower than that from the WT strain, indicating that the transloconindependent route is used in the absence of alternative.

The third model of T3ET (**Figure 1** left black pathway) was recently proposed to emphasize the effector endocytosis through direct interaction with receptors situated in eukaryotic PMs (Domingues et al., 2016). The molecular interaction may trigger the membrane trafficking mechanism (Allgood and Neunuebel, 2018) either by endoplasmic reticulum (ER) or vesicles (Wudick et al., 2015), providing a potential scheme for bacterial effector endocytosis (**Figure 2**). Protein and lipid trafficking via ER is universal (Cybulsky, 2017; Obacz et al., 2017), and vesicle-mediated PIP trafficking has been elucidated in roots of Arabidopsis following treatment with H2O<sup>2</sup> (Wudick et al., 2015). The treatment induces AtPIP2;1 accumulation in the late endosomal compartments, and increases stability of the PIP and its homologs in the cytoplasm (Wudick et al., 2015). Like AtPIP1;4 (Li et al., 2015, 2019), AtPIP2;1 also is an H2O/H2O2/CO<sup>2</sup> triple channel (Heckwolf et al., 2011; Rodrigues et al., 2017), but no study shows whether or not AtPIP2;1 resembles AtPIP1;4 to regulate bacterial effector translocation.

It is deserved of studying whether multiple substrate specificities of a PIP enable it to accommodate bacterial effectors.

There are two examples indicating the possibility that T3 effectors of plant-pathogenic bacteria are translocated along with membrane trafficking. One is the T3 effector HopZ1a of Pseudomonas syringae pv. syringae, bacterial pathogen of many plants. HopZ1a, HopZ1b, and HopZ1c are allelic forms, constitute the HopZ1 family of P. syringae T3 secretion system, and share a consensus myristoylation site required for membrane localization (Zhou et al., 2009). HopZ1a is an acetyltransferase, is activated by the eukaryotic co-factor phytic acid, and turns to acetylate itself and tubulin. Tubulin acetylation causes a decrease in microtubule networks, disrupts the secretory pathway, and suppresses cell wall-associated defense in plants (Lee et al., 2012). The defense is subject to complex regulatory networks, which involve vesicle trafficking linked to microtubules (Lehman et al., 2017). The other example is the T3 effector HopM1 of P. syringae pv. tomato. To infect tomato plants, the bacteria secrets HopM1, and delivers it into the plant PM-derived trans-Golgi network/early endosome (Nomura et al., 2011), suggesting a role of vesicle trafficking in HopM1 translocation.

The involvement of AQPs in T3 effector endocytosis can be speculated from independent studies summarized below. The trafficking of animal AQPs towards the cell interior is triggered by the AQP binding to a different protein (Zelazny et al., 2009; Ji and Dong, 2015a), such as vasopressin (Kamsteeg et al., 2006), or heat shock protein HSP70 (Lu et al., 2007). Nevertheless, molecular interactions at the PM transiently affect PM integrity (Laliberté and Sanfaçon, 2010; Guignot and Tran Van Nhieu, 2016; Santi-Rocca and Blanchard, 2017), which may extricate and internalize PM-associated proteins to accommodate foreign molecules like T3 effectors. It is possible that AvrBs3 and PthXo1 use this mechanism to enter rice cells together with OsPIP1;3 trafficking (**Figure 2**). Both effectors may be internalized through trafficking of OsPIP1;3 en route to degradation by the proteasome (Hirano et al., 2003; Centrone et al., 2017) or the autophagosome (Khositseth et al., 2017). This mode of trafficking and degradation has been shown to regulate animal AQP turnover (Khositseth et al., 2017; Shen et al., 2019) and may also apply to plant AQPs. It is necessary to verify whether OsPIP1;3, or any other PIPs, can interact with any of the bacterial effectors, in the absence of Hpa1, to cause the PIP and effector internalization.

How could PM binding lead to endocytosis of bacterial effectors? The binding of effectors or translocators to the PM induces transient damage to the integrity and function of PM compositions, providing an abnormal pathway for bacterial effector translocation (Guignot and Tran Van Nhieu, 2016). In addition to T3, other secretion systems, such as T4, may be involved also (Domingues et al., 2016). Salmonella enterica serovars are intracellular facultative pathogens with a wide host range, and cause serious diseases including typhoid fever and cholera in humans (Domingues et al., 2016; Piscatelli et al., 2016). About 40 different T3 effectors confer differential virulence to different serovars. For infection, Salmonella bacteria establish a bacteria-containing vacuole (BVC), induce tubules, and then deliver the T3 effector SteA onto the BCV and tubules. In both structures, SteA specifically interacts with PI4P to move into host cells (Domingues et al., 2016). Legionella pneumophila, the pathogen responsible for Legionnaire' disease, creates BCV through effectors secreted by the Dot/Icm T4 system (Finsel and Hilbi, 2015). In BCV, the pathogen hijacks host PM trafficking to induce BCV maturation (Hubber and Roy, 2010). The BCV membrane mainly contains PI4P (Weber et al., 2006; Finsel and Hilbi, 2015), which is important for anchoring many Dot/Icm effectors onto BCV (Dong et al., 2016). The T4 effector LepBd of L. pneumophila is a phosphatase (PP), and specifically converts PI3P into PI(3,4)P2. PI(3,4)P<sup>2</sup> is efficiently hydrolyzed into PI4P (Dong et al., 2016), which may be used to replenish the PI4P stock of BCV. This mechanism is also employed by the T3 effector SopB of cholera pathogen S. enterica serovar Typhimurium. Like the T4 effector LepBd of L. pneumophila, the T3 effector SopB of Typhimurium is also a PP, but possesses both 4-PP and 5-PP activities. This dual enzymatic function is essential for the formation of BCV membrane ruffles and subsequent bacterial invasion. The 5-PP activity of SopB is assumed to generate PI(3,4)P2, which is then recruited by sorting nexin 9 (SNX9), an actin-modulating protein. The 4-PP activity converts PI(3,4)P<sup>2</sup> to PI3P. Alone, neither activity is sufficient for

membrane ruffling. Instead, combined 4-PP and 5-PP activities induce SNX9-mediated membrane ruffling and bacterial invasion (Dong et al., 2016).

The three models of T3ET may be chosen to use circumstantially by bacteria with genetic variations in the T3 repertoire. For example, the translocon-dependent mechanism guarantees efficient translocation of AvrBs3 from the wild-type Xcv strain (Büttner et al., 2002), in contrast to insufficient translocation from the bacterial translocon mutants in a translocon-independent manner (Scheibner et al., 2017). An early report stated that the carboxy (C)-terminal region of HrpF is essential for the entry of Xcv AvrBs3 into plant cells, whereas the nitrogen (N)-terminal contains a secretion signal and has no effect on effector translocation (Büttner et al., 2002). This suggests that xanthomonads T3ET occurs in a translocondependent manner. By contrast, a recent report proposed a translocon-independent pathway (Scheibner et al., 2017). The N-terminal 10 and 50 amino acids are required for T3 secretion and AvrBs3 translocation, respectively. Additional signals in the N-terminal 30 amino acids and the region between amino acids 64 and 152 promote AvrBs3 translocation. AvrBs3 translocation occurs in the absence of the T3 secretion chaperon HpaB, and in the absence of HrpF, which is a predicted component of the T3 translocon assembly. The authors suggested that the delivery of AvrBs3 begins during the early stages of infection, before the activation of HpaB or translocon integration into the plant PM (Scheibner et al., 2017). It is more likely that a different translocator, present in reserve and lacking function when the bacteria possesses a workable HrpF, is employed when HrpF loses function or is removed from the bacterial proteome.

#### A CYTOLOGICAL GAP BETWEEN H2O<sup>2</sup> SIGNALING AND IMMUNITY PATHWAYS

H2O<sup>2</sup> is stable compared with other ROS molecules such as the superoxide anion O<sup>2</sup> <sup>−</sup> and hydroxyl radical OH−. In plants, H2O<sup>2</sup> is produced by the enzymatic activities via multiple biochemical mechanisms (Smirnoff and Arnaud, 2019). These mechanisms include electron leakage from the electron transport chain in chloroplasts and mitochondria, the activity of peroxisomal oxidases and peroxidases in cytoplasm or plant cell walls, as well as the activity of NADPH oxidases (NOXs) in the PM (Smirnoff and Arnaud, 2019). The rapid production of ROS, especially H2O2, indicates the successful recognition of pathogen infection and molecular patterns (Alvarez et al., 1998; Torres, 2009). Well-known examples of pathogenic patterns include invariant microbial epitopes, such as fungal chitin (Kaku et al., 2006) and bacterial flagellin (Zipfel et al., 2004) and harpin proteins (Sang et al., 2012; Choi et al., 2013). These pattern molecules can be recognized by pattern receptors within the PM, which induce immune responses, including H2O<sup>2</sup> production, in plants (Levine et al., 1994; Ausubel, 2005; Galletti et al., 2011).

The production of H2O<sup>2</sup> is typically apoplastic, resulting mainly from the enzymatic activity of NOXs located in PMs (Sagi and Fluhr, 2006; Kärkönen and Kuchitsu, 2015; Smirnoff and Arnaud, 2019). Then, there is crosstalk between H2O<sup>2</sup> and immunity pathways, such as systemic acquired resistance (SAR) and pattern-triggered immunity (PTI) to regulate plant disease resistance (Torres, 2009). SAR is characteristic of the induced expression of pathogenesis-related (PR) genes, typically PR-1 and PR-2, under the regulation of non-inducer of PR genes-1 (NPR1) (Cao et al., 1997; Kim et al., 2011). NPR1 functions through conformational changes under cytoplasmic redox conditions (Tada et al., 2008) and through proteasome-mediated turnover in the nucleus (Spoel et al., 2009). The PTI pathway activates a cytoplasmic MAPK cascade (Asai et al., 2002), including a branch in which MPK3 and MPK6 phosphorylate different substrates (Bigeard et al., 2015; Pitzschke, 2015) to activate immune responses, including H2O<sup>2</sup> and callose production (Bethke et al., 2012; Daudi et al., 2012). Callose is a β-1,3-glucan synthesized by glucan synthase-like (GSL) enzymes, with GSL5 playing a critical role in cellular immune responses (Lü et al., 2011). Therefore, both the SAR and PTI pathways comprise pivotal tiers of intracellular responses in the crosstalk with H2O<sup>2</sup> produced in the apoplast (Sagi and Fluhr, 2006). It is clear that a cytological gap exists between H2O<sup>2</sup> generation and functional performance. In fact, it remains unclear for a long time how apoplastic H2O<sup>2</sup> penetrates plant PMs to enter the cytoplasm and regulate immunity.

#### PIP-MEDIATED H2O<sup>2</sup> TRANSPORT AND ITS IMMUNOLOGICAL IMPORTANCE

Hpa1, applied to plants or produced in transgenic plants, functions as a bacterial pattern to activate the PTI and SAR pathways (Tian et al., 2016). Both pathways are activated by the generation of ROS, especially H2O2, in plant apoplasts. In Arabidopsis, inoculation with the bacterial pathogen Pseudomonas syringae pv. tomato or treatment with bacterial patterns, including Hpa1 and the flagellin functional fragment flg22, induce H2O<sup>2</sup> generation in the apoplast. This H2O<sup>2</sup> moves quickly into the cytoplasm, where H2O<sup>2</sup> associates with PTI and SAR signal transduction. AtPIP1;4 serves as a H2O<sup>2</sup> transport channel to facilitate apoplastic H2O<sup>2</sup> import into the cytoplasm (**Figure 3A**), bridging the cytological gap in immunity signaling cascades (Tian et al., 2016).

This finding validates the hypothesis that H2O<sup>2</sup> transport across a biomembrane is mediated by particular AQP isoforms in addition to certain membrane lipids (Bienert et al., 2006, 2007; Bienert and Chaumont, 2014; Aguayo et al., 2015). AQPs are transmembrane channels essential for the transport of H2O, H2O2, and other small substrates in all living cells (Maurel, 2007; Gomes et al., 2009). In this way, AQPs can modulate many physiological and/or pathological processes (Maurel, 2007; Ji and Dong, 2015a,b; Li et al., 2019; Pawłowicz and Masajada, 2019; Zhang et al., 2019). In most plant species, five major families of AQPs exist. The PIP family has 11 members, PIP1;1–5 and PIP2;1–8 (Gomes et al., 2009; Abascal et al., 2014; Maurel et al., 2015). These are believed to mediate the transport of different substrates across plant PMs in an overlapping or redundant substrate-specific manner (Maurel, 2007; Péret et al., 2012, 2013; Prado et al., 2013). To date, five AtPIP isoforms (2;1, 2;2, 2;4, 2;5,

FIGURE 3 | Crosstalk of AtPIP1;4-mediated H2O<sup>2</sup> transport with the intracellular immunity pathways and predicted mechanisms by which AtPIP1;4 fulfills the substrate transport function. (A) Plant sensing of a pathogen or microbial pattern not only is an essential step of apoplastic generation and cytoplasmic import of H2O2, but also induces damages to the PM integrity (Guignot and Tran Van Nhieu, 2016). Impairment of the PM integrity is likely to provide an abnormal channel, which is wider than the normal conduit, and capable of accommodating substrates larger than H2O. (B) Hypothetic determinants of AtPIP1;4 conformation for H2O<sup>2</sup> transport include amino acid compositions and locations in the NPA and SF regions. (C) Gating and trafficking regulation of the AtPIP1;4 channel for H2O<sup>2</sup> transport across plant PMs (left) may be subject to the annotated factors (right). The 3D-structure of AtPIP1;4 was predicted by using the PHYRE2 (Protein Homology/analogy Recognition Engine V 2.0) program

(http://www.sbg.bio.ic.ac.uk/phyre2/html/page.cgi?id~=~index). The diagrammatic transport of H2O<sup>2</sup> over H2O is a surmise, predicted to occur by the combined mechanisms indicated on right.

and 2;7) are assumed to mediate H2O<sup>2</sup> transport in engineered yeast cells (Bienert and Chaumont, 2014). The de novo expression of these PIPs can increase H2O<sup>2</sup> sensitivity and decrease the viability of yeast (Dynowski et al., 2008; Hooijmaijers et al., 2012). Based on incomplete literature search, not all PIPs whose expression increases H2O<sup>2</sup> sensitivity and decreases the viability of yeast have been verified for the H2O<sup>2</sup> transporting function. AtPIP2;1 was determined to increase H2O<sup>2</sup> uptake by yeast cells (Dynowski et al., 2008; Bienert and Chaumont, 2014) and by Arabidopsis guard cells (Rodrigues et al., 2017). AtPIP1;4 has been shown to function in H2O<sup>2</sup> transport from the apoplast into the cytoplasm of Arabidopsis (Tian et al., 2016). Many works are required to test in planta function of the AQP candidates in H2O<sup>2</sup> translocation.

#### CONSERVATIVE AQP FUNCTION FOR H2O<sup>2</sup> TRANSPORT

AtPIP1;4 was determined to have triple substrate specificities (Li et al., 2015; Tian et al., 2016). In addition to transporting H2O<sup>2</sup> (Tian et al., 2016), AtPIP1;4 partakes in the cellular hydraulic conductance (P<sup>f</sup> ) of roots, and in mesophyll conductance of CO<sup>2</sup> (gm); however, it does not affect stomatal CO<sup>2</sup> conductance (gs) or P<sup>f</sup> in leaves (Li et al., 2015). The interaction of AtPIP1;4 with Hpa1 at Arabidopsis PMs promotes substrate transport, increasing the net photosynthesis rate (AN), while P<sup>f</sup> is also increased in leaves and roots (Li et al., 2015). Therefore, a PIP can alter its physiological functions or effect extents in response to plant pathogens or bacterial patterns.

The function of AtPIP1;4 in immunity is an extension of its primary roles in substrate transport, which was initially assigned to AQPs in mammals (Preston and Agre, 1991) and subsequently in plants (Maurel et al., 1993). The functional extension of AtPIP1;4 has biological importance for at least two reasons. First, AtPIP1;4-dependent SAR responses induced by bacterial pathogens effectively repress pathogen virulence (Tian et al., 2016; **Figure 3A**). In this case, pathogen-associated repressors of plant immunity (Oh and Collmer, 2005; Zhang et al., 2007; Guo et al., 2009) may be inhibited, or their immunity-repressing functions may be counteracted by the role of AtPIP1;4 in H2O<sup>2</sup> translocation, which is linked to the immunity pathway. Second, AtPIP1;4 is an integral component of PTI in response to typical patterns, with conserved microbial cell-surface composition, i.e., flagellin (Zipfel et al., 2004) and chitin (Kaku et al., 2006). Despite their different biochemical nature, both patterns require AtPIP1;4 to induce PTI responses, except in the absence of induced MPK6 expression (Tian et al., 2016). This is consistent with previous findings that the MAPK cascade diverges at MPK3 and MPK6 (Asai et al., 2002; Bigeard et al., 2015) to regulate distinct substrates in response to different patterns (Galletti et al., 2011; Pitzschke, 2015). Moreover, the induction of MPK3 expression represents a circuit of the MAPK cascade in response to H2O<sup>2</sup> (Gudesblat et al., 2007). These sets of information suggest that AtPIP1;4 plays a prominent role in immunity signaling by mediating apoplastic H2O<sup>2</sup> translocation into plant cells.

AQP-mediated H2O<sup>2</sup> transport in immune signaling also occurs in mammals. Among 13 AQPs, AQP3 is a H2O<sup>2</sup> transport channel (Miller et al., 2010). AQP3-mediated H2O<sup>2</sup> transport is associated with necrosis factor-κB (NFκB) signaling in keratinocytes, and in the pathogenesis of psoriasis in response to cytokine regulation (Hara-Chikuma et al., 2015). The induction of psoriasis by cytokines, NFκB activation, and intracellular H2O<sup>2</sup> accumulation are concomitantly reduced in AQP3-knockout mice. In primary keratinocyte cultures, H2O<sup>2</sup> is generated by membraneassociated NOX2 in response to TNF-α, and moves into intracellular spaces. Cellular import of H2O<sup>2</sup> is facilitated by AQP3 and is required for NF-κB activation under PP2A regulation. Since AQP3 associates with NOX2 at PMs, this interplay may constitute H2O2-mediated signaling in response to TNF-α stimulation (Hara-Chikuma et al., 2015). Moreover, under oxidative stress, AQP3-mediated H2O<sup>2</sup> transport attenuates apoptosis by regulating the P38 MAPK pathway (Xu et al., 2018; He and Yang, 2019). Based on these findings, and those regarding PIPs, cytoplasmic import across the PM can reduce the cytological distance for H2O<sup>2</sup> generation, and functional performance (Bienert et al., 2006; Sang et al., 2012; Tian et al., 2016). Apocytoplastic signaling is conserved in plants and animals.

#### AQP STRUCTURE FOR H2O TRANSPORT

It is unclear how different AQPs function in the transport of corresponding substrates, and how an AQP, such as AtPIP1;4 (Li et al., 2015; Tian et al., 2016), can function as a triple substrate transport conduit. One hypothesis is that structural details allow for differences in selectivity and modes of regulation (Kreida and Törnroth-Horsefield, 2015). Regarding H2O<sup>2</sup> transport, the structures of AQP/PIP channels have not been studied, but can be inferred from information on structures of AQPs that function as water channels.

Plant aquaporins are predominant channels of H2O transport between the outside and inside of the cell, and between intracellular organelles (Huang et al., 2017). Although cotransporters and uniporters have been implicated in water homeostasis, AQPs have been accepted as intramolecular channels for the transmembrane movement of H2O down an osmotic gradient (Maurel et al., 2015; Huang et al., 2017; Yang, 2017; Pawłowicz et al., 2018). H2O transport by AQPs is determined by their three-dimensional structure.

Structural studies have characterized AQPs as homotetramers, which are integrated into the membrane with conserved configurations (Fu et al., 2000; Sui et al., 2001; Törnroth-Horsefield et al., 2006; Horsefield et al., 2008; Eriksson et al., 2013; Kirscht et al., 2016). Each monomer has a functional pore formed by six α-helical TM domains (TM1–TM6), five connecting loops (LA–LE), and two shorter helices (HB and HE). The outward end of HB and inward end of LE contain a pair of asparagine (N), proline (P), and alamine (A) tandem (NPA) motifs, which constitute the central channel through the membrane (Kirscht et al., 2016). Two NPAs form a conical funnel or traditional hourglass, which are linked at the tip and open outward from LE and inward from TM5 (Törnroth-Horsefield et al., 2010), and are essential for AQP function (Wree et al., 2011; Chen et al., 2018). Within LE, TM2 and TM5, the aromatic/arginine (Ar/R) selective filter (SF) is formed by four residues including aromatic amino acids and an arginine (R) residue; hence its name (de Groot et al., 2003). The SF is located in the outward opening of the channel and allows H2O to pass while repelling protons and cations (Eriksson et al., 2013). Multiple physical factors, such as hydrophilic and hydrophobic interactions, electrostatic repulsion, and dipole alignment between amino acid residues within or around the NPA and SF, influence substrate selectivity (Törnroth-Horsefield et al., 2006).

A pivotal step toward the substrate-transporting function of AQPs is the regulation of gating (opening and closing) and trafficking (substrate transport). This has been elucidated for water channels at angstrom (Å) or sub-Å resolution (Daniels et al., 1999; Fotiadis et al., 2001; Kukulski et al., 2005; Kreida and Törnroth-Horsefield, 2015). Considering spinach Spinacia oleracea SoPIP2;1, channel opening is triggered by the phosphorylation of conserved serine (S) 197 (Johansson et al., 1996; Kukulski et al., 2005), and is expedited by hydrogen bond networks in LD (Törnroth-Horsefield et al., 2010). Channel closure results from the dephosphorylation of S115 in LB and S274 in the C-terminal region of the AQP sequence under conditions of drought stress, or from the protonation of a conserved histidine (H) residue following a decrease in cytoplasmic pH due to anoxia during flooding. Dissection of SoPIP2;1 crystal structures, both the closed conformation at 2.1 Å and the open conformation at 3.9 Å, reveals the importance of LD displacement for gating and trafficking. The dephosphorylation of S115 and S274 prevents outward NPA entry from LB, and inward NPA exit in TM5. In the open conformation of SoPIP2;1, S197 is phosphorylated at LD, LD is displaced up to 16 Å, the nitrogen terminus of TM5 extends a further half-turn into the cytoplasm, and NPA entry and exit are promoted. In addition, H193 protonation and interactions between amino acids, including hydrogen bond networks and electrostatic repulsion, also influence the switch between opening and closing of the channel (Törnroth-Horsefield et al., 2010).

Crystal structure analysis of Aqy1, the only AQP in yeast Pichia pastoris, at a sub-Å (0.88 Å) resolution, provides evidence for tautomeric reactions to expedite H2O transport (Eriksson et al., 2013). Hydrophilic amino acids in NPA and SF interact to bind H2O molecules, which are then navigated through the channel. With polar hydrogen bond configurations, four H2O molecules per group pass the SF, and then divide into two pairs to pass through the inward NPA region. There are two types of tautomerism between hydrophilic amino acids in the SF. One is proton transfer – the atom Nδ, but not Nε, of H212 is protonated to provide a proton for L208, with the role of guiding H2O movement. The other one is covalent binding – atoms Cζ and Nη2 of R227 maximally bind to each other, Nη2 is closest to the central conduit, and its positive

charge repels cations, creating favorable conditions for H2O to travel through the SF. With this advantage, four compact H2O molecules are located within the full space of the SF, where they synchronize to move within and across the SF passage. Due to high impacts of atom tautomerism and hydrogen-bond interactions restricted to the H2O molecules in transport, other H2O molecules must wait for the next round of the channel opening and trafficking, and proton or cations are unable to enter the SF.

In addition to the structural configuration, biochemical regulation is also indispensable to the function of AQPs. In this aspect, channel gating and trafficking regulation by phosphorylation are ubiquitous for all AQPs (Li and Wang, 2017; Kapilan et al., 2018; Laloux et al., 2018; Nesverova and Törnroth-Horsefield, 2019). Additionally critical mechanisms underlying the functional regulation of different AQPs include biotic and abiotic signals. They induce the transport of different substrates (Tian et al., 2016; Ruiz-Lozano and Aroca, 2017; Balestrini et al., 2018; Smirnoff and Arnaud, 2019) by stimulating AQPs themselves with gradients over membranes and by interacting with other proteins (Ji and Dong, 2015a; Roche and Törnroth-Horsefield, 2017). These have been topics of many literatures (for example: Maurel and Plassard, 2011; Hara-Chikuma et al., 2015; Ji and Dong, 2015a; Maurel et al., 2015; Yang, 2017; Roche and Törnroth-Horsefield, 2017) and will not been discussed in this article.

#### CONTROL OF SUBSTRATE SPECIFICITIES

This is a question for AQPs capable of transporting substrates other than H2O, especially those that have multiple permeation properties. In addition to H2O, approximately 20 other substrates require AQPs to move between the exterior and interior of cells, and between organelles (Laloux et al., 2018). A fifth pore created by four AQP monomers of a homotetramer in the lipid bilayer (Wang et al., 2007) or yeast membrane (Otto et al., 2010) has been proposed for gas (CO<sup>2</sup> and O2) and ion transport (Kaldenhoff et al., 2014). Moreover, many AQPs have more than one substrate (Kreida and Törnroth-Horsefield, 2015; Maurel et al., 2015; Fox et al., 2017; Laloux et al., 2018). Examples include AtPIP2;1 for H2O/H2O<sup>2</sup> (Dynowski et al., 2008; Verdoucq et al., 2008), AtPIP1;4 for H2O/H2O2/CO<sup>2</sup> (Li et al., 2015; Tian et al., 2016), and TIPs for H2O, H2O<sup>2</sup> and/or ammonia (NH3; Maurel et al., 1993; Loque et al., 2005; Bárzana et al., 2014) transport. Regulation of gating and trafficking must differ considerably between specialist channels, different generalist channels, and channels for H2O and a different substrate. Variation in NPA diameter, the composition and width of SF, neighboring residues, and their interactions with each other and with the substrate might explain multiple functions of AQPs/PIPs in the transport of different substrates, and the multiple substrate transport capacities of a single AQP/PIP (Fox et al., 2017).

Recently, a smart solution was proposed in a study on the 1.18 Å crystal structure of AtTIP2;1 (Kirscht et al., 2016). That study characterized AtTIP2;1 as an NH<sup>3</sup> transport channel, which functions with an extended SF. The channel diameter in the NPA region is smaller than that of other AQPs, but remains constant at ∼3Å along the channel; this is in contrast to the narrowing of SF in other AQPs. The topological positions of four SF residues in TM2, TM5, LE, and HE are thought to determine substrate selectivity (de Groot et al., 2001). Consistent with this model, TIP2s deviate from other AQPs in terms of the wider SF, which is mainly caused by an isoleucine (I185) in TM5, replacing a histidine that is conserved in water-specific AQPs (Kirscht et al., 2016). The most striking feature of the SF in AtTIP2;1 is the R200 located in HE, while the arginine in HE is conserved in most AQPs. In AtTIP2;1, the R200 side chain is located at the edge of the channel due to the H131 situated in LC, making histidine the fifth residue of the extended SF. The position of this arginine is further stabilized by a hydrogen bond with histidine (H63) in TM2, which occupies the same space as the corresponding aromatic residues of water and glycerol channels without direct effects on the channel opening (Kirscht et al., 2016). Moreover, H131 in LC interacts directly with the substrate in the selectivity region. These structural features define the extended SF at five positions: I185, R200, H131, and H63, which have properties and configurations that establish the novel SF, plus G191 in LE, which is conserved in the canonical and extended SF. The concept of extended SF is instructive to conceiving study schemes before initiating analysis of APQ/PIP channels for transport of H2O<sup>2</sup> and more substrates other than H2O and NH3.

#### STRUCTURAL BASIS OF PIPS FOR MEDIATION OF H2O<sup>2</sup> TRANSPORT

Until the structural basis of PIP/AQP functions in H2O<sup>2</sup> transport is dissected, no more than inspiration can be deduced from referencing the crystal structures of SoPIP2;1 for NH<sup>3</sup> transport (Kirscht et al., 2016) and both Aqy1 (Eriksson et al., 2013) and AtTIP2;1 (Törnroth-Horsefield et al., 2010) for H2O transport. The topological displacement of the connecting loop (Törnroth-Horsefield et al., 2010) may have a broad importance for AQPs. Tautomeric reactions (Eriksson et al., 2013) and the SF extension (Kirscht et al., 2016) might be used by certain PIPs/AQPs to expedite H2O<sup>2</sup> transport. However, these features are likely to be insufficient to support H2O<sup>2</sup> transport, due to the difference in diameter/molecular mass of H2O<sup>2</sup> (3.70 Å/34), H2O (2.96 Å/33) and NH<sup>3</sup> (<2.96 Å/17), and in the Ar/R SF features. The location and composition of the SF is identical (F87, H126, T225, R231) in the H2O<sup>2</sup> channel AtPIP2;1 (Rodrigues et al., 2017) and the water channel SoPIP2;1 (Kirscht et al., 2016). However, the SF composition shared by AtPIP2;1 and SoPIP2;1 is distinct from that in the corresponding positions (G87, I126, L225, and T231) of the H2O<sup>2</sup> channel AtPIP1;4 (Tian et al., 2016). AtPIP1;4 is the same length as OsPIP2;1, but possesses six more residues than SoPIP2;1, with a predicted Ar/R SF comprising F95, H224, T231, and R239 (**Figure 3B**). If the SF extension permits AQPs to mediate H2O<sup>2</sup> transport, the degree of the SF extension must be considerably higher than that in the NH<sup>3</sup> transport channel (Kirscht et al., 2016).

Three issues are considered to infer the structural basis of the function of PIPs in H2O<sup>2</sup> transport between the outside and inside of plant cells. First, the apocytoplastic transport of H2O<sup>2</sup> is more intricate as compared to the signal shift ways by the cell-to-cell traveling via plasmodesmata (Wang et al., 2009) and via vesicle-aided trafficking between organelles through the ER system within the cell interior (Ashtamker et al., 2007; Melo et al., 2017). Second, H2O<sup>2</sup> transport in and out of plant cells is not constant throughout the life circle of plants (Dynowski et al., 2008; Tian et al., 2016). Third, H2O<sup>2</sup> trafficking across the PM is induced but is not constitutive, and occurs only when apoplastic H2O<sup>2</sup> is generated in response to pathogens, microbial patterns, or environmental signals (Levine et al., 1994; Xin et al., 2015; Liu and He, 2016; Tian et al., 2016).

Plasma membrane sensing of these distinct signals will promote H2O<sup>2</sup> generation in apoplasts and its immediate translocation into the cytoplasm (Ausubel, 2005; Ashtamker et al., 2007; Tian et al., 2016) by three possible mechanisms. One is inductivity (**Figure 3A**). When plants are infected by a pathogen or respond to a microbial pattern, such as Hpa1 or flg22, the enzymatic activity of NOX is induced to catalyze the generation of H2O<sup>2</sup> by peroxidation and superoxidation in PMs (Tian et al., 2016; Smirnoff and Arnaud, 2019). The generated H2O<sup>2</sup> accumulates, and the concentration increases temporarily in the apoplast. This creates a gradient from the outside to the inside of the cell (Tian et al., 2016), and induces the PIP channel to function (Tian et al., 2016).

The second mechanism is speculated to be the combination of factors (**Figure 3C**) found in SoPIP2;1 (Törnroth-Horsefield et al., 2010), Aqy1 (Eriksson et al., 2013), and AtTIP2;1 (Kirscht et al., 2016). Combined factors facilitate the passage of H2O<sup>2</sup> through the PIP channel, which could be established by SF extension (Kirscht et al., 2016), and optimized by amino acid protonation (Eriksson et al., 2013). H2O<sup>2</sup> generation (2O<sup>2</sup> <sup>−</sup> + 2H<sup>+</sup> = H2O2) requires protons, and may reduce the likelihood that amino acid residues near the SF and NPA regions are protonated. As the protonation navigates H2O movement along the channel (Eriksson et al., 2013), decreased protonation will disturb H2O transport. This might promotes the transport of H2O<sup>2</sup> over H2O through a PIP channel once a sufficient diameter is reached (>3.70 Å).

The third mechanism is supposed to be biochemical responses (**Figure 3C**) associated with the regulation of PM remodeling – injury and repair (Laliberté and Sanfaçon, 2010; Santi-Rocca and Blanchard, 2017). PM remodeling is triggered by the binding of an active extrinsic protein, including microbial patterns such as Hpa1 (Li et al., 2015; Tian et al., 2016), bacterial T3 translocators such as HrpF (Büttner et al., 2002; Li et al., 2011), and bacterial effectors (Hubber and Roy, 2010; Domingues et al., 2016; Dong et al., 2016). Binding of these bioactive proteins affects the PM integrity (Ji and Dong, 2015a; Guignot and Tran Van Nhieu, 2016). Reduced PM integrity is advantageous for solute influx, which, however, is strictly regulated by proteins and lipids that recognize microbial patterns, T3 effectors, or translocators (Gilbert et al., 2014).

The former two mechanisms may synergize in the gating and trafficking regulation, requiring AtPIP1;4 to transport H2O<sup>2</sup> in plants grown under regular conditions without any input signal, except for externally applied H2O<sup>2</sup> or H2O<sup>2</sup> induced by a pathogen or a microbial pattern (Tian et al., 2016). The third mechanism may occur in the presence of Hpa1 following application to plants or production in transgenic plants, in which AtPIP1;4 interacts with Hpa1 (Li et al., 2015) to increase the substrate transport function. Studies should aim to verify this hypothesis in order to elucidate the structures of PIP orthologs as transport channels for H2O<sup>2</sup> or different substrates.

#### CONCLUSION AND PERSPECTIVES

Finite research performed on these case studies is based on a solid foundation obtained through extensive studies; research on the structural regulation of PIP function in plant infection and immunity is invited. The first case study on Hpa1-mediated, OsPIP1;3-associated, and virulence-relevant PthXo1 translocation offers multiple experimental avenues to characterize interactions between T3 translocators and their receptors at target PMs, as well as the associated implications for effector translocation and virulence. The two subjects discussed here are yet to be thoroughly studied. First, which of the assumed delivery lanes is used by different effectors is a long-standing question for all plant-pathogenic bacteria. Xoo possesses more than 30 effectors secreted by the T3 system (White et al., 2009), similar to the number in other bacteria. Further study is needed to identify all T3 effectors in the three proposed mechanisms: translocon-independent pore formation (**Figure 1**), endocytosis with PM protein or lipid trafficking (**Figures 1**, **2**), and translocon-dependent delivery (**Figure 1**). The second subject includes the contribution of PM lipids and proteins to T3 effector translocation. T3 translocon assembly or pore formation must recruit both lipids and proteins situated in plant PMs (Büttner et al., 2008; Gilbert et al., 2014; Heilmann and Heilmann, 2015; Ji and Dong, 2015b; Guignot and Tran Van Nhieu, 2016). It would be of great interest to determine how effectors are internalized with PM protein or lipid trafficking, and how protein and lipid receptors of T3 translocators coordinate their actions to generate pores or translocons in plant PMs.

