# ROLE OF ENDOPHYTES IN PLANT HEALTH AND DEFENSE AGAINST PATHOGENS

EDITED BY : Massimiliano Morelli, Ofir Bahar, Kalliope K. Papadopoulou, Donald L. Hopkins and Aleksa Obradović 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 978-2-88966-098-8 DOI 10.3389/978-2-88966-098-8

#### 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

# ROLE OF ENDOPHYTES IN PLANT HEALTH AND DEFENSE AGAINST PATHOGENS

Topic Editors:

Massimiliano Morelli, Italian National Research Council, Italy Ofir Bahar, Agricultural Research Organization (ARO), Israel Kalliope K. Papadopoulou, University of Thessaly, Greece Donald L. Hopkins, University of Florida, United States Aleksa Obradović, University of Belgrade, Serbia

Citation: Morelli, M., Bahar, O., Papadopoulou, K. K., Hopkins, D. L., Obradović, A., eds. (2020). Role of Endophytes in Plant Health and Defense Against Pathogens. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88966-098-8

# Table of Contents

*05 Editorial: Role of Endophytes in Plant Health and Defense Against Pathogens*

Massimiliano Morelli, Ofir Bahar, Kalliope K. Papadopoulou, Donald L. Hopkins and Aleksa Obradović

*10 The Effect of Plant Genotype, Growth Stage, and* Mycosphaerella graminicola *Strains on the Efficiency and Durability of Wheat-Induced Resistance by* Paenibacillus *sp. Strain B2*

Erika Samain, Thierry Aussenac and Sameh Selim


Wenyan Cui, Pengjie He, Shahzad Munir, Pengbo He, Yueqiu He, Xingyu Li, Lijuan Yang, Biao Wang, Yixin Wu and Pengfei He


Eoghan King, Adrian Wallner, Isabelle Rimbault, Célia Barrachina, Agnieszka Klonowska, Lionel Moulin and Pierre Czernic


Alejandro del Barrio-Duque, Johanna Ley, Abdul Samad, Livio Antonielli, Angela Sessitsch and Stéphane Compant

*171 Anatomical and Biochemical Changes Induced by* Gluconacetobacter diazotrophicus *Stand Up for* Arabidopsis thaliana *Seedlings From*  Ralstonia solanacearum *Infection*

María V. Rodriguez, Josefina Tano, Nazarena Ansaldi, Analía Carrau, María S. Srebot, Virginia Ferreira, María L. Martínez, Adriana A. Cortadi, María I. Siri and Elena G. Orellano

*191 Biocontrol of Bacterial Wilt Disease Through Complex Interaction Between Tomato Plant, Antagonists, the Indigenous Rhizosphere Microbiota, and* Ralstonia solanacearum

Tarek R. Elsayed, Samuel Jacquiod, Eman H. Nour, Søren J. Sørensen and Kornelia Smalla


Ida Romano, Valeria Ventorino and Olimpia Pepe


Osama Abdalla Abdelshafy Mohamad, Jin-Biao Ma, Yong-Hong Liu, Daoyuan Zhang, Shao Hua, Shrikant Bhute, Brian P. Hedlund, Wen-Jun Li and Li Li

# Editorial: Role of Endophytes in Plant Health and Defense Against Pathogens

Massimiliano Morelli 1\*† , Ofir Bahar <sup>2</sup>† , Kalliope K. Papadopoulou3† , Donald L. Hopkins <sup>4</sup>† and Aleksa Obradovic´ <sup>5</sup>†

<sup>1</sup> Consiglio Nazionale delle Ricerche, Istituto per la Protezione Sostenibile delle Piante, Sede Secondaria di Bari, Bari, Italy,

<sup>2</sup> Department of Plant Pathology and Weed Research, Agricultural Research Organization (ARO), Rishon LeZion, Israel,

<sup>3</sup> Department of Biochemistry and Biotechnology, University of Thessaly, Larissa, Greece, <sup>4</sup> University of Florida, Gainesville,

FL, United States, <sup>5</sup> Faculty of Agriculture, University of Belgrade, Belgrade, Serbia

Keywords: endophytes, metagenomics, plant defense response, bioactive compounds, bioinoculants, plant growth-promoting bacteria

Editorial on the Research Topic

Role of Endophytes in Plant Health and Defense Against Pathogens

#### Edited by:

Essaid Ait Barka, Universite´ de Reims Champagne-Ardenne, France

#### Reviewed by:

Sameh Selim, UniLaSalle, France

#### \*Correspondence:

Massimiliano Morelli massimiliano.morelli@ipsp.cnr.it

† These authors have contributed equally to this work

#### Specialty section:

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

Received: 29 June 2020 Accepted: 11 August 2020 Published: 26 August 2020

#### Citation:

Morelli M, Bahar O, Papadopoulou KK, Hopkins DL and Obradovic´ A (2020) Editorial: Role of Endophytes in Plant Health and Defense Against Pathogens. Front. Plant Sci. 11:1312. doi: 10.3389/fpls.2020.01312

### INTRODUCTION

With the increasing social concern in avoiding, or at least reducing, the application of pesticides and chemical fertilizers, in favor of sustainable eco-friendly alternatives, the search for beneficial microorganisms and microbial-derived compounds has become one of the most popular Research Topics in the field of plant-microbe interactions (Cardoso Filho, 2019; Omomowo and Babalola, 2019).

Bacterial and fungal endophytes ubiquitously inhabit plant tissues without causing any adverse effect. On the contrary, their presence is often of benefit for the host, as they improve tolerance to abiotic adversities, enhance growth, and, relevantly, can modulate plant immune response and suppress pathogen colonization (Dini-Andreote, 2020). Since endophytic microorganisms typically cover the same ecological niches occupied by fungal and bacterial phytopathogens, they have been widely proposed as biocontrol agents that could be used as an alternative to pesticides (Compant et al., 2013).

Thanks to the multifaceted role they play, endophytic microbial resources are now considered crucial in the perspective of their potential use to achieve sustainable improvements in the agro-food system. As a consequence, there is now a scientific ferment trying to analyze every aspect of their interaction with plants and associated pathogens.

With 16 Original Research Articles and one Review, this Research Topic provides an overview of the current state of the art on the large research effort currently dedicated to understanding the role of endophytes in plant health and defense against pathogens.

### A CROSSTALK WITH PLANT DEFENSE PATHWAYS

Among the most challenging aspects resulting from the investigation on the application of endophytes, and in particular of plant growth-promoting bacteria (PGPB), there is the ability of

**5**

several strains to trigger plant defense mechanisms (Ma, 2017). Commonly, PGPB-induced systemic resistance (ISR) is found to be associated with the up-regulation of genes involved in the pathways of jasmonic acid and ethylene (Pangesti et al., 2016). Biochemical responses such as increased synthesis of reactive oxygen species (ROS) and phenolic metabolites (Benhamou, 1996; Samain et al., 2017) are often associated with ISR, as well as anatomical modifications like the deposition of callose and lignin in the endophyte-colonized tissues (Benhamou, 1996; Constantin et al., 2019).

The blurred distinction between ISR and the pathogeninduced systemic acquired resistance (SAR) (Van Loon et al., 1998) was manifested in the study of Samain et al., in which certain Paenibacillus strains (i.e. PB2), when used to control Mycosphaerella graminicola, induced up-regulation of genes, such as pathogenesis-related proteins (PR1) and chitinases, usually considered as markers of SAR. This peculiar induced resistance is of interest, as it may be a more usual phenomenon, previously observed in other PGPB genera, like Bacillus and Pseudomonas (Park and Kloepper, 2000; Trotel-Aziz et al., 2008; Samain et al., 2017) and confers wheat a durable resistance. Its mode of action, which appears to be significantly influenced by the pathogen strain, the plant growth phase, and its genotype, needs to be addressed in further detail.

The duration of the resistance effect to pathogen-induced biotic stresses that endophytes may activate is, indeed, a key point in management strategies. In highlighting the ability of Rhizobium etli, a common bean symbiont to activate robust defense responses against the pathogen Pseudomonas syringae pv. phaseolicola, Diaz-Valle et al. noted that R. etli-primed plants seem to develop a transgenerational defense memory. The persistence of this capability in F1 generation appears to be related to transcription factors, independent from ethylene signaling pathway, and again, involved in the activation of PR gene expression, as already proposed by (Huang et al., 2016).

Although in recent years associative symbioses have been widely studied in several beneficial bacteria (Ahemad and Kibret, 2014; Coutinho et al., 2015), relatively few studies analyzed their effects on the transcriptional response of plants. Relying on an established model of symbiosis, that constituted by rice and Burkholderia sensu lato (s.l.) (Cottyn et al., 2001; Mannaa et al., 2019), King et al. have described the differences in transcriptional regulations induced by two closely related PGPB with different phylogenetic and ecological backgrounds. Each strain induced a unique expression pattern in the jasmonic acid signaling pathway, and, interestingly, differences have been related to distinct colonization strategies.

Biochemical changes triggered in plants challenged with PGPB are often combined with anatomical alterations. Rodriguez et al. have shown that Gluconacetobacter diazotrophicus is capable of inducing a series of structural changes in inoculated Arabidopsis thaliana seedlings through the deposition of callose. As a result of this sclerosis in root, stem, and leaf tissues, the plant reinforces cell wall and withstands colonization by Ralstonia solanacearum responsible for wilt disease.

### COOPERATIVE ENDOPHYTE-MEDIATED RESISTANCE

The study by de Lamo and Takken focuses on an interesting example of endophyte-mediated resistance (EMR), a mechanism that seems distinct from ISR and SAR, as jasmonate, ethylene, and salicylic acid are not involved (Pieterse et al., 2014; Constantin et al., 2019). In this case, the target plant (tomato) established a tri-partite interaction with two different strains of the same root-invading fungus (Fusarium oxysporum). While pathogenic strains employ host-specific effectors to interfere with host immune signaling (van Dam et al., 2018), co-inoculation of pathogenic Fusarium strains with endophytic Fusarium strains induces resistance responses and reduces negative disease effects. This kind of tri-partite interactions that leads to EMR was also investigated by Del Barrio-Duque et al. In the attempt to identify bacterial strains that stimulate the growth of the beneficial fungus Serendipita indica (Varma et al., 1999; Gill et al., 2016), the authors found that strains belonging to Mycolicibacterium genus might boost the beneficial effects triggered by S. indica on tomato plants, while decreasing severity of the symptoms caused by F. oxysporum and Rhizoctonia solani. This example of cooperation in triggering the host response is intriguing, as it could be explained by the presence in the helper's genome of several genes involved in vitamin and secondary metabolite production that might supplement S. indica bioenergetic capacity. Thus, cooperation among cross-talking microbial players may be needed to restrain pathogen colonization (Zuccaro et al., 2011; Salvioli et al., 2016).

### ENDOPHYTES PRODUCE A VARIETY OF EXPLOITABLE BIOACTIVE METABOLITES

An interesting facet of the interaction between endophytes and their hosts is the capacity of many microorganisms to improve the plant's resistance by providing several bioactive metabolites (Gunatilaka, 2006). In some cases, the release of volatile compounds emitted by plant-associated bacteria has risen to the fore as a promising sustainable strategy to prevent the proliferation of above-ground fungal pathogens (Köberl et al., 2013; Bailly and Weisskopf, 2017; Garbeva and Weisskopf, 2020). Bruisson et al. have identified in grapevine leaf microbiome two Bacillus subtilis and B. cereus strains able to inhibit the growth of Phytophthora infestans, putatively through the emission of volatile compounds identified among pyrazines, chalconoids and tryptophan-derivatives.

The study by Teimoori-Boghsani et al. sheds light on another aspect of great interest, such as the potential of endophytes isolated from Salvia abrotanoides (Kar.) to induce synthesis of the bioactive diterpenoid cryptotanshinone by the plant and while doing that to produce the same molecule independent of the host. Their findings confirm the ability of endophytes to hijack the host's metabolic setups while providing an interesting basis for agricultural and pharmaceutical exploitation of medicinal plants for the production of higher amounts of bioactive constituents.

#### COMBINING CLASSICAL AND MODERN APPROACHES TO STUDY THE PLANT-ASSOCIATED MICROBIOTA

In recent years, it has become clear that the complex structure of plant-associated microbial communities is a major driver of plant health (Lebeis et al., 2012; Bulgarelli et al., 2013), and a deeper understanding of the endophytic microbiome has the potential to become a pivotal tool for reducing the incidence of plant disease (Bloemberg and Lugtenberg, 2001).

In recent decades, plant microbial ecology has experienced a relentless proliferation of available techniques, and this multiplicity of approaches is also reflected in the studies gathered in our Research Topic. In exploring a wide array of different plant environments, the proposed studies ranged from culturedependent screenings, where bacterial and fungal communities were profiled using PCR amplification of 16S ribosomal RNA gene or Internal transcribed spacer (ITS), as in Abdelshafy Mohamad et al., Teimoori-Boghsani et al., Bruisson et al., toward high-throughput technologies, as in the metagenomics strategy proposed by Liu et al., Araujo et al., Elsayed et al., and Anguita-Maeso et al.

This recourse to NGS technologies has provided a fundamental contribution, as they allow to reveal the presence of even rare microbial species and the interactions between these complex communities and their hosts, reaching a depth of resolution previously unimaginable (Bentley et al., 2008; Lebeis et al., 2012). Nevertheless, the importance of combining both culture-dependent and -independent methods to characterize the plants' microbiota was nicely illustrated by Anguita-Maeso et al., and provided support to the notion that no one method can capture the plant microbiome in its entirety. This notion appears to be especially relevant to the characterization of endophytic communities inhabiting nutritionally poor environments, such as the xylem vessels in perennial crops (Aranda et al., 2011; Mendes et al., 2011; Dissanayake et al., 2018).

#### THE RISING ROLE OF STREPTOMYCES AND SOIL-BORN ENDOPHYTES

As reviewed by Romano et al., the availability of the identification methods, not merely based on morphological characteristics, is of paramount importance to complement traditional methodologies in screening persistence of bioinoculants in the rhizosphere and their pattern of synergy with native microorganisms sharing the same niche. The interactions between rhizosphere and endosphere microbiomes and its dynamics have proved to play a critical role in shaping the agronomic traits of crop plants (Schlaeppi and Bulgarelli, 2015) and their study may help to identify effective biocontrol strains that can promote crop yield and reduce the consequences of serious infestations.

Among rhizosphere colonizers Streptomyces spp. are building a reputation as biocontrol agents against mycotoxigenic fungi, including numerous Fusarium spp. threatening cereal crops (Palazzini et al., 2007; Jung et al., 2013; Colombo et al., 2019). Colombo et al. identified an effective Streptomyces strain that was able to dramatically reduce Fusarium head blight symptoms in wheat and to prevent pathogen spread. Liu et al. provided evidence that Streptomyces strains isolated from Glycine max rhizosphere show excellent growth-promoting activity in soybean, in addition to delivering an important antagonistic activity against Sclerotinia stem rot disease. This two-side beneficial activity also resulted in the efforts of Araujo et al. who demonstrated the successful application of Streptomyces strains to accelerate the maturation of wheat heads and positively interfere with root microbiome challenged with severe R. solani infestation.