The second case study discusses AtPIP1;4-regulated, Hpa1 promoted, and immunity-linked H2O<sup>2</sup> transport, and establishes a cytological connection between the generation and function of H2O<sup>2</sup> in the apoplast and cytoplasm, respectively (Tian et al., 2016). The cytoplasmic import of H2O<sup>2</sup> bridges a physical gap, which was unknown for at least 20 years since the biphasic H2O<sup>2</sup> accumulation following induction was awarded biological significance (Levine et al., 1994). AtPIP1;4 mediated H2O<sup>2</sup> translocation is a pivotal step in apocytoplastic signal transduction for intracellular immunity pathways, which regulate SAR and PTI responses, leading to plant resistance against diseases (Dong et al., 1999; Chen et al., 2008a,b; Choi et al., 2013; Zhao et al., 2014). The future focus of studies will be difficult, highlighting the regulation of gating and trafficking of the AtPIP1;4 channel for H2O<sup>2</sup> transport. To date, the structures of AQP channels have only been determined for the transport of NH<sup>3</sup> (Kirscht et al., 2016)

and H2O (Daniels et al., 1999; Fotiadis et al., 2001; Kukulski et al., 2005; Kreida and Törnroth-Horsefield, 2015), and almost 20 substrates remain to be understood (Laloux et al., 2018). Rational hypotheses on structural themes in both gating and trafficking (Kreida and Törnroth-Horsefield, 2015) requires the efforts of researchers to explore structural mechanisms that govern diverse AQP channels. It is necessary to dissect the conformation of AtPIP1;4 (**Figures 3B,C**) involved in H2O<sup>2</sup> transport in response pathogens or patterns (**Figure 3A**). It is especially necessary to study whether the H2O<sup>2</sup> transport is facilitated by combined impetuses, including the SF extension, amino acid residue interactions (**Figure 3C**), and PM protein trafficking (**Figure 2**).

The two case studies have been designed to converge at the intersection Hpa1-PIP cooperation and branch into two directions. One targets plant immunity, for which Hpa1 functions as a bacterial pattern in a pathogen-independent manner. The other contributes to plant infection, in which Hpa1 acts as a T3 translocator after secretion by the bacteria, and mediates the translocation of virulent effectors that lead to disease. These findings provide insight into disease control either through induced immunity, or the prevention of bacteria from usurping the substrate transport gate. Practical application of both strategies to strengthen crop protection (Krinke et al., 2007; Chen et al., 2008b; Fu et al., 2014; Wang D. et al., 2014; Li et al., 2019)

#### REFERENCES


will integrate with crop involvement by using AQPs from plants themselves (Ruiz-Lozano and Aroca, 2017; Balestrini et al., 2018) and from symbiotic microbes as well (Kikuchi et al., 2016; Desaki et al., 2018).

#### AUTHOR CONTRIBUTIONS

LZ, LC, and HD drafted the manuscript. LZ predicted the 3D structure of AtPIP1;4. HD finalized the manuscript.

#### FUNDING

This study was supported by Natural Science Foundation of China (Grant No. 31772247) and China National Key Research and Development Plan (Grant No. 2017YFD0200901) to HD and Talent Recruitment Funding of Shandong Agricultural University (Grant No. 20171226) to HD, LC, and LZ.

#### ACKNOWLEDGMENTS

We thank Mr. Hao Wang (doctoral student in the laboratory) for his advice in the structural analysis.



to abscisic acid and hydrogen peroxide. New Phytol. 173, 713–721. doi: 10.1111/ j.1469-8137.2006.01953.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 © 2019 Zhang, Chen and Dong. 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.

# Comparative Transcriptomics and Proteomics of Atractylodes lancea in Response to Endophytic Fungus Gilmaniella sp. AL12 Reveals Regulation in Plant Metabolism

Jie Yuan<sup>1</sup> , Wei Zhang<sup>1</sup> , Kai Sun<sup>1</sup> , Meng-Jun Tang<sup>1</sup> , Piao-Xue Chen<sup>1</sup> , Xia Li<sup>2</sup> \* and Chuan-Chao Dai<sup>1</sup> \*

#### Edited by:

Marco Catoni, University of Birmingham, United Kingdom

#### Reviewed by:

Khondoker M. G. Dastogeer, Tokyo University of Agriculture and Technology, Japan Stefania Daghino, University of Turin, Italy

#### \*Correspondence:

Chuan-Chao Dai daichuanchao@njnu.edu.cn Xia Li jspplx@jaas.ac.cn

#### Specialty section:

This article was submitted to Plant Microbe Interactions, a section of the journal Frontiers in Microbiology

Received: 14 March 2019 Accepted: 13 May 2019 Published: 28 May 2019

#### Citation:

Yuan J, Zhang W, Sun K, Tang M-J, Chen P-X, Li X and Dai C-C (2019) Comparative Transcriptomics and Proteomics of Atractylodes lancea in Response to Endophytic Fungus Gilmaniella sp. AL12 Reveals Regulation in Plant Metabolism. Front. Microbiol. 10:1208. doi: 10.3389/fmicb.2019.01208 <sup>1</sup> Jiangsu Key Laboratory for Microbes and Functional Genomics, Jiangsu Engineering and Technology Research Center for Industrialization of Microbial Resources, College of Life Sciences, Nanjing Normal University, Nanjing, China, <sup>2</sup> Jiangsu High Quality Rice Research and Development Center, Nanjing Branch of Chinese National Center Rice Improvement, Institute of Food Crops, Jiangsu Academy of Agricultural Sciences, Nanjing, China

The fungal endophyte Gilmaniella sp. AL12 can establish a beneficial association with the medicinal herb Atractylodes lancea, and improve plant growth and sesquiterpenoids accumulation, which is termed "double promotion." Our previous studies have uncovered the underling primary mechanism based on some physiological evidences. However, a global understanding of gene or protein expression regulation in primary and secondary metabolism and related regulatory processes is still lacking. In this study, we employed transcriptomics and proteomics of Gilmaniella sp. AL12-inoculated and Gilmaniella sp. AL12-free plants to study the impact of endophyte inoculation at the transcriptional and translational levels. The results showed that plant genes involved in plant immunity and signaling were suppressed, similar to the plant response caused by some endophytic fungi and biotroph pathogen. The downregulated plant immunity may contribute to plant-endophyte beneficial interaction. Additionally, genes and proteins related to primary metabolism (carbon fixation, carbohydrate metabolism, and energy metabolism) tended to be upregulated after Gilmaniella sp. AL12 inoculation, which was consistent with our previous physiological evidences. And, Gilmaniella sp. AL12 upregulated genes involved in terpene skeleton biosynthesis, and upregulated genes annotated as β-farnesene synthase and β-caryophyllene synthase. Based on the above results, we proposed that endophyte-plant associations may improve production (biomass and sesquiterpenoids accumulation) by increasing the source (photosynthesis), expanding the sink (glycolysis and tricarboxylic acid cycle), and enhancing the metabolic flux (sesquiterpenoids biosynthesis pathway) in A. lancea. And, this study will help to further clarify plant-endophyte interactions.

Keywords: Atractylodes lancea, endophytic fungi, beneficial interaction, plant immunity, metabolism, terpenoid biosynthesis

## INTRODUCTION

fmicb-10-01208 May 25, 2019 Time: 16:29 # 2

Medicinal plants are rich in active compounds such as artemisinin (Sharma and Agrawal, 2013) and ginseng saponin (Wei et al., 2018), which are important sources of modern drugs. Within the increasing population pressure, costs and side effects of drugs, demands for the uses of medicines from plants for treatment of various human ailments are increasing. Compared with synthetic medicines, compounds from plants are thought to be safe to human beings and the ecosystem (Nema et al., 2013). To meet the market demand for these medicinal materials, artificial cultivation technology has been employed for planting medicinal herbs in some Asian countries (Zhou et al., 2005; Sharma et al., 2010; Oh et al., 2014; Lv et al., 2017). However, it is difficult to guarantee the quality and quantity of artificially cultivated medicinal plants because they are more vulnerable to infections by pests and pathogens. Thus, it is imperative to seek effective methods for medicinal plant cultivation. In the past few years, endophytic fungi possessing plant growth promoting properties have been an effective tool for medicinal plant cultivation (Mandal et al., 2013; Ming et al., 2013; Sharma and Agrawal, 2013; Zheng et al., 2016; Zhai et al., 2018).

Atractylodes lancea (Thunb.) DC., belonging to the Asteraceae family, is an endangered traditional Chinese medicinal herb (Wang et al., 2008). Its bioactive component, the sesquiterpenoids, possesses various pharmacology properties such as antibacterial, antitumour, and immunomodulation abilities (Wang et al., 2008; Koonrungsesomboon et al., 2014; Na-Bangchang et al., 2017). Over the past few years, natural sources of A. lancea have been in short supply because of the excessive exploitation and slow growth rate of the herb (Zhou et al., 2016). The medicinal source of A. lancea mainly derives from artificial cultivation, but the yield and quality of this herb are relatively low (Zhou et al., 2016). At present, it is urgent to improve the quality and quantity of the herb as the market demand for A. lancea is increasing on a daily basis. The endophytic fungus Gilmaniella sp. AL12 isolated from stem of A. lancea can establish a beneficial interaction with the host plant (Wang et al., 2012) and promote plant growth and sesquiterpenoid accumulation of tissue culture seedings, which is termed the "double promotion" effect of the endophyte on A. lancea (Yuan et al., 2016b). Consistent with this phenomenon, the endophytic fungi AL12 promotes plant growth and sesquiterpenoid accumulation within two years of growth in field experiments. Therefore, a beneficial interaction of Gilmaniella sp. AL12 with A. lancea is considered suitable for cultivation of A. lancea and will provide a theoretical reference for endophytic fungi-medicinal herb interactions.

In view of the limited carbon and energy source in plants, the accumulation of secondary metabolites occurs at the cost of primary metabolism, representing a discrepancy with the "double promotion" effect of Gilmaniella sp. AL12 on A. lancea. The plant growth promotion effect of the endophyte on A. lancea has been preliminarily ascribed to nutrient assimilation, photosynthesis, and phytohormone content regulation (Yuan et al., 2016b). Moreover, the enhanced sesquiterpenoids accumulation of A. lancea has been shown to be mediated by multiple defense related signals of the host induced by the endophyte (Wang et al., 2011; Ren and Dai, 2012, 2013; Yuan et al., 2016a). Given that primary metabolism-dependent terpenoid precursor biosynthesis and secondary metabolismrelated terpenoid skeleton biosynthesis and transformation are simultaneously involved in sesquiterpenoid synthesis (Dudareva et al., 2006; Chen W. et al., 2017; Sharma et al., 2017; Vattekkatte et al., 2018), the molecular and biochemical regulation of the plants relevant to primary and secondary metabolism should be considered. However, thus far, a global understanding of the -regulated expression of genes or proteins in primary and secondary metabolism and related regulatory processes is still lacking.

In this study, we employed transcriptomics and proteomics on endophyte-inoculated and endophyte free plants to better understand the impact of Gilmaniella sp. AL12 on plant metabolism and related regulatory processes of A. lancea at the transcriptional and translational level. The following four essential questions were addressed in this study: (1) Which plant metabolic or regulatory processes of A. lancea are affected by Gilmaniella sp. AL12? (2) What is the effect of the fungal endophyte on the regulation of primary metabolism-dependent terpenoid precursor biosynthesis in A. lancea? (3) What is the effect of the fungal endophyte on the regulation of secondary metabolism-related terpenoid skeleton biosynthesis and transformation in A. lancea? (4) Are other potential signals involved in the fungal endophyte-induced sesquiterpenoids biosynthesis of A. lancea?

#### MATERIALS AND METHODS

#### Plantlet Material and Fungal Inoculation

Atractylodes lancea meristem cultures were established using sterilized plantlets according to our previous studies (Wang et al., 2012). Firstly, meristem cultures were established using mature A. lancea planted in Maoshan, Jiangsu Province, China (Wang et al., 2012). Sterile adventitious buds (approximately 2–3 cm long) of young stems were collected and carefully washed under running tap water. They were surface sterilized by immersing in ethanol (75%) for 30 s, followed by soaking in mercury chloride solution (1%) for 10 min and rinsing in sterile distilled water five times (Wang et al., 2012). Subsequent procedures were conducted aseptically (Wang et al., 2012). The explants were transferred in 50 mL Murashige and Skoog medium containing sucrose (30 g L−<sup>1</sup> ), agar (10 g L−<sup>1</sup> ),

**Abbreviations:** A. lancea, Atractylodes lancea; AL12, Gilmaniella sp. AL12; AMF, arbuscular mycorrhizal fungi; ATP, triphosadenine; BAK1, brassinosteroid insensitive 1-associated receptor kinase 1; DEGs, differentially expressed genes; DPI, days post-AL12 inoculation; EFR, LRR receptor-like serine/threonine-protein kinase EFR; ETI, effector triggered immunity; FPKM, the fragments per kilobase of transcript per million mapped reads; GC, gas chromatography; GO, gene ontology database; KEGG, Kyoto Encyclopedia of Genes and Genomes database; LRR, leucine-rich repeat; MDH, malate dehydrogenase; NCBI, National Center for Biotechnology Information; OAA, oxaloacetate; PAMPs, pathogen-associated molecular patterns; Pi, orthophosphate; PLD, phospholipase D; PS II, photosystem II; PTI, PAMP triggered immunity; RT-qPCR, real-time quantitative PCR; TCA cycle, tricarboxylic acid cycle; 2-DE, two-dimensional gel electrophoresis.

naphthaleneacetic acid (0.3 mg L−<sup>1</sup> ), and 6-benzyladenine (2.0 mg L−<sup>1</sup> ) in 150-mL sealed Erlenmeyer flask to emerge adventitious buds for 4 weeks (Wang et al., 2012). Then, newborn adventitious buds were separated and grown in 50 mL Murashige and Skoog medium containing sucrose (30 g L−<sup>1</sup> ), agar (10 g L−<sup>1</sup> ), naphthaleneacetic acid (0.3 mg L−<sup>1</sup> ) and 6-benzyladenine (2.0 mg L−<sup>1</sup> ) in 150-mL sealed Erlenmeyer flask for further differentiation until newborn adventitious buds were suficient for transplantation. After separation, newborn axillary buds were transplanted into 50 mL Murashige and Skoog medium containing sucrose (30 g L−<sup>1</sup> ), agar (10 g L−<sup>1</sup> ), and naphthaleneacetic acid (0.25 mg L−<sup>1</sup> ) in a 150-mL sealed Erlenmeyer flask to develop into rooting plantlets. Each bud was cultured in an Erlenmeyer flask. All plantlets were grown in growth chamber (PGX-600A-12H, Ningbo Lifewww Technology Co., China) with a photoperiod of 12 h, a light density of 3,400 lm m−<sup>2</sup> , and a temperature cycle of 25/18◦C day/night. Thirtyday-old rooting plantlets were chosen for the endophytic fungus inoculation experiments. The endophytic fungus Gilmaniella sp. AL12 isolated from the stem of A. lancea was grown on potato dextrose agar medium (Chen et al., 2008; Wang et al., 2012). After five days of culture at 28◦C, 5-mm Gilmaniella sp. AL12 mycelial disks were placed near the plant caudexes on the medium for inoculation. Additionally, equal-sized potato dextrose agar disks were used as the control (Yuan et al., 2016a). All plantlets were grown in growth chamber and performed in triplicate.

#### Plant Growth and Sesquiterpenoid Content Analysis

Plantlets were harvested at 15 DPI for the shoot dry weight and sesquiterpenoid content analysis. Briefly, harvested plants were dried at 30◦C until the weight was constant, and then 1 g of the ground powder was extracted with 4 mL cyclohexane at room temperature for 10 h. After 15 min of sonication (60 Hz), the mixture was centrifuged for 5 min at 5,000 g at 4◦C. After filtering through 0.22-µm-diameter microporous membranes, the total cyclohexane extract was dried over anhydrous sodium sulfate and stored in a dark glass bottle at 4◦C before gas chromatography (GC) analysis (Yuan et al., 2016b).

A GC system (Agilent 7890A, United States) equipped with a fame ionization detector was used for sesquiterpenoid analysis (Yuan et al., 2016b). The GC column was DB-1ms (30 m × 0.32 mm × 0.10 µm) with high-purity nitrogen as a carrier gas at a flow rate of 0.8 mL min−<sup>1</sup> . For sesquiterpenoid analysis, 1 µl of cyclohexane was directly injected onto the GC column with the following temperature program: injection at 240◦C, an initial temperature of 100◦C (4 min hold) raised to 140◦C at a rate of 10◦C min−<sup>1</sup> (10-min hold), followed by an increase to 220◦C at a rate of 10◦C min−<sup>1</sup> (10-min hold), raised to 260◦C at a rate of 10◦C min−<sup>1</sup> (2-min hold). The detector temperature was then set at 350◦C. As previously described, qualitative and quantitative analyses of seven sesquiterpenoids were performed according to authentic standards (Yuan et al., 2016b).

#### Determination of Photosynthetic Parameters

Plantlets were harvested at 10: 00 AM at 15 DPI for Chlorophyll a fluorescence measurement using a Handy-PEA instrument (Handy PEA, Hansatech, United Kingdom) (Wang et al., 2017). Prior to measurement, A. lances leaves were dark-adapted for 30 min at room temperature to relax the reaction centers. The leaves were exposed to red light of 650 nm at an excitation irradiance of 3,000 µmol m−<sup>2</sup> s −1 for 800 ms. Then, fluorescence parameters were calculated according to **Supplementary Table S1**.

#### RNA Sequencing and Functional Annotation of Differentially Expressed Genes

Plantlets were harvested at 15 DPI. Shoot tissue of endophyteinoculated or endophyte-free plants were harvested for RNA extraction using TRIzol reagent (Invitrogen, CA, United States) (Chen F. et al., 2017). After quality and purity checking, equal amounts of RNA samples from three biological replicates of CK (endophyte-free plants for control) and AL12 (endophyteinoculated plants) were used for construction of six mRNA-Seq libraries. Libraries were indexed, pooled and then sequenced on an Illumina HiSeqTM 4000 sequencing platform (Vazyme Biotech Co., Nanjing, China). After filtering, the remaining high-quality reads were assembled de novo by using the Trinity program. Certain short reads with overlapping regions were assembled into longer clusters (contigs). Paired-end reads were used to fill scaffolds gaps to obtain unigenes. Based on a BLASTX search (E value < 10−<sup>5</sup> ), functional annotation of these unigenes was performed with the following public databases, including the National Center for Biotechnology Information (NCBI) non-redundant protein (NR) and nucleotide (NT) database, Swiss-Prot protein database, Gene Ontology (GO) database, Cluster of Orthologous Groups (COGs) database, and Kyoto Encyclopedia of Genes and Genomes (KEGG) database. The fragments per kilobase of transcript per million mapped reads (FPKM) method was used to determine the unigene expression. Differentially expressed genes (DEGs) between CK and AL12 samples were identified with log<sup>10</sup> (AL12FPKM/CKFPKM) ≥ 1 and a false discovery rate (FDR) ≤ 0.05. Subsequently, upregulated and downregulated DEGs were conducted by GO and KEGG enrichment analysis, respectively. The RNA sequencing data for this article were submitted to the NCBI Sequence Read Archive (SRA) under accession number SRP132616.

#### Real-Time Quantitative PCR Validation

Real-time quantitative PCR (RT-qPCR) was performed to determine the expression levels of differentially expressed genes of A. lancea cultured in vitro. Elongation factor 1 alpha gene (EF1a) was used as an internal reference (Yuan et al., 2016b). Primers for the selected DEGs are listed in **Supplementary Table S2**. One microgram of total RNA was transcribed into cDNA using HisCRIPT <sup>R</sup> II Q RT SuperMix for qPCR (+gDNA wiper) (Vazyme Biotech Co., Nanjing, China) according to the manufacturer's instruction. Subsequently, RT-qPCR was

conducted using the DNA Engine Opticon 2 Real-time PCR Detection System (Bio-Rad, Hercules, CA, United States). The reaction system consisted of a volume of 20 µL, which included 10 µL of 2 × AceQ qPCR SYBR <sup>R</sup> Green Master Mix (High ROX Premixed) (Vazyme Biotech Co., Nanjing, China), 2 µL of the cDNA template, 0.4 µL of each primer, and 7.6 µL ddH2O. The reaction conditions were 95◦C for 5 min, followed by 40 cycles of 95◦C for 30 s, 60◦C for 30 s, 95◦C for 15 s, 60◦C for 60 s, and 95◦C for 15 s. All assays were performed in triplicate. Relative expression levels for each cDNA sample were calculated by the 2 <sup>−</sup>11Ct method (Wang et al., 2015; Liu et al., 2018).

#### Total Protein Extraction, Two-Dimensional Gel Electrophoresis (2-DE) Separation, Image Analysis, In-Gel Digestion, Protein Identification and Functional Analysis

Plant shoots were harvested at 15 DPI for protein extraction. Protein was extracted using the trichloroacetic acid/acetone precipitation method (Wang et al., 2016). For this purpose, 2 g of shoot material was frozen in liquid nitrogen, ground into a fine powder, mixed with 10 mL cold trichloroacetic acid/acetone buffer (13% (w/v) trichloroacetic acid, 0.07% (v/v) 2-mercaptoethanol in acetone, and kept overnight at −20◦C. After shaking for 1 h, the proteins were separated by centrifugation (10,000 g, 15 min, 4◦C), and washed twice with cold acetone and once with 80% (v/v) acetone. The proteins were then air-dried and dissolved in protein lysis buffer (7 M urea, 2 M thiourea, 4% (w/v) 3-[(3-cholamidopropyl) dimethylam-monio]-1-propanesulphonate (CHAPS), 2% (v/v) IPG buffer, 1% (v/v) phenylmethanesulphonyl fluoride (PMSF), 10 mM Na3VO4, 10 mM NaF, 50 mM glycerophosphate, 5 µg mL−<sup>1</sup> antiprotease, 5 µg mL−<sup>1</sup> trasylol and 5 µg mL−<sup>1</sup> leupeptin). After sonication and centrifugation, the resulting protein extraction was quantified using a commercial dye reagent (Bio-Rad Laboratories, Hercules, CA, United States).

Proteins (900 µg) were separated using immobilized pH gradient (IPG) strips (pH 3–10, nonlinear gradient, 24 cm) and rehydrated for 12 h. First-dimension isoelectric focusing was performed at 20◦C for 1 h at 250 V, 2 h at 1,000 V, 5 h at 10,000 V, and 12 h at 500 V (Chen et al., 2016). The strips were then equilibrated with a buffer containing 6 M urea, 2% (w/v) sodium dodecyl sulphate (SDS), 75 mM Tris-HCl (pH 8.8), 30% (v/v) glycerol, and 1% (w/v) dithiothreitol (DTT). In the second dimension, 12% (w/v) sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) was used, and then coomassie brilliant blue stained (GE Healthcare). Gel images were scanned using an Image Scanner III (GE Healthcare) and analyzed using the Image Mater 2D Platinum 7.0 (GE Healthcare). Each spot was normalized to a relative volume. Quantitative analysis sets were created between each control and AL12-inoculated plants, and each treatment was performed in triplicate.

Differentially expressed spots were selected, manually excised, digested with trypsin (Tatli et al., 2017), and analyzed by UltrafleXtreme TOF/TOF (Bruker Daltonics, Germany). Reflector positive mode was used, with a wavelength of 355 nm and acceleration voltage of 2 kV. MS and MS/MS data were analyzed, and peak lists were generated using FlexAnalysis 3.1 (Bruker Daltonics, Germany). MS and MS/MS analyses were compared to the NCBI green plant protein database using the MASCOT 2.4 search engine (Matrix Science, London, United Kingdom) (Chen et al., 2016; Tatli et al., 2017). Search parameters were as follows: trypsin digestion with one missed cleavage, variable modifications (oxidation of methionine and carbamidomethylation of cysteine), and mass tolerance of a precursor ion and fragment ion at 0.2 Da for +1 charged ions. For all proteins successfully identified by a Peptide Mass Fingerprint and/or MS/MS, a Mascot score greater than 50 (the default MASCOT threshold for such searches) was accepted as significant (P < 0.05) (Tatli et al., 2017).

#### Statistical Analysis

The mean values and standard deviations were calculated using SPSS Statistics 17.0 software (SPSS Inc., Chicago, United States), and the statistical evaluation between two treatments was compared using the independent-samples t-test. Values are the means of three independent experiments. Bars represent standard deviations. Asterisks denote significant differences from the control (t-test; <sup>∗</sup>P < 0.05; ∗∗P < 0.01).

## RESULTS

#### Gilmaniella sp. AL12 Improved Shoot Growth and Sesquiterpenoid Accumulation in A. lancea

The content of β-caryophyllene, zingiberene, β-sesquiphellandrene, caryophyllene oxide, and atractylone in shoots of A. lancea all significantly increased after fungal endophyte inoculation (**Figure 1A**). The shoot dry weight of endophyteinoculated plants was higher than that of endophyte-free A. lancea (**Figure 1B**). And, we analyzed chlorophyll fluorescence parameter of endophyte-inoculated and endophyte-free A. lancea (**Figures 1C,D**). The active photosynthesis II (PSII) reaction center captures of light energy to transform into excitation energy, then converts part of excitation energy into chemical energy, thus promotes carbon assimilation reaction, and the rest of excitation energy dissipates (Li et al., 2005). Although the maximum quantum yield of primary photochemistry (Fv/Fm) of A. lancea was unchanged after endophyte inoculation (**Figure 1C**), the endophyte increased phenomenological fluxes per cross section including absorption flux of photons (ABS/CSm), phenomenological fluxes for trapping (TRo/CSm), and potential electron transport (ETo/CSm) of A. lancea (**Figure 1D**), indicating that the fungal endophyte improved PS II reaction center performance in A. lancea.

#### Gilmaniella sp. AL12 Induced Shoot Transcriptional Changes in A. lancea

We analyzed the transcriptome of A. lancea shoots inoculated or not inoculated with endophytic fungi AL12. A total of 1956 upregulated and 2063 downregulated genes were identified

(**Supplementary Figure S1**). To further elaborate the function of the DEGs, we first conducted GO enrichment analysis (**Supplementary Figure S2**). The most upregulated DEGs belonged to oxidation-reduction-related GO terms. Among the downregulated DEGs, stress response-related GO terms were the most enriched. We further performed KEGG pathway enrichment analysis for the upregulated and downregulated DEGs (**Figure 2**). For the upregulated DEGs, "metabolic pathway" and "biosynthesis of secondary metabolism" were the most enriched KEGG pathways (**Figure 2A**), while "plantpathogen interaction" and "plant hormone signal transduction" were significantly enriched pathways for the downregulated DEGs (**Figure 2B**).

The DEGs were functionally related to known metabolic pathways, secondary metabolic pathways, and regulatory processes using the online software iPath2: interactive Pathways Explorer. And, pathways including transcription, translation, replication and repair, protein folding, and transport tend to be upregulated, while protein degradation pathway tend to be downregulated in shoots of A. lancea after the fungal endophyte inoculation (**Supplementary Table S3**). Compared with non-inoculated plants, in endophyte-inoculated plants, nucleotide metabolism, glycan biosynthesis and metabolism, and lipid metabolism pathways tends to be downregulated (**Supplementary Table S4**). Additionally, the endophytic fungus AL12 upregulated the energy, carbohydrate, amino acid, cofactor and vitamin, and secondary metabolism pathways in shoots of A. lancea (**Supplementary Tables S5–S7**).

#### Gilmaniella sp. AL12 Upregulated Genes Involved in Primary Metabolism

Our results revealed that most DEGs annotated as functioning in photosynthesis and oxidative phosphorylation were globally upregulated after AL12 inoculation (**Supplementary Table S5** and **Figure 3**), including genes annotated as ferredoxin-NADP<sup>+</sup> reductase (FNR), photosystem II oxygenevolving enhancer protein 2 (OEE), ribulose-1,5-bisphosphate

http://pathways.embl.de/iPath2.cgi#

carboxylase/oxygenase activase (RCA), NADH dehydrogenase, and V-type H+-transporting ATPase, among others. However, three genes encoding chlorophyll A/B binding protein (LHCP) were down-regulated.

Endophyte inoculation enhanced plant carbon metabolism, with an upregulation of genes involved in carbohydrate metabolism and transport (**Figure 3**). For instance, most DEGs annotated as phosphate transporter (TPT), amylase, invertase, β-glucosidase (GlcCerase), fructokinase (FRK), glyceraldehyde 3-phosphate dehydrogenase (GAPDH), phosphopyruvate hydratase (enolase), and phosphoenolpyruvate carboxykinase (PEPCK) were upregulated. Interestingly, genes that function in controlling pyruvate and acetyl-CoA biosynthesis were upregulated. For example, AL12 upregulated genes encoding lactate dehydrogenase (LDHD), pyruvate decarboxylase (PDC), and aldehyde dehydrogenase (ALDH), suggesting that it promotes the conversion of lactate into pyruvate and the biosynthesis of acetic acid. Similarly, the expression of genes encoding alcohol dehydrogenase (ADH) and malate dehydrogenase (MDH) was increased after AL12 inoculation, indicating that oxaloacetate (OAA) and acetic acid biosynthesis were improved. Moreover, the endophytic fungus Gilmaniella sp. AL12 upregulated genes encoding MDH and pyruvate dehydrogenase (PDHE), thus contributing to acetyl-CoA biosynthesis and OAA regeneration. Consistent with the improved tricarboxylic acid cycle (TCA cycle), seven genes annotated as malate transporter were up-regulated.

The balance of carbon and nitrogen metabolism is very important for plant metabolism. Most DEGs annotated as functioning in nitrogen metabolism were upregulated after AL12 inoculation (**Figure 3**). These included genes encoding high affinity nitrate transporter (NRT2), glutamate dehydrogenase (GDH), and others. In particular, one gene encoding glutamine synthetase (GS) exhibited a 10-fold increase in expression in response AL12 inoculation. Additionally, most DEGs annotated as functioning in amino acid metabolism were upregulated (**Supplementary Table S6**).

Most DEGs annotated as functioning in lipid metabolism were downregulated after AL12 treatment (**Supplementary Table S4**). These DEGs compass fatty acid elongation, the biosynthesis of unsaturated fatty acids, wax biosynthesis, linoleic acid metabolism, alpha-linolenic acid metabolism, and others. In contrast, most genes involved in fatty acid metabolism, steroid biosynthesis, glycerophospholipid metabolism, ether lipid metabolism, and sphingolipid metabolism were upregulated. Among the upregulated genes, it is worth noting that 32 DEGs encoding phospholipase D (PLD) were upregulated after AL12 inoculation, and eight showed a greater than 10-fold increase in expression (**Supplementary Table S8**).

#### Gilmaniella sp. AL12 Upregulated Genes Involved in Secondary Metabolism

A total of 299 DEGs involved in 18 biosynthesis pathways of secondary metabolites were regulated in A. lancea after AL12 inoculation (**Figure 4**). Among these pathways, DEGs involved in eight biosynthesis pathways of secondary metabolites were significantly regulated in A. lancea after AL12 inoculation (**Supplementary Table S9**). The eight pathways include Phenylpropanoid biosynthesis, Zeatin biosynthesis, Flavonoid biosynthesis, Stilbenoid, diarylheptanoid and gingerol biosynthesis, Limonene and pinene degradation, Flavone and flavonol biosynthesis, Carotenoid biosynthesis, and Diterpenoid biosynthesis. The largest subcategory was phenylpropanoid biosynthesis, followed by zeatin biosynthesis. The upregulated DEGs involved in phenylpropanoid biosynthesis were associated with lignin biosynthesis. Additionally, most of the DEGs involved in zeatin biosynthesis were upregulated,


FIGURE 3 | Differentially expression genes involved in photosynthesis, carbon/nitrogen metabolism, and oxidative phosphorylation in shoots of A. lancea after the fungal endophyte Gilmaniella sp. AL12 inoculation. LHCP, light-harvesting chlorophyll a/b-binding protein; OEE, PSII oxygen-evolving enhancer protein; FNR, Ferredoxin-NADP reductase; RCA, ribulose bisphosphate carboxylase/oxygenase activase; RPLA, Ribose 5-phosphate isomerase; FBPA, fructose-bisphosphate aldolase; SS, starch synthase; SBE, 1,4-alpha-glucan branching enzyme; amylase, PHO, starch phosphorylase; TPT, Glucose-6-phosphate/phosphate-translocator; SUS, sucrose synthase; invertase, β-fructosidases; FRK, fructokinase; GlcCerase, Beta-glucosidase-related glycosidases; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; PEPCK, phosphoenolpyruvate carboxykinase (ATP); PK, pyruvate kinase; PPDK, Pyruvate orthophosphate dikinase; PDHE, pyruvate dehydrogenase; LDHD, lactate dehydrogenase; PDC, pyruvate decarboxylase; ALDH, aldehyde dehydrogenase (NAD+); ADH, alcohol dehydrogenase; MDH, malate dehydrogenase; NRT1, low-affinity nitrate transporter; NRT2, high-affinity nitrate transporter; NR, nitrate reductase; GS, glutamine synthetase; GOGAT, glutamate synthase; GDH, glutamate dehydrogenase; AST, asparagine transaminase; ASNS, asparagine synthetase; sugar, bidirectional sugar transporter; ALMT, aluminum-activated malate transporter; SLAH, S-type anion channel; MATE, MATE efflux family protein; CCR, cytochrome-c reductase; COX, Cytochrome c oxidase; PPases, Inorganic pyrophosphatase.

particularly adenylate dimethylallyl transferase (IPT). Consistent with the increased sesquiterpenoids content in A. lancea shoots (**Figure 1A**), the expression of genes associated with terpenoid backbone biosynthesis and sesquiterpene biosynthesis was increased (**Figure 4** and **Supplementary Figure S3**). Genes encoding farnesyl diphosphate synthase (FDPS) and β-farnesene synthase were increased after AL12 treatment (**Supplementary Figures S3, S4**).

#### Gilmaniella sp. AL12 Downregulated Genes Involved in Plant-Pathogen Interactions

Most DEGs associated with plant-pathogen interaction were downregulated after AL12 inoculation (**Figure 5**). Among the genes related to pathogen-associated molecular pattern (PAMP)-triggered immunity (PTI), DEGs annotated as leucinerich repeat (LRR) receptor-like kinase FLS2, brassinosteroid insensitive 1-associated receptor kinase 1 (BAK1), LRR receptorlike serine/threonine-protein kinase EFR, calcium-dependent protein kinase (CDPK), calmodulin/calcium-binding protein (CaM/CML), mitogen-activated protein kinase kinase kinase 1 (MEKK1), and WRKY29 tended to be downregulated. However, DEGs encoding NADPHox (Rboh) and mitogenactivated protein kinase kinase 1 (MKK1/2) tended to be upregulated in inoculated compared with control plants. Regarding effector-triggered immunity (ETI), the endophytic inoculation particularly decreased the expression of genes encoding the disease resistance proteins RPM1, RPS2, and RPS5. Additionally, defense genes such as WRKY and pathogenesisrelated protein1 (PR1) also tended to be repressed after AL12 inoculation.

#### Gilmaniella sp. AL12-Regulated Genes Involved in Signaling

Our result shows that DEGs involved in auxin, cytokinin, and ethylene biosynthesis tended to be upregulated (**Figure 6**), and DEGs involved in cytokinin and ethylene signal transduction also tended to be upregulated (**Figure 6**). In total, 99 TFcoding genes were differentially expressed between CK and AL12 samples (**Supplementary Figure S5**). These DEGs were

divided into 17 families, including WRKY, APETALA2/Ethylene-Response Factors (AP2/ERF), MYB, basic Helix-Loop-Helix (bHLH), DELLA, Heat stress, Trilelix, GATA, and NAC, among others. Of these families, DEGs encoding MYB, Trilelix, and NAC transcription factor tended to be upregulated, while many genes encoding AP2/ERF, WRKY, DELLA, Heat stress and GATA were downregulated.

#### RT-qPCR Validation of Gene Expression

To further validate the transcriptomic results, 15 transcripts related to primary metabolism, secondary metabolism, and defense were selected for RT-qPCR analysis (**Supplementary Table S2** and **Figure 7**). As shown in **Figure 7**, the results of the RT-qPCR analysis were largely consistent with the RNA sequencing analysis, supporting the high quality of the RNA sequencing datasets. In particular, the expression of genes annotated as GAPDH, GS, and β-caryophyllene synthase CPS1 markedly increased after endophytic fungus inoculation (**Figure 7**).

#### Shoot Proteome Changes by Gilmaniella sp. AL12 Inoculation Involved Diverse Biological Processes

To identify proteins that responded to the endophyte inoculation, we employed 2-DE to compare the shoot protein profiles of the CK and AL12 samples. A total of approximately 2136 protein spots were separated (**Supplementary Figure S6**). After optimization of the 2-DE gels, approximately 129 differentially expressed proteins with at least a two-fold change were identified (**Supplementary Figure S6**). Finally, 125 proteins were successfully identified via MALDI-TOF MS/MS (**Supplementary Figure S7** and **Supplementary Table S10**). These proteins were classified into nine functional categories (**Supplementary Figure S7**), including "energy metabolism," "carbohydrate metabolism," "amino acid metabolism," "lipid metabolism," "defense/stress," "plant secondary metabolism," "signal transduction," "transcription and translation," and "cell growth/division."

The expression of energy metabolism-related proteins, including ribulose bisphosphate carboxylase, ferredoxin-nitrite reductase, chloroplast oxygen-evolving enhancer protein, and phosphoenolpyruvate carboxylase, were upregulated (**Table 1**). Regarding carbohydrate metabolism, glyceraldehyde-3 phosphate dehydrogenase and NAD-dependent glyceraldehyde-3-phosphate dehydrogenase were up-regulated, whereas carbon catabolite repressor protein was down-regulated (**Table 1**). Four plant secondary metabolism-associated proteins were upregulated, including tropinone reductase, Type III polyketide synthase and cytochrome P450 CYP76F12 (**Table 1**).

Among the 30 proteins involved in signal transduction, 17 were upregulated. These proteins included membrane-associated

protein, protein kinase, protein phosphatase, abscisic acid (ABA) responsive protein, gibberellin (GA) biosynthesis related proteins, indoleacetic acid (IAA) biosynthesis related proteins, and a few transcription factors (**Supplementary Table S10**). The other 13 downregulated proteins incorporated pentapeptide, ethylene biosynthesis-related protein, and ubiquitin conjugate factors, among others (**Supplementary Table S10**). Additionally, AL12-regulated proteins involved in defense/stress response, amino acid and lipid metabolism, transcription and translation, cell growth, and other function were differentially regulated (**Supplementary Table S10**).

#### DISCUSSION

The pharmaceutical value of medicinal plants relies on the accumulation of active pharmaceutical ingredient, and guaranteeing the yield and quality of these herbs is the main challenge (Yang et al., 2012). Endophytes have been proven to exert multiple effects on their host plants, including plant growth promotion, secondary metabolite biosynthesis promotion, stress resistance enhancement (Ludwig-Müller, 2015; Yang et al., 2016; Dastogeer et al., 2018). The advantages provided by the endophyte fungus AL12 on the host plant A. lancea were an improved plant biomass and increased sesquiterpenoids content (Yuan et al., 2016b). In this study, transcriptomics and proteomics were employed to analyze how endophytic fungi affect the regulation of transcription and translation in A. lancea. Compared to the reported endophyte associations (Doehlemann et al., 2008; Kawahara et al., 2012; Morán-Diez et al., 2012; Perazzolli et al., 2012; De Cremer et al., 2013; Zouari et al., 2014; Adolfsson et al., 2017; Chen W. et al., 2017; Dinkins et al., 2017; Hao et al., 2017; Bajaj et al., 2018), the endophyte Gilmaniella sp. AL12 resulted in 2.7% differential gene expression, thus demonstrating a greater impact on their host than most other endophytes (**Supplementary Table S11**). During the plant life cycle, a dynamic trade-off between growth and defense is necessary for plant resource assignment in response to multiple developmental cues and environmental stimuli (Hou et al., 2013). In contrast to the common downregulated plant-pathogen interactions and most phytohormone signaling events, Gilmaniella sp. AL12 upregulated genes involved in primary and secondary metabolism (**Figures 2–4**, **8**). Consistent with the improved cytokinin content of A. lancea after AL12 inoculation (Yuan et al., 2016b), DEGs involved in cytokinin biosynthesis and signal transduction both tended to be upregulated (**Figure 6**), which may induce cell division, accelerate chlorophyll biosynthesis, and delay leaf senescence (Cortleven and Schmülling, 2015). In addition, ethylene induced by the endophyte mediates sesquiterpenoid biosynthesis in A. lancea (Yuan et al., 2016a). Based on these results, we propose that colonization by the endophytic fungus Gilmaniella sp. AL12 will shift plant metabolism from defense to growth, allowing the host plant to utilize limited energy for carbon assimilation and, thus, to achieve an increased biomass and sesquiterpenoid content in A. lancea.

It is acknowledged that plant primary metabolism and secondary metabolism required large amounts of energy. One way to support the increased energy demands is to enhance the carbon assimilation efficiency (Rozpa¸dek et al., 2015, 2019). We have previously observed that the net photosynthesis rate and contents of chlorophyll, rubisco, and soluble carbohydrate in A. lancea increase after AL12 inoculation (Yuan et al., 2016b). In the present study, phenomenological fluxes per cross section of the host plant was improved after AL12 inoculation (**Figure 1D**), indicating the improvement of PS II reaction center performance in A. lancea. As shown by the transcriptome and proteome results, endophyte inoculation would enhance NADPH production and CO<sup>2</sup> assimilation (**Table 1**, **Figures 3**, **8** and **Supplementary Figure S8**), thus explaining the positive effects of endophyte inoculation on plant photosynthesis. Similarly, Dactylis glomerate inoculated with Epichloë typhina, Medicago truncatula, and Populus alba inoculated with AMF all displayed a similar photosynthesis performance with improved plant light-driven energy production efficiency, Rubisco content, CO<sup>2</sup> assimilation and PS II photochemistry efficiency (Lingua et al., 2008; Aloui et al., 2009; Cicatelli et al., 2010, 2012; Rozpa¸dek et al., 2015, 2019). Although Gilmaniella sp. AL12, Epichloë typhina and AMF belong to distinct fungal species, they all provide a benefit, such as improving similar photosynthesis apparatuses of their host plant. Plant photosynthesis and

respiration processes exist with each other interdependently and share adenosine diphosphate (ADP) and NADP<sup>+</sup> or NAD+. Additionally, O<sup>2</sup> produced by photosynthesis can be utilized by the respiration process, and CO<sup>2</sup> produced by respiration can also be assimilated by photosynthesis. As speculated based on the results of this study (**Table 1** and **Figure 3**), NADPH and O<sup>2</sup> production by chloroplasts, as well as NADH dehydrogenation and triphosadenine (ATP) production by mitochondria, were improved after AL12 inoculation, supporting energy storage for plant metabolism. The TCA cycle mainly converts pyruvate to malate to produce ATP, incorporating citric acid biosynthesis, oxidation and decarboxylation, and OAA regeneration (Fernie et al., 2004). Epichloë typhina has been shown to markably strengthen NADPH-MDH enzyme activity in its host Dactylis glomerate (Rozpa¸dek et al., 2015). Likewise, the endophytic fungus Gilmaniella sp. AL12 upregulated genes encoding MDH and pyruvate dehydrogenase (PDHE) (**Figure 3**), thus contributing to acetyl-CoA biosynthesis and OAA regeneration (**Figure 8**). Acetyl-CoA and pyruvate biosynthesis were improved in A. lancea after endophyte inoculation (**Figure 8**), indicating that more precursor substances and energy were available for plant secondary metabolite biosynthesis (Nema et al., 2013). As the precursor substances of terpenoid biosynthesis (Chen F. et al., 2017), pyruvate and acetyl-CoA content increased after AL12 inoculation (Yuan et al., 2016b). Additionally, in contrast to the general downregulation of fatty acid biosynthesis (**Supplementary Table S4**), genes associated with terpenoid biosynthesis were upregulated after AL12 inoculation (**Supplementary Figure S3**), indicating that a greater amount of acetyl-CoA would be diverted into terpene biosynthesis.

protein RPS5; WRKY29, transcription factor WRKY29; PR1, pathogenesis-related protein1.

Plant terpenoids are derived from isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP) synthesized through the mevalonic acid (MVA) and the 2-C-methyl-D-erythritol-4-phosphate (MEP) pathway (Vattekkatte et al., 2018). The different molecular rates of isopentenyl diphosphate and dimethylallyl diphosphate condensation lead to geranyl diphosphate (GPP) for monoterpene, farnesyl diphosphate (FPP) for sesquiterpene and geranylgeranyl diphosphate (GGPP) for triterpene biosynthesis (Dudareva et al., 2013; Vattekkatte et al., 2018). Consistent with the increased sesquiterpenoids content in A. lancea shoots (**Figure 1A**), DEGs associated with terpenoid backbone biosynthesis and sesquiterpene biosynthesis were increased (**Supplementary Figure S3**). Unfortunately, we did not detect any spot identified associated with terpene backbone biosynthesis or sesquiterpenoids biosynthesis in proteome (**Table 1**). Considering that there is no database of A. lancea or related genus in NCBI, we identified proteins using NCBI green plant database. Thus, we considered that maybe some proteins related to secondary metabolism may not identified exactly. Yet despite that, comparative transcriptomics showed that genes related to terpene backbone biosynthesis and sesquiterpenoids biosynthesis tended to be upregulated in shoots of A. lancea after the endophyte inoculation (**Supplementary Figure S3**). In this study, high-quality reads were assembled de novo by using the Trinity program. The results showed that genes annotated as β-farnesene synthase and β-caryophyllene synthase were upregulated after AL12 inoculation, as shown by the transcriptome (**Supplementary Figure S4**) and RTqPCR results (**Figure 7**). It is known that one terpene synthase can catalyze formation more than one terpene (Dudareva et al., 2013). For example, TPS21 and TPS11 from Arabidopsis

TABLE 1 | Identification of differentially expressed proteins related to energy, carbohydrate, and secondary metabolism in shoots of A. lancea after endophyte inoculation.


The spot number corresponds to the number shown in Supplementary Figure S6. MS, mascot score; SC, sequences coverage; MW, molecular weight; pI, isoelectric points; FC, fold change.

FIGURE 8 | Simplified schematic summary of metabolic changes in shoots of A. lancea after the fungal endophyte Gilmaniella sp. AL12 inoculation. DEGs with significant differences in endophyte inoculated plants are shown as colored arrows (upregulated DEGs in red and downregulated DEGs in blue). The yellow ellipse indicates the phosphate translocator (TPT). Letters in green blocks represent precursor substances of terpenoid biosynthesis. C3 cycle, Benson-Calvin cycle; GS/GOGAT, glutamine synthetase (GS)/glutamate synthase (GOGAT); TCA cycle, tricarboxylic acid cycle; PS I, photosystem I; PS II, photosystem II; 3-PGA, glycerate-3-phosphate; OAA, oxaloacetate; 2-OG, 2-oxoglutarate; FNR, ferredoxin-NADP reductase; OEE, photosystem II oxygen-evolving enhancer protein.

thaliana are responsible for the biosynthesis of more than 20 sesquiterpenoids in flower fragrance (Tholl et al., 2005). Additionally, OkBCS of Ocimum can catalyze the production of β-caryophyllene and α-humulene (Jayaramaiah et al., 2016). The main sesquiterpenoids of A. lancea incorporate β-caryophyllene, zingiberene, β-sesquiphellandrene, caryophyllene oxide, hinesol, β-eudesmol and atractylone (Yuan et al., 2016b). Considering the multifunction of sesquiterpene synthase, we propose that β-farnesene synthase and β-caryophyllene synthase are crucial terpene synthases induced by AL12 to supply higher sesquiterpenoid contents in A. lancea. Further experiments will include the functional characterization of these sesquiterpene synthases, and how the fungal endophyte regulate genes coding for these sesquiterpene synthases.

The accumulation of secondary metabolites is regulated by cross-talking signaling cascades. Our previous studies have demonstrated that the endophyte AL12 activates multiple signals to induce plant sesquiterpenoids accumulation (Wang et al., 2011; Ren and Dai, 2012, 2013; Yuan et al., 2016a). In addition, mannan-binding lectin has been proposed to recognize mannan of Gilmaniella sp. AL12, thus causing signal transduction (Chen et al., 2016). It is worth noting that 32 DEGs encoding PLD tended to be upregulated after AL12 inoculation in the present study (**Supplementary Table S8**). PLD can hydrolyse phospholipids into phospholipid acid, which acts as an important molecular signal mediating heat stress-induced triterpene ganoderic acid biosynthesis in Ganoderma lucidum (Liu et al., 2017). These results indicate that phospholipid acid may be an essential signaling molecule in A. lancea after AL12 inoculation. Additionally, metabolomics and transcriptomics are required to study whether PLD or phospholipid signaling in A. lancea are activated and whether they mediate sesquiterpenoid biosynthesis after AL12 inoculation, which will help to understand their roles in inducing the biosynthesis of the secondary metabolites.

In contrast to the enhanced primary and secondary metabolism most genes involved in the plant-pathogen interaction were downregulated (**Figures 2**, **5**). Plants evolve pattern recognition receptors (PRRs) to recognize evolutionarily conservative PAMPs, triggering the host MAPK cascade (**Figure 5**). Then, MAPK activates plant hormone signal transduction to integrate diverse aspects of PTI (Cui et al., 2015; Pandey et al., 2016). These optional signaling events will give rise to various immune responses (Akira and Hemmi, 2003). Furthermore, effectors produced by the fungus are also known to inhibit MAPK cascade signaling or block plant-resistant protein expression, thus suppressing plant basic immune response PTI (Zhang et al., 2012; Meng and Zhang, 2013). Fungi deploy these co-evolved effectors to modify host plant cell processes or to establish biotrophic or other types of symbiotic relationship with the host (Le Fevre et al., 2015). In turn, fungal effectors can be recognized by plant disease resistance proteins (R proteins) and thus trigger plant ETI, which usually causes the plant hypersensitive response (HR) (Jones and Dangl, 2006; Liu et al., 2007). With endophytic fungal AL12 inoculation, genes annotated as plant PRRs, including LRR receptor-like kinase FLS2, BAK1 and EFR, tended to be downregulated (**Figure 5**), indicating a weakened PAMP recognition ability of the host plant. Furthermore, genes involved in the response of PAMPs or effectors were also repressed (**Figure 5**), suggesting that endophyte inoculation suppressed host immune reactions. As shown in our previous study (Yang and Dai, 2013), AL12 could successfully colonize the leaves of A. lancea, resulting in an altered hyphal morphology such as smooth and unbranched hyphae. In addition, slightly pointed hyphal tips of the endophyte also appears similar as the ends of a nematode (Yang and Dai, 2013). Considering that genes related to PTI and ETI of A. lancea were downregulated after AL12 inoculation (**Figure 5**) and that the hyphae of the endophyte were altered morphologically (Yang and Dai, 2013), we deduced that either masked fungal PAMPs were not detected by the host plant, or that the plant defense response was suppressed by the host or the endophyte (Dupont et al., 2015), potentially contributing to the compatible association of Gilmaniella sp. AL12 with A. lancea.

Given the apparent downregulation of genes involved in the plant-pathogen interaction, endophyte inoculation would enhance the pathogen susceptibility of the host plant. However, this is not the actual fact. Although plant defenserelated genes of Perennial ryegrass were down-regulated after Epichloë festucae var. lolii infection, its resistance to fungal pathogens was enhanced due to the production of phenolic compounds (Tian et al., 2008; Panka et al., 2013 ´ ; Dupont et al., 2015). Similarly, Gilmaniella sp. AL12 upregulated genes involved in the biosynthesis of phenylpropanoids and terpenes (**Figure 4**). Terpenoids are frequently used as phytoalexins, the accumulation of which are enhanced by biotic or abiotic stress (Schmelz et al., 2014; Vaughan et al., 2015). The sesquiterpenoid

phytoalexins such as gossypol, hemigossypolone and heliocides can provide defense mechanisms against pathogens and herbivores of cotton (Yang C. Q. et al., 2013). Additionally, β-caryophyllene can defend against Helicoverpa armigera and Pseudomonas syringae pv. tomato DC3000 (Huang et al., 2012; Singh et al., 2014). Similarly, the monoterpene S-limonene and triterpene momilactones or oryzalexins function as defensive metabolites against Magnaporthe oryzae or Magnaporthe grisea (Toyomasu et al., 2008; Chen et al., 2018). Therefore, although defense-related genes of A. lancea were downregulated after Gilmaniella sp. AL12 inoculation, the enhanced synthesis of phenylpropanoids and terpenoids may function as defensive strategy against other microbial pathogens or herbivores, functioning as a possible competitive exclusion tactic of the endophyte.

In this study, shoots of A. lancea were sterile, and were appropriate for investigating the effect of the fungal endophyte on plantlets without the distraction of other biotic or abiotic factors, thus helping to understand plant-endophyte interactions. The above results showed that the fungal endophyte Gilmaniella sp. AL12 weakened the plant immune response, which might contribute to its successful and stable colonization of the host plant (**Figure 9**). The decreased plant immunity resulted in large amounts of energy for plant primary or secondary metabolism. Moreover, the fungal endophyte improved photosynthesis, carbon fixation, carbohydrate transformation, and the conversion from pyruvate to acetyl-CoA in the host plant, indicating that the endophyte promoted the accumulation of precursors for terpenoid biosynthesis. In addition, the fungal endophyte specifically induced sesquiterpenoid biosynthesis by regulating genes involved in sesquiterpenoid biosynthesis and related signaling events in A. lancea. Thus, we propose that the fungal endophyte-plant association increased production (biomass and sesquiterpenoid content) by increasing the source (photosynthesis), expanding the sink (glycolysis and the TCA cycle), and enhancing metabolic flux (sesquiterpenoids biosynthesis-related genes) in A. lancea (**Figure 9**). Consistent with the "double promotion" in this study, the endophytic fungi AL12 promotes plant growth and sesquiterpenoid accumulation within two years of growth in field experiments. Further studies will be conducted to investigate whether pre-inoculation with the fungal endophyte Gilmaniella sp. AL12 regulate metabolism and immunity of A. lancea in the field experiments, which will help to understand plant-endophyte interactions, and will contribute to the application of the fungal endophyte in cultivating of A. lancea.

FIGURE 9 | Simplified schematic summary of the "double-promotion" effect on A. lancea bring by the fungal endophyte Gilmaniella sp. AL12. The fungal endophyte Gilmaniella sp. AL12 weakened the plant immune response of A. lancea, thus may contributing to plant-endophyte beneficial interaction. Additionally, endophyte-plant association increased production (biomass and sesquiterpenoid content) by increasing the source (photosynthesis), expanding the sink (glycolysis and the TCA cycle), and enhancing metabolic flux (sesquiterpenoids biosynthesis-related genes) in A. lancea.

Medicinal plants are rich in active compounds such as artemisinin (Sharma and Agrawal, 2013) and ginseng saponin (Wei et al., 2018), which are important sources of modern drugs. You-You Tu, the mother of artemisinin, was awarded Nobel Prize in medicine in 2015. The development of the medicinal plants gradually becomes a focused issue, and medicinal plantsendophytes interactions have received much attentions. A review of the medicinal plant microbiome has shown that the secondary metabolite content of medicinal herbs depends on where they are cultivated, which can be partly ascribed to different rhizospheric or endophytic microbes associated with their cultivation location (Köberl et al., 2013; Huang et al., 2018). Plantlets of A. lancea grown in the Maoshan area of southeast China represent the geo– authentic medicinal plant, with a much greater sesquiterpenoid content and diversity than A. lancea in other cultivation area (Zhou et al., 2018). The endophytic fungus Gilmaniella sp. AL12 was isolated from the stem of the geo–authentic A. lancea, and the compatible association contributes some benefits to the host plant, including the promotion of plant growth and sesquiterpenoid content, and the prevention of root rot (Wang et al., 2011; Ren and Dai, 2012, 2013; Wang et al., 2012; Yuan et al., 2016a,b; Ren et al., 2017). Whether Gilmaniella sp. AL12 is a key member of the core microbiome of geo–authentic A. lancea remain unanswered. In this regard, our previous studies have shown that prioritizing endophytic Gilmaniella sp. AL12 inoculation enhances the diversity and size of phyllospheric microbial communities of A. lancea by increasing the soluble sugar content in the rhizosphere (Yang T. et al., 2013). Given that the endophyte provided physiological and biochemical benefits to the host and affected host phyllospheric microbial communities, we more broadly propose a beneficial association of endophytic Gilmaniella sp. AL12 with A. lancea as a potential model for endophytic fungi-medicinal herb interaction. The endophyte-A. lancea association contributes to medicinal herb cultivation and helps to further clarify plant-endophyte interactions.

#### CONCLUSION

In summary, this study showed that the fungal endophyte Gilmaniella sp. AL12 weakened the plant immune response

#### REFERENCES


of A. lancea, thus may contributing to the beneficial plant-endophyte interaction. Additionally, endophyte-plant association increased production (biomass and sesquiterpenoid content) by increasing the source (photosynthesis), expanding the sink (glycolysis and the TCA cycle), and enhancing metabolic flux (sesquiterpenoids biosynthesis-related genes) in A. lancea. This study revealed the regulation of Gilmaniella sp. AL12 on plant metabolism and related regulatory processes in shoots of A. lancea at the transcriptional and translational level. This study provides a theoretical basis for medicinal herb cultivation and helps to further clarify plant-endophyte interactions.

#### DATA AVAILABILITY

The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

#### AUTHOR CONTRIBUTIONS

C-CD and JY designed the experiments, analyzed the data, and wrote the manuscript. P-XC and M-JT helped extract plant RNA and protein. WZ, KS, and XL helped supervise the manuscript writing. All the authors read and approved the final manuscript.

#### FUNDING

This work was supported by the National Key Research and Development Program of China (Grant No. 2017YFD0800705).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.01208/full#supplementary-material




sesquiterpenoid accumulation in Atractylodes lancea. Plant Physiol. Bioch. 130, 473–481. doi: 10.1016/j.plaphy.2018.07.016


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

Copyright © 2019 Yuan, Zhang, Sun, Tang, Chen, Li and Dai. 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.

# Diazotrophic Paenibacillus beijingensis BJ-18 Provides Nitrogen for Plant and Promotes Plant Growth, Nitrogen Uptake and Metabolism

Yongbin Li, Yunlong Li, Haowei Zhang, Minyang Wang and Sanfeng Chen\*

State Key Laboratory for Agrobiotechnology and College of Biological Sciences, China Agricultural University, Beijing, China

#### Edited by:

Marco Catoni, University of Birmingham, United Kingdom

#### Reviewed by:

Hassan Etesami, University of Tehran, Iran Christopher Peter Chanway, University of British Columbia, Canada Akshit Puri, University of British Columbia, Canada, in collaboration with reviewer CC

> \*Correspondence: Sanfeng Chen chensf@cau.edu.cn

#### Specialty section:

This article was submitted to Plant Microbe Interactions, a section of the journal Frontiers in Microbiology

Received: 21 January 2019 Accepted: 03 May 2019 Published: 29 May 2019

#### Citation:

Li Y, Li Y, Zhang H, Wang M and Chen S (2019) Diazotrophic Paenibacillus beijingensis BJ-18 Provides Nitrogen for Plant and Promotes Plant Growth, Nitrogen Uptake and Metabolism. Front. Microbiol. 10:1119. doi: 10.3389/fmicb.2019.01119 Diazotrophic bacteria can reduce N<sup>2</sup> into plant-available ammonium (NH<sup>4</sup> <sup>+</sup>), promoting plant growth and reducing nitrogen (N) fertilizer requirements. However, there are few systematic studies on the effects of diazotrophic bacteria on biological N<sup>2</sup> fixation (BNF) contribution rate and host plant N uptake and metabolism. In this study, the interactions of the diazotrophic Paenibacillus beijingensis BJ-18 with wheat, maize, and cucumber were investigated when it was inoculated to these plant seedlings grown in both low N and high N soils, with un-inoculated plants as controls. This study showed that GFP-tagged P. beijingensis BJ-18 colonized inside and outside seedlings, forming rhizospheric and endophytic colonies in roots, stems, and leaves. The numbers of this bacterium in the inoculated plants depended on soil N levels. Under low N, inoculation significantly increased shoot dry weight (wheat 86.1%, maize 46.6%, and cucumber 103.6%) and root dry weight (wheat 46.0%, maize 47.5%, and cucumber 20.3%). The <sup>15</sup>N-isotope-enrichment experiment indicated that plant seedlings derived 12.9–36.4% N from BNF. The transcript levels of nifH in the inoculated plants were 0.75–1.61 folds higher in low N soil than those in high N soil. Inoculation enhanced NH<sup>4</sup> <sup>+</sup> and nitrate (NO<sup>3</sup> <sup>−</sup>) uptake from soil especially under low N. The total N in the inoculated plants were increased by 49.1–92.3% under low N and by 13–15.5% under high N. Inoculation enhanced activities of glutamine synthetase (GS) and nitrate reductase (NR) in plants, especially under low N. The expression levels of N uptake and N metabolism genes: AMT (ammonium transporter), NRT (nitrate transporter), NiR (nitrite reductase), NR, GS and GOGAT (glutamate synthase) in the inoculated plants grown under low N were up-regulated 1.5–91.9 folds, but they were not obviously changed under high N. Taken together, P. beijingensis BJ-18 was an effective, endophytic and diazotrophic bacterium. This bacterium contributed to plants with fixed N2, promoted plant growth and N uptake, and enhanced gene expression and enzyme activities involved in N uptake and assimilation in plants. However, these positive effects on plants were regulated by soil N status. This study might provide insight into the interactions of plants with beneficial associative and endophytic diazotrophic bacteria.

Keywords: diazotroph, <sup>15</sup>N isotope enrichment, biological N<sup>2</sup> fixation, colonization, GFP, wheat, maize, cucumber

## INTRODUCTION

fmicb-10-01119 May 27, 2019 Time: 17:4 # 2

Biological nitrogen (N) fixation (BNF) is the major natural process through which atmospheric N<sup>2</sup> is reduced to bioavailable NH<sup>4</sup> <sup>+</sup>, providing a large amount of natural N into cultivated agricultural systems (Galloway et al., 2008). In addition to symbiotic N2-fixing Rhizobia associated with legumes, the non-symbiotic diazotrophic bacteria are also important contributors to the N nutrition of non-legumes (Gupta et al., 2006). It is estimated that the microbial N accounts for roughly 30–50% of the total N in crop fields (Liu et al., 2017). The non-symbiotic diazotrophic bacteria are highly diverse and associated with plants in different ways. Some bacteria live in the rhizosphere and are designated rhizobacteria (Kloepper and Beauchamp, 1992). Herbaspirilla seropedicae strain Z67 mainly colonize on the riceroot surface and are usually called associative diazotrophic bacteria (Monteiro et al., 2012). Paenibacillus polymyxa WLY78 live inside the plant without causing damage and are classified as endophytic diazotrophic bacteria (Hao and Chen, 2017). Endophytic diazotrophic bacteria may have an advantage over associative diazotrophic bacteria and rhizobacteria, since they live within plant tissues where better niches are established for N<sup>2</sup> fixation and assimilation of fixed N<sup>2</sup> by the plant (Reinhold-Hurek and Hurek, 1998, 2011).

The well-known associative and endophytic diazotrophic bacteria include Azospirillum (Boddey et al., 1986), Azoarcus (Hurek et al., 2002), Burkholderia (Baldani et al., 2000), Enterobacter (Magnani et al., 2010), Gluconacetobacter (James et al., 2001), Herbaspirillum (Boddey et al., 1995). BNF quantification experiments show that associative and endophytic bacteria can fix N<sup>2</sup> in plant tissues with higher efficiency (Carvalho et al., 2014). G. diazotrophicus inoculation enhanced sugarcane yield by providing 50–80% N from BNF (Boddey et al., 1995). It is estimated that an 18–28% of plant N derives from BNF of endophytic Enterobacter sp. strain (Mirza et al., 2001). Diazotrophic bacteria present in the mucilage of aerial roots contribute 29–82% of the N nutrition of Sierra Mixe maize (Van Deynze et al., 2018).

Although the positive effects of diazotrophic bacteria on plants are observed, little is known about plant response to inoculation with diazotrophic bacteria. Plant N metabolism is a complex process requiring some key enzymes. The plant genes NR (nitrate reductase), NiR (nitrite reductase), GS (glutamine synthetase) and GOGAT (glutamate synthase) play very important roles in N metabolism (Bloom et al., 1992; Lea and Miflin, 2003). The plant genes AMT (ammonium transporter) (Bloom et al., 1992) and NRT (nitrate transporter) (Sugiura et al., 2007) are involved in N uptake. It is shown that some endophytic fungi affect expression of N metabolism of plants (Yang et al., 2014).

Paenibacillus beijingensis BJ-18, isolated from wheat rhizosphere, was a N2-fixer (Wang et al., 2013). Inoculation with P. beijingensis BJ-18 promoted the growth of tomato seedlings (Xie et al., 2016) and increased wheat yield by 26.9% in field experiment (Shi et al., 2016), suggesting that this bacterium promotes plant growth. It was generally recognized that plant growth-promoting bacteria (PGPB) promoted plant growth by direct mechanisms (e.g., N fixation, phosphate solubilization, sequestering iron) and indirect mechanisms [e.g., indole-3-acetic acid (IAA), cytokinins, gibberellins] (Glick, 2012). However, the mechanisms utilized by P. beijingensis BJ-18 to promote plant growth were not clear. In this study, we investigated the colonization pattern and contributions of N<sup>2</sup> fixation by P. beijingensis BJ-18 to plants, and the plant responses (N uptake and metabolism processes) to the infection.

#### MATERIALS AND METHODS

#### Bacteria Strains and Growth Conditions

Paenibacillus beijingensis BJ-18 (accession number: JN873136), isolated from wheat rhizosphere, is a novel species with N2-fixing ability [1043 ± 12.9 nmol C2H<sup>4</sup> (mg protein h)−<sup>1</sup> ] (Wang et al., 2013). This bacterium has multiple antagonistic activities against plant pathogens and produces IAA (24.95 µg mL−<sup>1</sup> ) (Xie et al., 2016). The bacterial suspension of P. beijingensis BJ-18 used in inoculation was prepared as follows. The P. beijingensis 1–18 cells were cultured overnight in Luria-Bertani (LB) broth at 30◦C and 180 rpm, and then cells in the logarithmic growth phase were harvested by centrifugation and finally the pellet was suspended with sterile normal saline (0.89% w/v NaCl in double distilled water) to the final concentration at 10<sup>8</sup> cells mL−<sup>1</sup> .

#### Colonization of GFP-Tagged P. beijingensis BJ-18 in Wheat, Maize, and Cucumber Tissues

The recombinant plasmid pGFP300 carrying gfp gene (Hao and Chen, 2017) was introduced into P. beijingensis BJ-18 by electrotransformation (Zhang et al., 2013), yielding GFP-tagged P. beijingensis BJ-18. And the physiological ability of GFP-tagged P. beijingensis BJ-18 was not changed, compared with wild-type P. beijingensis BJ-18 (data not published). The GFP-tagged P. beijingensis BJ-18 suspension was obtained as described above. Plump seeds of wheat "Jimai 22" (Shandong Runfeng Seed Industry Co., Ltd), maize "Zhengdan 958" (Henan Shangke Seed Co., Ltd.) and cucumber "Zhongnong 8" (Beijing Shengfeng Garden Agricultural Technology Co., Ltd) were surface-disinfected with 10% sodium hypochlorite for 10 min, followed by rinsing with sterilized water three times, and grown on the sterile petri dishes containing moist filter papers in darkness at room temperature (25◦C) for 3–5 days, respectively. These plant seedlings had two treatments: inoculation with GFP-tagged P. beijingensis BJ-18 (E+) and mock inoculation (E-). For inoculation, the plant seedlings were soaked in bacterial suspension (10<sup>8</sup> cells mL−<sup>1</sup> ) of GFP-tagged P. beijingensis BJ-18 for 30 min to facilitate colonization. For mock inoculation, the plant seedlings were soaked in sterilized deionized water for 30 min. The germinated wheat, maize and cucumber seeds were, respectively, sown in sterile flask (3 seeds per flask, 6 cm in diameter and 10 cm in height) containing 100 mL 1/2 × Murashige and Skoog semisolid agar medium (Prod

No: M519, PhytoTechnology Laboratories, Shawnee Mission, United States) (Murashige and Skoog, 1962). Then these seedlings were grown in the light growth chamber (27◦C, 70% humidity and 16 h day/8 h night, with light at 250 µmol m−<sup>2</sup> s −1 ). The GFP-tagged P. beijingensis BJ-18 in plant tissues was observed at 2 weeks after inoculation by confocal laser scanning microscopy (CLSM, Olympus FluoViewTM FV1000 confocal microscope, Olympus Corporation, Tokyo, Japan). These images were collected using FV10-ASW software (03.01.02.02, Olympus Europa Holding GmbH, Hamburg Germany) and processed in Adobe Photoshop CC 2015 and Adobe Illustrator CS6 (Adobe, San Jose, CA, United States).

#### Plant Culture and Collection

Seedling growth assays were performed in plastic pots (diameter of 35 cm; height of 25 cm) filled with 5 kg non-sterile soil which was top soil (0–20 cm depth) taken from the Shangzhuang Experimental Station of China Agricultural University, Beijing, China (40◦ 080 12.15<sup>00</sup> N, 116◦ 100 44.83<sup>00</sup> E, 50.21 m above sea level). The soil was low N-content sandy loam (Nmin, 7.8 mg kg−<sup>1</sup> ; Olsen-P, 7.3 mg kg−<sup>1</sup> ; NH4OAc-K, 115.8 mg kg−<sup>1</sup> ; O.M., 7.2 g kg−<sup>1</sup> ; pH 7.7; E.C., 0.4 dS m−<sup>1</sup> ). The soil was air-dried, crushed, and screened by a 2 mm sieve to remove debris and reduce heterogeneity for cultivating plants: wheat, maize and cucumber. Before planting, P (Na2HPO4) and K (KCl) fertilizer were applied to soil as base fertilizers at amounts of 50 mg P and 17 mg K per kg soil, respectively, based on the recommendation by Ke et al. (2018). The microelements were not applied to soil during plant growth.

The experimental design was randomized with factorial arrangement (a 2 × 2 factorial design) in three replications with bacterial factor at two levels and N factor at two levels. Three different plants (wheat, maize, and cucumber) were chosen to obtain an objective conclusion. Therefore, the experiments had twelve treatments.

The seeds (wheat, maize, and cucumber) were germinated as described above. After germination, vigorous and homogenous seedlings were chosen for transplanting into plastic pots. These plant seedlings had two treatments: inoculation with P. beijingensis BJ-18 (E+) and mock inoculation (E-). For inoculation, the plant seedlings were soaked in bacterial suspension (10<sup>8</sup> cells mL−<sup>1</sup> ) of P. beijingensis BJ-18 for 30 min to facilitate colonization. For mock inoculation, the plant seedlings were soaked in sterilized deionized water. Then the inoculated seedlings and un-inoculated seedlings were, respectively, transplanted into pots (cucumber: 4 seedlings per pot; maize: 2 seedlings per pot; wheat: 4 hills per pot and 10 seedlings per hill). On day 7, 80 ml of the bacterial suspension was applied to pot containing inoculated seedlings and 80 mL of sterile water was applied to pot containing non-inoculated seedlings. Each of inoculation and mock inoculation treatments had three replicates.