#### CONCLUDING REMARKS AND FUTURE CHALLENGES

Summarizing its overall heterogeneous composition, this Research Topic well represents the variety of experimental approaches and possible directions in studying this broad and very attractive area of science. The collection of results, presented in the papers, opened a unique knowledge window to the determinants and mechanisms that regulate the dual plant-endophyte interplay, and at the same time, to increased levels of complexity of the tri-partite interaction with phytopathogenic agents that cause severe diseases. The proposed approaches provide insights crucial for the development of new agro-biotechnological strategies for plant protection that will improve food security and environmental sustainability.

We also see our effort as an opportunity to feed the debate on several open questions that clearly emerged from the proposed research. For instance, we may point out the need to complement traditional microbiological approaches with next-generation -omics technologies, to capture as much diversity as possible (Singh, 2019). There is also the necessity to strengthen the newly discovered evidence by providing bio-analytical methods that allow tracking the persistence of bioinoculants and to understand their relationships with the autochthonous microbial communities, in the environments where they are released. Last but not the least, the alert on possible biosafety issues. So far, we just gained a partial knowledge on metabolic profiles of the majority of the components of plant-associated above-ground communities. However, we should be aware that the absence of phylogenetic relationships with human pathogens does not imply that endophytes do not present any risk for our health. Biosafety characteristics should be well addressed before proceeding with the environmental application (Keswani et al., 2019).

Based on these grounds, we believe that the multidisciplinary approach, proposed in our compendium, may result in the best strategy stimulating fast scientific progress on this challenging issue.

#### AUTHOR CONTRIBUTIONS

MM wrote the draft and submitted the final version. OB, KP, DH, and AO edited the manuscript and provided critical review. All authors contributed to the article and approved the submitted version.

#### FUNDING

The participation of KP was partially funded by the General Secretariat for Research and Technology (GSRT) under the PRIMA Programme (INTOMED Project). AO was supported

#### REFERENCES


by the Faculty of Agriculture and Ministry of Education, Science and Technological Development, Republic of Serbia, contract no. 451-03-68/2020-14/200116.

#### ACKNOWLEDGMENTS

We would like to thank all authors and reviewers who have contributed to our Research Topic. We are sincerely grateful to Prof. Einat Zchori Fein for her initial involvement in the launch of this initiative. We acknowledge Prof. Brigitte Mauch-Mani for contributing as associate editor to the peer-review process.

in stems of grapevine (Vitis vinifera). Fungal Diversity 90 (1), 85–107. doi: 10.1007/s13225-018-0399-3


microbes. Annu. Rev. Phytopathol. 52, 347–375. doi: 10.1146/annurev-phyto-082712-102340


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 Morelli, Bahar, Papadopoulou, Hopkins and Obradovic. 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 Effect of Plant Genotype, Growth Stage, and Mycosphaerella graminicola Strains on the Efficiency and Durability of Wheat-Induced Resistance by Paenibacillus sp. Strain B2

#### Erika Samain1,2, Thierry Aussenac<sup>3</sup> and Sameh Selim<sup>1</sup> \*

#### Edited by:

Massimiliano Morelli, Institute for Sustainable Plant Protection, Italian National Research Council (IPSP-CNR), Italy

#### Reviewed by:

Esther Menendez, University of Évora, Portugal Joana Figueiredo, Universidade de Lisboa, Portugal

#### \*Correspondence:

Sameh Selim sameh.selim@unilasalle.fr; sameh.selim@lasalle-beauvais.fr

#### Specialty section:

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

Received: 07 February 2019 Accepted: 18 April 2019 Published: 09 May 2019

#### Citation:

Samain E, Aussenac T and Selim S (2019) The Effect of Plant Genotype, Growth Stage, and Mycosphaerella graminicola Strains on the Efficiency and Durability of Wheat-Induced Resistance by Paenibacillus sp. Strain B2. Front. Plant Sci. 10:587. doi: 10.3389/fpls.2019.00587 <sup>1</sup> AGHYLE, College of Agricultural Sciences, Institut Polytechnique UniLaSalle, Beauvais, France, <sup>2</sup> SDP, Laon, France, <sup>3</sup> UP Transformations & Agro-Ressources, Institut Polytechnique UniLaSalle, Beauvais, France

Plant-growth-promoting rhizobacteria are known as potential biofertilizers and plantresistance inducers. The current work aims to study the durability of the resistance induced as a response to the inoculation of wheat grains with Paenibacillus sp. strain B2 (PB2) and its influence by plant genotype, growth stage, and Mycosphaerella graminicola strain (the causal agent of Septoria tritici blotch or STB). The results of the plate-counting method showed that PB2 has high potential for wheat-root external colonization [>10<sup>6</sup> colony-forming unit (CFU)/g of root], and the quantitative real-time polymerase chain reaction (qPCR) analysis demonstrated its internal root-colonization capacity on all tested cultivars. However, the colonization seems to be dependent on wheat-growth stage. The durability of PB2-induced resistance (PB2-IR) was tested at the 3-leaf, tillering, and flag-leaf-growth stages. Additionally, the results showed that the PB2-IR is durable and able to protect the flag leaf, the most important leaf layer during grain fill. It conferred a high protection efficiency (55–94%) against four virulent strains of M. graminicola and over 11 wheat cultivars with different resistance levels to STB. Although, PB2-IR is dependent on M. graminicola strains, wheat genotypes and growth stages, its efficiency, under field conditions, at protecting the last wheat-leaf layers was not an influence. However, it showed 71–79% of protection and reached 81–94% in association with half of the recommended dose of Cherokee <sup>R</sup> fungicide. This may be explained using laboratory results by its direct impact on M. graminicola strains in these leaf layers and by the indirect reduction of the inoculum coming from leaves infected during the earlier growth stages. Gene expression results showed that PB2-IR is correlated to upregulation of genes involved in defense and cell rescue and a priming effect in the basal defense, jasmonic acid signaling, phenylpropanoids and phytoalexins, and reactive oxygen species gene markers. To conclude, PB2 induces a

**10**

high and durable resistance against M. graminicola under controlled and field conditions. The PB2-IR is a pathogen strain and is plant-growth-stage and genotype dependent. These results highlight the importance of taking into consideration these factors so as to avoid losing the effectiveness of induced resistance under field conditions.

Keywords: Mycosphaerella graminicola, Paenibacillus sp. strain B2, induced systemic resistance, pathogen strain, wheat genotype, wheat-growth stage

#### INTRODUCTION

Septoria tritici blotch (STB), caused by Zymoseptoria tritici (Teleomorph: Mycosphaerella graminicola), is regarded as the most important disease in wheat in Europe and many other countries; yield damage from this disease can reach 40% (Selim et al., 2011). To control this pathogen, the use of fungicides, mainly the sterol 14α-demethylase inhibitors (DMI) alone or mixed with succinate dehydrogenase inhibitors (SDHI), is privileged. In addition to its negative impact on the environment, animals, and human health, its application is fast, limited only by the emergence of fungicide-resistant genotypes (Selim, 2009). However, many alternatives to the integrated management of STB and other crop diseases are studied as the resistance inducers (RIs) of plant defense mechanisms such as elicitors and plantgrowth-promoting rhizobacteria (PGPR) (Selim et al., 2010; Samain et al., 2017). PGPR induce local and systemic resistance, which is commonly known as induced systemic resistance (ISR). PGPR-mediated ISR resembles pathogen-induced systemicacquired resistance (SAR) which enhances the plant's immunity system against biotic and abiotic stresses (van Loon et al., 1998; Samain et al., 2017). However, ISR is a quantitative non-specific resistance and broad-spectrum pathogen that is likely influenced by many factors such as environmental conditions, crop nutrition, microbial communities, pathogen strain, and host genotype (Walters et al., 2005; Samain et al., 2017). Unfortunately, almost all investigations under controlled conditions are realized with a limited number of strains and plant genotypes and even at the early plant-growth stage. This could explain why the high efficiency of the resistance induced under laboratory conditions is not maintained under field conditions.

Induced systemic resistance and SAR depend on different signaling pathways. Whereas, SAR acts through a salicylic acid (SA) pathway, followed by upregulation of pathogenesis-related (PR) genes such as PR1, PR2, and PR5, ISR depends on jasmonic acid (JA) and ethylene (ET) pathways (van Loon et al., 1998). PGPR-mediated ISR enhances the formation of physical barriers such as callose and lignin, the synthesis of plant defense chemicals as reactive oxygen species (ROS), phytoalexins and phenolic compounds (Benhamou, 1996; Samain et al., 2017), but certain PGPR do not induce PR proteins (Hoffland et al., 1995). On the other hand, the upregulation of genes, known as SAR pathway markers, such as PR2, chitinases and PR1, have been observed with PGPR including the genera Bacillus, Pseudomonas, Paenibacillus, and Paraburkholderia (Park and Kloepper, 2007; Samain et al., 2017; Baccari et al., 2019). The overexpression of PR1 induced by PGPR might be explained by the abiotic stress as a response to root colonization by bacteria (Timmusk and Wagner, 1999). Furthermore, the stimulation of both ISR and SAR at the same time as the simultaneous enhancement of the SA, JA, and ethylene pathways has been demonstrated without antagonism effects (van Wees et al., 2000). On this subject, we previously showed that Paenibacillus sp. strain B2 (PB2) protects wheat plants against STB, and this was associated with upregulation of PR1 and other genes implicated in the basal defenses, ROS, SA, and JA pathways. However, these results were wheat-genotype dependent (Samain et al., 2017).

Moreover, PB2 produces the cyclic lipo-polypeptides paenimyxin antibiotics, which are antagonistic to several plantpathogenic bacteria (Gram + and Gram −) and fungi, including M. graminicola (Selim et al., 2005; Samain et al., 2017), and have no negative effects on the genetic structure of soil microbial communities (Selim et al., 2007). Moreover, paenimyxin induces resistance in alfalfa (Medicago truncatula) against Fusarium acuminatum and in wheat against M. graminicola (Selim et al., 2010; Samain et al., 2017).

On the other hand, it has been demonstrated that PB2 promotes plant growth only in co-inoculation with another beneficial microorganism known as Funneliformis mosseae (previously Glomus mosseae) in tomato (Budi et al., 1999) and Curtobacterium plantarum in wheat (Samain et al., 2017), by its helper effect stimulating root mycorrhization and colonization, respectively.

The objectives of the current work are to study the durability and efficiency of the resistance induced by PB2 against STB over different wheat genotypes, growth stages and pathogen strains.

#### MATERIALS AND METHODS

The experimental design of this work (**Supplementary Figure S1**) covered the study of: (1) the impact of wheat genotypes and growth stage on the colonization of roots by PB2, (2) the impact of wheat genotypes on the resistance induced by PB2 against STB, (3) the impact of wheat-genotype-growthstage–M. graminicola strain interactions on durability of the resistance induced by PB2, (4) gene expression analysis of PB2-wheat-genotype–M. graminicola strain interaction, and (5) to confirm the PB2-resistance induced against M. graminicola under field conditions.

#### Microorganisms and Inoculum Preparation

Paenibacillus sp. strain B2 (Budi et al., 1999) was kindly provided by Dr. van Tuinen, INRA Dijon, France. Four highly virulent M. graminicola strains were used in this study: strain

IPO323 (provided by Dr. F. Suffert, INRA Grignon); strain 1193, characterized as a moderately resistant strain to DMI fungicides (TriMR), with three SNP mutations (M-281-V, A-379-G, I381- V) (Selim, 2017, NCBI GenBank database accession number KX356102); and strains TO256 (TriMr, Selim, 2009) and ST38 (low resistance to DMI fungicides, unpublished data), which were obtained from M. graminicola collection strains held in the authors' laboratory. PB2 inoculum was prepared as mentioned in Selim et al. (2005) and the M. graminicola inocula were prepared as mentioned in Selim et al. (2014). Briefly, to prepare the final inocula, bacterial cells, and fungal spores were collected from liquid cultures by centrifugation at 2655 × g for 5 min at 15◦C, washed twice with sterile distilled water, and then suspended in 10 mM MgSO<sup>4</sup> (Sigma <sup>R</sup> M-9397) containing 0.1% Tween 20 surfactant. Bacterial cells and fungal spore vitality were checked by spreading 100 mL of inoculum on Luria-Bertani (LB) or potato dextrose agar (PDA) media, respectively.

#### Plant Material and Growth Conditions

We used 11 winter-wheat cultivars – Alixan, Terroir, Altigo, Expert, Chevron, Complice, Hyking, Boregar, Cellule, Fructidor, and Hyfi – with approximately the same earliness and different levels of resistance against STB, 4, 5, 5.5, 5.5, 5.5, 6, 6, 6, 6.5, 6.5, and 7, respectively (Arvalis Institut du Végétal, 2017), on a scale from 1 (fully susceptible) to 9 (fully resistant) (**Table 1**). The Hyking and Hyfi cultivars are hybrids. The grains were disinfected according to Samain et al. (2017), with a few modifications, as follows: incubation in a solution of oxytetracyclin, streptomycin, penicillin, and ampicillin antibiotics (100 mg/L of each) overnight to ensure a large broad-spectrum activity against bacterial strains, then suspended in 10% calcium hypochlorite solution for 10 min and washed three times in autoclaved Milli-Q water after each disinfection step. The sterilized grains were pre-germinated on 0.5% water-agar medium and incubated in darkness at 4◦C for 24 h, 20◦C for 48 h, and 4◦C for 24 h. The germinated grains were transferred into an inoculum of PB2 adjusted to 10<sup>6</sup> CFU/mL of 10 mM MgSO4; 1 mL per grain for 1 h with light shaking.

TABLE 1 | Wheat cultivars used to study the impact of wheat genotypes on the resistance induced by Paenibacillus sp. strain B2.


<sup>∗</sup>The susceptibility rating is on a scale of 1–9, where 1 represents "susceptible" and 9 represents "resistant" (Arvalis Institut du Végétal, 2017). ∗∗Hybrid cultivars.

For the non-inoculated control, the grains were immersed in 10 mM MgSO4. After inoculation, the grains were transferred into 250-mL pots containing a sterilized soil mixture of silt-loam soil and sand (1:1, v/v). The pots were incubated in phytotron at 18◦C (±2 ◦C), 40% humidity, for a 16-h photoperiod with 185 µmol m−<sup>2</sup> s <sup>−</sup><sup>1</sup> photon flux density supplied by high-output white fluorescent tubes (Philips Master Cool White 80 W//865, Lamotte Beuvron, France). The plants were watered three times a week with 50 mL of distilled water per pot.

### Root Colonization by PB2

External and internal root colonization by PB2 was evaluated 7 days after sowing (das) by counting CFU in LB agar medium, as mentioned in Samain et al. (2017), using the four wheat cultivars Alixan, Altigo, Cellule, and Hyfi and using specific primers of the 16S rDNA of PB2 at 21 das. External and internal colonization by PB2 was also determined at the 3-leaf (3-L) and flag-leaf (FL) growth stages (GS) on Alixan and Cellule, by the quantitative real-time PCR (qPCR). Specific primers were designed based on the PB2 16S rDNA gene (GenBank accession No AJ011687) (**Table 2**). DNA was extracted from the plant roots using DNeasy 96 Plant kit (Qiagen, United States), according to the manufacturer's protocol. DNA quantity and quality were confirmed by Nanodrop (Thermo Fisher Scientific, Waltham, MA, United States). SYBR Green qPCR assays were carried out in a reaction mixture of 25 µL that contained the following: 12.5 µl Universal Quantifast SYBR Green PCR master mix (Qiagen, United States), 0.3 µM of each primer, 50 ng of DNA, and water up to a volume of 25 µl. The conditions of the quantitative PCR were as follows: 5 min at 95◦C, followed by 40 cycles of 10 s at 95◦C and 30 s at 60◦C. One final step from 60 to 95◦C with an increase of 0.2◦C s−<sup>1</sup> was added to obtain a specific melting curve (**Table 2** and **Supplementary Figure S3**). All quantitative PCR was carried out using StepOnePlus (Thermo Fisher Scientific <sup>R</sup> ).

The standard curve, obtained by plotting known amounts of PB2 DNA against Ct values, was used to determine the amplification efficiency (**Table 2** and **Supplementary Figure S5**). The resulting regression equations were used to calculate the amounts of PB2 DNA in the test samples.

#### PB2-Resistance Induced in Wheat Against M. graminicola

Grain sterilization, pre-germination, inoculation with PB2, and phytotron conditions were carried out, as mentioned above. The 11 wheat cultivars listed above were used and infected at 3-L GS with 0.5 mL/leaf of an inoculum of M. graminicola strain IPO323, consisting of 2 × 10<sup>6</sup> conidia/mL supplemented with 0.1% of Tween 20.

Susceptible and resistant non-hybrid wheat cultivars – Alixan and Cellule, respectively – were used to evaluate the effect of pathogen strains and growth stage on the efficiency of the resistance induced by PB2. They were inoculated with the four M. graminicola strains at 3-L, tillering (Ti), and FL GS. The inocula of each strain were prepared as mentioned above and applied as described in Selim et al. (2014). Briefly, 21-dayold plants were infected by spraying a 2 mL M. graminicola

#### TABLE 2 | Oligonucleotide primer sequences of wheat-defense genes and Paenibacillus sp. strain B2 16S rDNA.


<sup>∗</sup>Tm, primer's annealing temperature; MT, amplicon's specific melting temperature. ∗∗Primer pairs were designed using the Primer Express <sup>R</sup> program and tested for secondary structure using the AmplifX <sup>R</sup> program. All used primers did not show any form of dimerization.

inoculum (2 × 10<sup>6</sup> spores/mL) over the whole plant. Controls were sprayed with a solution of 10 mM MgSO<sup>4</sup> containing 0.1% Tween 20 surfactant. Five repetitions were carried out for each condition. Three control modalities were used as non-infected with M. graminicola and non-inoculated with PB2 (C−), inoculated with PB2 without pathogen infection (PB2), and non-inoculated with PB2 infected with pathogen (MG). Modalities inoculated with PB2 and infected with M. graminicola (PB2/MG) were compared to the control modalities. Seventeen days after infection with M. graminicola, leaves were collected and lyophilized to evaluate the protection efficiency as a response to PB2 using the qPCR. The DNA extraction was as mentioned above and the quantification of M. graminicola using qPCR was realized as mentioned in Selim et al. (2014). Briefly, primers and TaqMan minor groove binder probe (For: GCCTTCCTACCCCACCATGT; Rev: CCTGAATCGCGCATCGTTA; Probe: FAM-TTACGCCAAGACATTC-MGB) were used to target a 63-bp fragment of the M. graminicola β-tubulin specific gene (GenBank accession no. AY547264) (Bearchell et al., 2005). A TaqMan assay was carried out in 25 µL of a reaction mixture that contained 12.5 µL Universal TaqMan PCR Master Mix (Life Technologies SAS, Villebon-sur-Yvette, France), 0.3 µM of each primer, 0.2 µM probe, 200 ng DNA and water to a volume of 25 µL. The conditions of the qPCR determination were as follows: 10 min at 95◦C, followed by 40 cycles of 15 s at 95◦C and 1 min at 60◦C. All qPCR experiments were carried out using a StepOnePlus Real-Time PCR System (Thermo Fisher Scientific <sup>R</sup> ). qPCR analysis of the M. graminicola β-tubulin gene was calibrated from 10<sup>2</sup> to 10<sup>7</sup> copies by serial dilution of the appropriate cloned target sequence, as previously described (Selim et al., 2011).

#### RNA Extraction and Relative Gene Expression Quantification by Real-Time PCR

At the 3-L GS, aerial parts of the Alixan and Cellule plants were collected at the time of infection (T0) with M. graminicola, at 6, 12, 24, and 48 h after infection (hai), and 3, 5, 9, and 11 days after infection (dai) to study the evolution of the defense gene expression. Samples were stored directly in liquid nitrogen. RNA extraction and cDNA synthesis were carried out using RNeasy <sup>R</sup> Mini Kit and QuantiTect <sup>R</sup> Reverse Transcription Kit (Qiagen, United States), respectively, following the manufacturer's protocol. The gene expressions of 20 wheatdefense genes were studied using specific primers (**Table 2**). qPCR conditions were as described in Samain et al. (2017). Briefly, the Quantifast <sup>R</sup> SYBR <sup>R</sup> Green PCR Kit (Qiagen, United States) and the StepOnePlus Real-Time PCR Systems (Thermo Fisher Scientific <sup>R</sup> ) were used. Amplification conditions consisted of a denaturation cycle (95◦C for 5 min) and amplification and quantification cycles (95◦C for 10 s, 60◦C for 30 s) repeated 40 times. One final step from 60 to 95◦C with an increase of 0.2◦C s−<sup>1</sup> was added to obtain a specific melting curve for each studied gene (**Table 2** and **Supplementary Figures S2**, **S3**). The glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and β-tubulin (β-TUB) housekeeping genes were used to normalize results and to determine the expression ratio for each cDNA, as described by Ors et al. (2018). Briefly, expression ratios for each cDNA were calculated for each time point, relative to control at the same time using the 2−11Ct described by Livak and Schmittgen (2001), where 11Ct = [Ct Target (Sample) – Ct Reference (Sample)] – [Ct Target (Control) – Ct Reference (Control)] and Ct Reference = geometrical mean (Ct GAPDH: Ct β-TUB). Similar amplification efficiencies ranging between 90 and 110% were checked for all the tested primers (**Table 2** and **Supplementary Figures S4**, **S5**) and expression ratio values of two were considered as a minimum to be significantly different from the control.

#### Confirmation of the Efficiency of the PB2-Resistance Induced Under Field Conditions

The field site was located at Nanteuil-la-Fosse, France. Based on the susceptibility to M. graminicola, we used the moderately susceptible cultivars Expert and Chevron, with the resistance level of 5.5 (Arvalis Institut du Végétal, 2017). At sowing, wheat grains were coated with PB2 using 1 mL per 100 g of grains of an inoculum of 5 × 10<sup>7</sup> CFU/mL of water supplemented with 20% of the general commercial sugar. Trials were performed during the 2017–2018 growing season using a completely randomized block design with four replicate plots of 4 m × 12 m. Fertilization was according to plant requirements and protection against other diseases was carried out. Twenty leaves were randomly sampled from the third leaf layer (Fn-2) below the flag leaf (Fn) at Zadoks growth stage 49 (GS49) (Zadoks et al., 1974), 14 days after the date of Cherokee <sup>R</sup> (375 g/L chlorothalonil+62.5 g/L propiconazole+50 g/L cyproconazole, Syngenta, France) fungicide application at the recommended dose (RD) or half the RD (HD). The efficiency of PB2 at protecting wheat leaves against STB was compared to modalities that were non-treated, non-inoculated, fungicide-treated, or inoculated with PB2 in association with treatment with a HD of fungicide. Disease evolution was determined by assessing visual symptoms and by qPCR analysis, as mentioned above. The necrotic area related to Septoria blotch was recorded and then the leaves were stored at −80◦C until lyophilization.

#### Statistical Analysis

At least three biological replicates and five technical replicates were used for the experiments. For all experiments, significant differences were evaluated using ANOVA followed by Tukey's post hoc test (α = 0.05) and the XLSTAT <sup>R</sup> statistics program (version 2014, Addinsoft, Paris, France).

## RESULTS

#### Impact of Wheat Genotypes on Root Colonization by PB2

The results shown in **Figure 1**, of the two methods used to evaluate root colonization by PB2, confirmed the external colonization by PB2 on the four wheat cultivars. The plate

method showed higher levels of root colonization in Alixan and Cellule compared to Altigo and Hyfi, whereas, qPCR analysis did not show any significant differences between cultivars. For the internal colonization, Alixan and Cellule showed higher CFU values than Hyfi, and any CFU was detected with Altigo. Using qPCR, small amounts of PB2 DNA were detected in Altigo and at the same statistical level as in the other three cultivars.

#### Impacts of Wheat-Growth Stage on Root Colonization by PB2

The impact of the wheat-growth stage on root colonization by PB2 was investigated at 3-L and FL GS using the most susceptible and resistant cultivars, Alixan and Cellule, respectively, and using the qPCR method. For external root colonization with PB2, at 3-L GS, the results showed high root colonization in Alixan and Cellule, with, respectively, 34 and 26 pg of PB2 DNA/g of roots without significant differences. At FL GS, a non-significant increase in root colonization was observed in the cultivar Cellule, reaching 49 pg/g. Contrarily, a significant decrease (2 pg/g) was recorded in Alixan. Indeed, at FL GS, the external colonization of Alixan was significantly different from that of Cellule. For internal root colonization by PB2, an increase (2.5 times) was recorded in the two tested cultivars, at FL GS compared to 3-L GS (**Figure 2**). However, these differences were not significant between the two cultivars, nor between the two studied stages (**Figure 2**).

#### Impacts of Wheat Genotypes on the Resistance Induced by PB2 Against STB

The impact of wheat genotypes on the resistance induced by PB2 against STB was evaluated using 11 wheat cultivars with the same earliness but a different susceptibility rating given by the Arvalis Institut du Végétal. The plants had been inoculated at 3-L GS with the wild-type strain IPO323 of M. graminicola, and the disease level was quantified at 17 dai using qPCR and expressed in the β-tubulin copy number in 100 ng of leaf DNA (BCN100 ng), as previously mentioned in Selim et al. (2014). The average of leaves' BCN100 ng in the controls without PB2 were 510.7, 285.2, 202, 189.8, 262.8, 173.8, 213.5, 161.6, 331.5, 259, and 192.5, respectively, for Alixan, Terroir, Altigo, Expert, Chevron, Complice, Hyking, Boregar, Cellule, Fructidor, and Hyfi (**Figure 3**). The level of protection was determined as the percentage of the reduction of BCN100 ng as a response to root inoculation with PB2 compared to the control infected with M. graminicola and non-inoculated with PB2. The results in **Figure 3** show more than 56% of protection efficiency induced by PB2, over the 11 cultivars tested, with significant differences between cultivars. The two more susceptible cultivars, Alixan and Terroir, demonstrated 94 and 91% of protection, respectively; however, the protective effect induced by PB2 was not correlated with the resistance level of the cultivars.

#### Impact of Wheat-Genotype-Growth-Stage–M. graminicola Strain Interactions on Durability of the Resistance Induced by PB2

The most susceptible cultivar (Alixan) and the moderately resistant cultivar (Cellule) were used for this experiment. The resistance induced by PB2 against STB was analyzed against four M. graminicola strains, IPO323, 1193, ST38, and TO256, and its durability was followed from the earlier wheat-growth stages (3-L and Ti), until the FL GS, corresponding with 38, 59, and 153 das. The disease level was quantified at 17 dai using qPCR and expressed in BCN100 ng. The protection level against STB as a response to PB2 root inoculation was

FIGURE 2 | Evolution of root external and internal colonization of Paenibacillus sp. strain B2 (PB2) in Alixan and Cellule wheat cultivars, at 3-leaf growth stage and flag-leaf growth stage. Plants were inoculated with PB2 by immersing the pre-germinated seeds in a suspension of 10<sup>6</sup> CFU/mL. The PB2 colonization, represented as DNA amount of PB2 per gram of root, was determined by qPCR. The values shown are the means of three biological replicates and five technical replicates. Bars indicate means ± standard deviations. Different lower-case letters indicate significant differences between treatments, according to ANOVA followed by Tukey's post hoc test (α = 0.05).

estimated by comparing it to the BCN100 ng in the PB2 non-inoculated controls and we used the 40% protection level as a threshold to indicate the importance of protection as a response to root inoculation with PB2 when it is equal to, or greater than, this level (Selim et al., unpublished data). However, the global average of leaves BCN100 ng of the three tested growth stages in the controls without PB2 of Alixan were 413, 515, 2950, and 2000, and of Cellule were 245, 260, 2400, and 1200 for the IPO323, TO256, 1193, and ST38 strains, respectively. Remarkably, the BCN100 ng was approximately 50% less in the moderately resistant cultivar Cellule than in the susceptible cultivar Alixan, except for the strain 1193 (**Figures 4A–D**).

The results in **Figure 4A**, show high and stable protection effects against the IPO323 strain, where the BCN100 ng was reduced more than 59% in the two cultivars and at the three tested growth stages. However, the protection conferred on Alixan was significantly higher than in Cellule at the two first-growth stages.

**Figure 4B** also shows variable protection efficiencies, as a response to PB2 root inoculation, against strain TO256 depending on cultivars and wheat-growth stages. At 3-L GS, PB2 had no protective effect on the cultivar Alixan but induced

48% of protection efficiency in Cellule (**Figure 4B**). At the two subsequent growth stages, this protection situation was inverted for the two cultivars, giving more than 60% of protection to Alixan and no protection to Cellule (**Figure 4B**). Likewise, PB2 also seemed to induce growth-stage-dependent resistance against M. graminicola strain 1193 with high protective effect (>60%) at the earlier and later growth stages, but it had a very low protection efficiency at the Ti GS (**Figure 4C**). The obtained protection efficacy was significantly higher in Alixan than Cellule at the 3-L GS (**Figure 4C**). Furthermore, significant differences were observed between the two cultivars at 3-L GS but not at later growth stages with strains 1193 and TO256. The protection induced by PB2 against strain ST38 was stable over the three tested growth stages and for the two cultivars (**Figure 4D**). Indeed, a high and stable protection efficiency (>55%) was observed for the cultivar Alixan, but with a low protection efficiency (15–30%) for Cellule (**Figure 4D**).

#### Gene Expression Analysis of Paenibacillus B2-Wheat Genotype–M. graminicola Strain Interaction

To explain these results, the expression of genes implicated in the wheat-defense mechanisms were studied, in 3-week-old plants of the Alixan and Cellule cultivars, at time zero (T0, at the moment of infection with pathogen), and at 1 and 3 dai with M. graminicola IPO323 and TO256 strains. These strains were chosen to represent the two types of observed PB2 induced resistance (PB2-IR), the high and stable resistance in both cultivars against IPO323 and the growth stage cultivardependent resistance against the TO256 strain. Moreover, the earlier growth stage is important as the source of the inoculum for the upper-leaf layers.