There were two levels of N treatments: high N level (250 mg N kg−<sup>1</sup> soil) and low N level (83 mg N kg−<sup>1</sup> soil). The <sup>15</sup>N-labeled (NH4)2SO<sup>4</sup> (10.16% <sup>15</sup>N atom, Shanghai Research Institute of Chemical Industry, China) was applied to soils in all pots. The N fertilizer was added in three separate applications: the first application was done before planting as base fertilizer (approximately 33.3% of total N), and successive two applications (approximately 33.3% of total N per time) were made on day 7 and 14 after transplanting, respectively. Pots were placed in the greenhouse under optimum conditions (15 h light/9 h dark cycle, 25–30/15–20◦C day/night temperature and 40% day/60% humidity). The seedlings were regularly watered (tap water) to 40% relative soil moisture by weighing method every 5 days.

The samples of wheat, maize, and cucumber were harvested from each treatment on day 35 after transplanting, respectively.

Firstly, the whole seedling was uprooted, and then shoot and root samples were separated and washed with deionized water to remove the adhering soil. Shoot and root samples were ovendied at 105◦C for 30 min to inactivate the enzyme, respectively, followed by 65◦C until constant weight for dry weight analysis. Then, the oven-dried samples were grinded, screened by a 1 mm sieve, and stored in zip-lock bag for plant N content and <sup>15</sup>N enrichment determination. Afterward, the remaining samples were immediately frozen in liquid N and then maintained at −80◦C until further analysis.

## Bacterial Cell Concentration Within Plant Tissues

The cell densities of diazotrophic P. beijingensis BJ-18 within the inoculated plant tissues were estimated by qPCR according to the method described by Rasmussen et al. (2007). Primers for qPCR of the nifB from P. beijingensis BJ-18 included nifB F (5<sup>0</sup> -GAAGGTGAGAGTGAGGATGG-3<sup>0</sup> ) and nifB R (50 -TTGCTTCAGGCTCATCTCC-3<sup>0</sup> ). qPCR was performed with plant genomic DNA as template which was extracted form plant seedlings using DNA Kit [TianGen Biotech (Beijing) Co., Ltd.]. The 129 bp PCR product was ligated to the PMD 19-T vector (Takara, Otsu, Japan) and then introduced into Escherichia coli JM109. The introduced E. coli JM109 was grown in liquid LB medium, and the recombinant plasmids carrying nifB fragment were extracted and purified using TIANprep Mini Plasmid Kit [TianGen Biotech (Beijing) Co., Ltd.]. A standard curve was generated for each run 10-fold dilution series from 2 × 10<sup>1</sup> to 2 × 10<sup>7</sup> copies. The plant genomic DNA isolated from each of the different treatments was mixed with SYBR <sup>R</sup> Premix Ex TaqTM (Takara, Kyoto, Japan), primer pairs and ddH2O in a total volume of 20 uL for qPCR on a 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA, United States). Ct values were measured to quantify initial amounts of target DNA.

## Quantification of Biological N<sup>2</sup> Fixation Contribution

In this study, BNF contribution of P. beijingensis BJ-18 to plants was quantified by the method of <sup>15</sup>N isotope dilution technique. N content and <sup>15</sup>N enrichment in plant tissues were determined by DELTA V Advantage isotope ratio mass spectrometer (Thermo Fisher Scientific, Inc., United States). The plants without P. beijingensis BJ-18 inoculation were used as references to calculate the BNF contribution. The BNF

contribution was calculated according to formula 1 described by Boddey and Knowles (1987):

$$\% \text{Ndfa} = \left(1 - \frac{\%^{15} \text{Na.e.} \text{.}}{\%^{15} \text{Na.e.} \text{.} \text{.} \text{U}}\right) \times 100\tag{1}$$

Where, %Ndfa is the percentage of N derived from air and percent <sup>15</sup>Na.e. (%15N atom excess) is the enrichment of the inoculated (I) and un-inoculated (UI) plants, respectively.

#### Determination of the Concentration of Free NH<sup>4</sup> <sup>+</sup>, NO<sup>3</sup> <sup>−</sup>, and Activities of GS and NR

In order to determine the concentrations of the free ammonium (NH<sup>4</sup> <sup>+</sup>) and nitrate (NO<sup>3</sup> <sup>−</sup>), fresh plant tissues were ground with a mortar on ice in extraction buffer. The buffer consisted of 10 mM imidazole, 50 mM Tris-HCl (pH 8.0), and 0.5% (w/v) β-mercaptoethanol. After grinding, the samples were centrifuged at 12,000 × g for 20 min at 4◦C and the supernatant was collected for free NH<sup>4</sup> <sup>+</sup> and NO<sup>3</sup> <sup>−</sup> determination (Oliveira et al., 2002; Yang et al., 2014). Free NH<sup>4</sup> <sup>+</sup> concentration [µg g−<sup>1</sup> fresh weight (FW)] in the supernatant was assayed using the Berthelot color reaction method (Gordon et al., 1978), and free NO<sup>3</sup> − concentration (µg g−<sup>1</sup> FW) in the supernatant was determined using the Griess method (Eckhardt et al., 1999).

For analysis of GS activity, fresh plant tissues were ground with a mortar in pre-cold extraction buffer (containing10 mM MgSO4, 2 mM dithiothreitol, 70 mM 3 (n-morpholino) propane-sulfonic acid (pH 6.8), 5 mM glutamate, 10% (v/v) ethanediol, 0.1% (v/v) TritonX-100). The extracts were centrifuged at 12,000 × g for 30 min at 4◦C and the supernatant was collected for plant GS activity determination using NH2OH as a substrate, and the amount of γ-glutamyl hydroxamate (GHA, ug−<sup>1</sup> FW min−<sup>1</sup> ) released was determined spectrophotometrically at 540 nm according to the method of Yang et al. (2014).

To measure NR activity, fresh plant tissues were ground on ice in extraction buffer consisting of 25 mM phosphate buffer (pH 7.5, a mixture of K2HPO<sup>4</sup> and KH2PO4), 5 mM cysteine and 5 mM EDTA-Na2. The mixture was centrifuged at 4,000 × g and 4◦C for 10 min and the supernatant was collected. NR activity was measured spectrophotometrically at 540 nm according to Yu and Zhang (2012). NR activity was expressed as µg NO<sup>2</sup> <sup>−</sup> g <sup>−</sup><sup>1</sup> FW h−<sup>1</sup> .

#### Quantitative Real-Time PCR Analysis of Plant Genes and nifH in Plant Roots and Shoots

Total RNA was extracted from plant tissues using RNAiso Plus reagent (RaKaRa, Kyoto, Japan). Then, RNA was digested with DNase I and reversely transcribed into cDNA using PrimeScriptTM RT reagent kit (RaKaRa, Kyoto, Japan). Gene expression levels were determined by quantitative real-time PCR (qRT-PCR) analysis. The specific primers for qRT-PCR were shown in **Table 1**. The plant housekeeping gene actin was used as plant internal control, and the bacterial 16S rRNA was used as bacterium internal control. The relative expression of the target genes were calculated according to the standard comparative C(t) method (Livak and Schmittgen, 2001). Each treatment had three biological replicates, with three technical replicates for each biological replicate.

#### Statistical Analysis

Statistical tests were performed using SPSS software version 20 (SPSS Inc., Chicago, IL, United States). Two-way analysis of variance (ANOVA) was employed to check the significant differences between treatments. Means of different treatments were compared using the least significant difference (LSD) at 0.05 or 0.01 level of probability. Graphs were prepared using SigmaPlot software version 12.5 (Systat Software, Inc., CA, United States).

#### RESULTS

#### Colonization of GFP-Tagged P. beijingensis BJ-18 in Wheat, Maize, and Cucumber Tissues

The CLSM observation showed that the GFP-tagged P. beijingensis BJ-18 cells emitted bright green fluorescence (**Figure 1a**). The GFP-tagged P. beijingensis cells were found to colonize on the surface of the primary roots and the root hair zone of wheat (**Figures 1b,c**). The bacterial cells were found to be distributed within cortex of wheat primary roots (**Figure 1d**) and colonized in the wheat vascular bundle (**Figure 1e**). Moreover, bacterial cells were observed in the vascular bundle of wheat stem (**Figures 1f,g**) and in wheat leaf vein (**Figure 1h**). In the maize seedlings, the P. beijingensis cells were found to colonize the surface of the primary roots (**Figure 2a**), and on the junction of the primary and lateral roots (**Figure 2b**). The bacteria cells were found within the intercellular spaces and xylem vessels of maize roots and stems (**Figures 2c–e**) and leaves (**Figure 2f**). The colonization pattern in cucumber was similar to those obtained in wheat and maize seedlings. The bacterial cells were found to colonize the surface of the primary roots and the root hair zone (**Figure 3a**). The P. beijingensis cells were in cortex (**Figure 3b**) and vascular bundle (**Figure 3c**) of cucumber roots. As shown in **Figures 3d,e**, bacterial cells invaded the xylem vessels of cucumber stem. Moreover, the bacterial cells were found within the cucumber leaves (**Figure 3f**).

Taken together, P. beijingensis cells colonized on the surface of roots and within roots, stems and leaves of wheat, maize, and cucumber, indicating that the colonization patterns in the three plants were similar.

#### Concentration of Diazotrophic P. beijingensis BJ-18 in Wheat, Maize, and Cucumber Tissues

The concentration of P. beijingensis BJ-18 in the inoculated plants at 35 day after inoculation was assessed using qPCR and was expressed as the number of copies of the specific nifB genes per total (plant seedlings + bacteria) genomic DNA. As shown in **Figure 4**, copy numbers of nifB gene in the

#### TABLE 1 | Primers sequence and accession number in NCBI.

fmicb-10-01119 May 27, 2019 Time: 17:4 # 5


(Continued)


inoculated plant tissues under low N level were much higher than under high N level. Wheat had the highest copy numbers of P. beijingensis nifB gene, followed by cucumber and then maize under both low N and high N levels. Copy numbers of nifB gene in roots of the three plants were much higher (62.5–185.3%) than in shoots under both low N and high N levels. These data suggested that the concentrations of the bacterial cells in plant tissues were related to soil N levels, plant species and plant tissues.

#### Plant Growth Promotion

To assess the impacts of P. beijingensis inoculation on plant growth, the biomass (dry weight) was analyzed in inoculated and un-inoculated plants under low N and high N levels. Dry weights of wheat shoots and in roots under low N were increased by 86.1 and 46.0%, respectively. Dry weights of maize shoots and in roots under low N were increased by 46.6 and 47.5%, respectively. Dry weights of cucumber shoots and in roots under low N were increased by 103.6 and 20.3%, respectively. The data suggested that P. beijingensis BJ-18 could efficiently promote the growth of the three plants (**Figure 5**). The increased shoot and root biomass by P. beijingensis BJ-18 inoculation under low N condition were significantly higher than those under high N condition, consistent with the concentration of P. beijingensis BJ-18 in these plant tissues under different N conditions.

#### Quantification of BNF in the P. beijingensis-Inoculated Wheat, Maize, and Cucumber

To estimate the contribution of BNF, <sup>15</sup>N isotope enrichment analysis was conducted to analyze inoculated seedling of wheat, maize and cucumber grown in soil contain <sup>15</sup>N-labeled (NH4)2SO<sup>4</sup> as N fertilizer in greenhouse conditions, in comparison with un-inoculated plants. As shown in **Table 2**, the roots and shoots of seedlings inoculated with P. beijingensis BJ-18 had significantly lower δ <sup>15</sup>N value than un-inoculated seedlings under both low and high N conditions, suggesting that these plants derived a portion of N from atmospheric N2. The plant N derived from the atmosphere (%Ndfa) ranged from 18 to 36.4% under low N level and from 12.9 to 30.5% under high N level. Among the three plants, wheat showed maximum %Ndfa (30.5 and 36.4%) under both high N and low N levels, followed by cucumber (25.4 and 27.8%) and then maize (12.9 and 20.9%). These data suggest that the contribution of BNF was closely related to soil N levels and to plant species. These results were in agreement with the concentration of P. beijingensis BJ-18 within plants and the plant biomass under different N levels.

## Transcript Levels of nifH Gene

To investigate whether nif genes coding nitrogenase of P. beijingensis BJ-18 were expressed within plant tissues, the transcript levels of nifH gene, one of nif genes coding Fe protein of nitrogenase, were quantified. As shown in **Figure 6**, transcripts of nifH under low N were up-regulated in wheat shoots and roots by 1.09 and 1.61 folds, respectively, in maize shoots and roots by 0.77 and 1.0 folds, respectively, and in cucumber shoots and roots by 0.75 and 1.61 folds, respectively, compared to those under high N. The results suggested that soil N status affected transcript level of P. beijingensis nifH. The data were consistent with BNF rate and the concentration of P. beijingensis BJ-18 in plants grown in soils containing different N levels.

#### Promotion of N Uptake and Total N Content in Plants by Inoculation With P. beijingensis BJ-18

Compared to un-inoculated shoots and roots under low N level, a significant increase of free NH<sup>4</sup> <sup>+</sup> concentration was observed in shoots of wheat (26.6%), maize (21.9%) and cucumber (22.2%) and in roots of wheat (24.9%), maize (52.7%) and cucumber (32.2%) (**Figure 7A**). In contrast, inoculation did not significantly enhance free NH<sup>4</sup> <sup>+</sup> concentration in plant tissues under high N (**Figure 7B**).

Similarly, P. beijingensis inoculation significantly increased the free NO<sup>3</sup> <sup>−</sup> concentration in shoots of wheat (26.3%), maize (23.4%) and cucumber (43.7%) and in roots of wheat (60.7%), maize (62.6%) and cucumber (67.4%) under low N (**Figure 7C**). In contrast, inoculation did not significantly enhance free NO<sup>3</sup> <sup>−</sup> concentration in plant tissues under high N (**Figure 7D**).

Compared to un-inoculated shoots under low N level, a significant increase of total N content was observed in inoculated shoots of wheat (76.3%), maize (49.1%) and cucumber (88.8%) (**Figure 7E**). Similarly, inoculation greatly enhanced total N content in inoculated roots of wheat (61.6%), maize (68.9%) and cucumber (92.3%) (**Figure 7F**). In contrast, the increased levels of total N content in root and shoots were lower under high N (**Figures 7E,F**). The data were consistent with the change of the concentrations of the free NH<sup>4</sup> <sup>+</sup>, and free NO<sup>3</sup> <sup>−</sup> under low N and high N.

FIGURE 1 | Colonization of the GFP-tagged P. beijingensis BJ-18 in the seedlings of wheat. Wheat seedlings were grown in the presence of GFP-tagged P. beijingensis BJ-18 for 2 weeks. Images were taken with a fluorescent microscope. Excitation was at 488 nm. (a) Confocal image of the GFP-tagged P. beijingensis BJ-18 cells; (b–e) Colonization patterns in the root; (f,g) Colonization patterns in the stem; (h) Colonization patterns in the leaf. Bars represent 50 µm.

#### Enhancement of GS and NR Activities of Plants by Inoculation With P. beijingensis 1-18

Compared to un-inoculated plant shoots and roots under low N (**Figures 8A,B**), GS activities were significantly increased in inoculated shoots of wheat (89.7%), maize (19.0%) and cucumber (46.9%) and in inoculated roots of wheat (45.0%),

maize (85.5%) and cucumber (40.4%), but inoculation did not obviously enhance the GS activities of plants grown in high N soil.

Colonization patterns in the stem; (f) Colonization patterns in the leaf. Bars

represent 50 µm.

As shown in **Figures 8C,D**, compared to un-inoculated plant shoots and roots grown in low N soil, NR activities were significantly enhanced in inoculated shoots of wheat (20.9%), maize (42.0%) and cucumber (28.9%) and in inoculated roots of wheat (28.9%), maize (66.2%) and cucumber (44.3%). In contrast, inoculation did not obviously enhance the NR activities of plants grown in high N soil.

#### Up-Regulation of Expression of N Uptake and N Assimilation Genes in Plants by Inoculation With P. beijingensis

In maize, the transcript levels of three genes (ZmAMT1,1a, ZmAMT1,1b, and ZmNRT2.1) involved in N uptake and five genes (ZmNR, ZmNiR, ZmGS1-3, ZmGS1-4, and Zm CsGOGAT) involved N metabolism were analyzed (**Figure 9**). Under low

fluorescent microscope. (a–d) Colonization patterns in the root; (e) Colonization patterns in the stem; (f) Colonization patterns in the leaf. Bars represent 50 µm.

N level, the transcript levels of ZmAMT1,1a, ZmAMT1,1b, and ZmNRT2.1 were up-regulated by 1.83–3.93 folds in the inoculated shoots, while in inoculated roots, they were up-regulated 2.97–14.17 folds (**Figures 9A–C**), suggesting that inoculation promoted expression of plant N uptake genes. Similarly, inoculation significantly enhanced the transcript levels of ZmNR, ZmGS1-3, and ZmGS1-4 in shoots by 18.46, 7.44, and 11.79 folds, respectively, and in roots by 5.63, 6.64, and 6.59 folds, respectively (**Figures 9D,F,G**). In contrast, inoculation did not obviously affect the transcript levels of these genes in maize shoots and roots under high N condition. However, inoculation did not affect the transcript levels of ZmNiR and ZmGOGAT under both low N and high N conditions (**Figures 9E,H**).

In wheat, the transcript levels of four genes (TaAMT1.1, TaNRT1.1, TaNRT2.1, and TaNRT2.3) involved in N uptake and four genes (TaNR, TaNiR, TaGS1, and TaGOGAT) involved N metabolism were analyzed. Similar to maize, the eight genes in the inoculated shoots of wheat in low N were up-regulated by 1.56–46.49 folds, while they were up-regulated 1.98–91.93 folds in inoculated roots (**Figure 10**). In contrast, inoculation did not obviously affect the transcript levels of these genes under high N condition.

In cucumber, the transcript levels of four genes (CsAMT1, CsAMT3, CsNRT1.3, and CsNRT2.2) involved in N uptake and five genes (CsNR1, CsNR2, CsNiR, CsGS1, and CsGOGAT) involved N metabolism were also analyzed. Similar to wheat and maize, the nine genes in the inoculated shoots of cucumber in low N soil were up-regulated 1.60–5.82 folds, while they were up-regulated 1.47–11.85 folds in inoculated roots (**Figure 11**). However, these effects of inoculated were weak when cucumber was grown in high N soil.

#### DISCUSSION

The N2-fixing Paenibacillus strains have gained much attention due to their capacity of forming endospore to survive for long periods of time under adverse conditions (Grady et al., 2016). In this study, P. beijingensis BJ-18 was tagged by GFP and observation by laser confocal microscopy revealed that in seedlings of wheat, maize, and cucumber, bacterial cells could be found in the inner cortex and vascular bundle of roots and stems as well as within the leaves, suggesting P. beijingensis BJ-18 has similar invasion patterns in both monocotyledons and dicotyledons. The data indicated that P. beijingensis BJ-18 spread systemically from roots to stems and leaves of wheat, maize, and cucumber via xylem vessels. Therefore, P. beijingensis BJ-18 could be defined as a plant endophytic diazotrophic bacterium and it had a broad range of host plants. Similar colonization patterns were observed in the association of P. polymyxa WLY78 with wheat, maize, and cucumber (Hao and Chen, 2017) and in the association of P. polymyxa P2b-2R with lodgepole pine (a gymnosperm tree species) (Anand and Chanway, 2013). The colonization pattern of P. beijingensis BJ-18 was a little different form that of the associated diazotrophic A. brasilense Yu62 which colonized mainly on the surface of maize roots and only small part entered into maize tissues (Liu et al., 2003). These results indicated that on one side the diazotrophic Paenibacillus species/strains may fix N<sup>2</sup> inside plants and rapidly transfer the fixed product to plants, and on the other side they fix N<sup>2</sup> on the root surfaces and part of the fixed product may be remain in the soil.

In this study, the concentrations of P. beijingensis cells in inoculated plant tissues under high N and low N levels were determined by qPCR, with un-inoculated plant tissues as control. The results showed that the bacterial cell numbers were significantly higher in inoculated plant tissues grown under low N condition than those under high N condition, suggesting that soil N status controlled the concentration of bacterial cells in plants. Similar reports were found in sugarcane where high dose of N (ammonium nitrate) resulted in reduction of endophytic Acetobacter diazotrophicus concentration (Fuentes-Ramirez et al., 1999) and a higher number of endophytic diazotrophs were isolated from sugarcane plants under low N than under high N (de Oliveira et al., 2003). Moreover,

FIGURE 4 | The cell concentrations of P. beijingensis BJ-18 in the inoculated shoots and roots of wheat, maize, and cucumber seedlings in high N (HN) and low N (LN) levels. The concentrations of P. beijingensis BJ-18 are represented by nifB gene copies ng−<sup>1</sup> total genomic DNA. Values are given as mean of three independent biological replicates, and single asterisks or double asterisks (<sup>∗</sup> or ∗∗) indicate significant differences between HNE+ and LNE+ treatments determined by LSD at P < 0.05 or P < 0.01. The bars represent the standard error. LNE+ indicates plants grown in low N level of soils and inoculated with P. beijingensis BJ-18; HNE+ indicates plants grown in high N level of soils and inoculated with P. beijingensis BJ-18.

it was reported that the plant endogenous N status could also induce plant defense responses to regulate bacterial colonization (Carvalho et al., 2014). The increase in N compounds and amino acids (such as phenylalanine and hydroxyproline) was necessary to activate plant defense responses (Snoeijers et al., 2000). Amino acid transporters regulated by N status also have regulatory function in plant defense responses (Liu et al., 2010; Seifi et al., 2013). This study also showed that total N content and concentrations of NH<sup>4</sup> <sup>+</sup> and NO<sup>3</sup> <sup>−</sup> in un-inoculated plants under high N were much higher than those under low N. This finding may explain why high N caused a decrease in the numbers of P. beijingensis cells in plants tissues.

Furthermore, this study investigated whether nifH gene was expressed in the inoculated plants. The expression levels of nifH


Values are given as mean ± SE of three independent biological replicates. Different letters indicates indicate significantly differences in δ15N value inoculated (E+) and un-inoculated (E-) plants according to the LSD test (P < 0.05); Different letters indicate significantly differences in %Ndfa between high N and low N according to the LSD test (P < 0.05); δ <sup>15</sup>N value: percent atom excess <sup>15</sup>N; %Ndfa: percent N derived from atmosphere.

in the inoculated plant shoots and roots under low N were significantly higher than those under high N, consistent with the concentrations of P. beijingensis cells. It was well known that nif gene expression was inhibited by high concentrations of NH<sup>4</sup> <sup>+</sup> (Dixon and Kahn, 2004). GlnR mediated positive and negative regulation of nif gene expression in P. polymyxa WLY78 according to N availably (Wang et al., 2018). As mentioned above, there were high concentrations of NH<sup>4</sup> + in plants under high N. Thus, P. beijingensis cells in plant tissues sensed the N signal and then regulated nifH expression according to change of N levels. The expression of nif gene of Herbaspirillum seropedicae in maize, sorghum, wheat, and rice plants was reported (Roncato-Maccari et al., 2003). Similar report was found that N fertilizer application inhibited nifH gene expression of the endophytic diazotroph in sugarcane leaves (Jia Hui et al., 2017).

An important metric to evaluate the role of diazotrophic bacteria is whether they can provide fixed N<sup>2</sup> to the host plant. The <sup>15</sup>N isotope dilution analysis has been widely applied to quantify BNF in non-legume plant species such as rice inoculated with Herbaspirillum seropedicae Z67 (James et al., 2002), wheat with Azospirillum brasilense Wa5 (Christiansenweniger and Vanveen, 1991), Kallar grass with Azoarcus sp. BH72 (Hurek et al., 2002) and maize with P. polymyxa P2b-2R (Puri et al., 2015). Here, <sup>15</sup>N isotope enrichment analysis method was used to estimate the contribution of BNF by P. beijingensis BJ-18 inoculation to plants. The BNF rates in the three different plants were higher under low N level than under high N level, indicating that BNF was affected by N levels. The data were consistent with the nif gene expression and the concentrations of P. beijingensis BJ-18 within the plant tissues under different N levels. Similar results were reported that sugarcane plants inoculated with diazotrophic strains gained higher N from BNF in the N-deficient soil (de Oliveira et al., 2003). This study also revealed that wheat gained the highest N from BNF under both low N and high N levels, followed by cucumber and then by maize, suggesting that BNF rate was related to host plant species. The shoots and roots of palm inoculated with Bacillus sphaericus UPMB-10 gained 13.2–13.4% N from BNF (Zakry et al., 2012), which was lower than that gained by inoculation with P. beijingensis BJ-18.

In this study, the effects of P. beijingensis BJ-18 inoculation on plant N uptake and metabolism were investigated. The

level of soils and inoculated with P. beijingensis BJ-18.

FIGURE 7 | Ammonium (NH<sup>4</sup> <sup>+</sup>) (A: shoot; B: root), nitrate (NO<sup>3</sup> <sup>−</sup>) (C: shoot; D: root) and total N (E: shoot; F: root) content in shoots and roots of wheat, maize and cucumber seedlings inoculated with P. beijingensis BJ-18 under high N (HN) and low N (LN) levels. Values are given as mean of three independent biological replicates, and single asterisks or double asterisks (<sup>∗</sup> or ∗∗) indicate significant differences between inoculated (E+) and un-inoculated (E-) plants determined by LSD at P < 0.05 or P < 0.01. The bars represent the standard error. LNE- indicates plants grown in low N level of soil and un-inoculated with P. beijingensis BJ-18; LNE+ indicates plants grown in low N level of soils and inoculated with P. beijingensis BJ-18; HNE- indicates plants grown in high N level of soil and un-inoculated with P. beijingensis BJ-18; HNE+ indicates plants grown in high N level of soils and inoculated with P. beijingensis BJ-18.

concentrations of NO<sup>3</sup> <sup>−</sup>, NH<sup>4</sup> <sup>+</sup> and total N were higher in roots and shoots of inoculated plants than in un-inoculated plants, and were higher in roots than in shoots. The positive effects were also controlled by soil N status. The results indicated that inoculation with P. beijingensis BJ-18 promoted plants to uptake NO<sup>3</sup> <sup>−</sup> and NH<sup>4</sup> <sup>+</sup> from soil. The increased concentrations of NH<sup>4</sup> <sup>+</sup> and total N in inoculated plants were at least partially resulted from BNF. Studies on the effects of diazotrophs on plant N uptake and metabolism were rare. However, inoculations with some endophytic fungi significantly

improved N accumulation and metabolism were observed in rice (Yang et al., 2014), sugar beet (Shi et al., 2009), and tall fescue (Lyons et al., 1990).

NO<sup>3</sup> <sup>−</sup> was mainly absorbed via NRT protein family members, and then transformed into NH<sup>4</sup> <sup>+</sup> by NR. In this study, P. beijingensis BJ-18-inoculated plants showed significantly higher expression levels of NRT genes (CsNRT1.3 and CsNRT2.2 in cucumber; TaNRT1.1, TaNRT2.1, and TaNRT2.3 in wheat; ZmNRT2.1 in maize) in both shoots and roots under low N condition. Similar reports were found that under low N condition NRT genes were significantly up-regulated in rice inoculated with endophytic fungus Phomopsis liquidambari (Yang et al., 2014) and in tomato inoculated with diazotrophic Enterobacter radicincitans (Berger et al., 2013).

Higher NR activities were observed in the P. beijingensis BJ-18-inoculated plants. It was reported that inoculation with endophytic fungus Plectosphaerella cucumerina F11 greatly increased the activity of NR in sugar beet (Shi et al., 2009). To investigate whether the changes of NR activities in the P. beijingensis BJ-18- inoculated plants were closely related to differential expression of plant NR genes, the expression levels of NR genes were quantified. qRT-PCR analysis indicated that the expression levels of NR genes (CsNR2 in cucumber, TaNR in wheat and ZmNR in maize) were significantly higher in inoculated plants than in un-inoculated plants under low N. It was reported that the endophytic fungus Piriformospora indica inoculation promoted N accumulation in Arabidopsis and tobacco seedlings by inducing the expression of NR (Sherameti et al., 2005). In contrast, endophytic fungus Phomopsis liquidambari inoculation significantly reduced NO<sup>3</sup> − concentration in rice shoots under low N condition, since the higher NR activity made more NO<sup>3</sup> <sup>−</sup> to be transformed into NH<sup>4</sup> <sup>+</sup> (Yang et al., 2014).

This study demonstrated that P. beijingensis BJ-18-inoculated plants showed higher NH<sup>4</sup> <sup>+</sup> concentration in roots and shoots, compared with those in un-inoculated plants. NH<sup>4</sup> <sup>+</sup> was absorbed via AMT protein family members mainly, and then transformed into organic molecules by GS and GOGAT. GS is an important rate-limiting enzyme in NH<sup>4</sup> <sup>+</sup> assimilation. The higher GS activities were observed in the P. beijingensis BJ-18-inoculated plants than those in un-inoculated plants. GS activities were closely related to soil N status. Transcript levels of

FIGURE 9 | The transcript levels of genes involved N uptake (A: ZmAMT1, 1a; B: ZmAMT1, 1b; C: Zm NRT2.1) and assimilation (D: ZmNR; E: ZmNiR; F: ZmGS1-3; G: ZmGS-4; H: ZmGOGAT) in shoots and roots of maize seedlings inoculated with P. beijingensis BJ-18 under high N (HN) and low N (LN) levels. Values are given as mean of three independent biological replicates, and single asterisks or double asterisks (<sup>∗</sup> or ∗∗) indicate significant differences between inoculated (E+) and un-inoculated (E-) plants determined by LSD at P < 0.05 or P < 0.01. The bars represent the standard error. LNE- indicates plants grown in low N level of soil and un-inoculated with P. beijingensis BJ-18; LNE+ indicates plants grown in low N level of soils and inoculated with P. beijingensis BJ-18; HNE- indicates plants grown in high N level of soil and un-inoculated with P. beijingensis BJ-18; HNE+ indicates plants grown in high N level of soils and inoculated with P. beijingensis BJ-18.

GS genes were also measured to confirm whether the changes of GS activities in plants caused by P. beijingensis BJ-18 inoculation were related to GS genes transcription. qRT-PCR indicated that the higher expression levels of GS genes (CsGS1, TaGS1, ZmGS1-3, and ZmGS1-4) in inoculated plant tissues than in un-inoculated ones under low N. Similarly, it was also reported that the expression levels of GS genes were also significantly higher in inoculated plants with the endophytic fungus P. liquidambari than in un-inoculated plants under low N (Yang et al., 2014).

This study demonstrated that inoculation with P. beijingensis BJ-18 promoted dry weight of plant roots and shoots grown under low N to be increased by 20.3–103.6%. The current results were in agreement with previous reports that inoculation with P. beijingensis BJ-18 increased wheat yield by 26.9% in field experiment (Shi et al., 2016) and increased tomato shoot length, fresh weight, and dry weight in the pot experiments (Xie et al., 2016).

The plant-growth-promoting rhizobacteria facilitate plant growth by several direct and indirect mechanisms. Direct mechanisms include P solubilization, N fixation and hormone (e.g., IAA, cytokinins and gibberellins) production. Indirect mechanisms include controlling phytopathogens by producing antibiotics or lytic enzymes (Glick, 2012). As mentioned above, P. beijingensis BJ-18 provided N for plants by BNF and thus promoted plant growth. Also, this bacterium may promote plant growth by producing IAA and antimicrobial compounds (Xie et al., 2016). Compared to A. brasilense Yu62 which produced high amount of IAA, P. beijingensis BJ-18 produced a little amount of IAA. Thus, IAA produced by P. beijingensis BJ-18 might be not the major factor promoting plant growth. Since the soil used in greenhouse study was not sterile, the BNF/plant growth promotion observed in this study could have been in part due to one or several indigenous microbes present in that soil. This study for the first time revealed that P. beijingensis BJ-18 promoted plants to uptake N from soil and enhanced gene expression and enzyme activities involved in N uptake and assimilation in plants. In addition to BNF, these endogenous changes in plants induced by P. beijingensis BJ-18 might be another major factor promoting plant growth.

#### REFERENCES


#### CONCLUSION

This study demonstrated that P. beijingensis BJ-18 was an effective and endophytic diazotrophic bacterium which has similar colonization patterns in monocotylous and dicotyledonous plants. This bacterium promoted plant growth by direct mechanisms through BNF. Also, this bacterium might promote plant growth by indirect mechanisms through inducing endogenous changes in plants, including enhancement of N uptake and enzyme activities, and expression of N uptake and assimilation genes. The bacterial density within plant was closely related to the BNF efficiency and the endogenous changes in plants. However, the bacterial density, the BNF efficiency and the endogenous changes in plants during the association with P. beijingensis BJ-18 were controlled by the soil N status. These data suggested that successful colonization of P. beijingensis BJ-18 on plant was the first key step for this bacterium to promote plant growth by BNF and by inducing endogenous changes in plants. How soil N level affects bacterial colonization deserves further study.

#### AUTHOR CONTRIBUTIONS

YbL and SC designed and wrote the manuscript. YbL, YlL, HZ, and MW conducted the experiments.

## FUNDING

This work was funded by the National Key Research and Development Program of China (No. 2017YFD0200807) and by National Nature Science Foundation of China (Grant No. 31770083).

#### ACKNOWLEDGMENTS

We would like to thank Caixia Wang for helping in writing early manuscript.


Dixon, R., and Kahn, D. (2004). Genetic regulation of biological nitrogen fixation. Nat. Rev. Microbiol. 2, 621–631. doi: 10.1038/nrmicro954


fmicb-10-01119 May 27, 2019 Time: 17:4 # 17

in cucumber (Cucumis sativus L.) seedlings response to abiotic stress. Front. Plant Sci. 6:982. doi: 10.3389/fpls.2015.00982


isotope dilution technique. Microbes Environ. 27, 257–262. doi: 10.1264/jsme2. ME11309

Zhang, W., Ding, Y., Yao, L., Liu, K., and Du, B. (2013). Construction of gene knock-out system for Paenibacillus polymyxa SC2. Wei Sheng Wu Xue Bao 53, 1258–1266.

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

Copyright © 2019 Li, Li, Zhang, Wang and Chen. 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.

# Arbuscular Mycorrhizal Symbiosis: Plant Friend or Foe in the Fight Against Viruses?

Laura Miozzi<sup>1</sup> \*, Anna Maria Vaira<sup>1</sup> , Marco Catoni<sup>2</sup> , Valentina Fiorilli<sup>3</sup> , Gian Paolo Accotto<sup>1</sup> and Luisa Lanfranco<sup>3</sup>

1 Institute for Sustainable Plant Protection, National Research Council of Italy (IPSP-CNR), Turin, Italy, <sup>2</sup> School of Biosciences, University of Birmingham, Birmingham, United Kingdom, <sup>3</sup> Department of Life Sciences and Systems Biology, University of Turin, Turin, Italy

Edited by:

Didier Reinhardt, Université de Fribourg, Switzerland

#### Reviewed by:

Manuela Giovannetti, University of Pisa, Italy Sergio Saia, Council for Agricultural Research and Economics, Italy

> \*Correspondence: Laura Miozzi laura.miozzi@ipsp.cnr.it

#### Specialty section:

This article was submitted to Plant Microbe Interactions, a section of the journal Frontiers in Microbiology

Received: 26 February 2019 Accepted: 17 May 2019 Published: 04 June 2019

#### Citation:

Miozzi L, Vaira AM, Catoni M, Fiorilli V, Accotto GP and Lanfranco L (2019) Arbuscular Mycorrhizal Symbiosis: Plant Friend or Foe in the Fight Against Viruses? Front. Microbiol. 10:1238. doi: 10.3389/fmicb.2019.01238 Plant roots establish interactions with several beneficial soil microorganisms including arbuscular mycorrhizal fungi (AMF). In addition to promoting plant nutrition and growth, AMF colonization can prime systemic plant defense and enhance tolerance to a wide range of environmental stresses and below-ground pathogens. A protective effect of the AMF against above-ground pathogens has also been described in different plant species, but it seems to largely rely on the type of attacker. Viruses are obligate biotrophic pathogens able to infect a large number of plant species, causing massive losses in crop yield worldwide. Despite their economic importance, information on the effect of the AM symbiosis on viral infection is limited and not conclusive. However, several experimental evidences, obtained under controlled conditions, show that AMF colonization may enhance viral infection, affecting susceptibility, symptomatology and viral replication, possibly related to the improved nutritional status and to the delayed induction of pathogenesis-related proteins in the mycorrhizal plants. In this review, we give an overview of the impact of the AMF colonization on plant infection by pathogenic viruses and summarize the current knowledge of the underlying mechanisms. For the cases where AMF colonization increases the susceptibility of plants to viruses, the term "mycorrhiza-induced susceptibility" (MIS) is proposed.