First, the expression levels of the selected 20 defense genes were analyzed in Cellule and compared to that of Alixan in the controls (C−). The results showed that all the studied defense pathways, except these related to defense and cell rescue, and the allene oxide synthase (AOS) gene showed 1.8-fold less in Cellule compared to Alixan (**Figure 5** and **Supplementary Table S1**).

At T0, significant upregulations (≥1.9-fold), of the pathogenesis-related protein (PR1), chitinase (CHIT), lipase (LIP), lipoxygenase (LOX), phenylalanine ammonia-lyase (PAL), chalcone synthases (CHS), flavonoid 7-O-methyltransferase-like (FLAV), oxalate oxidase (OXO), glutathione-s-transferase (GST), and glutathione peroxidase (GPX) genes, were observed in leaves of the cultivar Alixan as a response to PB2 root inoculation and 2.1- and 3.8-fold of OXO and GPX genes, respectively, in the cultivar Cellule (**Figure 5** and **Supplementary Table S1**).

For the MG modalities, almost all of the tested genes were upregulated as infection with IPO323 and TO256 strains in both tested cultivars Alixan and Cellule. However, the genes

showed more significant upregulations in the MG/PB2 modalities than in the MG modalities labeled with stars in **Figure 5**, and they were used to discriminate the effect of PB2 of that on that of the MG strains. Concerning the strain TO256, where a significant protection was observed in the Cellule cultivar compared to no protection in Alixan, significant upregulations of the PR1, CHIT, and related protein kinase (rpK) genes with 2.9, 5.0, and 9.2-fold, respectively, were observed at 3 dai of the PB2/MG modality. These genes were, respectively, 2.5, 1.6, and 7 times more than that of the MG modality (**Figure 5** and **Supplementary Table S1**). However, none of these genes were upregulated at 1 dai. In Alixan, these genes were not upregulated at 3 dai, but at 1 dai significant upregulations were observed in the PR1, β-1,3-glucanase (GLU), and GPX genes with 3.5, 2.4, and 61.8-fold, respectively, and 3.5, 1.4, and 19 times more than in the MG modality.

In the second type of cultivar-independent resistance, where both cultivars showed significant protection against the IPO323 strain, in Alixan at 1 dai, the PR1, CHIT, thaumatin-like protein (TLP), and peroxidase (POX) genes were upregulated with 3.7, 2.2, 1.8, and 2.5-fold, respectively, in the PB2/MG modality, corresponding to 1.7, 1.6, 4.0, and 8.0 times, respectively, more than that of the MG IPO323 modality. At 3 dai, only the LOX gene showed significant upregulation with 2.8-fold in the PB2 modality, and it was not affected in the MG modality. In the PB2 modality of the Cellule cultivar at 1 dai, the PR1, FLAV, catalase (CAT), rpK, and WRKY1 transcription factor (WRKY) genes were upregulated with at least 2.8-fold, representing ≥1.5 times more than that of the MG modality (**Figure 5** and **Supplementary Table S1**). At 3 dai, the overexpression of PR1, CHIT, GLU, TLP, MAP kinase (WCK1), PAL, FLAV, and POX were at least 2.0 times more than that of the MG modality (**Figure 5** and **Supplementary Table S1**).

#### Gene Expression Time-Course Analysis

For better understanding of the defense mechanisms implicated in wheat genotypes-M. graminicola-PB2 interactions, the expression of the 20 defense-related genes, as a response to root inoculation with PB2, was studied using Alixan (**Figure 6** and **Supplementary Table S2**) and Cellule (**Figure 7** and **Supplementary Table S3**) cultivars, at 6, 12, 24, and 48 hai, and 3, 5, 9, and 11 dai with the IPO323 strain at 3-L GS. The strain IPO323 was chosen to represent the high and stable PB2-IR in both cultivars for the objective of collecting more information about genes that could be implicated in the protection against M. graminicola. The genes showed a significant upregulation in the PB2/MG modalities compared to the MG modalities labeled with stars. In Alixan, this case was observed in the PR1, CHIT, GLU, TLP, LOX, AOS, FLAV, POX, and germin-like-protein (GLP) genes with an average expression level over studied

timings of ≥2.4-fold, representing ≥2 times more than that of the MG modality (**Figure 6** and **Supplementary Table S2**). In Cellule, the PR1, CHIT, GLU, TLP, WCK1, PAL, FLAV, POX, CAT, and WRKY genes had an overexpression average ≥2.2-fold, representing ≥1.5 times more than that of the MG modality (**Figure 7** and **Supplementary Table S3**).

#### Confirmation of the PB2-Resistance Induced Against M. graminicola Under Field Conditions

The moderately susceptible cultivars Expert and Chevron (resistance level = 5.5) were chosen to evaluate the protection efficiency of PB2 under field conditions. The disease-infection level was determined on May 17, 2018 corresponding to the GS49 of Chevron and Expert cultivars. At this date, the STB disease pressure was very weak, and no visual symptoms were observed on the three end-leaf layers (Fn, Fn-1, and Fn-2). qPCR analysis was realized on Fn-2-extracted DNA. The results showed 80% of protection efficiency in the fungicide RD and HD modalities. The PB2 modalities showed 71 and 79% protective effects for Expert and Chevron, respectively. These protection levels increased to 81 and 94%, respectively, in Expert and Chevron in modalities where PB2 was associated with an HD of fungicide (**Figure 8**). As the disease pressure was still weak until the end of experimentation, the results of the yield did not show significant differences between modalities and the PB2 non-inoculated controls with an average of 10.54 and 10.96 tons/hectare in Chevron and Expert, respectively (**Supplementary Figure S6**).

### DISCUSSION

With the increasing social interest in avoiding, or at least reducing, pesticide applications, the use of natural RIs as PGPRs has been one of the most important strategies studied over the last 20 years. Unfortunately, knowledge of their use is not yet sufficient to transfer laboratory results into the field in such a way as to maintain their high-protection efficiency. We showed previously that the plant genotype is one of the main factors that might influence the efficiency of RIs under field conditions (Samain et al., 2017; Ors et al., 2018). Here, we studied the impact of wheat genotypes and M. graminicola strains on the durability of the resistance induced over different growth stages as a response to root inoculation with Paenibacillus sp. strain B2.

In the current work, the protection efficiency of PB2 against STB was observed under laboratory conditions and confirmed under field conditions, where PB2 showed protection percentages similar to RD and HD of the applied fungicide. Unfortunately, the STB disease pressure was not sufficient to evaluate the impact of

FIGURE 7 | Time course of relative expression of wheat-defense genes in the moderate resistant cultivar Cellule at the time of infection with M. graminicola (MG) strain IPO323 (T0), 6, 12, 24, and 48 h after infection (hai), and 3, 5, 9, and 11 days after infection (dai). Gene expression levels of the following tested modalities, Paenibacillus sp. strain B2-inoculated and MG-non-infected (PB2), PB2-non-inoculated and MG-infected (MG), and PB2-inoculated and MG-infected (PB2/MG), were determined comparing to the PB2-non-inoculated and MG-non-infected control modalities. Stars indicate gene induction ≥twofolds and significant differences between PB2/MG and MG modalities, according to ANOVA followed by Tukey's post hoc test (α = 0.05). The values shown are means of three biological replicates and five technical replicates.

replicates and five technical replicates. Bars indicate means ± standard deviations. Different lower-case letters indicate significant differences between treatments, according to ANOVA followed by Tukey's post hoc test (α = 0.05).

PB2 on yield production compared to the PB2 non-inoculated controls. On the other hand, results confirm the absence of any negative impact on the yield as a response to wheat-PB2 symbiotic relationship. Indeed, the high correlation between disease control and yield benefit usually found when fungicides are applied, is often not significant when applying RIs (Selim et al., 2010). This is probably related to the fact that plant defenses stimulation is energetically costly.

However, the laboratory results highlight the importance of factors such as M. graminicola strains, wheat genotypes, and growth stages in the protection induced by PB2, as well as the response of wheat-defense mechanisms to PB2-wheat genotype-M. graminicola strain interactions.

#### Impact of Wheat Genotype and Growth Stages on Root Colonization With PB2 and the Resistance Induced Against M. graminicola

The results of the plate-counting method showed that root colonization with PB2 is wheat-genotype dependent. Indeed, PB2 CFU varied between cultivars external as well as internal to the roots, even though PB2 was totally absent as endophytic in the cultivar Altigo. To confirm these results, we designed specific primers that proved, using qPCR, to be highly efficient in detecting and quantifying the PB2 16S rDNA gene. On the other hand, the qPCR results confirmed the impact of wheat genotypes on the root internal amount of PB2 DNA. However, the effect of plant genotypes on the endophytic and ectophytic root colonization by PGPR was observed previously in tomato–Pseudomonas spp. interaction (Pillay and Nowak, 1997). This finding might be explained by the physical and chemical properties of root exudates, which vary between plant genotypes, and also by the soil microbial communities that interact during the specific dialogue between plants and PGPR (Kumar et al., 2007; Samain et al., 2017). Root exudates, which have a high diversity of organic nutriments such as sugars, vitamins, organic acids, and amino acids, represent an important source of carbon supply in the soil, attracting up to 10<sup>10</sup> bacteria per gram of soil (Roesch et al., 2007; Badri and Vivanco, 2009). However, the correlation between root exudates and the endophytic PGPR is unclear (Fan et al., 2012). Furthermore, the plant root exudate composition may be modified according to plant-growth stages (Aira et al., 2010). Concerning this, our qPCR results did not show any significant effect on wheat-growth stages on the internal root colonization with PB2. Contrarily, the external root colonization by PB2 was significantly reduced at the FL GS compared to the 3-L GS and only in the Alixan cultivar. Interestingly, this reduction was not correlated with the protection conferred against STB, which maintained a higher or equal level to that of the Cellule cultivar and against all tested M. graminicola strains. These results indicate that the highprotection level conferred by PB2 is not proportional to the level of external colonization with PB2 but might be more correlated to internal root colonization.

#### Impact of Wheat Genotype, Growth Stage, and M. graminicola Strain on the Durability of the Resistance Induced by PB2

The PB2-IR was influenced strongly by M. graminicola strains, wheat cultivars, and growth stages. Indeed, the PB2-IR was high and significant against strain IPO323 and was not dependent on either wheat genotypes or growth stages. In the case of the strain ST38, PB2-IR was influenced by wheat cultivars over all tested growth stages, since it was very low with Cellule and very high with Alixan. However, the DNA amount in the ST38 strain in the MG modality of Cellule, without root inoculation with PB2, was already 50% less than in Alixan. In fact, the populations of M. graminicola are qualified as highly recombinant, rapidly evolved, and locally adapted to their environment. As observed previously by Selim (2009), no significant effect of wheat cultivar was observed and the population structure was stable over all the analysis dates in the same season (Selim, 2009; Selim et al., 2014). However, under field conditions, it has never been found that there is only one genotype on the wheat-leaf samples analyzed by qPCR (Selim, 2009), and approximately two to six M. graminicola genotypes were determined in the same STB disease lesion (Linde et al., 2002). These observations indicate that the final protection level might be the sum of the resistance induced against the different pathogen strains. On the same hand, the impact of the wheat-growth stage was despite the case of strains 1193 and TO256. In the 1193 strain, typical responses were observed in both cultivars, where highly significant protection at 3-L and FL GS and no protection at TI GS were recorded. In the TO256 strain, no protection at 3-L GS and high protection at TI and FL GS were observed with Alixan and the opposite situation was observed with Cellule.

However, these results, as well as the results of the field trials, showed that the resistance induced is durable and maintains its high efficiency over all wheat-growth stages and especially for the last three leaf layers, which participate directly in the grain filling. The fact that PB2 controls some strains in the earlier growth stage and not in later stages, as in Cellule against strain TO256, is still indirectly important in the control of M. graminicola by reducing the inoculum of the upper leaf layers, which is the conidia transported from the bottom leaf layers by the impact of rain splash (Selim et al., 2014).

On the other hand, while the quantitative induced resistance mediated by PB2 is normally controlled by multiple genes and is non-specific, our results showed that it is strain-dependent and influenced strongly by plant genotype and growth stage. These results are in agreement with those reported by Ryan et al. (2010), namely that Streptomyces scabies isolates potato cultivars and growing season influences the resistance induced by Streptomyces spp. They also agree with the findings of Mazzola et al. (1995), who showed variability of protection conferred in wheat against Gaeumannomyces graminis var. tritici strains as a response to Pseudomonas spp.

#### Defense Mechanisms Induced in Wheat as a Response to PB2–Wheat Genotype–M. graminicola Strain Interactions

Gene expression analysis showed that PR1, CHIT, LOX, PAL, CHS, FLAV, OXO, GST, and GPX genes were basically upregulated in Cellule cultivar compared to Alixan. These gene upregulations might explain the 50% reduction in the STB infection level in the Cellule compared to Alixan. However, gene expression results also confirmed the impact of wheat genotype-M. graminicola strain interaction on the resistance induced by

PB2. At 3-L GS, only, the Cellule cultivar showed a high resistance to the TO256 strain as a response to PB2. This resistance was correlated to a significant upregulation at 3 dai of the PR1, CHIT, and rpK genes in the PB2/MG modality compared to the MG modality. In Alixan, where no protection was observed, almost all of the upregulated genes (PR1, CHIT, TLP, LOX, PAL, CHS, FLAV, OXO, GST, and GPX), at T0 as a response to PB2, were inhibited by the TO256 strain except the PR1, GLU, and GPX genes at 1 dai and the GLU, TLP, and LOX genes at 3 dai, which were significantly upregulated in PB2/MG modality compared to the MG modality. These results do not neglect the role of PR1 as a protection gene marker of the wheat resistance to M. graminicola, as proposed previously by Adhikari et al. (2007), and highlight its possible important role in association with other defense genes. Moreover, its induction in Alixan without a protection effect against the TO256 strain may indicate that it is isolate dependent. At the same time, the protection conferred on Cellule could be related to CHIT and rpK with the possible later integration of PR1 at 3 dai.

The importance of CHIT as the key to the resistance against STB has been confirmed in the second kind of resistance induced in both cultivars (Alixan and Cellule) against the IPO323 strain where it was associated with significant upregulations of CHIT in the PB2/MG modality compared to that of the MG modality. In addition, the PR1, TLP, and POX genes were upregulated at 1 dai and the LOX gene at 3 dai in Alixan, and the PR1, FLAV, CAT, rpK, and WRKY genes at 1 dai and the PR1, GLU, TLP, WCK1, PAL, FLAV, and POX genes at 3 dai in Cellule. To obtain more information about genes that could be implicated in this PB2-MG-wheat genotype interaction, we studied their expression over a time course from 6 to 11 dai, covering the biotrophic symptomless period of the infection process of M. graminicola until the passage into the necrotrophic phase (Selim et al., 2014). However, the results of the time course have confirmed the overexpression of genes mentioned above for both cultivars, with the GLU, FLAV, AOS, and GLP genes added for Alixan.