Keywords: arbuscular mycorrhiza, plant virus, mycorrhiza-induced resistance, plant-AMF-pathogen interaction, priming

#### INTRODUCTION

#### Effect of Mycorrhizal Colonization on Plant Responses to Biotic Stress

In natural environments, plants interact with pathogenic and beneficial microorganisms that might affect their growth, performance and survival. Arbuscular mycorrhizal fungi (AMF) (subphylum Glomeromycotina) (Spatafora et al., 2016) establish a mutualistic association with c. 85% of land plants, providing substantial benefits to plant growth and fitness (Jung et al., 2012; Auge et al., 2015). As a consequence of the improved mineral nutrition, AMF colonized plants often display

increased biomass and productivity (Bona et al., 2016; Fiorilli et al., 2018). AMF root colonization induces a systemic effect also evident on epigeous portions of the plant (Fiorilli et al., 2009; Zouari et al., 2014) and exerts beneficial impacts beyond the nutritional status improvement, i.e., an enhanced ability to cope with biotic and abiotic stresses. This advantage relies on physiological and metabolic changes that take place in the plant upon AMF colonization (Fritz et al., 2006; Auge et al., 2015; Fiorilli et al., 2018) and proposes AM symbiosis as a biocontrol agent, impacting on the outcome of below- and above-ground interactions with other organisms.

Enhanced resistance of mycorrhizal plants against soilborne pathogens was often observed (Whipps, 2004), while contrasting results have been obtained for above-ground attackers (Pozo and Azcón-Aguilar, 2007). In roots, the bio-protective effect exerted by AMF seems to rely on several biotic factors such as fungal/host genotypes, mycorrhization degree and soil microbiota alteration, including development of pathogen antagonism and accumulation of defensive compounds (Pozo et al., 2002; Vierheilig et al., 2008; Cameron et al., 2013). The effects on above-ground pathogens seems to greatly depend on the pathogen lifestyle (Shaul et al., 1999; Fiorilli et al., 2011; Miozzi et al., 2011; Campos-Soriano et al., 2012; Jung et al., 2012; Song et al., 2013, 2015; Sanchez-Bel et al., 2016).

The boost of basal defenses in mycorrhizal plants was defined mycorrhiza-induced resistance (MIR) and several studies pointed to priming (Martinez-Medina et al., 2016) as a main mechanism operating in MIR (Pozo and Azcón-Aguilar, 2007; Cameron et al., 2013). Cameron et al. (2013) proposed that MIR is a cumulative effect of plant responses to mycorrhizal colonization, able to confer protection against a wide range of challengers, including biotrophic and necrotrophic pathogens, nematodes and insects. MIR, at least in shoots, seems to be a two-step process with a preliminary induction of a broad range of defense genes (including chitinases, glucanases and Pathogenesis Related (PR) proteins) (Spanu et al., 1989; Liu et al., 2007; Fiorilli et al., 2009) during AMF colonization, followed by a faster and stronger activation of pathogen-specific defense genes upon pathogen challenge (Campos-Soriano et al., 2012; Fiorilli et al., 2018). The main actors proposed to be involved in this process are plant hormones, i.e., salicylic acid (SA), jasmonic acid (JA) and its derivates oxylipins, ethylene and probably abscisic acid (ABA), whose level changes during the different steps of mycorrhizal symbiosis (Foo et al., 2013; Pozo et al., 2015; see section "Conclusion and Perspectives"). It is tempting to speculate that, beside the genetic, molecular and physiological mechanisms, other factors could affect MIR such as AMF associated endobacteria and virome (Bonfante and Desirò, 2017; Turina et al., 2018). Pioneering studies indicate that AMF endobacteria may improve the fungal ecological fitness (Salvioli et al., 2016) and promote antioxidative responses in both fungal and plant hosts (Vannini et al., 2016). However, data on the impact of endobacteria and mycoviruses (Ezawa et al., 2015) on AMF phenotypic expression and higher order biological interactions are scarce and deserve further investigations.

#### Virus Infection, Damage in Agriculture and Management at Various Scales

Viruses are obligate pathogens able to infect virtually all organisms, including plants. Their infection process depends on the host machinery, allowing the virus to multiply and spread in the host. In plants, virus infections generally induce a disease syndrome, with symptoms such as developmental abnormalities, necrosis and chlorosis. For all major agronomical crops, viral diseases cause huge losses in production and quality, representing a serious threat to global food security (Varma and Malathi, 2003). Virus infection has been often associated to a general reduction of plant performance, i.e., inhibition of photosynthesis (Rahoutei et al., 2000), decrease of biomass (van Mölken and Stuefer, 2011) and pollen production (Harth et al., 2016), although recent evidences suggest that virus infection may improve drought tolerance (Xu et al., 2008; Pantaleo et al., 2016). Majority of viruses spread among plants very efficiently exploiting as vectors other organisms (mostly insects) characterized by a high level of mobility. Since climate changes can favor insect colonization of new habitats (Pureswaran et al., 2018), many viral diseases are representing an emerging problem in agriculture (Varma and Malathi, 2003; Rojas and Gilbertson, 2008; Ghini et al., 2011).

Sustainable and effective approaches to limit viral diseases include the development of viral-resistant/tolerant crop, the integration of crop management strategies to reduce the disease spreading (Nicaise, 2014), the introgression of resistance genes (e.g., NBS-LRR) from wild accessions, the use of transgenic plants expressing viral components, able to interfere with viral infection mechanisms at RNA or protein level. Unfortunately, these strategies are not immediately applicable to uncharacterized emerging viral pathogens.

AMF inoculation has been proposed as a cost-effective and sustainable solution for plant virus control. However, despite some information were available from the early 70's, the studies on the effects of AMF on plant-virus interactions are surprisingly low and contradictory. The aim of this review is to summarize the actual knowledge on the effect of AMF on virus infection and the underlying mechanisms. We propose the term "mycorrhizainduced susceptibility" (MIS) for the cases where a better performance of the virus, defined by replication efficiency and induced symptomatology, is observed in mycorrhizal plants. Furthermore, the state of the art on the effect of virus infection on mycorrhization will be reported; finally, we suggest different aspects that deserve further investigations.

#### PLANT PROTECTIVE EFFECT OF THE AMF COLONIZATION AGAINST VIRAL INFECTION

Up to now, three studies highlighted a plant protective effect of the AMF colonization against viral infection. All of them

considered Solanaceae or Cucurbitaceae plant species and positive single stranded RNA viruses, with the exception of Maffei et al. (2014) that focused on a single-stranded circular DNA geminivirus (**Table 1**). In Maffei et al. (2014), previously AMF-colonized tomato plants displayed attenuated symptoms and reduced virus titre when infected by Tomato yellow leaf curl Sardinia virus (TYLCSV) although AMF colonization could not contrast the reduction of root biomass induced by the virus. Since, TYLCSV encodes proteins able to interact with the plant hormone pathways (Lozano-Durán et al., 2011) and particularly with JA, a key hormone in MIR (Cameron et al., 2013), the authors hypothesized that the high JA level in mycorrhizal plants creates an unfavorable environment for TYLCSV, limiting its replication and reducing symptoms severity. This hypothesis is in agreement with the priming effect induced by JA exogenous application during geminivirus infection, which was sufficient to reduce symptoms and viral titre in Beet curly top virus-infected plants (Lozano-Durán et al., 2011).

Differently from Maffei et al. (2014), Thiem et al. (2014) investigated the effect of mycorrhizal colonization on potato plants already infected by Potato virus Y (PVY): milder symptoms and a significant stimulation of shoot growth were observed in PVY-infected plants inoculated with Rhizophagus irregularis.

Finally, tobacco and cucumber plants colonized by R. irregularis and infected by Tobacco mosaic virus (TMV) and Cucumber green mottle mosaic virus (CGMMV), respectively, showed reduced disease symptoms and virus titre if compared to non-mycorrhizal plants (Stolyarchuk et al., 2009); the same authors, in the TMV-tomato system, observed that the content of viral antigens in mycorrhizal plants in respect to non-mycorrhizal ones changed overtime, being equal at 14 days post viral inoculation (dpi), then increasing and subsequently decreasing from 21 to 49 dpi and 56 dpi, respectively.

#### NEGATIVE IMPACT OF AMF COLONIZATION ON PLANT RESPONSE TO VIRUS INFECTION

Several studies report increased virus multiplication and/or symptom severity in infected mycorrhizal plants. They considered plants belonging to the Solanaceae, Rosaceae and Poaceae families and mostly dealt with single stranded RNA viruses (**Table 1**). Results indicate that mycorrhizal colonization facilitates or enhances virus multiplication, suggesting a prevailing detrimental effect of AMF on plant virus infection, for which we propose the term "mycorrhiza-induced susceptibility" (MIS). Even if, in the first days after inoculation by Tomato aucuba mosaic virus (now a TMV strain), the virus titre in tomato plants colonized by Funneliformis macrocarpa (formerly Endogone macrocarpa) was lower in respect to the control ones, at 8–12 dpi, it became higher in mycorrhizal plants and increased over time (Daft and Okusanya, 1973). Similar results were obtained in leaves and roots of both tomato and strawberry plants inoculated with Potato virus X (PVX) (Daft and Okusanya, 1973). No data on plant biomass or performance were reported. These authors observed a similar virus titre increment in infected non-mycorrhizal plants grown with increased concentration of soluble phosphate, and suggested that enhanced viral multiplication could be a general consequence of increased phosphorus availability provided by the symbiosis. This hypothesis relies on the established correlation between phosphate nutrition and TMV infection in tobacco plants (Bawden and Kassanis, 1950; Kassanis, 1953) and was also suggested by Sipahioglu et al. (2009) to explain the increase in PVY titre and symptomatology in potato plants colonized by R. irregularis. It is interesting to note that the results of Sipahioglu et al. (2009) were in contrast with those of Thiem et al. (2014) even if the authors considered a similar biological system (potato, R. irregularis, PVY) and experimental design (mycorrhization of already PVY-infected plants). Sipahioglu et al. (2009), also observed a reduction in the length, fresh and dry weight of shoots, and in tubers weight, as well as a slight reduction in leaves chlorophyll content in virus-infected mycorrhizal plants.

The MIS outcome, consisting in viral titre increase and worsening of symptoms, was confirmed by Jabaji-Hare and Stobbs (1984) and Shaul et al. (1999) respectively in TMVinfected tomato and tobacco plants colonized by Glomus sp. Jabaji-Hare and Stobbs (1984) also observed an increase of roots fresh weight in the virus-infected mycorrhizal plants when compared to non-infected mycorrhizal ones. The results of Shaul et al. (1999) suggest that MIS is not entirely dependent by the improved plant nutritional status; his study excluded any effect of improved phosphorus nutrition and linked the increased plant susceptibility to viral infection with the delay in PR proteins induction in mycorrhizal plants.

More recently, Miozzi et al. (2011) observed increased Tomato spotted wilt virus (TSWV) titre in infected mycorrhizal tomato plants at 34 and 56 dpi, but not at 14 dpi, compared to non-mycorrhizal controls. A delay in recovery (symptoms disappearance/reduction in plants initially showing severe disease; Pennazio, 2010) was observed in TSWVinfected mycorrhizal plants at 34 dpi, but disappeared later (56 dpi). Similarly to Shaul et al. (1999), these authors observed a reduction in the number and fold-change of PR proteins coding genes in TSWV-infected mycorrhizal plants when compared with TSWV-infected non-mycorrhizal plants. Since the MIR-related JA-dependent defense priming is hypothesized to be linked to the partial suppression of the salicylic acid (SA)-dependent response (Pozo and Azcón-Aguilar, 2007), it was proposed that the SA level increase induced by TSWV infection may prevent the MIRmediated response. Within the TSWV-AMF-plant interaction an involvement of ABA and a more complex cross-talk among phytohormones, not limited to SA and JA, were also postulated (Miozzi et al., 2011). Indeed, the pretreatment with ABA can suppress the non-pathogenesis related protein 1 (NPR1) gene, an important regulatory component of SA signaling involved in PR genes activation (Dong, 2004; Yasuda et al., 2008). The long-term changes in virus titre and symptomatology observed by Miozzi et al. (2011) are in agreement with the observation that the protective or

TABLE 1 | AMF-plant-virus biological systems investigated; in the upper section are listed the case studies reporting a protective effect of AMF against viral infection while in the lower section are listed those reporting a detrimental effect.


AMF-Plant-Virus Interaction

# Frontiers in Microbiology | www.frontiersin.org



<sup>∗</sup>a strain of tobacco mosaic virus.

AMF-Plant-Virus Interaction

detrimental effect of mycorrhizal colonization on viral infection can substantially change over time (Daft and Okusanya, 1973; Stolyarchuk et al., 2009) underling that timing is a key parameter in the complex interaction among plant, viruses and AMF. In agreement with other studies here reported, Miozzi and co-authors (2011) observed that the AMF colonization failed to compensate the biomass reduction induced by the virus (**Table 1**).

Rùa et al. (2013), addressing the possible consequences of climate changes and increase of atmospheric CO2, observed that, under elevated CO<sup>2</sup> concentration, the titre of both Barley yellow dwarf virus (BYDV) and Cereal yellow dwarf virus (CYDV) increases in the grasses Bromus hordeaceus L. and Avena fatua L. colonized by R. irregularis; no differences were observed under normal CO<sup>2</sup> concentration.

Less investigated is the effect of mycorrhization on fruit trees. Nemec and Myhre (1984), studying the changes induced by Claroideoglomus etunicatum (formerly Glomus etunicatum) on sour orange and Duncan grapefruit seedlings infected by Citrus tristeza virus and Citrus leaf rugose virus, respectively, observed that AMF colonization did not significantly reduce the pathogenic effect caused by virus infection. A reduction of roots fresh weight and plant growth in the virus-infected mycorrhizal plants in respect to non-infected mycorrhizal plants was found. Similarly to Thiem et al. (2014) and Sipahioglu et al. (2009), in this work, plants were inoculated with AMF after virus infection.

#### THE IMPACT OF VIRAL INFECTION ON MYCORRHIZATION

Most studies addressing the plants-viruses-AMF interaction mainly focused on the effect of AMF on virus infection; however, viral infection can impact mycorrhization. In this regard, Nemec and Myhre (1984) observed that the number of fungal spores and the percentage of mycorrhization were generally higher in not-infected plants in respect to infected ones. Maffei et al. (2014) observed that the frequency of mycorrhization moderately but significantly increased in TYLCSV-infected mycorrhizal plants compared to not-infected mycorrhizal ones. However, no differences were observed in intensity of mycorrhization, abundance of arbuscules within colonized areas and percentage of the root system with arbuscules, suggesting that the onset and spread of TYLCSV throughout the whole plant may not significantly interfere with the F. mosseae intraradical development. This result is consistent with the up-regulation, in both mycorrhizal and TYLCSVinfected mycorrhizal plants, of five selected plant genes previously described as mycorrhiza-responsive and preferentially expressed in arbuscule-containing cells. Similarly, Sipahioglu et al. (2009) reported equal degree of mycorrhizal colonization in PVY-infected and healthy mycorrhizal potato plants. Finally, Rùa et al. (2013) observed that, only under elevated CO<sup>2</sup> concentration, BYDV and CYDV infection increased the fungal colonization of roots, suggesting that, in this case and condition, AMF and virus interacted to stimulate each other success.

#### CONCLUSION AND PERSPECTIVES

The picture arising from the reported studies, even if very complex, suggests a prevailing detrimental effect of AMF on plant virus infection, for which we propose the term "mycorrhizainduced susceptibility" (MIS). Indeed, the interaction among virus, AMF and plant is a complex system where several factors, including viral pathogen lifestyle, plant nutritional status and timing of interaction, can move the dynamic equilibrium toward the final establishment of a MIS or MIR outcome. A key role is probably played by the hormonal crosstalk that finely tunes both plant-AMF and plant-virus interactions (**Figure 1**). In the early stage of AMF colonization, MIR has been associated with SAR (Systemic Acquired Resistance)-like priming of SA-dependent genes, while in the later stage, MIR coincides with priming of JA- and ethylene-dependent defenses. In addition, ABA has been proposed as a new candidate acting as a complementary longdistance signal controlling MIR, and several reports considered JA and its derivates (i.e., oxylipins) as key signals operating in this process (Cordier et al., 1998; Van Wees et al., 2008; Van der Ent et al., 2009; Jung et al., 2012; Song et al., 2013, 2015). In parallel, SA-dependent defenses are the major defensive pathway against viruses while ABA may act either positively and negatively on plant defenses against viruses, respectively, limiting cell-to-cell movement by inducing callose deposition on plasmodesmata, or inhibiting SA-mediated defenses (Alazem and Lin, 2015). Interestingly, SA and ABA may also interact with the plant RNA silencing machinery, respectively inducing dicer-like 1 and 2 and RNA-dependent-RNA polymerase 1 and 2 in virus infected plants (Campos et al., 2015) and regulating the expression of argonaute genes involved in plant antiviral defense (Várallyay et al., 2010; Harvey et al., 2011; Alazem and Lin, 2015). This mechanism could interfere with the siRNA-mediated antiviral plant defense (Alazem and Lin, 2015) and adding further level of complexity to the mechanisms regulating the plant-virus-AMF interaction. Indeed, in this context, the role of siRNAs/miRNAs has not been explored so far, but may be a key element in determining the final interaction outcome. Beyond plant siRNAs, fungal siRNAs have been recently proposed having a functional significance in the trans-kingdom communication between the AMF and its host plant (Silvestri et al., 2019), suggesting that their possible role in the final outcome of viral infection in mycorrhizal plants should be addressed.

The complexity of the AMF-plant-virus interaction may further increase if considering that mycorrhiza can impact other trophic levels such as the interaction between plants and insects (Gehring and Bennett, 2009), including those acting as viral vectors. Indeed, AMF R. irregularis can improve the survival of the aphid Macrosiphum euphorbiae, vector of Cucumber mosaic virus, thus possibly improving viral spread, but also activate indirect defenses, attracting the aphid parasitoid Aphidius ervi (Volpe et al., 2018).

These observations highlight the intricate network of processes that regulate the plant-virus-AMF interaction, and, far to be conclusive, indicate that several factors able to direct the dynamic equilibrium of the system toward a MIS or MIR outcome remain to be evaluated such as the different changes induced by AMF colonization performed before or after viral infection, and the importance of timing in evaluating the interaction outcome. Moreover, since only few AMF have been analyzed so far, future studies should consider different AMF species and isolates (Sikes, 2010; Turrini et al., 2018) and even AMF consortia for detecting possible synergistic/antagonistic effects. Finally, it must be emphasized that, based on current knowledge, drawing conclusions on the efficacy of AMF to act as biocontrol agents in agricultural environments is extremely difficult, since, in these conditions, many other biotic and abiotic factors have the potential to interfere with the final MIS or MIR outcome.

#### AUTHOR CONTRIBUTIONS

fmicb-10-01238 June 1, 2019 Time: 10:29 # 8

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

#### REFERENCES


#### FUNDING

Research in LL's laboratory is supported by the Project Università di Torino – Compagnia di San Paolo Bando ex-post – 2018.



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

Copyright © 2019 Miozzi, Vaira, Catoni, Fiorilli, Accotto and Lanfranco. 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.

# Streptomyces Strains Induce Resistance to Fusarium oxysporum f. sp. lycopersici Race 3 in Tomato Through Different Molecular Mechanisms

#### Sakineh Abbasi<sup>1</sup> , Naser Safaie<sup>1</sup> \*, Akram Sadeghi<sup>2</sup> \* and Masoud Shamsbakhsh<sup>1</sup>

<sup>1</sup> Department of Plant Pathology, Faculty of Agriculture, Tarbiat Modares University, Tehran, Iran, <sup>2</sup> Department of Microbial Biotechnology, Agricultural Biotechnology Research Institute of Iran (ABRII), Agricultural Research, Education and Extension Organization (AREEO), Karaj, Iran

#### Edited by:

Luisa Lanfranco, University of Turin, Italy

#### Reviewed by:

Soner Soylu, Mustafa Kemal University, Turkey Marta Berrocal-Lobo, Polytechnic University of Madrid, Spain

> \*Correspondence: Naser Safaie nsafaie@modares.ac.ir Akram Sadeghi aksadeghi@abrii.ac.ir

#### Specialty section:

This article was submitted to Plant Microbe Interactions, a section of the journal Frontiers in Microbiology

Received: 30 January 2019 Accepted: 17 June 2019 Published: 03 July 2019

#### Citation:

Abbasi S, Safaie N, Sadeghi A and Shamsbakhsh M (2019) Streptomyces Strains Induce Resistance to Fusarium oxysporum f. sp. lycopersici Race 3 in Tomato Through Different Molecular Mechanisms. Front. Microbiol. 10:1505. doi: 10.3389/fmicb.2019.01505 Plant growth promoting rhizobacteria (PGPR) are potential natural alternatives to chemical fungicides in greenhouse production via inducing plant immune system against biotic stresses. In this research, 126 Streptomyces isolates were recovered from rhizosphere soils of 13 different commercial vegetable greenhouses in Iran. Streptomyces isolates were screened for in vitro Plant growth promoting (PGP) traits and ability to antagonize Fusarium oxysporum f. sp. lycopersici race 3 (FOL), the causal agent of Fusarium wilt of tomato (FWT). Six isolates with the highest antagonistic activity and at least three PGP traits were selected and compared with chemical fungicide Carbendazim <sup>R</sup> in a greenhouse experiment. All bacterial treatments mitigated FWT disease symptoms like chlorosis, stunting and wilting at the same level or better than Carbendazim <sup>R</sup> . Strains IC10 and Y28 increased shoot length and shoot fresh and dry weight compared to not inoculated control plants. Phenotypic characterization and 16S rRNA gene sequencing showed, strains IC10 and Y28 were closely related to S. enissocaesilis and S. rochei, respectively. The ability of the superior biocontrol strains to induce antioxidant enzymes activity and systemic resistance (ISR) was investigated. Increased activity of catalase (CAT) in plant treated with both strains as well as an increase in peroxidase (POX) activity in plants treated with Y28 pointed to a strain specific-induced systemic resistance (ss-ISR) in tomato against FOL. The differential induced expression of WRKY70 and ERF1 (two transcription factors involved in plant defense) and LOX and TPX by the analyzed Streptomyces strains, especially after inoculation with FOL, suggests that ss-ISR is triggered at the molecular level.

Keywords: Fusarium wilt, induced systemic resistance, Streptomyces, tomato growth promoting, siderophore production

#### INTRODUCTION

In recent decades, overuse of chemical fertilizers and fungicides to proliferate and protect greenhouse vegetables has been a threat to human safety. Using biological agents is the best alternative for chemical treatments (Pilkington et al., 2010). Soil-borne fungus Fusarium oxysporum Schlecht. f. sp. lycopersici (Sacc.) Snyder and Hansen (FOL) causing Fusarium wilt of tomato

reduces yield in greenhouses (Fravel et al., 2003; McGovern, 2015). Invading through the vascular tissue and soil- borne feature of the pathogen make it difficult to control this disease. Besides, new races of the mentioned pathogen (e.g., race 3) have emerged that can overcome host resistance (Reis et al., 2005). The symptoms of Fusarium wilt disease caused by race 3 are yellowing the lower leaves, vascular necrosis, epinasty, defoliation, plant stunting and ultimately plant death (Sanchez-Pena et al., 2010). Biological control of this pathogen has been the subject of many studies (Ben Abdallah et al., 2016). Plant growth promoting rhizobacteria (PGPR) colonize rhizosphere or plant root and improve plant health and growth. Some of the most important plant growth promoting (PGP) activity include diazotrophic nitrogen fixation, siderophore production, solubilization of mineral phosphates and production of hormones such as indole-3-acetic acid (IAA) (Sadeghi et al., 2012; Dahal et al., 2017; Gouda et al., 2018). Plant defense strategies including physical barriers, numerous secondary metabolites and antimicrobial agents, which the effective, aggressive pathogens have to overcome them (reviewed by Bruce and Pickett, 2007). Some plant defense strategies are constitutive while others are inducible and only launch in response to a stimulating pathogen and/or beneficial microbes (Pieterse et al., 2012). Plant defense strategies keep damage of specific pests below an economic threshold; however, maintain the beneficial organisms involved in integrated pest management (IPM) (Chellemi et al., 2016). IPM considers available pest control techniques that prevent the development of pest populations and keep pesticides to levels that are economically reasonable and reduces risks to human health and the environment. Developing sustainable biocontrol measures for managing Fusarium wilt disease requires a comprehensive understanding of the molecular basis of plant–pathogen–biocontrol interactions. Hormones, PR proteins, terpenoid synthases, polyketide terpene synthases, peroxidases, lignin synthases, transcription factors, calcium signal transducers, and UDP-glucosyltransferases and ubiquitinprotein ligases are components of the plant defense-related genes (Wang et al., 2015). Generally, the plant responses to microbes are regulated through signaling pathways including salicylic acid (SA), jasmonic acid (JA), and ethylene (ET). Jasmonic acid pathway is required for defense against necrotrophic pathogens and chewing insects, while SA pathway is involved in a wide range of plant defense responses, which ends to systemic acquired resistance (SAR) and occurs following the exposure to many biotrophic and some necrotrophic pathogens. Induced systemic resistance (ISR) is also an activated resistance process elicited by contacting with non-pathogenic microorganisms. This procedure is independent of SA and is synchronized by JA and ET (Walters and Heil, 2007). Activated induced resistance (via ISR or SAR) is a broad-spectrum and long-term resistance, which usually suppresses a disease up to 20–85% (Walters et al., 2013). Thus, inducing plants by direct interaction with rhizobacteria prior to pathogen infection, so-called priming, decreases disease severity (Beckers and Conrath, 2007). Streptomyces species, are Grampositive filamentous bacteria, reported as PGP and biocontrol agent of Alternaria alternata (Verma et al., 2011), Rhizoctonia solani (Goudjal et al., 2014), F. oxysporum (Kim et al., 2011), Phytophthora drechsleri (Sadeghi et al., 2017), and Verticillium dahliae (Cao et al., 2016). Little is known about bio-suppression of tomato wilt by Streptomyces in greenhouse conditions.

The aims of this study were to (1) isolate and characterize Streptomyces from vegetable greenhouse soils (2) detect PGP characteristics and antagonistic activity of isolates against FOL race (3) evaluate Fusarium wilt biocontrol in tomato by superior isolates in greenhouse condition (4) establish induced systemic resistance (ISR) of tomato through inducing antioxidant enzymes and defense-related genes by inoculation of plants with biocontrol Streptomyces.

#### MATERIALS AND METHODS

#### Microorganisms

Rhizosphere soil samples (500 g soil with pH = 6.2–7 and EC < 2.5 dS/m) were collected from 18 cucumber and tomato commercial greenhouses in Yazd, Isfahan and Kerman provinces of Iran in 2016. There were no symptoms of wilt and damping off diseases developing in these greenhouses. The samples were placed in plastic bags, taken to the laboratory and then air-dried for 7 days. For isolation of Streptomyces, 2 g rhizosphere soil was suspended in 100 mL of sterile saline solution (0.9% NaCl) and shaken for 30 min. Two dilutions (1:100 and 1:1000) were prepared using sterile saline solutions in a final volume of 1 mL. An aliquot of 0.1 mL of each dilution was spread on 1.8% water agar supplemented with 300 mM NaCl. The plates were incubated at 29◦C for 7 days. Representative colonies were selected and streaked on plates of "International Streptomyces Project media 2" (ISP2) medium (Kampfer et al., 1991) containing 10 g/L malt extract, 4 g/L yeast extract, 4 g/L glucose, and 18 g/L agar, adjusted to pH 7.2. The plates were incubated at 29◦C for 7 days. Then morphologically (color, size, and shape) distinct colonies were stored in 30% glycerol solution at −70◦C (**Table 1**).

The fungal pathogen was isolated from tomato plants displaying disease symptoms and pathogenicity test was conducted using root-dipping inoculation method on tomato seedlings. The fungus was re-isolated from the vascular tissue of a symptomatic plant. Fungal DNA extraction was carried out by the methods described by Zhang et al. (2010), Salehi et al. (2018, 2019). Polymerase chain reaction (PCR) was performed with universal pair primers (ITS4-ITS5). Race determination of fungus was done using four specific primer pairs designed by Hirano and Arie (2006) based on the sequence of the exo-polygalacturonase gene (**Table 2**). ITS4-ITS5 nucleotide sequence obtained by DNA amplification of fungal pathogen was deposited in NCBI GenBank with accession number MG670445. Blast analysis showed that this pathogen had 99 % similarity to F. oxysporum. The result of the race determination using four specific primer pairs showed that the fungal isolate belongs to F. oxysporum f.sp. lycopersici race 3 (**Supplementary Figure S1**).

#### Growth on Nitrogen-Free Medium

Screening of the free-living (non-symbiotic) diazotroph isolates was carried out according to the procedure described by Xu and Zheng (1986). Each colony was spot-inoculated on nitrogen-free


The means halo zone diameter/ colony diameter of three replications ± standard deviation is represented. 1, growth; 0, no growth. The data indicated for IAA production was in µg/mL.

agar medium incubated at 29◦C for 14 days, and then growth or lack of growth was compared to ISP2 medium.

#### Inorganic Phosphate Solubilization

Pikovskaya's medium (PVK) was used to measure calcium phosphate [Ca<sup>3</sup> (PO4)2]-solubilizing activity. Sterilized PVK medium with pH 7.2 was poured into petri plates. After solidification of the medium, bacterial isolates were spotinoculated onto the center of the plates and incubated at 29◦C for 7 days. Solubilization index was evaluated according to the ratio of the clear zone diameter to colony diameter (Soltani et al., 2010).

#### Siderophore Production

Siderophore production was evaluated as described by Alexander and Zuberer (1991) on Chrome Azurol agar (CAS) medium. Streptomyces isolates were spot-inoculated onto the center of the plate and incubated at 29◦C for 7 days. The formation of orange halo around the colonies were considered as siderophore-producing isolates. After 3 days, the ratio of the halo zone diameter to colony diameter was calculated as siderophore production.

#### IAA Production

To determine amounts of IAA produced by each isolate, 1 mL of bacterial culture in ISP2 broth was inoculated in Tryptic Soy Broth supplemented with 150 mg/L L-tryptophan. Approximately 1 mL of culture solution was centrifuged at 12000 rpm for 5 min, and 1mL of the supernatant was mixed with 2 mL of Salkowski's reagent (150 mL concentrated H2SO4, 250 mL distilled water, 7.5 mL 0.5 M FeCl3.6H2O) and incubated for 20 min in darkness at room temperature (de Oliveira-Longatti et al., 2014). IAA production was qualitatively assayed as pink color development and quantified by measurement of absorbance at 530 nm using a spectrophotometer infinit M200 (Tecan, Switzerland). The calibration plot was constructed using dilutions of a standard IAA (Fluka, Switzerland) solution and the uninoculated medium with the reagent as a standard curve (0, 5, 10, 20, 50, and 100 µg/mL). The quantity of IAA in the culture was expressed as µg/mL.

#### Antagonistic Effect of PGPRs

Bacterial suspension of candidate PGPR isolates (**Table 1**) (20 µL of the 10<sup>8</sup> cfu/mL sterile saline solution) was inoculated at 29◦C linearly at the two opposite sides (1 cm from the plate edge) of potato dextrose agar (PDA) plates for 48 h. Then, one fungal


plug (0.5 cm diameter) was inoculated at the center of PDA plate (Kunova et al., 2016) and then incubated for 4 days. The growth inhibition percentage was calculated using the formula (a − b)/a × 100, where "a" is the fungal growth radius of a control culture (in cm) and "b" is the distance of the pathogen growth in the direction of bacteria (in cm).

#### Chitinase Activity

Chitinase production was determined according to the method of Hsu and Lockwood (1975). Streptomyces isolates were grown on chitin agar containing 0.4% colloidal chitin and 1.5% agar adjusted to pH 7.2. Colloidal chitin was prepared according to Berger and Reynolds (1958). Plates were incubated for 5 days at 29◦C. The ability of chitinase production was shown by a clear halo around the colonies. The ratio of the clear zone diameter to colony diameter was calculated as chitinase activity.

#### Cellulase Activity

Carboxymethyl cellulase (CMCase) activity was determined by Mandels-Reese medium with carboxymethyl cellulose (CMC) as sole carbon source (Majidi et al., 2011). The bacteria were grown on CMC agar containing 0.4 g/L KH2PO4, 0.02 g/L CaCl2, 0.02 g/L NaCl, 0.02 g/L FeSO<sup>4</sup> 7H2O, 2.5 g/L CMC, and 15.0 g/L agar. The pH was adjusted to 7.2 with 1 M NaOH. The CMC agar plates were incubated at 29◦C for 7 days to allow the secretion of cellulase. To visualize the hydrolysis zone, agar medium was flooded with an aqueous solution of Congo red (1 mg/mL) for 20 min. Congo red solution was then poured off, and the plates were further treated by flooding with 1 M NaCl for 15 min. To indicate cellulase activity, the diameter of the clear zone around each colony was measured. The ratio of the clear zone diameter to colony diameter was calculated as cellulase activity.

#### Protease Activity

Extracellular protease activity of Streptomyces isolates was assayed by a modification method of Mehrotra et al. (1999). Each bacterial isolate was streaked on skim milk agar containing 15 g/L skim milk powder, 0.5 g/L yeast extract and 10 g/L agar. After incubation at 29◦C for 48 h, the ratio of the clear zone around bacterial colony to colony diameter was measured as protease activity.

#### Phenotypic and Molecular Characterization of the Selected Isolates

The potent PGP and antagonist isolates were further characterized by differential morphological traits on ISP2, ISP3, and ISP4 media (Shirling and Gottlieb, 1966), melanin formation (Suter, 1978), growth in high temperatures (37◦C and 42◦C), and growth on medium supplemented with 6, 10, and 12% NaCl (Kutzner, 1981).

Molecular characterization of the selected isolates was done using PCR amplification of 16S rRNA gene sequence. DNA was extracted according to the method described by Tripathi and Rawal (1998). PCR amplification was performed using the primers 27F: 5<sup>0</sup> -AGAGTTTGATCCTGGCTCAG-3<sup>0</sup> and 1525R: 5 0 -AAAGGAGGTGATCCAGCC-3<sup>0</sup> as described by Chun and Goodfellow (1995). Almost-complete 16S rRNA gene sequences (1400 nt) were deposited in the GenBank database under the accession numbers of MG722685 (strain IT25), MG786938 (strain TO612), MG786894 (strain Y17), MG786896 (strain Y28), MG654776 (strain IC10), and MG676358 (strain IC13). The sequences were aligned manually with corresponding sequences of available Streptomyces species deposited in the GenBank, EMBL and DDBJ databases using BLAST search tool (Altschul et al., 1997). Phylogenetic tree was constructed using the MEGA 6.0 software package (Tamura et al., 2013) based on neighbor-joining method. Bootstrap analysis was used to evaluate the stability of relationships based on 1000 resampling. Strains IC10 and Y28 were preserved and deposited in the Agricultural Biotechnology Research Institute of Iran Culture collection (ABRIICC) under accession numbers of ABRIICC 20108 and ABRIICC 20114, respectively.