These results show also that the protein kinases are cultivardependent as the rpK (related protein kinase) and WCK1 [Mitogen-Activated Protein kinase (MAPK) of wheat] genes were not upregulated in Alixan. On the other hand, they are isolate dependent, as rpK was upregulated in Cellule-TO256 and WCK1 with Cellule-IPO323. It was shown that the protein kinases play important roles in promoting plant-defense reactions against biotic and abiotic stresses (Zhang et al., 1998; Ahlfors et al., 2004; Mayrose et al., 2004; Reyna and Yang, 2006). Normally, the induction of ROS and MAPK are known to occur during the earlier contact with pathogens (Tavernier et al., 1995; Pugin and Guernb, 1997; Lebrun-Garcia et al., 1998) or as a response to general elicitors, pathogen-associated molecular patterns (Kroj et al., 2003) and various fungal toxins (Rasmussen et al., 2004). As observed in the current study, the later upregulation of MAPK was observed by Rudd et al. (2008) in wheat resistance against M. graminicola and explained by the response to the slow growth of the pathogen within leaf; interestingly, they showed that MAPK's activation was isolate dependent. The transcription factor WRKY is known by its implication in the plant tolerance to abiotic stress (Davletova et al., 2005) and also in defense against pathogens by regulating plant-defense genes directly or via association with MAPK (Pokholok, 2006).

The CHIT gene codes for the hydrolytic enzyme of the chitin in the pathogen cell wall, and many studies have shown its induction in resistant cultivars to M. graminicola contrarily to susceptible cultivars where it was downregulated (Somai-Jemmali et al., 2017; Ors et al., 2018). However, the association between the gene markers of the basal defenses, CHIT, GLU, TLP, and PR1 was observed previously. The PR-2 gene codes for the β-1,3 glucanase (GLU) also has an inhibition mode of action on the fungal development by degrading the β-1,3-glucan of the fungal cell wall and, in addition, has a signaling role in the elicitation of other plant-defense mechanisms (Ham et al., 1991; Wessels, 1993). The association of these pathogen-related proteins may be responsible for the synergistic effect on the plant defense activities against M. graminicola, as observed previously with TLP (PR5) (Leah et al., 1991), and the delay of the Phaeosphaeria nodorum incidence in wheat as a response to the combination of CHIT and GLU (Anand et al., 2003). In addition, GLU and TLP are induced in leaves by both SA and JA and exhibit glucanase activity, contrary to GLU alone, leading to degraded fungal plasma membrane and decreased disease incidence in potato and wheat (Vigers et al., 1992; Grenier et al., 1999; Anand et al., 2003).

Moreover, the upregulation of the PAL and FLAV genes involved in the phenylpropanoid and phytoalexin pathways, POX, CAT, and GLP, the gene markers of the ROS pathway, and AOS and LOX, the gene markers of the JA pathway, confirm our previous results concerning the association of these defense pathways with the basal-defense genes in the resistance induced in wheat against M. graminicola as a response to PB2 (Samain et al., 2017) where we discussed the possible defense mechanisms related to these pathways and especially against M. graminicola.

The upregulation of the CAT and GLP genes code for catalase and germin-like protein, respectively, indicates the plant protection against oxidative stress results in ROS pathway stimulation (Khuri et al., 2001). Catalase detoxifies H2O2, resulting in water and oxygen (Smirnoff, 1993). In addition to its role in oxygen burst detoxification, GLP also has protease inhibitor activity against M. graminicola (Segarra et al., 2003).

However, several factors may influence resistance inducer– plant–pathogen interactions, including the efficiency of the biocontrol strain, pathogen aggressiveness, host susceptibility, and environmental conditions (Francés et al., 2006; Recep et al., 2009).

## CONCLUSION

STB is the most economic important disease in wheat. The current work shows the efficiency of Paenibacillus sp. strain B2 as a biological control agent against STB, under controlled and field conditions, by inducing wheat-resistance mechanisms combined with basal defenses, ROS, phenylpropanoid and phytoalexin, SA, JA pathways, and the important role of the chitinase gene. This resistance is durable and able to protect the last wheat-leaf layers, which are the most important in yield production. Although

PB2-induced resistance depends on M. graminicola strains, wheat genotypes, and growth stages, its efficiency against STB, under field conditions, is less influenced by these factors. This may be explained by its direct impact on M. graminicola inoculum in the upper-leaf layers, which usually consist of a mixture of genotypes, or by its indirect impact on reducing the inoculum coming from the oldest leaves infected during the earlier growth stages. Interestingly, the wheat-PB2 symbiotic relationship is not energetically costly and without negative impact on the yield production. However, results under field conditions will be confirmed during, at least, two wheat growth seasons more.

### AUTHOR CONTRIBUTIONS

ES carried out the experimental work and wrote the manuscript. SS managed and supervised the project and the Ph.D. research programs, and revised the manuscript with TA.

#### FUNDING

This work was supported by SDP society and ANRT in France. We are grateful to SDP for financing this work with the support of the French "Association Nationale de la Recherche et de la Technologie (ANRT)," under convention N◦ 56/2016.

#### ACKNOWLEDGMENTS

We would like to thank the excellent contribution of the SDP Department of Research, Innovation and Technics, and especially Cédric Ernenwein, Amandine Hahn, and Franck Vasseur. Special thanks to Emilie Parmentier and Selin Altintas for their excellent assistance.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.00587/ full#supplementary-material

FIGURE S1 | The experimental design of this work covered the study of: (1) the impact of wheat genotypes and growth stage on the colonization of roots by PB2, (2) the impact of wheat genotypes on the resistance induced by PB2 against STB, (3) the impact of wheat-genotype-growth-stage–M. graminicola strain interactions on durability of the resistance induced by PB2, (4) gene expression analysis of PB2-wheat-genotype-M. graminicola strain interaction, and (5) to confirm the PB2-resistance induced against M. graminicola under field conditions. Wheat grains' inoculation with PB2 was at sawing. Under controlled conditions, wheat leaf infection with M. graminicola was realized using 10<sup>6</sup> spores/leaf at 3-leaf (3-L),

### REFERENCES

Adhikari, T. B., Balaji, B., Breeden, J., and Goodwin, S. B. (2007). Resistance of wheat to Mycosphaerella graminicola involves early and late peaks of gene expression. Physiol. Mol. Plant Pathol. 71, 55–68. doi: 10.1016/j.pmpp.2007.10.004

tillering (Ti), or flag-leaf (FL) growth stage (GS). Leaf infection level was determined using quantitative real-time PCR (qPCR), at 17 days after infection. Under field conditions, leaf infection was by the natural inoculum and disease level in the third leaf under the FL was quantified using qPCR at GS 49. The highly virulent strains of M. graminicola, IPO323, TO256, 1193, and ST38 were used and at least two wheat cultivars with different resistant level to M. graminicola.

FIGURE S2 | PCR's melting curve for each primer pair used in the gene expression study. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH), β-tubulin (B-TUB), pathogenesis-related protein (PR1), Chitinase (CHIT), β-1,3-glucanase (GLU), thaumatin-like protein (TLP), lipase (LIP), lipoxygenase (LOX), allene oxide synthase (AOS), phenylalanine ammonia-lyase (PAL), chalcone synthases (CHS), and flavonoid 7-O-methyltransferase-like (FLAV).

FIGURE S3 | PCR's melting curve for each primer pair used in the gene expression study and in the quantification of Paenibacillus strain B2. Peroxidase (POX), oxalate oxidase (OXO), glutathione-s-transferase (GST), germin-like-protein (GLP), glutathione peroxidase (GPX), catalase (CAT), superoxide dismutase (SOD), related protein kinase (rpK), WRKY1 transcription factor (WRKY), MAP kinase (WCK1), and Paenibacillus strain B2 16S ribosomal DNA (16S rDNA).

FIGURE S4 | PCR's amplification efficiency (E), for each primer pair used in the gene expression study, is deducted from the slopes (S) of the standard curves based on E = 100<sup>∗</sup> (10−1/ s −1 ). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH), β-tubulin (B-TUB), pathogenesis-related protein (PR1), Chitinase (CHIT), β-1,3-glucanase (GLU), thaumatin-like protein (TLP), lipase (LIP), lipoxygenase (LOX), allene oxide synthase (AOS), phenylalanine ammonia-lyase (PAL), chalcone synthases (CHS), and flavonoid 7-O-methyltransferase-like (FLAV).

FIGURE S5 | PCR's amplification efficiency (E), for each primer pair used in the gene expression study and in the quantification of Paenibacillus strain B2, is deducted from the slopes (S) of the standard curves based on E = 100<sup>∗</sup> (10−1/ s −1 ). Peroxidase (POX), oxalate oxidase (OXO), glutathione-s-transferase (GST), germin-like-protein (GLP), glutathione peroxidase (GPX), catalase (CAT), superoxide dismutase (SOD), related protein kinase (rpK), WRKY1 transcription factor (WRKY), MAP kinase (WCK1), and Paenibacillus strain B2 16S ribosomal DNA (PB2 16S rDNA).

FIGURE S6 | Field trials grain yield production the two cultivars, Expert and Chevron, as a response to wheat grains' inoculation with Paenibacillus sp. strain B2, Cherokee <sup>R</sup> fungicide application, in recommended dose (RD) and half the recommended dose (HD), an association between PB2 and Cherokee <sup>R</sup> in HD (PB2+HD), and in PB2-non-inoculated and fungicide-non-treated controls (C−). The values shown are the means of one biological replicates and five technical replicates. Bars indicate means ± standard deviations. Different lower-case letters indicate significant differences between treatments, according to ANOVA followed by Tukey's post hoc test (α = 0.05).

TABLE S1 | Gene expression ratio of some wheat-defense-related genes encoding proteins from different classes, estimated by real-time PCR.

TABLE S2 | Gene expression ratio in the susceptible cultivar Alixan as a response to Paenibacillus B2 (PB2), M. graminicola strain IPO323 (MG) and Paenibacillus B2 and M. graminicola strain IPO323 (PB2/MG), at the time of infection with IPO323 (T0), 6, 12, 24, and 48 h after inoculation (hai), 3, 5, 9, and 11 days after inoculation (dai).

TABLE S3 | Gene expression ratio in the moderate cultivar (Cellule) as a response to Paenibacillus B2 (PB2), M. graminicola strain IPO323 (MG) and Paenibacillus B2 and M. graminicola strain IPO323 (PB2/MG), at the time of infection with IPO323 (T0), 6, 12, 24, and 48 h after inoculation (hai), 3, 5, 9, and 11 days after inoculation (dai).

Ahlfors, R., Macioszek, V., Rudd, J., Brosché, M., Schlichting, R., Scheel, D., et al. (2004). Stress hormone-independent activation and nuclear translocation of mitogen-activated protein kinases in Arabidopsis thaliana during ozone exposure: ozone activation of MAP kinases in Arabidopsis. Plant J. 40, 512–522. doi: 10.1111/j.1365-313X.2004.02 229.x



gene expression: a possible connection between biotic and abiotic stress responses. Mol. Plant. Microbe Interact. 12, 951–959. doi: 10.1094/MPMI.1999. 12.11.951


**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 Samain, Aussenac and Selim. 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.

# Deciphering Microbiome Related to Rusty Roots of Panax ginseng and Evaluation of Antagonists Against Pathogenic Ilyonectria

Defei Liu1,2, Huanjun Sun<sup>1</sup> and Hongwu Ma<sup>1</sup> \*

<sup>1</sup> Key Laboratory of Systems Microbial Biotechnology, Tianjin Institute of Industrial Biotechnology, Chinese Academy of Sciences, Tianjin, China, <sup>2</sup> University of Chinese Academy of Sciences, Beijing, China

Plant roots host diverse microbes that are closely associated with root fitness. Currently, the relationship between microbes and rusty roots of Panax ginseng remains unclear. Here, we described the root-associated microbiome in rusty and healthy ginseng by metagenomic sequencing of 16S rRNA and ITS regions. Being enriched in Diseasedroots (Dr) of ginseng and their rhizosphere soil, the fungus of Ilyonectria, was identified as the most probable cause of the disease after ITS analysis. Meanwhile, an increase of Mortierella was observed in Healthy-roots (Hr). Surprisingly, an enriched Fusarium was found in both Hr and their rhizosphere soil. Besides, in comparison with Hr, decreased relative abundance of Actinomycetales and increased relative abundance of Pseudomonadales was observed in Dr after 16S rRNA analysis. What's more, we isolated several microorganisms as antagonists that showed strong inhibiting effects on Ilyonectria in plate assays. In field trials, inoculation of Bacillus sp. S-11 displayed apparent suppression effect against Ilyonectria and shifted microbial communities in rhizosphere soil. Our research identified key microbiota involved in rusty roots of P. ginseng and offered potential biocontrol solutions to rusty disease.

Keywords: biocontrol, Ilyonectria, Panax ginseng, rusty roots, microbiome

### INTRODUCTION

Terrestrial plants are colonized by diverse microorganisms which include commensals, symbionts, and opportunistic pathogens (Muller et al., 2016). The colonized microbes present in endosphere and rhizosphere have a direct relationship with the fitness of plant roots. With recent advances in metagenomic sequencing and computational analysis, it is now possible to unravel rootassociated microbiome (Lebeis et al., 2015; Castrillo et al., 2017). A previous study elucidated the microbes in disease-suppressive soils against pathogen in sugar beet (Mendes et al., 2011). They found several key bacterial taxa and genes involved in suppression of Rhizoctonia solani. Besides, a recent study revealed that Flavobacteriia was enriched in tomato rhizosphere, which was resistant to the soil-borne pathogen Ralstonia solanacearum. Pot trials showed that the inoculation of Flavobacteria had a beneficial role in defending R. solanacearum (Kwak et al., 2018). Therefore, metagenomics is a powerful tool for revealing the microbiota related to pathogentriggered root diseases, offering possible solutions to preventing pathogen-infection in plants.

#### Edited by:

Donald L. Hopkins, University of Florida, United States

#### Reviewed by:

Muhammad Saleem, Alabama State University, United States Frédérique Reverchon, Instituto de Ecología (INECOL), Mexico

> \*Correspondence: Hongwu Ma ma\_hw@tib.cas.cn

#### Specialty section:

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

Received: 11 March 2019 Accepted: 31 May 2019 Published: 18 June 2019

#### Citation:

Liu D, Sun H and Ma H (2019) Deciphering Microbiome Related to Rusty Roots of Panax ginseng and Evaluation of Antagonists Against Pathogenic Ilyonectria. Front. Microbiol. 10:1350. doi: 10.3389/fmicb.2019.01350

Panax ginseng Meyer (Araliaceae) has been an important herb of traditional medicine for thousands of years in East Asia. Mainly cultivated in China, Korea and Japan (Yun, 2001), P. ginseng is considered as an important cash crop due to its pharmaceutical properties contributed by ginseng saponins (Attele et al., 1999; Shin et al., 2015). As a kind of herbaceous perennial plant, it takes at least 5–6 years for ginseng to reach a marketable size. For those cultivated in mountain forest, more than 10 years is needed to grow to a good shape.

During continuous cultivation, there is a greater chance for ginseng to suffer from soil-borne diseases that lead to dramatic yield losses and quality decline. Major root diseases include root rots and rusty roots (Farh et al., 2018). The latter disease, also known as rusted roots or "the rust," is characterized by small or large reddish-brown spots on the surface of ginseng roots (**Supplementary Figure S1**). It is widely reported in China, South Korea and Canada as a major cause of ginseng root deterioration at all stages (Hildebrand, 1934; Lee et al., 2011; Lu et al., 2015). Given the fact that ginseng roots are commercially graded according to their sizes, shapes and overall appearances, rusty root disease severely limits the output and quality of ginseng worldwide.

Presently, the causes of P. ginseng rusty roots are still controversial. It is mentioned in some studies that there was a close relationship between rusty roots and rhizosphere soil properties (Liu et al., 2014). Besides, microorganisms also play a key role in this disease as suggested in other studies that ginseng roots could be infected by putative pathogens, including both bacteria (Choi et al., 2005; Lee et al., 2011) and fungi (Rahman and Punja, 2005; Reeleder et al., 2006; Lu et al., 2015, 2019), and showed rusty symptom after inoculation of the isolates. However, there is no systematic research on the microbiome involved in the disease. Therefore, it is necessary to study the links between microbiome and rusty roots with metagenomic sequencing.

We used 16S rRNA and ITS sequencing to identify endophytic and rhizospheric microbiome involved in rusty roots of P. ginseng. After that, microbiota diversity and structure of healthy and rusty ginseng roots were evaluated. Moreover, pureculture was used to isolate antagonists and their biocontrol abilities were assessed in pot and field trials. We also investigated the changes of microbial community in rhizosphere soil after inoculation of antagonists by metagenomic sequencing in field trials.

#### MATERIALS AND METHODS

#### Sample Processing

Ginseng roots were sampled from Tonghua and Baishan, Jilin Province, China in August 2016 and July 2017. Twelve sites (2 m × 2 m) with different rusty root indices were selected, and 5–7 ginseng roots randomly collected from each site. Root samples were kept in a sterile plastic bag and transported to the laboratory in cold chain. The detail information of sampling sites was described in **Supplementary Table S1**.

Ginseng organs above the ground were aseptically removed and loose soil was physically removed until only soil adhered to the root surface remained. After that, loose soil was airdried and crushed to pass through 2 mm nylon sieve to test the chemical properties. Soil pH was determined with a digital pHmeter (FE20-K, Mettler-Toledo, Switzerland) in a suspension of 1:2.5 soil/water ratio (w/v). Soil organic matter (SOM), available nitrogen (N), available phosphorus (P), and available potassium (K) were determined by using commercial chemical assay kits (Suzhou Comin Biotechnology Co., Ltd., China), respectively, according to the instructions. Soil chemical properties are shown in **Supplementary Table S2**.

The soil adhered to the root surface was processed according to the method of Lebeis et al. (2015). The roots were placed in a clean and sterile 50 mL conical tube containing 25 mL of phosphate buffer (6.33 g of NaH2PO4·H2O, 16.5 g of Na2HPO4·H2O, and 200 µL Silwet L-77 in 1 L of water). Rhizosphere soils were separated from the roots by vortexing the root system in buffer at the maximal speed (3000 rpm) for approximately 15 s. The turbid obtained was filtered through a sterile 100 µm nylon mesh cell strainer (Biologix Group Limited) into another sterile 50 mL conical tube to remove plant materials, sand, and other large debris. The filtrate was centrifuged in two steps to form a tight pellet and defined as rhizosphere soil sample. The rhizosphere soils from the same sampling site were mixed as one sample.

The rusty root indices of soil were calculated as followed:

$$\text{Rusty root index} = \frac{\Sigma \left( n \times \text{rusty root grade} \right)}{N \times \text{ the highestrusty root grade}} \tag{1}$$

where n equals to the number of ginsengs with different rusty root grades, N equals to the number of total ginsengs. Rusty-root grades range from 0 to 4, with 0 standing for healthy ginseng roots without rust, 1 for rust areas <10%, 2 for those covering 10–25%, 3 for those covering 25–50%, and 4 for those >50%. According to the rusty root index, we sorted out soil samples with index ≤0.5 as the Healthy-soil (Hs) group. While rest of the samples were classified as the Diseased-soil (Ds) group.

To obtain endophytic microbial communities, 3–5 roots from each site were selected according to the symptoms on the surface of ginseng roots. Roots were subsequently placed in new sterile phosphate buffer for sonication to remove soil and microbial aggregates left on the root surface by using an ultrasonic cleaner set on low frequency for 5 min (30 s bursts followed by 30 s rests). After that roots were washed again with sterile water. Then, Healthy-root (Hr) samples and Diseased-root (Dr) samples were excised (about 1–2 cm long by 0.5–1 cm wide and about 1 mm deep) from main-roots of ginseng with a sterile scalpel. For healthy roots without rusty spots and early lesion roots with small rusty parts, one piece of Hr sample was excised per root from asymptomatic parts. For those rusty ones, one Dr sample was taken per root from rusty parts. The details of samples were described in **Supplementary Table S3**.

Soil and root samples were flash frozen and stored at −80◦C until DNA extraction with the Qiagen DNeasy PowerSoil Kit (Qiagen, Germany) following the manufacture's protocol. For the root samples, we performed a pre-homogenization step using liquid nitrogen grinding.

### Pyrosequencing of the 16S rRNA and ITS Genes

The libraries were constructed and sequenced based on Illumina Miseq PE300 platform (Illumina, United States) according to the manufacturer's protocols. Three sets of primers were used to amplify different regions of genes. For soil samples, the amplification of V3-V4 region of the 16S rRNA gene was achieved by using 338F and 806R. As for root samples, we amplified V5-V7 region of the 16S rRNA gene with 799F and 1193R, which displayed very low amplification of non-target DNA, such as plastid (mostly chloroplast) DNA and mitochondrial DNA (Beckers et al., 2016). While ITS1 and ITS2 were used concerning ITS sequence for both soil and root samples. The primer sequences were listed in **Supplementary Table S4** and sequence data were deposited in GenBank (accession number: PRJNA512054).

#### Bioinformatics Analysis of 16S rRNA and ITS Gene Sequences and Data Statistic Analysis

The 16S rRNA and ITS gene sequences generated were processed through the open-source software pipeline QIIME2 (Version 2018.4) (Caporaso et al., 2010). Quality control was performed using DADA2 plug-in for quality filtering, chimera removal and feature table generation (Callahan et al., 2016). As for data sequenced at different time, we used "qiime feature-table merge" to combine the results. Raw data sequence and quality filtering parameters were shown in **Supplementary Table S5**. The command of training feature "qiime feature-classifier" using Greengenes (McDonald et al., 2012) and UNITE (QIIME release, Version 01.12.2017<sup>1</sup> ) as reference was utilized to classify representative sequences from our datasets. Features of 16S rRNA which were assigned as "Chloroplast" and those of ITS with no taxonomic assignment at kingdom level (27.9% for root samples and 0.3% for soil samples) were subsequently removed from the datasets.

Alpha-diversity, represented by Shannon index and Pielou's evenness, was calculated using QIIME2 pipeline based on the feature table. Before the calculation of alpha-diversity, samples were rarefied to the same sequence depth (**Supplementary Table S6**), which also applied to beta-diversity calculation. Principal coordinate analysis (PCoA) was performed on the basis of Bray-Curtis distance matrix calculated in QIIME2 and visualized by EMPeror (Vázquez-Baeza et al., 2013). Permutational multivariate analysis of variance (PERMANOVA) (Anderson, 2001) with 999 permutations was used to detect statistical significances in QIIME2.

In this work, we presented linear discriminant analysis (LDA) effect size (LEfSe) to perform differentially abundant taxa between healthy and rusty groups (Segata et al., 2011). Redundancy analysis (RDA) was performed via Canoco for Windows 5 (Microcomputer Power, NY, United States).

Alpha-diversity indices and relative abundance of abundant taxa were tested with methods described below. Those of root and soil samples were tested by Student's t-test using R (version 3.5.0), while those of biocontrol treatments were tested using one-way ANOVA followed by LSD post hoc test by IBM SPSS Statistics 19 (SPSS Inc.).

#### Isolation of Microorganisms

Rusty root tissues were used in attempts to isolate causal agents. To begin with, roots were washed again with sterile water, and sections of rusty tissues were cut off. Root pieces were surfacesterilized by being first immersed in 1% sodium hypochlorite for 30 s, then in 70% ethanol for 30 s and finally rinsed in sterile water for three times. After that, disinfected periderm tissues (about 1 mm deep) were torn off with a pair of sterilized forceps and cut into small pieces (0.5 mm × 0.5 mm). Then the pieces were placed on a plate with potato dextrose agar (PDA) (6 g potato dextrose, 20 g glucose, 18 g agar, and 1 L distilled water) amended with 50 µg/mL tetracycline. Then, inoculated PDA plates were kept in darkness at 22 ± 1 ◦C. After 7 to 14 days, fungal hyphae appeared. Pieces of hyphae were excised with an inoculating needle, placed on a new PDA plate and incubated at 22 ± 1 ◦C.

After that, we isolated antagonists against fungal pathogens from ginseng root, Hs and uncultivated soil. First, asymptomatic sections from healthy roots or regions of apparently healthy tissue on diseased roots were conducted as described above. Three to five disinfected periderm pieces were placed separately on nutrient agar (NA) plates (3 g beef extract, 5 g peptone, 5 g NaCl, 18 g agar, 1 L distilled water, and pH 7) for bacteria isolation, Gauze's No. 1 agar plates (20 g soluble starch, 1 g KNO3, 0.5 g K2HPO4, 0.5 g MgSO4·7H2O, 0.5 g NaCl, 0.01 g FeSO4·7H2O, 18 g agar, 1 L distilled water, and pH = 7.4 – 7.6) for actinomycete isolation and PDA plates for fungi. Apart from antagonists from root, soil microbes that might inhibit pathogens were also isolated. Serial dilutions were prepared up to 10−<sup>4</sup> using sterile phosphate-buffered saline (PBS). Next, 100 µL of each diluted sample was spread onto plates with NA, Gauze's No. 1 agar medium and PDA, respectively. Plates with NA and Gauze's No. 1 agar were incubated at 30 ± 1 ◦C for 7 days. While plates with PDA were incubated at 22 ± 1 ◦C for 7 days. Bacteria and actinomycete colonies were transferred to new NA and Gauze's No. 1 plate and purified by streak plate.

Identification of the purified isolates was achieved using multilocus sequence analysis. Firstly, Total genomic DNA was extracted from bacteria colonies or fungi fresh mycelia by using bacterial genomic DNA extraction kit (BioTeke Corporation, China, Cat#: D3350-01) and rapid fungi genomic DNA isolation Kit (Sangon Biotech, China, Cat#: B518229-0100), respectively. Afterward, PCR was performed in 50 µL reaction mixture containing 1 µL of DNA, 2 µL forward primer (10 mM), 2 µL reverse primer (10 mM), 25 µL 2 × Taq Plus MasterMix (Dye) (CWbio Corporation, China, Cat#: CW2849M), and 20 µL sterile PCR-grade water. Bacterial 16S rRNA genes were amplified by PCR with the primer pair of 27F and 1492R (Lane, 1991) for an initial denaturation at 94◦C for 2 min, following by 30 cycles of 94◦C for 30 s, 56◦C for 1 min, 72◦C for 1 min, and a final elongation step of 72◦C for 10 min. As for GyrB gene of Bacillus and Streptomyces, primers UP1 and UP2R (Yamamoto and Harayama, 1995) and gyrBPF and gyrBPR (Guo et al., 2008)

<sup>1</sup>https://unite.ut.ee

were used, respectively. The amplification was performed under the amplification conditions described in references. In the case of ITS gene of fungi, we used primers ITS5 and ITS4 (White et al., 1990) with the following modifications to the amplification protocol: 35 cycles, annealing at 52◦C for 30 s. We also amplified intergenic spacer (IGS) sequences of fungi by PCR using the primers LR12R and invSR1R (**Supplementary Table S4**) with modifications to protocol: 35 cycles, annealing at 52◦C for 1 min. Then the PCR amplified products were separated by agarose gel electrophoresis and sequenced at Genewiz (Tianjin, China). Bacteria were identified on the basis of similarities to 16S rRNA and gyrB sequences while fungi identification according to ITS and IGS sequences with their morphological characteristics. The sequences of the isolates were searched in NCBI<sup>2</sup> and have been deposited in GenBank under the accession MK459368, MK459369, MK512096, MK512097, MK459418-MK459425, and MK512088-MK512095.

#### Pathogenicity Test

To confirm the pathogenicity of isolated pathogens, a test was carried out via bare-root inoculation as described by Reeleder et al. (2006) with some modifications. Healthy ginseng roots (3-year-old) were washed with tap water and root surfaces were sterilized as described. Ilyonectria robusta (I. robusta) 4D-1 (4D-1) and I. mors-panacis TH5 (TH5) were randomly selected from pathogen isolates obtained from rusty ginseng roots. 4D-1 and TH5 were grown on PDA for 14 days at 22 ± 1 ◦C. The endoconidia were washed and adjusted to 1 × 10<sup>7</sup> endoconidia/mL. Sterilized filter paper strips (about 1 cm wide by 8 cm long) were immersed in endoconidia suspension for 30 s. Meanwhile, paper strips immersed in sterile water for 30 s were used as control. Then, treated paper strips were placed on the top and middle parts of the roots (4 replicates for each treatment). The inoculated parts in the middle were wounded with a sterile scalpel (about 1 cm long and 1 mm deep). Then, every inoculated root was wrapped with four layers sterile moistened gauze and placed in a folded plastic bag. Bags were incubated at 22 ± 1 ◦C in dark for 3 weeks. After that, pathogens were recovered from symptomatic roots and confirmed by analyzing the ITS sequence.

### In vitro Inhibition of Ilyonectria

Microorganisms isolated from both soil and healthy ginseng roots were evaluated for their activities against Ilyonectria. In order to do that, PDA disks (5-mm diameter) containing 14-day-old 4D-1 mycelia were placed at the center of PDA plates. While isolates with possible antagonism were placed 2.5 cm away from them (3 replicates per isolate). Three weeks after inoculation, putative suppressive isolates were selected based on the width of the inhibition zone or inhibition rate. The activities of selected isolates against TH5 were also evaluated.

## In vivo Evaluation of Antagonistic Microbes

The pathogen 4D-1 was grown in potato dextrose broth (PDB) for 7 d at 22 ± 1 ◦C, while three kinds of antagonisms were

<sup>2</sup>https://www.ncbi.nlm.nih.gov/

cultivated, respectively. For Bacillus sp. S-11 (S-11), it was cultured in nutrient broth (NB) for 36 h at 37 ± 1 ◦C. Besides, Streptomyces sp. S6-31 (S6-31) was cultured in Gauze's No. 1 broth for 5 days at 30 ± 1 ◦C. As for Trichoderma koningiopsis S7-1 (S7-1), it was grown in PDB for 7 days at 25 ± 1 ◦C. After that, concentrations of 4D-1 and S7-1 were observed with a hemocytometer, while concentrations of S-11 and S6-31 were determined by flat colony counting. Concentrations of all isolates were adjusted before use.