#### Greenhouse Experiments

Tomato (Solanum lycopersicum L.) cv. Rio Grande susceptible to FOL races 2 and 3 (Barker et al., 2005), was used in greenhouse experiments. Seeds were surface sterilized with 1% sodium hypochlorite (NaOCl) and incubated in a growth chamber at 25◦C for 7 days. Germinated seedlings were placed into 84 cell plug tray (50 × 30 × 5 cm deep) filled with sterilized soil and peat moss (1:2), with one seedling occupying each cell. Seedlings were watered every day with tap water and kept at 27◦C and 16 h brightness/8 h darkness. After 21 days, the seedlings were transferred to pots (15 × 20 cm) filled with a sterile mixture of field soil and peat moss (2:1). For bacterial treatments, Streptomyces cell and spore were centrifuged at

8000 rpm for 15 min and then the pellet re-suspended in 10 mL sterile saline solution. Bacterial suspension was added to autoclaved sand and final cfu/g sand adjusted to 10<sup>6</sup> . Seven gram of sand containing bacteria was added to the surface of each cultivated pot immediately after transferring of plant to the pot. Sterilized sand was used as a negative control. The pots were watered every 2 days. For FOL inoculation, 10 days after bacterial treats, tomato seedlings were uprooted and inoculated with conidial suspension for 30 min (root-dip method). Seedlings submerged in sterile distilled water (mock inoculation) were used for bacterial treatments and negative control. To prepare FOL suspension (pathogen inoculum), fungi was cultivated on PDA for 10 days. Conidia were harvested by scraping in sterile water (10 mL/plate) and final conidia/mL adjusted to 5 × 10<sup>7</sup> using hemocytometer. The treatments were negative control plants (mock inoculation), positive control (FOL inoculated), Carbendazim <sup>R</sup> (soil drenched with fungicide in a concentration of 1.5 g/L) and six Streptomyces treatments inoculated or noninoculated with FOL. The plants were harvested 60 days after transferring to the pots. The growth parameters comprising of shoot length, fresh and dry weight of shoot and root were measured and recorded. Further, disease severity (DS) was assessed on a scale from 0 to 5: 1 = symptoms free = 0%; 2 = slight chlorosis, stunting, or wilting = 25%; 3 = moderate chlorosis, stunting, or wilting = 50%; 4 = severe chlorosis, stunting, or wilting = 75%; 5 = death = 100% (Marlatt et al., 1996). Percentage of control value was calculated using the formula (DC − DT)/DC × 100, where "DC" is disease index of inoculated control (FOL) and "DT" is disease index of inoculated treatment (%) (Song et al., 2004).

A greenhouse experiment was conducted separately for the next two tests. The procedure of this experiment was similar to the first one, but the experiment duration was shorter and the plants were harvested 14 days after transfer to the pots, also 1 cm of the root end of the seedlings was cut before FOL/mock inoculation. Foliar was sprayed with plant hormones (1 mM SA or 10 mM MeJA) 48 h before FOL inoculation. The treatments were negative control plants (mock inoculation), positive control (FOL inoculated), Carbendazim <sup>R</sup> (soil drenched with fungicide in a concentration of 1.5 g/L) and two Streptomyces treatments (strains IC10 and Y28) inoculated or non-inoculated with FOL. Plants leaves were harvested 2 days after FOL inoculation, frozen in liquid nitrogen and kept at −70◦C for the following analysis.

#### Antioxidant Enzymes Activity

Frozen leaf samples (100 mg fresh weight) with 10 mg polyvinyl pyrrolydone (PVP) was transferred to 1.5 mL tube and homogenized in 1 mL Na-Pi buffer (1 mM, pH 7). The homogenate was centrifuged at 15000 × g for 15 min. All operations were performed at 4◦C. The supernatant was used as a crude enzyme extract. The activities of antioxidant enzymes including peroxidase (POX: EC 1.11.1.7) and catalase (CAT: EC 1.11.3.6) were measured in a reaction containing 250 µL 0.1 M phosphate buffer (pH 7.0), 250 µL 10 mM guaiacol, 30 µL H2O2, and 40 µL crude enzyme extract (Chance and Maehly, 1955). The enzyme activity (U/mL) was measured spectrometrically (Cary 300, United States) by monitoring of the degrading H2O<sup>2</sup> by the increase in the absorbance at 470 nm and the decrease in the absorbance at 290 nm over 3 min.

#### Quantitative Real-Time PCR Analysis of the Defense-Related Genes

Total RNA was extracted from shoots using RNeasy plant mini kit (QIAGEN). cDNA was synthesized using 1 µg of each RNA sample after treating with RNase-free DNase I (Invitrogen) using iScript cDNA synthesis kit (BioRad) according to manual description. Quantitative PCR was performed in a 25 µL reaction containing 1 µL of template cDNA, 0.5 µL of 10 pM of each forward and reverse specific primer (**Table 2**) and iQ SYBR Green Supermix kit (BioRad) on BioRad multicolor real-time PCR detection system. The PCR profile included an initial denaturation step at 95◦C for 3 min, followed by 40 cycles of denaturation (95◦C/30 s), primer annealing (60◦C/30 s) and primer elongation (72◦C/30 s), by a final elongation step (72◦C/ 5 min) and recording melting curves. Results were expressed as the normalized ratio of mRNA level of target gene to internal control tubulin gene (TUB). Changes was estimated by using the Relative Expression Software Tool (REST 2009) (Pfaffl et al., 2002).

#### Statistical Analysis

Statistical analysis was performed using analysis of variance (ANOVA) by Microsoft Excel (Microsoft Corporation, United States) and SPSS version 16.0 (SPSS Inc. Chicago, IL, United States). All data shown are average value of three (in vitro experiments) biological replicates ± SE. The greenhouse experiments were carried out in randomized blocks design with four blocks and there were four biological replications for each treatment. The significance differences of the treatments were evaluated using multivariate generalized linear model (GLM) with Duncan multiple range test post hoc analysis at level of P < 0.05. The relationship between in vitro and in vivo data was studied using bivariate Pearson test at P ≤ 0.01.

## RESULTS

#### Isolation and Selection of Streptomyces Bacteria

A total of 126 colonies displaying Streptomyces appearance including compact colored heaped and wrinkled, waxy, chalky, powdery or velvety colony, were isolated from tomato and cucumber rhizosphere soil collected from Isfahan, Yazd and Kerman (**Figure 1A**). Totally, 106 isolates (84%) were able to grow on solid nitrogen-free medium (**Figure 1B**) of which thirty-eight isolates were able to solubilize inorganic phosphate. Three isolates comprising of IS8, Y7, and Y28 with the ratio 1.9, 0.5, and 0.4 had the most potential to solubilize tricalcium phosphate, respectively (**Figure 1C**). Further, sixtysix percent of 106 isolates were able to produce siderophore (**Figure 1D**). The maximum orange halo due to iron chelation was recorded for isolate TO612 after 7 days of incubation. All 106 isolates produced IAA in a range of 7.0–40.9 µg/mL

of which twenty percent produced IAA more than 27 µg/mL. Thirty-two isolates from 106 showed proteolytic enzyme activity. The greatest halo zone/colony diameter ratio for protease activity was 1.0 and recorded for isolate IC10 (**Table 1**). Fortyfour isolates were found to produce cellulose. The greatest halo zone/colony diameter ratio for cellulase activity was 1.4 which recorded for Y16. Furthermore, biosynthesis of the chitinolytic enzyme was detectable for 22 isolates. Isolates IC13, K40, SS12, IC6, and Y17 were able to produce all three examined hydrolytic enzymes. Overall, 24 isolates containing at least three PGP traits and hydrolytic enzymes activity were selected for evaluation in the following experiments as shown in **Table 1**.

#### Growth Inhibition of the Selected Isolates Toward FOL

Six isolates (25% of the selected isolates) showed more than 30% inhibitory effect against FOL in dual culture test. The highest percentages of growth inhibition were 69, 49, 48, 42, 39, and 38, which were recorded for IC10, IT25, Y17, TO612, Y28, and IC13, respectively (**Table 3**). The biocontrol potential of these isolates (superior PGP antagonistic isolates) was evaluated in a greenhouse experiment.

#### Phenotypic and Genotypic Characterization of the Superior Isolates

Phenotypical characterization was performed using aerial hyphae and spore chains color of 10-days-old bacterial culture on ISP media. On the medium ISP2, isolates TO612 and Y17 were differentiated from IC10, IC13, IT25, and Y28 according to the color of spore chains. On the medium ISP3, isolates Y28, IT25, and IC10 were distinct from the others based on the color of aerial hyphae. Further, on the medium ISP4, isolate IC13 was differentiated from IC10 by its color of aerial hyphae. Also, these two isolates were recognizable on ISP3 according to the color of their aerial hyphae. Isolates IC10 and Y28 were well distinguished from TO612, Y17 and IC13 using ISP media and from IT25 based on melanin production. Isolates IC10 and Y28 slightly were different from each other on ISP2 medium (**Supplementary Figure S2**). Physiological tests showed only IC13 had potential

TABLE 3 | In vitro antagonistic activity against Fusarium oxysporum f.sp. lycopersici by selected isolates.


Values are the means (averaged from three replicates) ± SE. <sup>∗</sup>Same letters within each column represent non-significant difference according to Duncan's Multiple Range Test (P < 0.05).

to grow at 42◦C (**Table 4**). All six isolates were able to grow on NaCl 6 % while only isolates Y28 and Y17 were able to grow on medium containing NaCl 10 % (**Table 4**). The preliminary phylogenetic analysis of the 16S rRNA gene sequences showed IC10 and Y28 were closely related to species of the genus Streptomyces. The phylogenetic tree constructed from 16S rRNA sequences showed that isolates IC10 and Y28 are two members of Streptomyces genus with more than 99.5% sequence similarity to S. enissocaesilis NRRL B-16365<sup>T</sup> and S. rochei NBRC 12908<sup>T</sup> , respectively (**Figure 2**).

#### Tomato Growth Promotion Activity of the Superior Isolates

Strains TO612, Y28, IC10, and IC13 significantly increased shoot length compared to control plants. Strains Y28 and IC10 increased shoot length up to 20% compared to control (**Figure 3** and **Table 5**). Strains Y17, Y28, IC10 and IC13 significantly (p < 0.05) increased fresh shoot weight while only strains IC10 and IC13 increased shoot dry weight. The increases in the shoot fresh and dry weight were in a range 29–36% and 22–37%, respectively (**Table 5**). Streptomyces strains did not increase root fresh and dry weight while induced root lateral branching (hairy roots) compared to control (data not shown).

#### Biocontrol Potential of the Superior Isolates

Biocontrol efficacy of the superior isolates against FOL causing wilt disease was evaluated in the greenhouse and compared to the chemical fungicide Carbendazim <sup>R</sup> . The high level of disease severity (4.3) was observed in positive control. In Carbendazim <sup>R</sup> treatment, disease severity was 2.1. Disease severity in the plants treated with IC10 and TO612 were 1.6 and 1.9, respectively. Strain IC10 increased control value by 12.5% compared to the chemical fungicide (**Figure 3**). Minimum shoot length, fresh and dry weight were related to positive control and minimum fresh and dry root weight were related to Carbendazim <sup>R</sup> . All superior isolates significantly increased shoot length, fresh and dry weight of shoot, compared to positive control (**Table 6**). Strains IC13 increased tomato total dry weight by 30 and 51% compared to control and Carbendazim <sup>R</sup> respectively. Strains Y28 and IC10 with the highest PGP and biocontrol activity were selected to determine their role in the induction of tomatosystemic resistance through antioxidant enzymes and defenserelated genes.

#### Antioxidant Enzymes Activity

Peroxidase activity significantly increased 24 h after inoculation of plants with FOL. Further, MeJA individually increased the enzyme activity at the same time interval. Plants inoculated with FOL in SA or Y28 treatment increased POX activity after 24 h like positive control (FOL). In SA treatment, POX activity increased 48 h after exogenous application. The level of POX activity remained constant in plant treated with FOL and MeJA and also SA and Y28 inoculated with FOL. In soil treated with strain IC10 and inoculated or non-inoculated with the pathogen, POX activity did not increase during the experiment time (48 h) (**Figure 4**). CAT activity of tomato plants increase 48 h after FOL inoculation. Treatments including bacterial strains and plant hormones increased CAT activity after 24 h. Pathogen inoculation slightly suppressed CAT activity in the SA treatment (**Figure 5**).

#### Quantitative Real-Time PCR Analysis of the Defense-Related Genes

Plant inoculation with FOL and MeJA increased UDPglycosyltransferase (UDP) transcripts by 103- and 98-fold,


branching points. Bar indicates 0.01 substitutions per nucleotide position.

respectively, compared to control. Although Streptomyces strains induced UDP transcription, they significantly suppressed gene expression in the plants inoculated with FOL (**Figure 6A**). Treatments including Streptomyces strains and plant hormones and not FOL significantly increased phenylalanine ammonialyase (PAL) transcripts compared to control. Salicylic acid and strain IC10 also retained their induction effect in plants inoculated with FOL (**Figure 6B**). The results revealed that plants treated with SA, IC10, Y28, SA-FOL, and IC10-FOL significantly stimulated (up-regulated) WRKY70 expression. Treatments of FOL and MeJA did not increase relative expression of WRKY70 in tomato plants (**Figure 6C**). Tomato peroxidase (TPX1) transcripts was significantly up-regulated in all treatments including Streptomyces strains and plant hormones. Salicylic

FIGURE 3 | Biocontrol of tomato wilt and stunt caused by FOL using S. enissocaesilis strain IC10 and S. rochei strain Y28 in greenhouse conditions. Data recorded 60 days after bacterial treatment. Stunting is seen in infected plants. Arrows showing wilting symptoms (chlorosis and dried leaves) in FOL inoculated plants.



Data recorded 60 days after bacterial treatment. <sup>∗</sup>Same letters within each column represent non-significant difference at 5%, according to Duncan's tests using the GLM procedure (df = 9, F = 7.16, and p < 0.05).

TABLE 6 | Biocontrol of tomato wilt and stunt caused by F. oxysporum f.sp. lycopersici race 3 (FOL) using superior PGP antagonistic isolates (IT25, TO612, Y17, Y28, IC10, and IC13) in greenhouse conditions.


Data recorded 60 days after bacterial treatment. Disease severity was assessed on a scale of 1 to 5 (1, symptoms free; 5, severe symptoms). <sup>∗</sup>Same letters within each column represent non-significant difference at 5%, according to Duncan's tests using the GLM procedure (df = 11, F = 3.63, and p < 0.05).

acid and strain Y28 also induced TPX1 gene expression in plants inoculated with FOL (**Figure 6D**). The highest lipoxygenase (LOX) gene expression level (38- fold) was detected in Y28. There was a significant down-regulation of LOX gene in FOL and IC10 treatments. On the contrary, MeJA and Y28 stimulated LOX expression. Induced LOX expression was also observed in plants treated with bacteria and inoculated with FOL (**Figure 6E**). In a similar pattern, SA and Y28 significantly increased pathogenesis related protein 1 (PR1) and ethylene response factor 1 (ERF1) transcripts. Strains IC10 and Y28 also

were able to stimulate the PR1 expression in the presence of FOL (**Figures 6F,G**).

#### Correlation Analysis

There was a significant relationship between siderophore and IAA production (r = 0.36). Besides, a correlation was found between growth inhibition and protease activity (r = 0.31). Moreover, growth inhibition of FOL in dual culture assay was significantly correlated to siderophore production (r = 0.64) (**Table 7**). A significant negative correlation (−0.63) was observed between growth inhibition (in vitro assay) and disease severity (in vivo assay).

#### DISCUSSION

Almost half of tomato and cucumber greenhouses in Iran (more than 7000 hectares) is located in Yazd, Isfahan, and Kerman

provinces. Amount and type of organic fertilizer and soil used to prepare the greenhouse growth bed are different in these provinces. Farmers in Yazd and Kerman use sand from the bed of the nearest rivers to their greenhouse and farmers in Isfahan use soil of the greenhouse floor as a raw material to prepare culture bed (private conversation). Our results showed distribution of the bacterial isolates with the Streptomyces appearance in the three provinces is slightly different. Among the isolates



ns, non-significant; <sup>∗</sup>P < 0.05; and ∗∗P < 0.01.

that showed a PGP trait, a relatively stable share (42–46%) belonged to Isfahan. The contribution of the Streptomyces isolated from Yazd or Kerman in each PGP trait was not always constant showing Streptomyces with several PGP traits in the field soil is greater than that of rivers sand. A previous study on Streptomyces showed that bacterial diversity was more affected by the habitat than the population (Davelos et al., 2004). The restriction of some nutrients in soil may be a major contributor to the change in the microbial diversity due to natural selection.

Growth ability on a nitrogen-free medium and IAA and siderophore production were three dominant characteristics of Streptomyces isolated from vegetable rhizosphere soils. A lower percentage (30%) of the isolates had the potential for solubilizing phosphate. Besides, the location of isolation was also effective on this ratio. There was a significant positive correlation between IAA and siderophore production, which may remind a complementary role of the siderophore in plant growth promotion activity (i.e., in addition to its role in antagonistic activity). Such correlation was not observed between other traits involved in antagonist activity and IAA production.

The ability of FOL growth inhibition was very different between Streptomyces isolates showing PGP properties. Isolate CU122 with four PGP traits, was not able to inhibit FOL growth. Failure to inhibit the pathogen growth was correlated to the inability of isolate CU122 to produce chitinase and protease enzymes. On the contrary, isolate IC10 with same in vitro PGP features as well as enzymes activity inhibited FOL growth very well. The correlation of protease and chitinase activity with inhibition of FOL growth was positive and respectively, quantitative and non-quantitative. Only two exceptions did not adhere to this relationship. Isolates TO612 and KH12, which did not show enzyme activity, inhibited FOL growth in an acceptable range (equal to or greater than 30%).

This finding revealed that although chitinolytic activity of Streptomyces and digestion of the fungal cell wall is an effective mechanism of growth inhibition (Gherbawy et al., 2012), other mechanisms such as antibiosis and even an enzyme like protease (Xu et al., 2016) that is less studied are also involved in antagonism. Interestingly, there was a significant correlation between IAA and siderophore productions and FOL growth inhibition, showing relationship between PGP features and antagonistic activities in Streptomyces. IAA production is a common characteristic among antagonistic species of Streptomyces (Sreevidya et al., 2016). It seems that IAA production alone does not cause plant growth, and other traits, such as siderophore production, also has an important role in increasing plant growth. In this regard, Kunova et al. (2016) showed that the antagonistic isolates producing IAA did not increase growth of different vegetable species. These isolates were not able to produce siderophore. The significant negative correlation that was observed between FOL growth inhibition (in vitro assay) and disease severity (in vivo assay) confirmed a need for in vitro tests to select the best biocontrol strains. Recently, Wu et al. (2019) reported that a decreased percentage disease index of rice sheath blight caused by Rhizoctonia solani was associated with an increase in Streptomyces derived antifungal agent concentration. They concluded that the antifungal agent reduces sheath blight symptoms via exerting a strong antagonistic activity against R. solani both in vivo and in vitro conditions. S. rochei ACTA1551 from Greek (Kanini et al., 2013), S. miharaensis KPE62302H (Kim et al., 2012), and S. psammoticus KP1404 (Kim et al., 2011) from Korea, S. plicatus from Egypt (Abd-Allah, 2001) and S. griseorubens E44G from Saudi Arabia (Rashad et al., 2017) were reported as successful biocontrol agents to control tomato Fusarium wilt. In the previous reports, the pathogen race has not been determined and to our knowledge, this is the first report of biocontrol of Fusarium wilt caused by F. oxysporum Schlecht. f. sp. lycopersici (Sacc.) race 3 in tomato with Streptomyces strains.

Strain Y28 and IC10 were close to S. rochei and S. enissocaesilis, respectively and are grouped in a clade on the phylogenetic tree. These PGP strains were isolated from the rhizosphere soils of two greenhouses in two different locations with a distance of 300 kilometers. The high similarity in PGP, and biocontrol activity of these two strains revealed that evolutionary process can keep PGP traits and biological control activities together.

All plants treated with SA, MeJA and Y28 or inoculated with FOL increased POX activity during 48 h after FOL inoculation. The POX activity in hormone or PGPR treated plants was not affected by FOL inoculation. Despite the similar biocontrol activity, isolates IC10 and Y28 had a different effect on the induction of plant POX activity. None of the plants treated with IC10, inoculated or non-inoculated with FOL, increased POX activity. Peroxidase is considered an important pathogen-related protein (PR-protein) or defense protein involved in many physiological responses of plants to biotic stresses. Contribution to biosynthesis of lignin (Chittoor et al., 1997) and antimicrobial compounds such as phytoalexins and quinones (Mayer, 2006) are two well-known POX roles associated with ISR. Increased peroxidase activity of cucumber was reported in plants showing Fusarium wilt (Zhao et al., 2012) and also plant treated with Streptomyces as biocontrol agent (Salla et al., 2016).

Peroxidase, as well as CAT, are involved in plant antioxidant defense system and reduce the harmful effects of stresses by scavenging of ROS (Das and Roychoudhury, 2014). In this study, the effects of FOL, plant hormones and PGP strains

to induce CAT activity were the same. Induction of systemic resistance through defense-related enzymes POX and CAT in tomato plants treated with salicylic acid or Pseudomonas fluorescens was reported by Nikoo et al. (2014). They showed that pathogen inoculation of plants treated with bacterial or hormonal elicitors increased both POX and CAT activity in higher levels compared to non-inoculated treated plants or plants only inoculated with pathogen. This association was not observed in our experiment.

Relative expression of several candidate genes encoding transcription factors (WRKY70 and ERF1), PR1, TPX1, UDP, LOX, and PAL was evaluated in this study. The plant-specific transcription factor WRKY70 is an important factor in Arabidopsis signaling pathways and its expression is activated by SA and repressed by JA. Ren et al. (2008) showed that overexpression of WRKY70 reduced JA responses and mutually JA treatment inhibited WRKY70 expression. On the contrary, WRKY70 is known as a positive regulator of SA-mediated defense because increases PR1 which is often studied as a marker gene for SA-dependent defense signaling (Li et al., 2004). It is reported that PGP Bacillus cereus AR156 stimulated the transcription of WRKY70 in Arabidopsis leaves. According to their study, WRKY70 modulated B. cereus-triggered ISR through activating SA signaling pathway (Wang et al., 2018). Our results are in accordance with the reported studies and showed induction of WRKY70 transcription upon treatment with SA, and not with MeJA, in FOL inoculated and noninoculated plants. IC10 and Y28 also significantly induced transcription of WRKY70 and IC10 continued to induce gene expression after FOL inoculation. Following increased WRKY70 transcripts, expression of PR1 increased in SA and Y28 treatments. Great increase in PR1 transcript abundance by Y28 and not IC10 indicates that these two PGPRs stimulate different pathways to elicit defense priming in tomato plant.

UDP-glucose dependent hydroquinone: O-glucosyltransferase (arbutin synthase) is a member of glycosyltransferases catalyze production of arbutin in higher plants (Arend et al., 2000a). Arbutin is a phenolic glycoside that has antimicrobial and antifungal activity (Kundakovic et al., 2014). Arbutin synthase is a multifunctional enzyme converting various natural products, xenobiotics and toxins (Arend et al., 2000b). There is no report about induction of UDP in response to F. oxysporum or Streptomyces PGPRs in tomato plant. A previous report revealed that induction of members of UDP family, UGT73B3 and UGT73B5 is necessary during the hypersensitive response of Arabidopsis to plant pathogen Pseudomonas syringae (Langlois-Meurinne et al., 2005). Also treatments of Arabidopsis with SA and MeJA induced the expression of another gene family member UGT73C5 involved in mycotoxin detoxification (Poppenberger et al., 2003). Increased UDP transcript in FOL inoculated tomato can be attributed to the role of arbutin in detoxification of fungal toxin or its antimicrobial effect (Kuzniak et al., 2015). The increased expression of this gene in treatment of MeJA and not SA represents its role in ISR. Compared to FOL inoculation, a significant lower level of UDP expression was observed in plants treated with IC10 and Y28 highlighting the potential role of these PGP strains in tomato defense priming.

FOL, SA, MeJA and PGPRs induced expression of TPX1. TPX1, peroxidase encoding gene, is involved in the synthesis of lignin and suberin (Quiroga et al., 2000). Here, TPX1 expression induction by Y28 and IC10 revealed their role in ISR. TPX1 expression at the lower level was observed 48 h after FOL inoculation. Different POX activities in plants treated with PGPRs could be related to the TPX1 transcript abundance.

Phenylalanine ammonia lyase encoded by PAL is involved in the biosynthesis of salicylic acid and defense compounds including flavonoids, phenylpropanoids and lignin. Induction of PAL activity in response to various stimulants such as tissue plant pathogens and hormones was reported (Edwards et al., 1985). In this study, induced expression of PAL in bacterial and hormonal treatments was observed and there was no difference between the induction effects of bacterial strains. The induction of PAL enzymatic activity and PAL gene expression in tomato plants upon pre-treatment with Bacillus thuringiensis (Akram et al., 2013) and the mycorrhizal fungus Funneliformis mosseae (Song et al., 2015) were also reported.

Members of the ERF protein family are transcription factors involved in ET/JA or SA signaling pathways and cause moderate disease resistance responses in various plant species (Liang et al., 2008). Induced expression of ERF1 in Arabidopsis plants treated with a PGP biocontrol strain of Paraburkholderia phytofirmans in response to a model plant pathogen P. syringae pv. tomato DC3000 was reported recently (Timmermann et al., 2017). Likewise, induction of ERF1 by treatment of Y28 tomato was observed in the present study too. On the contrary, plants treated with IC10 decreased ERF1 expression. Our data supports the hypothesis that studied PGP strains regulate tomato defense responses through different molecular pathways possibly acting at transcriptional level.

LOX is a signaling molecule involved in pathogen plant resistance. Strain Y28 slightly induced LOX expression in non-pathogen inoculated plants. FOL inoculation and IC10 treatment caused a significant decreased in transcript level of LOX compared to the control plants. FOL inoculation of both bacterial treatments caused an induction of gene expression. Our finding is consistent with Mariutto et al. (2011) that showed tomato plants treated with PGP Pseudomonas putida did not induce LOX transcription before inoculation with the pathogen B. cinerea. Although transcripts of LOX for both strains were higher in FOL inoculated plants, gene expression levels in non-inoculated bacterial treatments were different significantly.

#### CONCLUSION

The superior PGP antagonistic Streptomyces strains, show biocontrol activities against Fusarium wilt of tomato caused by F. oxysporum Schlecht. f. sp. lycopersici (Sacc.) race 3. Significant positive correlation between siderophore and IAA production and siderophore accumulation and growth inhibition showed a relationship between PGP and antagonistic traits. A significant

negative correlation between growth inhibition (in vitro assay) and disease severity (in vivo assay) confirmed a need for in vitro tests to select the best biocontrol strains. The correlation of PGP traits with biocontrol activity in phylogenetically close species isolated from distant habitats refers to the role of the natural selection in preserving traits that give superiority to bacteria in the rhizosphere. Here we showed biocontrol Streptomyces stimulate plant defense system through different molecular pathways at the transcriptional level.

#### AUTHOR CONTRIBUTIONS

NS and AS designed and directed the research. MS gave advice during experiments. SA carried out all experiments. All authors contributed to the interpretation of the results. SA wrote the manuscript with help from NS and AS. All authors read and approved the final manuscript.

#### REFERENCES


#### FUNDING

This work was supported by the Agricultural Biotechnology Research Institute of Iran (ABRII) (Grant No. 12-05-05-034- 09454-950941).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.01505/full#supplementary-material

FIGURE S1 | Identification of Fusarium oxysporum f. sp. lycopersici (FOL) physiological race 3 by selective primers. Polymerase chain reactions carried out using unif, sp13, sp23, and sprl primer sets. 1 kb: I kb ladder; C, negative control.

FIGURE S2 | Bacterial colonies, S. enissocaesilis strain IC10 and S. rochei strain Y28, on ISP2 medium 10 days after cultivation.


lycopersici. J.Gen. Plant Pathol. 72, 273–283. doi: 10.1007/s10327-006- 0287-7



Arabidopsis thaliana similarly to flg22. Mol. Plant. Microbe Interac. 31, 311–322. doi: 10.1094/MPMI-10-17-0240-R


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

Copyright © 2019 Abbasi, Safaie, Sadeghi and Shamsbakhsh. 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.

# Plant Host-Associated Mechanisms for Microbial Selection

Piet Jones 1,2, Benjamin J. Garcia<sup>1</sup> , Anna Furches 1,2, Gerald A. Tuskan<sup>1</sup> and Daniel Jacobson1,2 \*

*<sup>1</sup> Oak Ridge National Laboratory, Biosciences Division, The Center for Bioenergy Innovation, Oak Ridge, TN, United States, <sup>2</sup> The Bredesen Center for Interdisciplinary Research and Graduate Education, University of Tennessee, Knoxville, Knoxville, TN, United States*

Plants serve as host to numerous microorganisms. The members of these microbial communities interact among each other and with the plant, and there is increasing evidence to suggest that the microbial community may promote plant growth, improve drought tolerance, facilitate pathogen defense and even assist in environmental remediation. Therefore, it is important to better understand the mechanisms that influence the composition and structure of microbial communities, and what role the host may play in the recruitment and control of its microbiome. In particular, there is a growing body of research to suggest that plant defense systems not only provide a layer of protection against pathogens but may also actively manage the composition of the overall microbiome. In this review, we provide an overview of the current research into mechanisms employed by the plant host to select for and control its microbiome. We specifically review recent research that expands upon the role of keystone microbial species, phytohormones, and abiotic stress, and in how they relate to plant driven dynamic microbial structuring.

#### Edited by:

*Nicolae Radu Zabet, University of Essex, United Kingdom*

#### Reviewed by:

*Laila Pamela Partida-Martinez, Center for Research and Advanced Studies (CINVESTAV), Mexico Rong Li, Nanjing Agricultural University, China*

> \*Correspondence: *Daniel Jacobson jacobsonda@ornl.gov*

#### Specialty section:

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

Received: *05 February 2019* Accepted: *14 June 2019* Published: *03 July 2019*

#### Citation:

*Jones P, Garcia BJ, Furches A, Tuskan GA and Jacobson D (2019) Plant Host-Associated Mechanisms for Microbial Selection. Front. Plant Sci. 10:862. doi: 10.3389/fpls.2019.00862* Keywords: microbial community, jasmonic acid, salicylic acid, ethylene, keystone species, abiotic stress, biotic stress, microbiota

#### 1. INTRODUCTION

The sessile nature of plants limits their capacity to deal with an immediate and localized disturbance, irrespective of whether the disturbance is caused by biotic or abiotic stress. It therefore stands to reason that plants have evolved systems to manage the impact of these collective and respective stresses. From a biotic microbial view point, plants play host to a number of organisms that reside in the phyllosphere, endosphere, and rhizosphere, influencing how a plant reacts to its environment. If viewed in the context of an ecological unit, the community of organisms is known as the holobiont. Further incorporating the environment results in what is collectively known as the phytobiome, where the possible plant-microbe-stress interactions are given in **Figure 1**.

The holobiont has a much greater evolutionary potential for dealing with biotic and abiotic stress than the plant itself. Therefore, it is potentially more sustainable to manage abiotic/biotic stresses in a holistic and multifaceted manner. The plant employs a combinatorially complex system of receptors and signals to adapt to different stressors (Hacquard et al., 2017). Unraveling the complexity of the system is not a trivial task, with researchers providing different perspectives for elucidating a contextual understanding of the dynamics of plant-microbiome interaction.

The improved understanding of the interactions between the plant and its microbiome has broadened our knowledge on the capabilities of the plant to influence its microbiome and vice versa. In interacting with its microbiome, plants have the capacity to release chemical signals into their environment. The signals can either have a positive or negative effect on other plants or members of the microbiome. Root exudates, comprised of allelochemicals, have been associated with signaling in plant-microbe interaction and can also facilitate plant to plant communication (Bais et al., 2004). Exudates with potential allelopathic properties can help the plant both positively and negatively select for members of their phytobiome (Bertin et al., 2003; Sasse et al., 2018), allowing the plant to establish a rhizosphere and soil microbiome that may also be beneficial or detrimental to other plants and microbes. The concept of influencing the plant phytobiome has also been explored in biocontrol strategies, e.g., strategies against nematodes (Stirling, 2017). The ability of the plant, together with individual members of its microbiome, to control and shape the overall microbiome influences a plant's growth and stress response. A better understanding of the resultant interplay between defense and control may allow for an optimized holobiont that can benefit, among others, agricultural and bioremediation efforts (Ojuederie and Babalola, 2017; Pappas et al., 2017; Ab Rahman et al., 2018).

The microbes that a plant hosts are broadly classified as pathogenic or non-pathogenic. The nature of the nonpathogenic interaction may be beneficial, mutualistic, commensal, or neutral and pathogenic interactions may be parasitic or amensal. The plant can play host to biotrophs, who receive nourishment from a living host cell, necrotrophs, who receive nourishment from a dead host cell, and hemitrophs, who switch between the different nourishment states (see **Figure 2**). The nature of plant-microbe relationships and the ability of the plant to influence select neighbors may potentially benefit the plant's own growth or defense (Bulgarelli et al., 2013; Mikicinski et al., 2016 ´ ). Nutritional deficiencies in the plant can also alter the transcriptional profile of its microbiome and thereby mitigate the impact of nutritional stress (Carvalhais et al., 2013). Additionally, the plant's abiotic stress tolerance and disease resistance may not be a mutually exclusive processes. For example, common mechanisms exist between phytohormone-based pathogen defense responses and the plant's ability to tolerate drought and salt stress (Cho et al., 2013; Huang et al., 2018). At a community level, evidence suggests that the ability of the plant host to shape its microbial community may also serve as an additional layer of defense to disease and stress (Hacquard et al., 2017; Berendsen et al., 2018).

Interaction between plants and microbes has generally been studied with respect to individual (or small collections of) microbes. Additionally, the focus has often been on how the microbe negatively or positively impacts the plant and not how a plant may influence the selection of microbes under biotic or abiotic stress. However, there is dynamic feedback between a plant and a microbe that may impact the plant's microbial community (Bever et al., 2012). The reciprocal nature of the interactions between the plant and its microbial community is poorly understood. It is therefore not only important to determine how individual microbes may impact the plant, but also how the plant may impact its community. There is also evidence to suggest that the plant itself may not be the only determinant in microbial diversity, suggesting the environment and abiotic factors are also important to consider when identifying plant-microbial interactions (Kennedy et al., 2004; Whitaker et al., 2018). There is still much that we can learn about the interplay between the different elements of the phytobiome, including (1) how do the individual microbes influence the plant-mediated structure, (2) how can the plant shape its microbiome through signaling and nutrient availability, and (3) what role does the abiotic stress play in the plant's ability to influence microbial community members and structure? We review recent contributions to the understanding of the impact that keystone members of the microbiome may have on the plant and other community members. We also review the current understanding of the extent to which a plant may shape its microbial community. Finally we look at how abiotic stress on the plant influences the microbial community. Each of the mechanisms that a plant employs to interact with its microbial community and abiotic stresses is part of a complex dynamic system that influences plant growth and stress tolerance.

#### 2. KEYSTONE MICROBIAL SPECIES

Improving our ability to manipulate plant phenotypes, including growth, is of great interest for agricultural, industrial, and ecological restoration efforts, among others. Understanding the mechanisms that underpin pathogen resistance, abiotic stress tolerance, and range shifts (including the ability to establish crops in marginal agricultural land) is of particular importance in the face of shifting climate conditions. An important mechanism through which a plant shapes its microbiome and larger ecosystem is through interactions with keystone microbial species. The concept of keystone species was developed by ecologist Robert T. Paine (Paine, 1966, 1969) and has since been applied to many different fields and with varied meanings. The classical idea of a keystone is attractive because it represents a top-down mechanism of ecosystem control that humans can potentially understand and manipulate (Busby et al., 2017; Trivedi et al., 2017; Hamonts et al., 2018). A few well-known macroecological examples in which keystones play important roles in stabilizing species diversity and ecosystem function are marine rocky intertidal zones dominated by the top predator starfish Pisaster ochraceus (Paine, 1966), riparian habitats dominated by extended Populus phenotypes (Whitham

et al., 2003), kelp forests that disappear without otters (Estes and Palmisano, 1974), and the very landscape of Yellowstone National Park following the reintroduction of gray wolves (Ripple et al., 2001).

Interest in applying the keystone species concept to microbial communities and the plant microbiome has grown in recent years (Meyer and Leveau, 2012; Ze et al., 2012; Berry and Widder, 2014; Copeland et al., 2015). In the phytobiome, the concept of a keystone is attractive as a means of understanding how the plant controls its microbiome. Several authors have used computational network approaches and experimental methods to identify putative keystones. However, few studies have attempted to validate the role of putative keystones. Moreover, investigation of the mechanisms by which a plant recruits a keystone microbe for the purpose of manipulating its microbiome and larger ecosystem has only recently begun. Early studies show promising results and suggest that further research into keystone species recruitment would contribute significantly to the fields of sustainable agriculture, conservation, and ecological restoration.

#### 2.1. Identification and Validation of Keystone Species

Identifying and validating keystone species is inherently complex. By definition, classical keystones increase alpha and beta diversity and stabilize ecosystem function via top-down mechanisms and are typically described as being rare or low in abundance, having few direct interactions with other community members. In the phytobiome, keystones are theorized to play vital roles in helping establish the core microbiome (Harrison et al., 2018). In contrast to members of the core microbiome, for which low-level taxonomic composition may vary but ensemble function is conserved (Lemanceau et al., 2017), keystones perform unique functions inherently tied to their identity. While core microbes also perform important functions for the host plant and its community, removal of a core member does not necessarily lead to dissolution of the community. The keystone's status as a community linchpin means its functional role and very presence are often difficult to separate from those of its community members—a "control" community without the keystone simply does not exist or function in any comparable fashion. Furthermore, identifying low abundance and rare microbes that exert top-down control or have an otherwise outsized effect on community structure requires deep or singlecell sequencing of soil and plant samples (Oberholster et al., 2018). However, currently used statistical methods filter out low abundance taxa and those not present in the majority of samples to avoid biasing analyses (Berry and Widder, 2014; Agler et al., 2016; Duhamel et al., 2018; Oberholster et al., 2018), thus rare taxa recovered by deep sequencing are likely to be pruned a priori.

Some studies have taken the approach of focusing on organisms previously theorized to be keystones and then attempting to characterize the putative functional role they fulfill in the community (Agler et al., 2016; Duhamel et al., 2018), while others approach the problem from the opposite direction by seeking out rare organisms that could be candidates for carrying out well-studied functions of interest (Ze et al., 2012). These approaches essentially utilize the classical keystone concept to narrow the hunt for the needle in the haystack. While these studies clearly demonstrate that microbes do fulfill "classical" keystone roles, it is likely that a broader or more flexible definition will become necessary as research in this arena progresses; the "classical" keystone model was developed within the macroecological realm and therefore may not sufficiently encapsulate uniquely microbial keystone characteristics (Meyer and Leveau, 2012). This unexplored frontier means that attempting to find new keystone taxa particularly in uncharacterized systems—is difficult because the characteristics and roles unique to microbial keystones are largely unknown at this stage. In contrast to description of keystones at the macroecological scale, recent studies have found evidence that microbial keystones may be ephemeral, performing a specific function in a particular environment or developmental stage, but playing a less prominent role in others (Duhamel et al., 2018; Oberholster et al., 2018). For example, many vertically transmitted microbes that are dispersed on seed surfaces (such as Alternaria fulva on Astralagus lentiginosusseeds) fulfill important roles in early holobiont development that affect the long-term composition of the phytobiome (Harrison et al., 2018).

Interest in using co-occurrence networks to identify putative microbial keystones has been rapidly growing (Berry and Widder, 2014; Copeland et al., 2015; Agler et al., 2016; Banerjee et al., 2018). Such networks are often created by calculating the correlation of pairwise OTU abundances. Several papers have used computational methods to identify network microbial "hubs" (OTUs that have a high number of connections within the community). However, hub taxa are not necessarily keystones (Berry and Widder, 2014; Rottjers and Faust, 2018). Using co-occurrence as a proxy for biologically meaningful interactions is problematic. Understanding microbiome interactions relies upon properly identifying the true sphere and scale in which interactions take place. The lack of directly observed interactions presents an immense challenge because the niches which partition true interaction from mere co-occurrence are an unknown prior and vary greatly, often in ways not understood by researchers. For example, the interior of a single leaf presents a multitude of ecologically distinct microhabitats. Among niches, microbes may have few or zero interactions but still "co-occur" because they are present in the same macrohabitat, i.e., the leaf (Berry and Widder, 2014; Hamonts et al., 2018). Nevertheless, in uncharacterized microbiomes, network methods can offer valuable insight regarding community dynamics and serve as a first step in the process of identifying the keystone taxa that serve as the linchpins in community structure.

#### 2.2. Evidence for Plant-Driven Keystone Interactions

The majority of studies on microbial keystones have not investigated the mechanisms underpinning host interactions, perhaps because methods to explicitly identify and validate the effects of keystone species are still largely under development. However, there appears to be promising evidence that plants are actively recruiting keystone microbes. Using a combination of field experiments and network analyses to investigate the rhizosphere communities of sunflower and sorghum crops, Oberholster et al. (2018) identified 47 hub taxa and putative keystone microbes deemed to be necessary for maintaining network structure, including Proteobacteria, Rhizoplanes, Flavisolibacter, Povalibacter, Nitrososphaera, Lysobacter, and Sphingomonas. Although investigating plant-keystone-species interactions was outside the scope of the study, Oberholster et al. (2018) found that the abundance of taxa in the rhizosphere varied with soil chemistry and plant developmental stage. Furthermore, changes in soil chemistry were correlated with plant species and plant developmental stage. Phylogenetic diversity of the sorghum rhizosphere was significantly correlated with soil carbon and nitrogen concentrations, whereas sunflower rhizosphere diversity was correlated with potassium, calcium, magnesium, and phosphorus. Together they suggest that differences in plant root exudates likely contributed to the structure of rhizosphere communities, potentially through putative keystone taxa.

Despite the fact that formal investigation of keystone species is still in early days, mycorrhizal fungi have long been recognized as playing an essential role in structuring the microbiome of plant hosts. Both arbuscular mycorrhizal (AM) and ectomycorrhizal (ECM) fungi have been described as keystone species (Gamper et al., 2010; Soka and Ritchie, 2015; Oberholster et al., 2018), and much can be drawn from mycorrhizal fungal host interactions. Duhamel et al. (2018) found a strong association between plant species and rhizosphere communities in greenhouse and field experiments. In greenhouse experiments in which three plant species (Pinus muricata, Baccharis pilularis, and Ceanothus thyrsiflorus) were planted in media containing a tripartite mix of soils from each plant's native range (and thus each plant's native microbial community), the final composition of each plant's microbiome most strongly resembled that of each host plant's native range. The significant similarity in community structure between greenhouse and native soils supports the idea that plants can play an active role in selecting their microbiome structure. In agreement with Oberholster et al. (2018), Duhamel et al. (2018) found that community composition and keystone prominence changed over time. For example, members of the genera Rhizopogon and Suillus (both belonging to Suillineae) are keystone mutualists; these ECM basidiomycetes are abundant during development of P. muricata seedlings and under conditions similar to those that occur during range expansion, but are rare during other phases such as in well-established monodominant stands. Although the mechanism by which P. muricata recruits keystones such as Suillus pungens was not investigated by Duhamel et al. (2018), the obligate relationship of Pinus species with ECM has long been of interest. Suillus species in particular have been shown to facilitate Pinus invasions, as was demonstrated for S. luteus and P. contorta (Hayward et al., 2015). Kikuchi et al. (2007) found that S. bovinus germinated when co-cultured with P. densiflora, and could be induced to germinate in the absence of the host by treatment with flavonoids previously reported from root exudates, including hesperidin, morin, rutin, quercitrin, naringenin, genistein, and chrysin. Liao et al. (2016) found that relationships between members of Pinus and Suillus were species-specific, and that compatible Pinus and Suillus pairings elicited expression of unique gene sets including genes related to production of fungal small secreted proteins and host leucine-rich repeat-containing R proteins. Moreover, the JA and ET pathways were found to be upregulated during incompatible pairings (but interestingly, SA was not).

In contrast to beneficial keystones that increase microbial alpha and beta diversity (Herren and McMahon, 2018), pathogenic keystones tend to reduce diversity. Agler et al. (2016) found a significant correlation between the obligate biotrophic oomycete Albugo laibachii and host genotype in Arabidopsis thaliana. The authors found that phyllosphere alpha and beta diversity were dramatically reduced following infection. Similar to the effects of beneficial keystones, community structure was stabilized in infected hosts relative to pathogen-free hosts. Agler et al. (2016) did not investigate the mechanisms involved in susceptibility or resistance to A. laibachii; however, closely related A. candida has been shown to alter host metabolism in A. thaliana (Chou et al., 2000). Ruhe et al. (2016) showed that rather than killing the host or severely decreasing fitness, A. laibachii maintains a level of infection that is tolerable to the host. Additionally, there is evidence that A. laibachii is largely able to tolerate host defense mechanisms and suppresses only a small portion of host activity. Ruhe et al. (2016) reasoned that the immune tolerance is an adaptation over non-host evolved pathogens that would serve as competitors to A. laibachii or kill the host. Even though A. laibachii is classified as a pathogen due to predominantly negative effects on the host, leaving the host's defense response intact to compete against more virulent pathogens has a less negative impact on A. thaliana. Furthermore, the "keystone" role of A. laibachii is in stabilizing the post-infection community composition, which prevents other pathogens from invading while plant immunity is already compromised.

In consideration of the challenges associated with identifying and validating keystone species, it is not a surprise that there is a shortage of studies investigating plant-driven mechanisms for recruiting or maintaining relationships with keystones. However, some inferences can be made based on previous studies that investigated host mechanisms with regards to a single organism or a simple community (many of which will be described in subsequent sections and should influence the future direction of keystone research). Synthetic (constructed) communities offer a promising avenue for isolating host mechanisms (Bodenhausen et al., 2014). Niu et al. (2017) constructed a seven member synthetic community on maize roots and discovered through iterative removal of members that the community collapsed without Enterobacter cloacae. In addition, when the community was intact, it functioned to suppress Fusarium verticillioides. Niu et al. (2017) did not investigate the host mechanisms involved in these interactions, but we argue doing so is a natural next step that will yield valuable fundamental information. Although simplified communities are likely to miss some important hostmicrobe dynamics, they are a good place to start for gaining basic understanding. In addition, the use of microfluidics can facilitate the dissection of complex plant-microbe interactions by facilitating the fine-scale manipulation and imaging of real-time plant recruitment of and colonization by microbes (Massalha et al., 2017; Stanley and van der Heijden, 2017). Furthermore, microfluidic methods allow plant exudates, phytohormones, and internal signaling cascades to be characterized using proteomic, transcriptomic, and other techniques to gain new insight into host mechanisms that operate at specific spatial-scales. To date, these methods have largely been used to study culturable microbes. However, the use of ensemble culturing techniques such as those used by Agler et al. (2016) could facilitate study of unculturable and rare taxa, which would provide more nuanced and realistic insights into holobiont dynamics. Furthermore, the use of such techniques in combination with time-series experiments would inform how plant recruitment of keystones varies with developmental stage, which would in turn improve our understanding of the core microbiome changes over time.

#### 3. PLANT DEFENSE MODULATION OF THE MICROBIOME FROM A HOST GENE AND PHYTOHORMONE PERSPECTIVE

Concepts such as evolutionary pressure and dynamical feedback have shaped our understanding of plant-microbe interactions. A recognized hypothesis of the plant immune system, referred to as the zigzag model, characterizes the defense response as a successive pattern-triggered immunity (PTI) and effectortriggered immunity (ETI), set of responses. PTI is viewed as the first line of defense and involves protein recognition receptors (PRRs) on the cell surface. PRRs bind often conserved microbial compounds referred to as microbe (pathogen) associated molecular patterns, or MAMPs (PAMPs). The binding of microbial compounds by PRRs then elicits a signal cascade of defense responses that inhibit microbial growth. ETI is the second line of defense, comprised of intracellular resistance (R) genes that contain nucleotide binding leucine rich repeat (NB-LRR) domains. Resistance genes code for proteins that bind microbial virulence effector proteins. Binding of the effector proteins then triggers a signal cascade that often results in cell death. PTI can be evaded by the microbe, eliciting successive ETI responses. ETI responses can also be evaded by the microbe, creating an evolution of responses by the plant and evasion by the microbe. However, the model views the PTI and ETI responses as distinct, and implicitly views the PTI responses as more conserved evolutionary than the ETI. The standard PTI-ETI model contradicts observations indicating that PRR may evolve similarly to R genes, and that certain R genes may play a similar role to PRR genes (Cook et al., 2015). The review by Cook et al. (2015) suggests that the plant immune system is an interacting set of co-evolving responses that occur both within and outside the cell, and is a response that involves multiple signal cascades. Phytohormones are a fundamental part of the resultant defense signal.

General defense related phytohormones form part of what is referred to as the plant's systematic acquired resistance (SAR) and induced systemic resistance (ISR) (Pieterse et al., 2012; Fu and Dong, 2013).

Of the various phytohormones, ethylene (ET), jasmonic acid (JA), and salicylic acid (SA) have been classically characterized in some plant defense role (Pieterse et al., 2012) and have been shown to preferentially impact certain bacterial phylla in a community (Carvalhais et al., 2014). There is an emerging interest in the potential of phytohormones to shape the plant's microbial community. Given the importance of the microbial community in plant defense, growth, and sustainability, it is therefore important to understand the hormone-microbial dynamic. We therefore describe recent evidence toward the role of ET, JA, and SA in shaping the microbiome community. Additionally, we look at research into the interplay of the respective hormone biosynthetic pathways and how they may assist in microbial colonization of the plant.

#### 3.1. Ethylene

Originally shown in oats Avena sativa and broad bean, Vicia faba, the volatile hormone ET influences plant growth (Laan, 1934), with many further studies further characterizing the role of ET on plant growth and development (Burg and Burg, 1966; Smalle et al., 1997; Sukumar, 2010). The role of ET in plant defense was suspected due to a measured increased in ET biosynthesis during early PTI response in Nicotiana tabacum (Bailey et al., 1990; Sharon et al., 1993). It later became evident that ET signaling was required for the expression of the receptor kinase (FLS2) which binds bacterial flagelin (flg22) in A. thaliana and thereby triggers the defense response (Mersmann et al., 2010). ET has also been shown to be involved with stress tolerance (Thao et al., 2015). There has been emerging interest in characterizing the role that ET may play in not only defense from an individual microbe, but also in how ET influences the community (Nascimento et al., 2018). Mutant A. thaliana lines have provided an ideal framework to work toward characterization of ET.

A synthetic community approach was used in A. thaliana to determine host genetic factors that may influence phyllospheric bacterial community structure (Bodenhausen et al., 2014). Bodenhausen et al. found that ET-insensitive mutants, which possessed a mutation in the EIN2 gene, displayed a significant shift in the bacterial community structure at the genus level. They identified an increase in the relative abundance of Variovorax, a genus consisting of the metabolically diverse gram negative Variovorax pardoxus (Han et al., 2011). It is difficult to determine if increased abundance is directly associated with ET, or if it is mediated by pathway cross talk, especially given that Bodenhausen et al. observed a significant decrease in Variovorax abundance in the SA-insensitive mutant.

Another experiment involving ein2 mutants also showed a shift in the rhizosphere bacterial community composition in nonautoclaved soil (Doornbos et al., 2011). However, the result was not observed in autoclaved recolonized soil. The recolonization of the autoclaved soil in particular consisted of either species that survived the autoclave process or those species in the surrounding environment, indicating that the initial microbial community composition may play a role in the capacity of ET to influence microbial structure. While the aforementioned hypothesis was not explored in Doornbos et al. (2011), they did observe supporting evidence in that bacterial community shifts were observed prior to disease symptoms, and no significant differences were observed in the absences of defense signaling. The latter indicates that a potential selective pressure triggering a defense response may be required to observe ET-mediated microbial community shifts. An early shift in community before disease symptoms may be justified in that ET is known as a potential early response signal (Mersmann et al., 2010). Therefore, an existing microbial community may provide the necessary pressure to elicit an ET response that can shape the community structure.

There is evidence to suggest that genotype effect on root microbiome is much weaker than the potential effect on the leaf microbiome (Wagner et al., 2016). While the full genotypic differences were not fully characterized in the Wagner et al. study, it was determined that leaves and roots differ in glucosinolate concentrations. Glucosinolate may be regulated by JA and ET signaling during rhizobacterial colonization (Pangesti et al., 2016). Therefore, glucosinolate secondary metabolites may provide a possible strategy for microbial community selection.

#### 3.2. Jasmonic Acid

The role of JA in plant defense was first described as part of an infection-mediated wound response (Farmer and Ryan, 1992). Other associations to wound healing and herbivoryrelated defense have since been observed for components of the lipid-derived hormone's biosynthetic pathway (Li et al., 2005; Schilmiller et al., 2007; Koo et al., 2009; Christensen et al., 2013). JA has also been associated with plant necrotroph defense (Plett et al., 2014; Wei et al., 2016).

There is evidence to suggest that the phytohormone JA may have the capacity to shape the root microbial community by means of root exudates (Bertin et al., 2003; Sasse et al., 2018). In particular, evidence of root-associated allelopathic and chemotactic negative and positive selection for constituents of the microbiome has been discussed in Bais et al. (2004). Arabidopsis thaliana knock-out mutants myc2 and med25 were shown to have disrupted JA signaling pathways that result in attenuated wounding, herbivory, and defense responses as well as altered root exudate profiles (Carvalhais et al., 2015). Carvalhais et al. found correlations between specific exudate concentrations and the abundances of several bacterial microbes. While the roles of these exudate compounds in shaping the microbial community are not yet fully understood, compounds such as tryptophan and fructose are chemotactic to several bacteria (Ordal et al., 1979; Yang Y. et al., 2015). One way JA might facilitate hostdriven selection of the plant microbiome is by fine tuning the concentrations of root exudates that attract various microbes.

Furthermore, a recent root exudate study in maize found benzoxazinoids, which is regulated by JA, exhibited the capacity to alter the composition of the microbial community (Oikawa et al., 2001; Hu et al., 2018). Here, Hu et al. also experimented with the effect of benzoxazinoid inoculation on soil, which identified improved herbivory defense, exhibited genotype dependent growth reduction, and increased levels of JA. It has previously been shown that benzoxazinoids are chemotactic for Pseudomonas putida KT2440, which elicit JA priming and thereby resistance to particular fungi (Neal et al., 2012; Neal and Ton, 2013). A differential secondary metabolite analysis of genotype root exudates in Monchgesang et al. identified differential concentration of glucosinolate, SA catabolites, and dihydrohydroxy JA, indicating JA associated genotypic influences on root exudation (Mönchgesang et al., 2016). JA's influence of root exudates may in turn influence the rhizosphere microbiota, given the strong correlation between genotype root exudation and the rhizosphere bacterial community structure (Micallef et al., 2009). However, direct experimentation is needed to test JA's influence and better understand the potential mechanisms involved.

#### 3.3. Salicylic Acid

The role of SA in plant defense was first described in tobacco, against tobacco mosiac virus (White, 1979), where supplementation of diluted aspirin induced a defense response. SA has since been described as an important component in plant defense signaling (Shah, 2003). It is believed that SA forms part of the plant's defense strategy against biotrophes, as opposed to necrotrophes which are more associated with the JA and ET pathways (Glazebrook, 2005).

In its capacity to regulate the microbiome, SA has been shown to modulate the composition of the root microbiome at the family level in A. thaliana (Lebeis et al., 2015). Arabidopsis knockout mutant lines were used, where essential components of the SA, JA, and ET pathways were targeted. Expectedly, mutants with the three respective pathways knocked out showed a lower survival rate. Apart from the microbial compositional shift, it was shown that certain bacterial endophytic families may actually require SA-related processes to colonize. Exogenous supplementation of SA resulted in an observed altered microbial community profile indicating potential SA-mediated preferential selection for microbial families (Lebeis et al., 2015). Treating ginsing with phenolic acids over a six-year period resulted in rhizosphere fungal community shifts (Li et al., 2018). Li et al. observed dramatic relative abundance changes of taxa at both the genus and phylum levels, with SA-associated changes significantly different from control. However, a study in wheat observed no significant SA-induced root microbial diversity shifts in a 72 h time window (Liu et al., 2018). There is recent evidence examining the heterogeneity of the SA pathway, whereby in wheat the induction of SA may result in various different chemical and even physiological responses (Gondor et al., 2016). It is therefore unclear whether the underlining mechanism is temporal-, genotype-, or community-specific.

#### 3.4. Cross-Talk and the Interplay Between Pathways

Given the potential capacity of ET, JA, and SA to modulate the microbiome, it is often unclear how much cross-talk there is between pathways; however, there is evidence of significant interplay between the respective defense pathways (Koornneef et al., 2008; Diezel et al., 2009; Song et al., 2014; Yang Y.X. et al., 2015). The interplay of phytohormones can be antagonistic, with microbes being able to exploit the antagonism to facilitate colonization and thus evade host defense responses (Jacobs et al., 2011; Plett et al., 2014; Jha et al., 2018). The microbes themselves may utilize effector molecules to actively manipulate the phytohormone pathways and elicit antagonism between the pathways (Kazan and Lyons, 2014). The classical interpretation of the interplay between JA, SA, and ET is reviewed in Yang Y.X. et al. (2015). Unfortunately, little has been done thus far in untangling the potential pathway interaction and their effect on the microbial community. Here we therefore discuss recent evidence that may suggest the capacity of pathway cross-talk to manipulate and shape the microbiome.

As indicated above in the study of Bodenhausen et al. (2014), a statistically significant difference was observed in Variovorax abundance in both the ET and SA associated A. thaliana knockout mutants. There is the potential that Variovorax abundance is actually managed by the interaction between these respective pathways, given that ET and SA generally have a positive interaction and that Variovorax has been shown to be positively correlated with SA (Badri et al., 2013; Yang Y.X. et al., 2015).

The phytohormone abscisic acid (ABA) is well-known in its role in drought stress, salinity stress and as a modulator of plant defense signaling. It has been shown to negatively impact the SAassociated defense pathway, both positively and negatively affect JA-associated defense, and has been shown to affect ET-associated pathogen defense (Pieterse et al., 2012; Takatsuji and Jiang, 2014). In potting soil, exogenous application of ABA has resulted in preferential selection for Cellvibrio, Limnobacter, and Massilia microbes at the genus level (Carvalhais et al., 2014). However, cross-talk between ABA and the other phytohormone defenses, in terms of the effect on whole microbial communities, is largely still unexplored. Additionally, the mechanism of ABA's affect on whole microbiomes is poorly understood. Certain microbial species have been shown to leverage ABA cross-interaction either by producing ABA or effecting ABA biosynthesis, thereby effecting plant-microbe dynamics (Jiang et al., 2010; Ho et al., 2013; Takatsuji and Jiang, 2014).

#### 4. ABIOTIC PLANT STRESS AND THE IMPACT ON MICROBIAL COMMUNITIES

Environmental stressors, including drought stress, temperature stress, and salinity stress, impact plant development, metabolic activity, and the ability for the plant to interact with its phytobiome. The altered phytohormonal signaling and community structure alters the plant's ability to resist stress, resist disease, and alters the capacity for nutrient acquisition (Hawkes and Connor, 2017). While many studies have been performed on the response of the microbiome to abiotic stress and the potential beneficial and deleterious effects on the host, it is less clear how the host influences its microbiome under abiotic stress conditions.

#### 4.1. Drought Stress

Drought stress has a significant impact on plant growth, development, metabolism, and mortality (Allen et al., 2010). Changes in the host in response to drought, in addition to changes in environmental conditions, induce plant-specific (Naylor et al., 2017) and compartment-specific (Santos-Medellín et al., 2017) selection of microbial communities; however, many drought responses, including changes in the microbiome, are conserved across species (Naylor et al., 2017) and soil types (Santos-Medellín et al., 2017).

Actinobacteria is commonly enriched in drought across a wide range of different compartments and plant species (Naylor et al., 2017; Santos-Medellín et al., 2017; Garcia et al., 2018; Timm et al., 2018; Xu et al., 2018). In Sorghum, drought causes developmental delays in the root microbiome, selecting for monoderms (Xu et al., 2018). During drought, there was an association between increased carbohydrates in the roots and increased carbohydrate transporters in Actinobacteria (Xu et al., 2018), suggesting altered root metabolites may play a role in selecting certain species. Additionally, monoderms are less affected by the increase in ROS by the plant during drought stress, relative to diderms (Shade et al., 2012). Host ROS metabolism genes were shown to be associated with Streptomyces (a genus of Actinobacteria) in Populus leaves (Garcia et al., 2018), potentially showing a more universal drought association between the host and its phytobiome. ROS metabolism has been shown to be a general change across species, omics levels, and compartments in drought (Fang et al., 2015; Abraham et al., 2018; Garcia et al., 2018; Zandalinas et al., 2018) that has impacts beyond that of Actinobacteria.

ROS metabolism transcription and defense response transcription are correlated during drought with a variety of taxa including Rhizophagus and nematodes (Garcia et al., 2018). ROS have been shown to modulate the host microbiome, including mitigating nematode infection (Nath et al., 2017) in soybeans (Beneventi et al., 2013) and in tomatoes (Vos et al., 2013). ROS have also been show to be beneficial in regulating rhizobial symbiosis in Medicago truncatula (Andrio et al., 2013). In addition to ROS, other hormones with defense affecting properties, such as ABA, are able to alter the host microbiome. ABA is upregulated in drought in a variety of plants, including Populus, Arabidopsis, Sorghum, etc. (Daszkowska-Golec, 2016; Sah et al., 2016; Kalladan et al., 2017; Garcia et al., 2018). Upregulation of ABA is associated with increased disease susceptibility in a variety of plants (Xiong and Yang, 2003; Gao et al., 2016; Pye et al., 2018). However, the plant's ability to withstand disease under stress has been shown to be plant- and disease-specific (Sinha et al., 2016). Furthermore, combined drought and disease stress has been shown to have an increased ability to mitigate disease (Pandey et al., 2015).

Metabolite production and exudates of the plant, including carbohydrates, amino acids, and other nutrients, are altered in response to drought stress (Bouskill et al., 2016; Tripathi et al., 2016; Abraham et al., 2018; Timm et al., 2018). Under more mild drought conditions, rhizodeposition is increased, while under more severe drought conditions, rhizodeposition is decreased, causing the exudate profile to be related to the severity of the drought experienced (Preece and Peñuelas, 2016). The change in metabolite profile with the plant also correlates with changes in the bacterial community, with root community composition in Arabidopsis shown to be dependent on the exudate profiles of the host plant (Badri et al., 2013). Under drought, an increase in hydrolytic enzymes responsible for breaking down complex carbohydrates such as lignin, cellulose, and other plant metabolites within the microbial communities has been shown (Bouskill et al., 2016). Additionally, bacteria can alter ethylene production within the plant with ACC deaminase (Arshad et al., 2008), which in turn alters plant growth and metabolite profiles to the benefit of plants and microbes (Mayak et al., 2004; Zhang et al., 2018). Not only can the host plant alter its exudate profile to recruit organisms, the microbial community can influence what compounds are being exuded, potentially creating a reciprocal relationship between the community and exudate profile. It is currently unknown how much of the exudate profiles are a plantdriven process and how much the microbial community can influence that process.

#### 4.2. Temperature and Salinity Stress

Temperature stress can often accompany drought stress and has an impact on the fluidity of plant membranes (Sangwan et al., 2002), plant metabolism (Koscielny et al., 2018), ROS activation (Kotak et al., 2007), and protein misfolding (Scharf et al., 2012). Under colder temperatures, root nodulation is decreased in beans, lentils, and peas (Junior et al., 2005), and many organisms that live in the nodules have lower survivability (Singh et al., 2012). Under higher temperatures, a plant is less able to combat pathogens (Mendes et al., 2011), allowing for colonization of disease-causing organisms. However, some of the disease suppression lost during high heat can be associated with the loss of microorganisms that naturally inhibit plant diseases (van der Voort et al., 2016). Conversely, in a wheat high temperature seedling plant, high temperatures induce a WRKY transcription factor that promotes resistance to Puccinia striiformis infection (Wang et al., 2017).

In addition to drought and temperature stress, salinity stress also limits the types of plants that can grow in a given area. High salinity can cause ionic and osmotic stress that limits plant growth and damages plant cells (Zhu, 2002). Also, under salinity stress, plants have increased ROS generation, ABA synthesis, and accumulation of carbohydrates within the plant (Gupta and Huang, 2014), resulting in a similar response to water stress. Plants also respond similarly to drought in that ACC deaminase also can confer salinity resistance to the host plant (Qin et al., 2014). Many studies have been performed that have identified potential plant growth promoters under high salinity stress (Yuan et al., 2016; Hussain et al., 2018; Fouda et al., 2019). For example, 14 halotolerant microbes were shown to improve canola root growth under salt stress by decreasing ethelyne production (Siddikee et al., 2010). Additionally, Piriformospora indica was correlated with an increase in barley antioxidants under salt stress (Baltruschat et al., 2008). Salt tolerance can also be promoted by fungi, such as Montagnulaceae potential improving nitrogen availability in Suaeda salsa under salt stress (Yuan et al., 2016). Despite many studies identifying potential plant growth promoters under saline stress, there is limited knowledge of host plants influencing selection of community structure under salinity stress.

#### 5. REMAINING CHALLENGES

It is clear that there are recent efforts toward identifying plant specific modes of action to control its microbiome. However, in a number of cases it is challenging to determine whether any mode of action was plant-mediated, microbial-mediated, or environment-mediated. Furthermore, the complex and potentially reciprocal nature of interaction adds to the challenge of identifying plant control mechanisms, which is evident when identifying and understanding phytohormone-based control mechanisms. Phytohormone pathways are interconnected, and they mediate abiotic stress responses in complex and sometimes antagonistic manners. The multi-layer interplay must then be delineated to understand the plant-microbe dynamic. Apart from the possible plant-associated complexity, there is the dynamic among the microbial community that must also be understood. Individual taxa in a microbiome may have a large effect on the whole microbiome structure, and the plant-based mechanisms may effect influential members.

#### 5.1. Toward Understanding Combinatorial Plant Mechanisms

Diverse plant accession libraries provide a good framework to begin to compartmentalize and understand the dynamic of combinatorial interactions. For example, the A. thaliana mutant lines have provided a wealth of information on plant mechanisms that may influence microbial populations; however, they do not address the potential for confounding interactions between pathways. Additionally, the mutant lines on their own do not address the question of whether or not a respective pathway is necessary for the observed microbial shift. Approaches that use the supplementation of plant associated compounds identifying interplay between pathways and microbes need to be further addressed. Mutant lines with a combination of potential pathways effected, or general quantitative trait loci (QTL) studies, may be able to address the interaction confounding factors. Unfortunately, using model organisms such as Arabidopsis may not provide generalized results to all species given the variety in hormone pathways and physiological-based mechanisms that plants have for interacting with their microbial communities. There can be significant difference in the phytohormonal pathways between plant species (De Vleesschauwer et al., 2014), with the differences being associated with the biosynthetic pathway itself or the associated function of the pathway (Gondor et al., 2016). Additionally, certain plant species may be more amenable to hormonal amendment than others (De Vleesschauwer et al., 2014). It would therefore be prudent to investigate other, non-model, organisms in a variety of environments with the above mentioned techniques. Furthermore, multiple aspects of a study would need to be manipulated, such as hormone abundance, abiotic stressors, biotic stressors, and presence or absence of a microbial community members (including putative keystone species), to elucidate potential combinatorial effects between genotype, hormones, stress, and microbes.

#### 5.2. Approaches to Unravel Microbial Community Dynamics

Synthetic or constructed communities provide an efficient approach to model microbial diversity. In contrast to axenic controls that allow for the ability to hypothesize about the colonization capacity or pathogenicity of an individual microbe, synthetic communities allow for the discovery of higher level interactions between plants and the microbial community. To therefore begin to understand the complex ecosystem may require an ecosystem point of view, including those based upon constructed communities in controlled environments (http:// eco-fab.org/). However, synthetic communities may not be able to capture all of the complex site to site variation observed in natural environments, given that site variation can be a dominant factor in microbial diversity for certain plants (Whitaker et al., 2018). Nevertheless, constructed communities provide a model from which we can begin to reason, hypothesize, and understand the plant's role in dynamic microbial community interactions.

Using computational biology approaches in combination with experimental field and lab methods, including tools such as microfluidics and constructed communities will help advance understanding regarding plant recruitment of keystone microbes. Further understanding of host-mediated recruitment of its microbiome will in turn improve our ability to effectively and efficiently construct or manipulate plant-microbe systems for improved agricultural and ecological restoration efforts.

#### 5.3. The Holobiont

A significant body of literature focuses on the soil or rhizosphere; however, we know that other compartments, including the root microbiota, can also influence above-ground phenotypes (Pangesti et al., 2017). Root exudates have a reciprocal impact on the microbial community and are influenced by the abiotic stress, biotic stress, and phytohormones. Some drought stresses cause irreversible changes to root exudates (Gargallo-Garriga et al., 2018), which can be important when trying to engineer a community to promote plant growth under a variety of environmental conditions. Therefore, it is important to study plant-mediated effects on other compartments, but determining if effects are compartmental cross-talk, abiotic stress, or direct plant associated is still an open problem.

#### AUTHOR CONTRIBUTIONS

PJ, AF, and BG wrote the manuscript. DJ and GT edited the manuscript and provided guidance on content.

#### FUNDING

Funding was provided by The Center for Bioenergy Innovation (CBI). U.S. Department of Energy Bioenergy Research Centers are supported by the Office of Biological and Environmental Research in the DOE Office of Science. Further funding came

#### REFERENCES


from the Plant-Microbe Interfaces SFA (https://pmiweb.ornl. gov/), and the Oak Ridge National Laboratory's, Laboratory Directed Research and Development (LDRD) program, project 8321.

#### ACKNOWLEDGMENTS

We would like to acknowledge the support and guidance of the members of the JAIL CompSysBio team. This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http:// energy.gov/downloads/doe-public-access-plan).


Rev. Cell Dev. Biol. 28, 489–521. doi: 10.1146/annurev-cellbio-092910- 154055


light. Proc. Natl. Acad. Sci. U.S.A. 94, 2756–2761. doi: 10.1073/pnas.94. 6.2756


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

Copyright © 2019 Jones, Garcia, Furches, Tuskan and Jacobson. 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) 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.