Biocontrol activity was determined in pot experiments with five treatments: CK: sterilized water; TI: 4D-1 (1 × 10<sup>5</sup> endoconidia/g of dry soil); TB: S-11 (1 × 10<sup>6</sup> CFU/g of dry soil) with 4D-1 (1 × 10<sup>5</sup> endoconidia/g of dry soil); TS: S6-31 (1 × 10<sup>5</sup> CFU/g of dry soil) with 4D-1 (1 × 10<sup>5</sup> endoconidia/g of dry soil); and TT: S7-1 (1 × 10<sup>5</sup> endoconidia/g of dry soil) with 4D-1 (1 × 10<sup>5</sup> endoconidia/g of dry soil). All treatments were arranged according to the principle of randomized complete blocks. Each treatment contained five pots with each pot serving as a replicate. Every pot was planted with 4 1-year- old ginsengs and placed in a greenhouse at 22 ± 2 ◦C with 14 h of sunlight (without direct sunlight) and 10 h darkness. Soil water content ranging from 30 to 50%. During our experiments, root colonization of inocula was assessed at 0, 10, 20, 30, and 40 days after inoculation. Every time, three roots were sampled randomly in each treatment for the extraction of genomic genes of rhizosphere soil as described above for qPCR analysis. Forty days later, incidence and severity of the disease was evaluated.

Biocontrol studies were also conducted in field with 4-yearold ginsengs severely impacted by rusty disease in Jilin province. During our trial, four treatments were applied by root watering: CK: sterilized water; TB: S-11 (1 × 10<sup>6</sup> CFU/mL); TS: S6-31 (1 × 10<sup>5</sup> CFU/mL); and TT: S7-1 (1 × 10<sup>5</sup> endoconidia/mL). Each treatment consisted of three replicates arranged in a randomized complete block design. Each block consists of about 50 plants. Five roots were sampled randomly in each block and mixed as one sample at 40 days after inoculation. Then root colonization of isolates and biocontrol activities against Ilyonectria were evaluated by qPCR. The effects of inoculum on soil microbiome were also evaluated by pyrosequencing of 16S rRNA gene and ITS gene.

### Real-Time qPCR

A quantitative real-time PCR (qPCR) assay based on standard curves that used C<sup>t</sup> value was developed for the detection and quantification of I. robusta, I. mors-panacis, S-11, S6-31, S7-1, and genus of Ilyonectria. In order to develop a standard curve for target genes, we first amplified target fragments cloned in PMD-19 Simple T vector (Takara, Japan) with primers listed in **Supplementary Table S4**. Then, concentration of linear recombinant plasmids was determined by NanoDrop 2000 spectrophotometer (Thermo Scientific, MA, United States).

All PCRs were performed by Faststart Universal SYBR Green Master (ROX) kit (Roche, Switzerland) in LightCycler 96 System (Roche Diagnostics) with reaction volume being 25 µL. Each reaction contains 1-fold FastStart Universal SYBR green Master, PCR-grade water, 150 nM forward and reverse primer, and 2.5 µL soil DNA extract. After preparation, PCRs were operated as

followed: denaturation for 10 min at 95◦C followed by 40 cycles of denaturation for 15 s at 95◦C, annealing and extension at 60◦C for 60 s. Then melting curve analysis at the temperature ranging from 60◦C to 95◦C at the rate of 0.2◦C/s was conducted after amplification.

## RESULTS

#### Metagenomic Sequence Analysis

When preparing for sequencing, root samples with low DNA concentrations or qualities after PCR were discarded. After quality filtering, 766 561 high-quality reads of total 16S rRNA V5- V7 sequences from 31 root samples were obtained. As for fungi sequencing, 736 894 high-quality ITS reads from 36 root samples were acquired. We also sequenced bacterial 16S rRNA V3-V4 and fungal ITS sequences of 12 ginseng rhizosphere soil samples, thus obtaining 260 541 and 406 810 high-quality reads, respectively. Concerning soil samples treated with inoculation in field trials, we obtained a total of 227 322 and 358 510 high-quality sequences for 16S rDNA V3-V4 and ITS sequences, respectively.

## Microbiota Differences Between Healthy and Rusty Ginseng Roots

To elucidate the distribution and assembly patterns of microbial communities related to rusty ginseng roots, a high-throughput sequencing approach was used. Compared with the Hr group, Dr had a significantly decreased bacterial Shannon diversity (**Figure 1A**), and no significant difference was detected in evenness (pielou\_e) index (**Figure 1B**), which meant that Dr had a lower species richness. The dissimilarity between samples was explored using PCoA and the result revealed that the bacterial compositions of the two groups distinct with each other (**Figure 1C**, F = 2.185, p = 0.002). After sequence annotation, we found that Pseudomonadales increased from 1.03 to 5.15% (t-test, p = 0.02) (**Figure 1D**) among the identified orders in Dr

group, whereas Actinomycetales decreased from 23.5 to 10.0% (ttest, p = 3.402 × 10−<sup>4</sup> ). Moreover, on order level, LDA effect size (LEfSe) indicated that Pseudomonadales enriched in Dr while Actinomycetales enriched in Hr (**Figures 1E,F**).

As for fungal community diversity in root samples, a marked drop in fungal alpha-diversity was observed in Dr group (**Figures 2A,B**), which revealed Dr group had decreased species richness and evenness. Meanwhile, PCoA result demonstrated dramatic discrepancies in the structure of fungal community (**Figure 2C**, F = 2.227, p = 0.005) between groups of Hr and Dr. After assignment, we found that the fungal microbiome showed severe dysbiosis in rusty roots. On genus level, some genera of fungi predominated (30–90%) in these samples. The predominated fungi were mainly (71%) assigned as the genus of Ilyonectria. Compared with Hr, obviously higher relative abundance of Ilyonectria was observed (t-test, p = 5.924 × 10−<sup>4</sup> ) (**Figure 2D**). LEfSe results showed that the genera of Ilyonectria and Basidiomycota gathered (LDA > 4) in Dr while the fungal genera enriched (LDA > 4) in Hr were Mortierella and Fusarium (**Figures 2E,F**).

#### Microbiota Differences Between Rhizosphere Soil of Healthy and Rusty Ginseng Roots

We also studied the diversity of microbiome in ginseng rhizosphere soil (**Supplementary Figure S2**). After analysis, no significant differences were detected in Shannon diversity or evenness (pielou\_e) between Hs and Ds (**Supplementary Figures S2A,B,E,F**). For bacteria, LEfSe results revealed that Ellin309 and Xanthomonadales enriched in Ds while Alteromonadales enriched in Hs on the level of order (LDA > 2, **Supplementary Figure S2D**). On the other side, ITS sequencing

results showed that Ilyonectria took up 0.02% of the relative abundance in Hs on genera level. While that percentage in Ds was 15.1%, presenting a significant difference (t-test, p = 2.145 × 10−<sup>3</sup> ) when compared with Hs. This result was also confirmed by LEfSe (LDA > 2) (**Supplementary Figure S2H**). Besides, an enrichment of Fusarium was observed in Hs.

After comparison, we found that Ilyonectria, as a genus of fungi colonizing in ginseng roots, was the most abundant fungus in Dr and enriched in Dr and Ds as revealed by LEfSe analysis. In the following experiments, we isolated I. robusta and I. morspanacis from rusty roots and the pathogenicity was confirmed by bare-root inoculation (**Supplementary Figure S3**). In order to figure out the dominant specie, the qPCR approach was used to detect copies of I. robusta and I. mors-panacis in rusty ginseng roots (relative abundance of Ilyonectria >29%) and their rhizosphere soil (**Figure 3**). The results showed that both I. robusta and I. mors-panacis were found in 12 rusty samples and three soil samples. While in the remaining samples, either I. robusta or I. mors-panacis was observed. Thus, we speculated that both the two species contributed to rusty disease. Based on these results, we conclude that the infection of Ilyonectria from rhizosphere soil is very likely to be the predominant cause of the rusty disease.

#### Relationship Between Soil Physicochemical Properties and Microbial Abundances

Soil characteristics have a great influence on soil microbial communities. To clarify the relationships between microbiota and soil properties, we used RDA with SOM, N, P, K, and pH being environmental variables (**Figure 4**). The most 10 best fitting genera were shown (relative abundance >0.001). The total explanatory variables account for 57.8 and 71.9% of bacterial and fungal variations, respectively.

Soil pH was considered as a dominant factor of soil bacterial communities (Girvan et al., 2003). According to our findings, pH plays a crucial part in shaping the microbiota composition of bacteria orders (Monte-Carlo test, p < 0.05). Also, the bacterial community was strongly affected by SOM (Monte-Carlo test, p < 0.05) (**Figure 4A**). Regarding to fungi, the result showed that fungal community was obviously impacted by P (Monte-Carlo test, p < 0.05), N (Monte-Carlo test, p < 0.05), SOM (Monte-Carlo test, p < 0.05), and pH (Monte-Carlo test, p < 0.05). Among fungal genera, Ilyonectria was positively related with SOM, N and pH, but negatively related to P (**Figure 4B**).

#### Biocontrol Activity Evaluation of Antagonists and Root Colonization

In order to find a biocontrol strategy against Ilyonectria, pureculture was used to isolate pathogens and antagonists. Four I. robusta and 3 I. mors-panacis from rusty roots were obtained as possible pathogens, while 205 bacteria and 82 fungi were isolated from ginseng roots and soils as potential antagonists for biocontrol activity experiments. Among all isolates, 11 bacteria and 12 fungi showed antagonistic activities and three of them showed strong inhibiting effects on 4D-1 and TH5 on PDA plates (**Supplementary Figure S4**), namely Bacillus sp. S-11 (S-11), Streptomyces sp. S6-31 (S6-31) and T. koningiopsis S7-1 (S7-1).

To investigate the colonization of antagonists and their suppression of Ilyonectria in rhizosphere soil under pot experiments, we tracked population dynamics of antagonistic microbes and 4D-1 by qPCR. The results (**Supplementary Figure S5**) showed that S-11 and S7-1 increased on day 10 and a population density above 3.86 × 10<sup>3</sup> and 1.50 × 10<sup>4</sup> (copies/g soil) was maintained until day 40, respectively, while S6-31 declined all the way along. Meanwhile, all antagonists showed antifungal activities in the soil with 4D-1 being significantly suppressed on day 30 or 40 (**Figure 5A**). Besides, we counted rusty spots on ginseng root skin on day 40 as the indicator of rusty disease and the result showed that treatment TT had the least rusty spots (**Figure 5B**).

To study the role of antagonists in area suffering rusty disease, biocontrol studies were conducted in a field severely impacted

by rusty disease. Results showed that the numbers of S-11, S6- 31, and S7-1 in rhizosphere soil were 4.47 × 10<sup>4</sup> (copies/g soil), 4.19 × 10<sup>4</sup> (copies/g soil), 4.89 × 10<sup>4</sup> (copies/g soil) on day 40, respectively. The inoculation of antagonists inhibited Ilyonectria in soil, especially in treatments TB and TS (**Figure 6**).

#### Changes of Microbial Community Structure in Response to Antagonists Inoculation

To further assess the effect on soil microbiota after inoculation of antagonists in field trials, we used the same sequencing strategy described above for soil samples. The results exhibited that the treatment TB showed the highest bacterial diversity among all treatments (**Figures 7A,B**). For PCoA results, the inoculations became significant factors in shaping composition and structure of bacterial (F = 1.616,

p = 0.001) and fungal (F = 1.641, p = 0.001) community in ginseng rhizosphere soil (**Supplementary Figures S6A,B**). After one-way ANOVA analysis, the result showed that all inoculated treatments had lower relative abundance of Ilyonectria than CK, while that of TB was the lowest (p < 0.05) among all treatments. Other statistically significant differences in bacterial order and fungal genera were shown in **Supplementary Table S7**.

#### DISCUSSION

Plants harbor a wide variety of microorganisms in both endosphere and rhizosphere. These microorganisms, including bacteria and fungi, have a strong impact on the fitness of plants (Vandenkoornhuyse et al., 2015). Most of previous studies about ginseng rusty root were based on culture-dependent methods (Reeleder and Brammall, 1994; Reeleder et al., 2006; Lu et al., 2014, 2015). However, the relationship between

microbial community of ginseng and rusty disorder is still unclear. Here, we systematically researched the microbiome related to rusty roots of P. ginseng by metagenomic sequencing. Preliminary understanding of microbial compositions about ginseng roots were obtained.

To get insights into the microbes in rusty ginseng roots, we compared the bacteria and fungi communities between healthy and rusty periderm tissues. Our results revealed that the fungal microbiome showed severe imbalance in rusty tissues that mainly attribute to the abundance of Ilyonectria. Ilyonectria, a species complex, is commonly associated with root-rot diseases of a wide range of hosts (Cabral et al., 2011). Among the species complex, I. mors-panacis was genetically distinct from the other isolates and was reported as an aggressive pathogen to cause rot ginseng roots, while I. robusta was clustered into a different species with weak aggressiveness (Seifert et al., 2003; Cabral et al., 2011). Pathogenicity of Ilyonectria was also reported in many studies (Rahman and Punja, 2005; Lu et al., 2015, 2019).

The abundance of Ilyonectria found in endosphere showed that this genus was highly specialized to its ecological niche. Once the infection occurred, possible pathogen effectors produced by Ilyonectria may provide a fitness benefit to the pathogen during host colonization (Guttman et al., 2014). The invaded pathogen outcompeted other fungal genera, such as Mortierella, Fusarium, reducing the relative abundance of these genera and lowering the diversity in fungal microbiome (**Figures 2A,B**). Generally, most soil-borne pathogens grow saprophytically in rhizosphere so as to reach their hosts and proliferate to a certain level before infecting host tissues and escaping from rhizosphere battle zone (Berendsen et al., 2012). Compared with Hs, the Ds has an enriched Ilyonectria, suggesting the pathogen that infected ginseng roots originated from the rhizosphere soil.

As we know, there is a fierce battle among microorganisms in the rhizosphere and endosphere due to resource heterogeneity and availability (Raaijmakers et al., 2008; Saleem, 2015).

Fusarium, a major fungal pathogen in ginseng, was more abundant in Hr and Hs than in Dr and Ds (**Figures 2E,F** and **Supplementary Figure S2H**). We assume this situation is attributed to the competition between Fusarium and Ilyonectria for ecological niche and nutrients. Besides, Mortierella was found enriched in Hr (**Figures 2E,F**) and its relative abundance was positively associated with P (**Figure 4B**). Mortierella spp. is ubiquitous in the bulk and rhizosphere soils (Wang et al., 2018) with an important role in keeping plant healthy by suppressing soil-borne pathogens and assisting plant with phosphorus uptake (Osorio and Habte, 2001; Miao et al., 2016). With these functions, Mortierella may play a key role in promoting the health of ginseng roots.