# Improved Powdery Mildew Resistance of Transgenic Nicotiana benthamiana Overexpressing the Cucurbita moschata CmSGT1 Gene

Wei-Li Guo1,2, Bi-Hua Chen1,2, Yan-Yan Guo1,2, He-Lian Yang1,2, Jin-Yan Mu1,2 , Yan-Li Wang1,2, Xin-Zheng Li1,2 \* and Jun-Guo Zhou1,2

<sup>1</sup> School of Horticulture Landscape Architecture, Henan Institute of Science and Technology, Xinxiang, China, <sup>2</sup> Henan Province Engineering Research Center of Horticultural Plant Resource Utilization and Germplasm Enhancement, Xinxiang, China

Powdery mildew (PM), which is mainly caused by Podosphaera xanthii, is a serious biotrophic pathogen disease affecting field-grown and greenhouse-grown cucurbit crops worldwide. Because fungicides poorly control PM, the development and cultivation of PM-resistant varieties is critical. A homolog of SGT1 (suppressor of the G2 allele of skp1), which encodes a key component of the plant disease-associated signal transduction pathway, was previously identified through a transcriptomic analysis of a PM-resistant pumpkin (Cucurbita moschata) inbred line infected with PM. In this study, we have characterized this SGT1 homolog in C. moschata, and investigated its effects on biotic stress resistance. Subcellular localization results revealed that CmSGT1 is present in the nucleus. Additionally, CmSGT1 expression levels in the PM-resistant material was strongly induced by PM, salicylic acid (SA) and hydrogen peroxide (H2O2). In contrast, SA and H2O<sup>2</sup> downregulated CmSGT1 expression in the PM-susceptible material. The ethephon (Eth) and methyl jasmonate (MeJA) treatments upregulated CmSGT1 expression in both plant materials. The constitutive overexpression of CmSGT1 in Nicotiana benthamiana (N. benthamiana) minimized the PM symptoms on the leaves of PM-infected seedlings, accelerated the onset of cell necrosis, and enhanced the accumulation of H2O2. Furthermore, the expression levels of PR1a and PR5, which are SA signaling transduction markers, were higher in the transgenic plants than in wild-type plants. Thus, the transgenic N. benthamiana plants were significantly more resistant to Erysiphe cichoracearum than the wild-type plants. This increased resistance was correlated with cell death, H2O<sup>2</sup> accumulation, and upregulated expression of SA-dependent defense genes. However, the chlorosis and yellowing of plant materials and the concentration of bacteria at infection sites were greater in the transgenic N. benthamiana plants than in the wild-type plants in response to infections by the pathogens responsible for bacterial wilt and scab. Therefore, CmSGT1-overexpressing N. benthamiana plants were hypersensitive to these two diseases. The results of this study may represent valuable genetic information for the breeding of disease-resistant pumpkin varieties, and may also help to reveal the molecular mechanism underlying CmSGT1 functions.

Edited by:

Valentina Fiorilli, University of Turin, Italy

#### Reviewed by:

Yan Xu, Northwest A&F University, China Stefano Pavan, University of Bari Aldo Moro, Italy

> \*Correspondence: Xin-Zheng Li lxz2283@126.com

#### Specialty section:

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

> Received: 24 April 2019 Accepted: 09 July 2019 Published: 25 July 2019

#### Citation:

Guo W-L, Chen B-H, Guo Y-Y, Yang H-L, Mu J-Y, Wang Y-L, Li X-Z and Zhou J-G (2019) Improved Powdery Mildew Resistance of Transgenic Nicotiana benthamiana Overexpressing the Cucurbita moschata CmSGT1 Gene. Front. Plant Sci. 10:955. doi: 10.3389/fpls.2019.00955

Keywords: Cucurbita moschata Duch., powdery mildew, CmSGT1, functional analysis, N. benthamiana

#### INTRODUCTION

fpls-10-00955 July 24, 2019 Time: 14:55 # 2

The genus Cucurbita is composed of several species, including the cultivated Cucurbita moschata Duch., Cucurbita pepo L., Cucurbita maxima Duch., and several wild species. Pumpkins (C. moschata) are valued for their fruit and seeds. Additionally, they are rich in vitamins, amino acids, flavonoids, phenolics, and carbohydrates, and possess medicinal properties, including antidiabetic, anti-oxidant, anti-carcinogenic, and anti-inflammatory activities (Wang et al., 2002; Yadav et al., 2010). In China alone, the annual yield of pumpkin, squash, and gourds is 8,051,495 t (i.e., approximately 22.68% of the global yield) from a harvested area of 438,466 ha (i.e., 17.42% of the global area) (Food and Agriculture Organization, 2017)<sup>1</sup> . Cucurbit powdery mildew (PM) is a serious disease affecting field-grown and greenhousegrown cucurbit crops worldwide. The disease is mainly caused by Podosphaera xanthii (formerly known as Sphaerotheca fuliginea), which is a biotrophic plant pathogen (Perez-Garcia et al., 2009; Fukino et al., 2013). Fungicide applications poorly control PM and the long-term use of pesticides may lead to increased environmental pollution and the residual chemicals on food crops may be harmful for humans and animals. Therefore, studying the mechanism underlying PM resistance and exploiting the resistance genes to breed resistant varieties represents an effective way to control PM in pumpkin.

To date, there has been relatively little research on pumpkin (2n = 2x = 40), especially at the molecular level, which has seriously hindered developments in the fields of molecular biology and genetics. We previously conducted a RNA sequencing analysis of pumpkin inbred line highly resistant to PM (inbred line "112-2") and identified 4,716 differentially expressed genes, including genes encoding broad-spectrum disease-resistance/susceptibility proteins [PR protein, ubiquitin, heat shock protein and Mildew Locus O (MLO)], reactive oxygen scavengers (peroxidase, superoxide dismutase, and catalase), signal transduction molecules serine/threonine protein kinase, and transcription factors (WRKY, TGA, and MYB). Additionally, candidate disease-resistance genes, such as WRKY21, MLO3, and SGT1, were identified (Guo et al., 2018). One of the genes cloned from the transcriptome was a homolog of Cucumis melo SGT1 (suppressor of the G2 allele of skp1). The expression of this SGT1 homolog was highly upregulated in PM-resistant material at 6 h after an inoculation with the PM fungus, but not in PM-susceptible material. However, pumpkin SGT1 has not been functionally characterized regarding its potential involvement in the defense response to PM.

The SGT1 protein was originally defined in yeast, in which it interacts with SKP1, which is a component of the Skp1/CDC/Fbox protein E3 ubiquitin ligase complex (Kitagawa et al., 1999). The SGT1 protein contains three functional domains, namely the tetratricopeptide repeat (TPR), CHORD SGT1 (CS), and the SGT1-specific sequence (SGS) (Kumar and Kirti, 2015). SGT1 is closely related to the disease resistance mediated by the plant resistance (R) genes. The silencing or mutation of NbSGT1 from Nicotiana benthamiana (N. benthamiana) can lead to reduction

<sup>1</sup>http://www.fao.org/faostat/en/#data/QC

of steady-state levels of R proteins and the loss of the mediated resistance, and AtSGT1a overexpression can contribute positively to resistance triggered by the NB-LRR type R proteins, and can complement for loss of AtSGT1b in auxin signaling (Muskett and Parker, 2003; Azevedo et al., 2006). A previous study revealed that NbSGT1 overexpression in N. benthamiana accelerates the development of the hypersensitive response (HR) during R-mediated disease resistance (Wang et al., 2010). Additionally, Hv-SGT1 overexpression in wheat enhances the resistance to PM, which is correlated with increased levels of reactive oxygen intermediates at the pathogen entry sites (Xing et al., 2013). Furthermore, PsoSGT1 from Prunus sogdiana appears to interact with molecular chaperones (RAR1 and HSP90) to activate a nucleotide-binding domain and leucine-rich repeat-containing (NB-LRR)-type protein that confers disease resistance (Zhu et al., 2017). The regulation of NB-LRR-type protein stability or substrate degradation to maintain the balance between the activation and inhibition of plant defense responses (Meldau et al., 2011; Hoser et al., 2013), contributes to the RXLR elicitorinduced HR-related cell necrosis (Xiang et al., 2017).

The recent release of the pumpkin genome provided an opportunity for an additional screening for disease-resistance genes (Sun et al., 2017). In this study, on the basis of the abovementioned transcriptome analysis that identified differentially expressed genes responsive to PM, we functionally characterized the pumpkin homolog of SGT1 (designated as CmSGT1). The transcription of CmSGT1 in the PM-resistant inbred line "112- 2" was strongly induced by PM, salicylic acid (SA), hydrogen peroxide (H2O2), ethephon (Eth), and methyl jasmonate (MeJA). Transgenic N. benthamiana plants that constitutively overexpressed CmSGT1 exhibited increased resistance to PM and were hypersensitive to bacterial wilt and scab.

#### RESULTS

#### Cloning of CmSGT1 and Subcellular Localization the Encoded Protein

Pumpkin PM-related candidate genes identified in a transcriptome were reported previously (Guo et al., 2018). One of the isolated clones exhibited 89% identity at the nucleotide level to C. melo SGT1. A full-length clone of this homolog was obtained, and the gene was named CmSGT1 and submitted to GenBank (accession number MH105820). The CmSGT1 gene comprised 1,206 bp, which included a 1,080-bp open reading frame (ORF) encoding 360 amino acids. The predicted polypeptide was basic, with a pI of 5.32, and a molecular mass of 40.3 kDa. An alignment of the deduced CmSGT1 amino acid sequence with homologous sequences is presented in **Supplementary Figure S1**. At the amino acid level, CmSGT1 was highly similar to the SGT1 from the following plant species: Cucumis melo (CmSGT1, 84.4% identity), and Cucumis sativus (CsSGT1, 82.5% identity), N. benthamiana (NbSGT1.1 and NbSGT1.2, 89.5% identity), and Arabidopsis thaliana (AtSGT1a and AtSGT1b, 75.2% identity). The CmSGT1 sequence contained three conserved domains, namely TPR, CS, and SGS.

The subcellular localization of CmSGT1 was assessed with a CmSGT1-GFP fusion protein that was produced in A. thaliana protoplasts under the control of the 35S CaMV promoter. The GFP signal in protoplasts producing GFP alone was detected in the cytoplasm and nucleus (**Figure 1**), whereas the signal from the CmSGT1-GFP fusion protein was detected exclusively in the nucleus.

#### Expression of CmSGT1 in Pumpkin Seedlings in Response to PM and Exogenous Treatments

A qRT-PCR assay was completed to analyze the CmSGT1 expression patterns in PM-resistant inbred line "112-2" and PMsusceptible cultivar "JJJD" treated with PM, H2O2, SA, abscisic acid (ABA), Eth, and MeJA (**Figure 2**). The expression data were normalized against that of the β-actin gene and were recorded relative to the CmSGT1 transcript level in the water-sprayed control "JJJD" plants at 0 hour post-inoculation (hpi).

The CmSGT1 expression level in "112-2" plants were upregulated by PM (except at 24 h), H2O2, and SA treatments, with the PM-induced expression level at 3 hpi upregulated by 6.62-fold. In contrast, the ABA treatment essentially had no effect. The expression of CmSGT1 in "JJJD" plants was inhibited by H2O2, SA, and ABA (except at 24 h) treatments, with irregular expression-level changes induced by PM. After the Eth treatment, the CmSGT1 expression level was significantly higher in the "112- 2" and "JJJD" seedlings than in the water-treatment control (CK) seedlings (except at 48 h), with peak levels occurring at 24 h (i.e., 10.2-fold and 4.5-fold increases in "112-2" and "JJJD" seedlings, respectively). In response to the MeJA treatment, CmSGT1 expression levels were higher in "112-2" (except at 48 h) and "JJJD" (except at 3 h) seedlings than in CK seedlings. The results indicated that SA, and H2O<sup>2</sup> upregulated CmSGT1 expression in inbred line "112-2" (PM-resistant material), but downregulated CmSGT1 expression in cultivar "JJJD" (PM-sensitive material). Moreover, Eth and MeJA induced CmSGT1 expression regardless of PM susceptibility.

#### Improved PM Resistance of CmSGT1-Overexpressing N. benthamiana Plants

Firstly, transcript level of the high homology (NbSGT1) modulated by the overexpression of CmSGT1 under normal conditions was determined by qRT-PCR. Compared to wild type (WT) plants, the expression of the homologous genes was not basically altered in transgenic plants when grown in normal condition (**Figure 3**), indicating that overexpression of pumpkin CmSGT1 gene has no obvious effect on NbSGT1 transcript in N. benthamiana. The CmSGT1 gene was not detected in WT plants.

The disease severity at 10 days post-inoculation (dpi) was 78% lower for the transgenic plants than for the CK plants (**Table 1**). Powdery mildew symptoms were detectable on infected WT seedlings at 7 dpi, and the infected spots were chlorotic at 28 dpi. In contrast, the PM symptoms were undetectable and relatively slight at 7 and 28 dpi, respectively, in the CmSGT1 overexpressing transgenic plants (**Figure 4A**). A comparison of the leaves of PM-infected WT and transgenic plants revealed more blue spots displayed by the trypan blue staining, on the transgenic leaves than on the WT leaves at 4 dpi, although blue spots were developing on the WT leaves. The blue spots on the transgenic leaves expanded at 5 and 7 dpi, and were bigger than those of WT leaves, implying that the overexpression of CmSGT1 in transgenic plants accelerated cell death following a PM infection (**Figure 4B**). Moreover, there were more brown spots manifested by the DAB staining, on the transgenic leaves than on the WT leaves at 1 dpi. These brown spots reflected the accumulation of H2O2, and were darker brown and larger at 3 dpi. At 5 dpi, the brown spots lightened in the infected plants, but were more intense in the transgenic plants than the spots in the WT plants. The results indicated that the overexpression of CmSGT1 promoted the accumulation of H2O<sup>2</sup> in transgenic plants infected with PM (**Figure 4C**).

#### Expression of Signal-Related Genes in Transgenic N. benthamiana Plants

To investigate the signal transduction pathways affected by CmSGT1 during plant defense responses to PM, the expression levels of five signaling-associated genes (NPR1, PR5, PR1a, PAL, and PDF1.2) in the SA, JA, and ET signal transduction pathways were analyzed by qRT-PCR for the CmSGT1-overexpressing transgenic and WT N. benthamiana plants, with water-treated WT plants at 0 hpi serving as the control samples (**Figure 5**). The NPR1, PAL (except at 120 h), and PR5 expression levels were lower in the transgenic and WT plants infected with PM than in the CK plants, implying that PM inhibited the expression of these genes. The PR1a expression level was higher in the PMinfected transgenic plants than in the CK plants (i.e., 68.3-fold at 120 h), whereas the PR1a level was lower in the PM-infected WT plants than in the CK plants. The PDF1.2 expression levels in the PM-infected transgenic plants were lower at 12 hpi and higher at 48 hpi than in the CK plants, with no differences thereafter. Thus, in response to the PM infection, the NPR1, PAL (except at 24 and 48 h), and PDF1.2 (except at 48 h) expression levels were significantly lower in the transgenic plants than in the WT plants, whereas the opposite pattern was observed for the PR1a, and PR5 expression levels. These results implied that the overexpression of CmSGT1 in N. benthamiana upregulates the expression of PR1a, and PR5, which are involved in SA defense signal transduction. Furthermore, the increased PM resistance of the transgenic N. benthamiana plants appears to be related to the upregulated expression of these genes.

#### Compromised Resistance of Transgenic N. benthamiana Plants to Bacterial Diseases

To analyze the effect of CmSGT1 on other plant diseases, two common bacterial pathogens causing bacterial wilt (Ralstonia solanacearum) and scab (Xanthomonas euvesicatoria) were injected into N. benthamiana plants (**Figure 6**). At 6 days after inoculation with R. solanacearum and X. euvesicatoria,

FIGURE 3 | Relative expression of CmSGT1 and NbSGT1 in transgenic or wild-type plants under normal growth conditions. N. benthamiana encodes NbSGT1 isoform. L3, L5, and L11 are three independent transgenic lines that overexpress CmSGT1. Three biological triplicates were averaged and Bars indicate standard error of the mean. Data (NbSGT1 expression in transgenic lines vs. NbSGT1 expression in wild-type plants) were analyzed by one-way ANOVA and unmarked <sup>∗</sup> denotes no statistical differences at p < 0.05.

TABLE 1 | Disease severity of leaves of N. benthamiana seedlings infected with powdery mildew.


Data are mean values (±SD) of at least three independent experiments. "<sup>∗</sup> " indicates significantly different values between treatments (p < 0.05). L3, L5, and L11 are three independent transgenic lines that overexpress CmSGT1.

the chlorosis and yellowing of the sixth leaf veins was greater for transgenic N. benthamiana plants than for the WT plants. Additionally, there were 5.94- and 21.1-times more the concentration of bacterial wilt and scab bacteria, respectively, in the transgenic plants than in the WT plants. Yellowing was also observed between the veins of the 12th leaf in transgenic N. benthamiana plants, and there were 13.3- and 8.28-times more R. solanacearum and X. euvesicatoria bacteria, respectively, in the transgenic plants than in the WT plants. These observations suggested that the overexpression of CmSGT1 in N. benthamiana decreases the resistance to bacterial wilt and scab.

#### DISCUSSION

In this study, we isolated a novel pumpkin SGT1 gene, which was designated as CmSGT1. The predicted amino acid sequence was 89 and 82% identical to the NbSGT1.1 and CsSGT1 sequences, respectively. In A. thaliana and N. benthamiana, SGT1 encodes three functional domains (TPR, CS, and SGS) that are essential for SGT1 activity (Peart et al., 2002; Noël et al., 2007). In this study, the CmSGT1-GFP fusion protein was localized to the nucleus in A. thaliana protoplasts, which was inconsistent with the results of earlier studies involving SGT1 in other plant species (Xing et al., 2013; Liu et al., 2016). This inconsistency may be related with low identity (Haynaldia villosa HvSGT1, 64.0% identity; pepper CaSGT1, 63.7% identity) or the number of transformed cells examined. Arabidopsis SGT1b fused to Cerulean localized to the cytosol, but could be seen in nuclei of 25% of 55 transformed cells examined (Noël et al.,2007), and translocation of the SGT1/SRC2-1 complex from the plasma membrane and cytoplasm to the nuclei is required in pepper upon the inoculation of Phytophthora capsici (Liu et al., 2016), suggesting the movement of SGT1 between the cytosol and nucleus.

The interplay among complex signaling networks, including various pathways regulated by phytohormones, such as SA, JA, ethylene (ET) and ABA, considerably influences plant resistance to diseases. An earlier investigation proved that HvSGT1 expression levels substantially increase following treatments with H2O<sup>2</sup> and MeJA, slightly increase following exposure to ET or ABA, and are unchanged in response to SA (Xing et al., 2013). In the current study, SA and H2O<sup>2</sup> treatments considerably upregulated CmSGT1 expression in the PM-resistant inbred line "112-2", but downregulated CmSGT1 expression in the PM-susceptible material. The expression of CmSGT1 in both materials was upregulated by Eth and MeJA treatments. Recent reports indicated that SGT1 expression may be induced by phytopathogens, such as Blumeria graminis in wheat and P. capsici in pepper (Xing et al., 2013; Liu et al., 2016). In this study, the CmSGT1 expression in inbred line "112-2" was induced by PM. Additionally, the overexpression of CmSGT1 in transgenic N. benthamiana plants decreased the disease index by 78%, accelerated cell necrosis, and increased the accumulation of H2O2. These results indicated that the PM resistance of the transgenic N. benthamiana plants was enhanced, likely because of the changes to the HR-related cell necrosis and H2O<sup>2</sup> accumulation. Our findings are consistent with the results of an earlier study that revealed that H2O<sup>2</sup> accumulation and the subsequent cell death usually lead to the resistance to diseases caused by biotrophic pathogens (Li et al., 2011). Moreover, SGT1 helps mediate cell death during compatible and incompatible plant–pathogen interactions, suggesting that SGT1 is an essential component of common signaling pathways responsible for cell death (Cuzick et al., 2009; Uppalapati et al., 2010; Wang et al., 2010). The silencing of the pepper SGT1 gene adversely affects HR-related cell death, prevents H2O<sup>2</sup> accumulation, and downregulates HR-related and SA/JA-dependent marker gene expression levels, and influences the PcINF1/SRC2-1-induced pepper defense response by SGT1 interacting with SRC2-1 (Liu et al., 2016).

The PDF1.2 gene is important for the JA/ET-dependent signaling pathway. The disease-associated NPR1 gene (nonexpresser of PR1) affects various disease-resistance signal transduction pathways, and encodes one of the important transcription factors downstream of SA. The SA-dependent disease-resistance signal transduction pathway can be divided into NPR1-dependent and NPR1-independent transduction pathways (Gao and Zhang, 2012). The PAL, PR1a, and PR5 expression levels are markers of the SA signaling pathway. In wheat, HvSGT1 activates PM resistance mechanisms through JA-dependent defense pathways and suppresses the activities

of SA-dependent defense pathways (Xing et al., 2013). In the current study, following a PM infection, the NPR1, PAL, and PDF1.2 expression levels were lower in the transgenic plants than in the WT plants, whereas the opposite pattern was observed for the PR1a and PR5 expression levels. This suggests that in the SA pathway, the transactivation of NPR1 is unaffected by CmSGT1, whereas the transactivation of PR1a, and PR5 is dependent on CmSGT1. Additionally, CmSGT1 does not directly affect the JA/ET-dependent defense pathway to regulate PDF1.2 expression. We propose that CmSGT1 activates stress-resistance mechanisms through SA-dependent defense pathways without inducing NPR1 and PAL expression levels, but suppresses the activities of JA/ET-dependent defense pathways. Therefore, we speculate that CmSGT1 positively regulates the H2O<sup>2</sup> and SA pathways. Moreover, H2O<sup>2</sup> might directly transfer the SA signal to regulate the expression of downstream response genes in the CmSGT1-overexpressing transgenic plants infected with PM. In NPR1-dependent SA signal transduction pathways, the activation of PR genes needs NPR1 and TGA transcription factor binding to identify its promoters (Fan and Dong, 2002; Spoel and Dong, 2012). The equivalent mutation in NPR1 abolishes its ability to bind SA and promotes SA-induced defense gene expression (Ding et al., 2018). We speculated that whether there is the relation of the downregulation of NPR1 and SA-mediated transcriptional activation of PR genes or not. Notably, the phenotypes and genes in CmSGT1-overexpressing transgenic N. benthamiana plants infected with N. benthamiana PM might not be exactly the same as those regulated by CmSGT1 in C. moschata in response to cucurbit PM. Further studies will be necessary to reveal biological functions for CmSGT1 in C. moschata infected with PM.

Two globally important diseases that affect N. benthamiana, bacterial wilt and scab, are caused by the necrotrophic R. solanacearum and the hemi-biotrophic X. euvesicatoria, respectively. There are differences in the resistance mechanisms and patterns of fungal development for biothophic (P. xanthii) and necrotrophic (R. solanacearum) pathogens. Recent studies confirmed that SGT1 promotes the resistance to biotrophic pathogens, while suppressing the resistance to necrotrophic and hemibiotrophic pathogens. Silencing of SGT1 compromised resistance to the barley and H. villosa biothoph B. graminis (Shen et al., 2003; Xing et al., 2013) and the wheat biotroph Puccinia striiformis (Scofield et al., 2005), while enhancing the resistance of N. benthamiana to the necrotrophic pathogen Botrytis cinerea (El Oirdi and Bouarab, 2007). A recent study involving N. benthamiana indicated silencing of NbSGT1 compromised the protective effect of systemic acquired resistance-induced plants on neighboring plants against bacteria wilt caused by R. solanacearum (Cheol Song et al., 2016). In the current study, the chlorosis and yellowing of the infection sites on leaves were greater in transgenic N. benthamiana plants than in WT plants at 6 dpi. Additionally, there were substantially more concentration of bacterial wilt and scab bacteria in the transgenic plants than in the WT plants, indicating that the overexpression of CmSGT1 in N. benthamiana adversely affects the resistance to bacterial wilt and scab, which differs from the effects of CmSGT1 overexpression on the resistance to PM. Consequently, SGT1 may function differently depending on the plant–pathogen combinations with diverse effectors and R proteins.

In conclusions, the results of this study indicate that the overexpression of CmSGT1 may increase the PM resistance, but decrease bacterial wilt and scab resistance, in transgenic N. benthamiana plants. Additionally, CmSGT1 overexpression may improve PM resistance by enhancing HR-related cell death and H2O<sup>2</sup> accumulation and upregulating the expression of SA-mediated defense-response genes. Further analyses of the CmSGT1 gene may be useful for characterizing biotic stress signaling pathways and for the genetic engineering of novel pumpkin cultivars. The results described herein may be relevant for future biotechnology-based

investigations and the molecular breeding of pumpkins and related plant species.

#### MATERIALS AND METHODS

#### Plant Materials and Treatments

Pumpkin (C. moschata) inbred line "112-2" and cultivar "Jiujiangjiaoding" (abbreviated "JJJD"), which are resistant and susceptible to PM, respectively, were provided by the Henan Institute of Science and Technology, Xinxiang, Henan, China (Zhou et al., 2010). Pumpkin seeds were germinated and the resulting seedlings were grown as previously described (Guo et al., 2018). Seedlings at the three-leaf stage were treated as follows. The PM infection was initiated as described by Guo et al. (2018). Specifically, conidia were collected from the leaves of pumpkin plants naturally infected with PM in a local greenhouse. Seedlings were sprayed with a freshly prepared spore suspension

(10<sup>6</sup> spores/ml) or an exogenous signaling molecule or hormone, including 1.5 mM H2O2, 100 µM SA, 100 µM ABA, 100 µM MeJA, 0.5 g/L Eth. Moreover, water alone was used for the control treatment (CK). The treated seedlings were maintained in a growth chamber with a 15-h light (28◦C)/9-h dark (18◦C) cycle (5,500 lux light intensity) and harvested after 0, 3, 6, 12, 24, and 48 h to examine the CmSGT1 expression pattern. At each time point, two young leaves were collected from the upper parts of four seedlings (i.e., one sample), wrapped in foil, frozen in liquid nitrogen, and stored at −80◦C. The treatments were arranged in a randomized complete block design, with three biological replicates.

#### Isolation of CmSGT1 cDNA Clone and Sequence Analysis

The SGT1 homolog expressed sequence tag (GenBank accession number SRR5369792) was identified in a PM-resistant pumpkin seedling transcriptome by Guo et al. (2018). The full-length ORF of the SGT1 homolog was obtained with the cDNA fragment of this homolog used as a probe, as previously described (Guo et al., 2014). The theoretical molecular weight (Mw) and isoelectric point (pI) were calculated with the ExPASy Compute pI/Mw tool (Bjellqvist et al., 1993). Sequence data were analyzed with the ClustalW program (Thompson et al., 1994). The NCBI databases were screened for homologous sequences with the default parameters of the BLAST program<sup>2</sup> (Altschul et al., 1997).

#### Subcellular Localization Analysis of CmSGT1

The CmSGT1 ORF (without the termination codon) was ligated into the pBI221-GFP vector for the subsequent production of a green fluorescent protein (GFP)-tagged CmSGT1 fusion protein. Polyethylene glycol was used during the transformation of A. thaliana protoplasts with the recombinant plasmid (Lee et al., 2013). The subcellular localization of CmSGT1 was determined based on the GFP signal, which was detected with the confocal fluorescence microscope (UltraVIEW VoX, Olympus, Japan) under excitation wavelength 488 nm and captured light wavelength range 448–508 nm. The cell membrane and nucleus were stained with 1,1<sup>0</sup> -dioctadecyl-3,3,3<sup>0</sup> ,30 -tetramethylindocarbocyanine perchlorate (Dil) and 2-(4-Amidinophenyl)-6-indolecarbamidine dihydrochloride (DAPI), respectively.

#### Generation of CmSGT1-Overexpressing Transgenic N. benthamiana Plants

Pumpkin are known to be one of the plants most refractory for transformation. To date, only two reports on transformation in C. moschata existed using a combined method of vacuum infiltration and Agrobacterium infection (Nanasato et al., 2011; Nanasato and Tabei, 2015). So, we choose a heterologous overexpression assay in N. benthamiana instead of generating C. moschata transgenic plants. Forward and reverse primers with an added BamH I site and Kpn I site, respectively, were used to amplify CmSGT1. The amplified sequence was then inserted into the pVBG2307 vector for the subsequent expression of CmSGT1 under the control of the 35S cauliflower mosaic virus (CaMV) promoter. The recombinant plasmid was introduced into Agrobacterium tumefaciens GV3101 cells as previously described (Guo et al., 2014). The resulting A. tumefaciens cells were used to transform N. benthamiana plants according to a previously described leaf disk method (Li et al., 2012). The transgenic N. benthamiana plants were confirmed by examining the segregation ratio of the kanamycin selectable marker and by PCR analysis of NPTII and CmSGT1. T2 lines that produced 100% kanamycin-resistant plants in the T3 generation were considered as homozygous transformants. In each experiment, T2 generations of homozygous transgenic lines (L3, L5 and L11) were selected for further analysis. Similar phenotypes and results used for this study were observed in more than three independent lines of transgenic plants.

#### Primer Design

Details regarding all of the primers designed and used in this study are provided in **Supplementary Table S1**.

#### Analysis of Transgenic N. benthamiana Plants Infected With Powdery Mildew

Conidia were collected from N. benthamiana leaves naturally infected with Erysiphe cichoracearum DC., which causes PM. Transgenic N. benthamiana seedlings at the five-leaf stage were used for phenotypic analyses. Specifically, the petiole of the second leaf from the top of the seedlings was wrapped with cotton moistened with water. The leaf was placed on a porcelain tray containing filter paper, after which it was sprayed with a spore suspension (10<sup>6</sup> spores/ml). The tray was then covered with film to maintain humidity. At 10 dpi, disease severity of leaves in vitro was calculated as [(5A + 4B+ 3C+ 2D+ E)/5F] × 100 according to Ishii et al. (2001). Additionally, the first fully expanded true leaf of the transgenic and WT plants were sprayed with the abovementioned spore suspension and sampled after 0, 12, 24, 48, 72, and 120 hpi, frozen in liquid nitrogen, stored at −80◦C and used for extraction of total RNA. On the other hand, these leaves were harvested symmetrically along the sides of the main vein after 0, 1, 3, 4, 5, and 7 days to examine of cell death and H2O<sup>2</sup> accumulation. The treatments were arranged in a randomized complete block design with three replicates.

Leaves from the transgenic N. benthamiana seedlings were stained with 3, 3<sup>0</sup> -diaminobenzidine (DAB) and trypan blue staining as previously described (Choi et al., 2012) to analyze H2O<sup>2</sup> accumulation and cell death. For DAB staining, whole leaves were immersed in DAB solution (1 mg/ml, pH 5.7) and incubated overnight (almost 8 h) in darkness. Leaves were then de-stained three times with 95% ethanol. Regarding trypan blue staining, whole leaves were boiled in a lactophenol-ethanol trypan blue solution (10 ml lactic acid, 10 ml glycerol, 10 g phenol, 30 ml absolute ethanol, and 10 mg trypan blue dissolved in 10 ml distilled water) for 10 min and then maintained at room temperature overnight. Leaves were de-stained with 2.5 g/ml chloral hydrate in distilled water.

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

#### Analysis of Transgenic N. benthamiana Plants Infected With Bacterial Wilt and Scab

The sixth and twelfth leaf from the top of transgenic N. benthamiana seedlings were inoculated with bacterial solutions (10<sup>8</sup> cfu/ml). Specifically, the bacterial solutions were injected into the underside of leaves between the lateral veins with a syringe lacking a needle. Each leaf was inoculated in four places. The petiole of leaves placed on a porcelain plate was wrapped with water-saturated degreased cotton, after which the plate was covered with film to maintain humidity and then incubated at 28◦C with a 16-h light/8-h dark cycle. Water was periodically added to the cotton to maintain an appropriate moisture level. Experiments were done in triplicate for each line. After a 6-day incubation, the concentration of bacteria at the injection sites was determined as follows. The injection sites were sampled with a circular perforator (1 cm diameter) and ground in aseptic water (0.1 g added to 0.9 ml water). The ground material was serially diluted to produce the 10<sup>2</sup> -fold, 10<sup>3</sup> -fold, and 10<sup>4</sup> -fold diluents. A 0.1-ml aliquot of the 10<sup>3</sup> -fold, 10<sup>4</sup> -fold, and 10<sup>5</sup> -fold diluents were added to a petri dish containing NA medium. The medium was incubated upside down at 28◦C. Plates with 30–300 colonies were considered ideal. The bacterial concentration was calculated with the following formula: colony average × 10 × dilution/g.

#### Quantitative Real-Time PCR Analysis

The RNA extraction, first-strand cDNA synthesis and quantitative real-time (qRT-PCR) were completed as described by Guo et al. (2015). Relative gene expression levels were determined with the 2−11CT method. Total RNA was extracted from the leaves of pumpkin seedlings treated with various stresses or distilled water for 0, 3, 6, 12, 24, or 48 h as described above. The β-actin gene was used as an internal control, because it was confirmed as a suitable reference gene for normalizing of gene expression levels in pumpkin (Wu and Cao, 2010). On the other hand, total RNA was extracted from CmSGT1-overexpressing transgenic and WT N. benthamiana seedlings to examine the expression of five hormone-related genes (NtNPR1, NtPR1a, NtPR5, NtPDF1.2, and NtPAL) and N. benthamiana isoform (NbSGT1). The N. benthamiana NtEF1-α gene (AF120093) was included in the assays as an internal control.

#### Statistical Analyses

Values were expressed as the mean ± standard error of three independent determinations. Data were compared by analysis of variance (ANOVA) using a one-way ANOVA, and differences

#### REFERENCES


between WT and transgenic plants were tested by a post hoc comparison test (Student–Newman–Keuls) at p < 0.05 with SPSS 19.0 for Windows (SPSS Inc., Chicago, IL, United States).

#### DATA AVAILABILITY

The datasets generated for this study can be found in NCBI, MH105820.

#### AUTHOR CONTRIBUTIONS

W-LG and Y-YG conceived and designed the experiments. W-LG and B-HC performed the experiments. J-YM and H-LY analyzed the data. Y-LW, J-GZ, and X-ZL contributed reagents, materials, and analysis tools. W-LG wrote the manuscript.

#### FUNDING

This work was supported by National Natural Science Foundation of China (No. 31401876), Henan High-level Talent Introduction and Training Special Foundation (No. 103020218003/008), and Natural Science Foundation of Henan Province (No. 162300410110).

#### ACKNOWLEDGMENTS

We thank Guizhou Academy of Tobacco Sciences researcher Hancheng Wang for providing tobacco E. cichoracearum and Liwen Bianji, Edanz Editing China (www.liwenbianji.cn/ac) for editing the English text of a draft of this manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.00955/ full#supplementary-material

FIGURE S1 | Amino acid sequences alignment of pumpkin CmSGT1 with others. The three conserved domains (TPR, CS, and SGS) are shown by thin underlines. The genes included are CmSGT1 (Cucumis melo L., XP\_008439299.1) and CsSGT1 (Cucumis sativus L., XP\_004140745.1) AtSGT1A (Arabidopsis, AT4G23570.3) and AtSGT1B (Arabidopsis, AT4G11260.1), NbSGT1.1 (N. benthamiana, AF516180) and NbSGT1.2 (N. benthamiana, AF516181).

TABLE S1 | Primers used in this investigation.



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

Copyright © 2019 Guo, Chen, Guo, Yang, Mu, Wang, Li and Zhou. 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.