A recent study showed that infection caused by Ilyonectria led to the upregulation of salicylic acid (SA) in ginseng roots (Farh et al., 2019). SA, typically effective against infection caused by pathogens, is a major signaling regulator in plant (Pieterse et al., 2012). SA also influences the microbial community of roots (Lebeis et al., 2015). In our research, a decreased Shannon diversity was observed in Dr (**Figure 1A**). In comparison with Hr,

decreased relative abundance of Actinomycetales and increased relative abundance of Pseudomonadales was observed in Dr (**Figures 1E,F**), suggesting that the balance between these two orders could be important to the health of ginseng roots. As one of the dominant bacterial orders, Actinomycetales has a powerful ecological interaction with pathogens because of its abundance in soil and various broad-spectrum antibiotics it produces (Cha et al., 2016). Herein, Actinomycetales with antagonistic activities could be the key to keep ginseng root free of rust. Pseudomonadales, another important bacterial order, is typically considered as plant growth-promoting bacteria in soil (Bertrand et al., 2001; Zamioudis et al., 2013). Besides, Xanthomonadales, the cause of a wide variety of serious diseases in plants (Naushad and Gupta, 2013), was enriched in Ds. Plants have been shown to adjust their root microbiome upon pathogen infection and specifically recruit a group of growth-promoting and disease resistance-inducing microbes (Berendsen et al., 2018). This could explain the enriched abundance of Pseudomonas in Dr and increased Xanthomonadales in Ds. In future, we should pay more attention to the function of these species relative to ginseng in that many species are still not classified as either pathogenic or non-pathogenic using metagenomic sequencing.

Fungicides and host resistance often cannot offer adequate and sustainable control on soil-borne diseases (Weller et al., 2002). Therefore, we isolated three antagonists with antifungal activities against Ilyonectria. As is known to all, Bacillus and Streptomyces are producers of antibiotics against pathogens in different plant species (Chowdhury et al., 2015; Cha et al., 2016). Bacillus has a good potential as a microbial agent for the biocontrol of the ginseng root rot caused by Fusarium (Song et al., 2014) and Cylindrocarpon destructans (Kim et al., 2017). With regard to Trichoderma, it has been one of the most popular genera of fungi commercially available as a plant growth promoting fungus and a biological control agent with the abilities to inhibit phytopathogens in many ways (Tahia et al., 2004; Keswani et al., 2014). In field trials, the relative abundance and copies of Ilyonectria in rhizosphere soil obviously decreased in TB after the inoculation of S-11. This result might be explained by the presence of S-11 and the remarkable raising in bacterial diversity, as bacteria could protect plants against root-derived fungi considering the negative interactions between them (Durán et al., 2018). The reduction of Ilyonectria may help ginseng suffering from rusty diseases recovery in the long term.

In field trials, S7-1 did not perform as well as it did in pot conditions. Differences in biocontrol abilities between pot and field trials maybe due to the differential climate conditions and soil properties. We speculated that S7-1 could protect ginseng roots away from the rusty disease at the early stage, while S-11 might be the best choice in severe rusty areas. These isolates are potential biocontrol candidates for the rusty disease. Further, disease suppressiveness is generally attributed to microbial consortia rather than to one microbial species only. Hence, application of synthetic communities has been suggested as an alternative to improve the consistency of pathogen control (Grosskopf and Soyer, 2014). In addition, RDA analysis showed that Ilyonectria presented different relationships with different soil properties with both positive and negative links. Therefore, changing physical and chemical properties of soil to shift the structure of microbiome and reduce the abundance of Ilyonectria for disease management will be a promising strategy (Vida et al., 2016). In the future, more investigation is needed to enhance our understanding of the relationship between ginseng roots and root-associate microbiome to improve ginseng yields and increase resilience toward biotic and abiotic stresses (Haney and Ausubel, 2015; Saleem et al., 2018).

## CONCLUSION

In conclusion, utilizing metagenomic tool, we uncovered the differences of microbial community between rusty and healthy P. ginseng roots. Microbial diversity is higher in healthy tissues than in diseased ginseng root tissues. Moreover, the structure of root microbiome demonstrates that the genus of Ilyonectria is the main microorganism that causes rusty ginseng roots. Several other bacterial and fungal taxa are also differentially distributed in healthy and diseased groups. Such differences play an important role in keeping ginseng root healthy. At last, we developed a biocontrol practice to reduce the amount of Ilyonectria in soil and studied the microbiota changes after treatment. This offers promising solutions to the biocontrol of rusty disease.

## DATA AVAILABILITY

No datasets were generated or analyzed for this study.

## AUTHOR CONTRIBUTIONS

HM and DL designed the research. DL performed the experiments and analyzed the data. DL and HS wrote the manuscript. All authors critically revised the manuscript and approved the final version.

## FUNDING

This work was funded by the National Key Basic Research Program of China (2015CB755704), the Key Research Program of the Chinese Academy of Sciences (ZDRW-ZS-2016-3), and the International Partnership Program of Chinese Academy of Sciences (153D31KYSB20170121).

## ACKNOWLEDGMENTS

We thank Jian Ding for assistance in the sampling and field experiments.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2019. 01350/full#supplementary-material

#### REFERENCES

fmicb-10-01350 June 14, 2019 Time: 17:26 # 11


microbial community data. GigaScience 2:16. doi: 10.1186/2047-21 7X-2-16


**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 Liu, Sun and Ma. 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.

fmicb-10-01350 June 14, 2019 Time: 17:26 # 12

# Biocontrol of Soft Rot of Chinese Cabbage Using an Endophytic Bacterial Strain

Wenyan Cui1,2, Pengjie He1,2 \*, Shahzad Munir<sup>2</sup> , Pengbo He<sup>2</sup> , Yueqiu He3,4, Xingyu Li3,5 , Lijuan Yang<sup>4</sup> , Biao Wang<sup>2</sup> , Yixin Wu3,4 and Pengfei He2,3 \*

<sup>1</sup> Guizhou University of Traditional Chinese Medicine, Guiyang, China, <sup>2</sup> Faculty of Plant Protection, Yunnan Agricultural University, Kunming, China, <sup>3</sup> National and Local Joint Engineering Research Center for Screening and Application of Microbial Strains, Kunming, China, <sup>4</sup> Faculty of Agronomy and Biotechnology, Yunnan Agricultural University, Kunming, China, <sup>5</sup> Faculty of Science, Yunnan Agricultural University, Kunming, China

Soft rot caused by Pectobacterium carotovorum subsp. carotovorum (Pcc) is a major constraint in the production of Chinese cabbage. The objective of this study was to demonstrate that the causative agent Pcc may be successfully managed by Bacillus amyloliquefaciens KC-1, both in vitro and in vivo. Chinese cabbage seedlings were cultivated in organic substrate termed bio-organic substrate using a floating-seedling system with B. amyloliquefaciens KC-1. This approach was applied in a greenhouse to evaluate the management of soft rot. The results showed that the extent of soft rot, as well as the transmission of Pcc to the stem progeny and its survival in the rhizosphere, was reduced following inoculation with B. amyloliquefaciens KC-1. In contrast, the population diversity of B. amyloliquefaciens KC-1 persisted in the Chinese cabbage stems after germination. These findings revealed that B. amyloliquefaciens KC-1 was able to survive and suppress the growth of Pcc in Chinese cabbage and its rhizosphere, protecting the host from the pathogen. The use of B. amyloliquefaciens KC-1 throughout the growth period of plants may be an effective strategy for the prevention of soft rot in Chinese cabbage.

Keywords: Pectobacterium carotovorum subsp. carotovorum, Bacillus amyloliquefaciens, bio-organic substrate, rhizosphere competence, colonization, biocontrol

### INTRODUCTION

Pectobacterium carotovorum subsp. carotovorum (Pcc) is a common cause of soil-borne soft rot, in a broad range of vegetable and flower hosts such as Chinese cabbage, tomato, potato, cucumber, Amorphophallus konjac, and Zantedeschia hybrida (Des Essarts et al., 2016; Kang et al., 2016; Shao et al., 2016; Garge and Nerurkar, 2017; He et al., 2018). The typical symptoms include maceration and rotting of leaves and other organs of plant, resulting in loss of the marketable yield. Furthermore, soft rot may occur during transit, storage, or marketing (Bhat et al., 2010).

Soil-borne diseases are difficult to overcome. Chemical methods – though generally effective – are not desirable due to concerns regarding the development of resistance and environmental pollution (Gill and Garg, 2014). Effective and environmentally friendly methods for controlling these diseases are required to reduce the use of chemical pesticides. Biocontrol is one of the most effective and promising approaches for the control of soft rot and other plant diseases.

#### Edited by:

Aleksa Obradovic,´ University of Belgrade, Serbia

#### Reviewed by:

Chunhao Jiang, Nanjing Agricultural University, China Qian Guoliang, Nanjing Agricultural University, China

#### \*Correspondence:

Pengjie He 2497032366@qq.com Pengfei He nanhudaozhu@sina.com

#### Specialty section:

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

Received: 30 March 2019 Accepted: 12 June 2019 Published: 03 July 2019

#### Citation:

Cui W, He P, Munir S, He P, He Y, Li X, Yang L, Wang B, Wu Y and He P (2019) Biocontrol of Soft Rot of Chinese Cabbage Using an Endophytic Bacterial Strain. Front. Microbiol. 10:1471. doi: 10.3389/fmicb.2019.01471

**38**

Previous studies revealed a limited number of bacteria (i.e., Bacillus, Actinomycete, Pseudomonas, Lactobacillus, and Serratia), which may be useful as biocontrol agents against soft rot pathogens (El Karkouri et al., 2010; Czajkowski et al., 2012; Des Essarts et al., 2016; Tsuda et al., 2016; Garge and Nerurkar, 2017; Gerayeli et al., 2018). Among these, Bacillus subtilis and Bacillus amyloliquefaciens (B. amyloliquefaciens) have attracted considerable attention owing to its strong biological control activity and remarkable environmental suitability to survive under adverse conditions (Romero et al., 2007; Ongena and Jacques, 2008).

Bacillus amyloliquefaciens suppresses the pathogens through competition, promotion of growth, antibiosis, and induction of systemic resistance (Diallo et al., 2011). The polyketides (i.e., difficidin, bacillaene, and macrolactin) of non-ribosomal synthesis are the most effective antibacterial compound constituents by B. amyloliquefaciens and B. subtilis (Chen et al., 2009; Chowdhury et al., 2015). Another product of non-ribosomal synthesis, the dipeptide bacilysin consisting of anticapsin and alanine moieties, together with the polyketides, was found as being involved in contribute the control of bacterial plant diseases (Bais et al., 2004; Chen et al., 2009; Zeriouh et al., 2011). However, the effectiveness of endophytes as biological control agents (BCAs) is dependent on efficient colonization of the plant environment. The extent of endophytic colonization in plant rhizosphere and tissues reflects the ability of bacteria to selectively adapt and compete in those specific ecological niches (Chowdhury et al., 2015; Des Essarts et al., 2016; Abdallah et al., 2018; Munir et al., 2018).

The majority of biocontrol agents against Pcc are applied to the rhizosphere using water. Rhizosphere competence is a prerequisite for the effective biological control of soft rot. Furthermore, stable biological control against soil-borne pathogens cannot be expected merely through the application of an antagonist without a suitable management method. This is because the introduced biocontrol agents compete for niches and nutrients with other native microbes for their survival (Kokalis– Burelle et al., 2002; El-Hassan and Gowen, 2006).

In this study, we isolated B. amyloliquefaciens KC-1 from an infected field of Chinese cabbage (cv. "Qingdao 83-1") to control soft rot. We used biological and molecular tools to identify and characterize the bacterium. In addition, we assessed its effectiveness in terms of biological control against the transmission of soft rot pathogen Pcc through experiments in the greenhouse. The aim of this research was to establish effective management strategies for the development of a safe microbial pesticide applicable to the production of Chinese cabbage in the greenhouse and field.

### MATERIALS AND METHODS

#### Bacterial Strains, Growth Conditions, and Preparation of Inoculum

All bacterial strains were grown in Luria-Bertani (LB) medium (NaCl, 10 g/l; bacto tryptone, 5 g/l; yeast extract, 5 g/l; AoBoXing Bio-tech Co., Ltd., Beijing, China) at 30 (B. amyloliquefaciens) and 28◦C (Pcc) for 48 h under shaking (160 rpm). Whenever necessary, bacteria were washed thrice using isotonic saline solution (NaCl, 8 g/l) through centrifugation at 5,000 rpm for 5 min, followed by resuspension of the pellets in saline. Pcc strain E1 had been isolated from symptomatic Chinese cabbage and identified (MH934929).

### Isolation and Identification of B. amyloliquefaciens KC-1

Bacillus amyloliquefaciens KC-1 was isolated from the rhizosphere of an asymptomatic Chinese cabbage plant infected with soft rot in Kunming City, China (25◦ 13<sup>0</sup> N, 102◦ 750 E). The physiological and biochemical properties as well as the morphological characteristics were analyzed as previously described (Logan and De Vos, 2009). The amplification and sequencing of the gyrB gene was conducted using the following universal primers: GyrB-F (5<sup>0</sup> -GAA GTC ATC ATG ACC GTT CTG CAY GCN GGN AAR TTY GA-3<sup>0</sup> ) and GyrB-R (5<sup>0</sup> -AGC AGG GTA CGG ATG TGC GAG CCR TCN ACR TCN GCR TCN GTC AT-3<sup>0</sup> ) (Abdallah et al., 2018). The obtained sequences were compared with known sequences of the bacterial gyrB gene available in the National Center for Biotechnology Information database using the BlastT program<sup>1</sup> . Furthermore, a phylogenetic tree was constructed using the neighbor-joining method with the MEGA (version 7.0) software (MEGA, United States) (Kumar et al., 2016).

#### In vitro Antibacterial Activity of B. amyloliquefaciens KC-1 Against Pcc

Testing of in vitro antibacterial activity was performed using LB agar plates (Des Essarts et al., 2016) in quadruplicates. Pcc E1 cell suspension [10<sup>6</sup> colony-forming units (CFU)/ml] (200 µl) was evenly spread onto the surface of the LB plates (φ9.0 cm). Subsequently, 10 µl of B. amyloliquefaciens KC-1 culture and sterile distilled water (SDW) were added onto the edge of the plate. The plates were incubated at 30◦C for 48 h.

#### Biocontrol Activity of B. amyloliquefaciens KC-1 Against Pcc in Chinese Cabbage

The biological assays were performed as previously described (Dong et al., 2004). B. amyloliquefaciens KC-1 and Pcc E1 (10<sup>8</sup> CFU/ml, each) were mixed as an inoculant. Healthy Chinese cabbages (1-month-old seedlings of Chinese cabbage, cv. "Qingdao 83-1") were purchased. After sterilization of the surface, cabbage petioles were cut into pieces (∼3.5–4.0 × 5.5– 6.5 cm in size). Subsequently, inoculant (10 µl) was inoculated onto the surface of the cabbage petioles (i.e., quadruplicate per condition) through a pipette without wounding. SDW, B. amyloliquefaciens KC-1, and Pcc E1 cell suspension were used separately here, respectively, as a control. The infected discs were placed in 17-cm sterile Petri dishes and incubated at 30◦C for 24 h. The extent of soft rot on each disc was evaluated by detecting the maceration area and the percentage maceration

<sup>1</sup>http://blast.ncbi.nlm.nih.gov/Blast.cgi

(PM) (Dong et al., 2004). The maceration area in all cases was analyzed using Mshot Digital Imaging System (version 9.3.3.1) software (Mingmei, Guangzhou, China), PM = the weight of macerated tissue after inoculation × 100/the weight of tissue before inoculation. The data for the values of maceration area and PM were compared according to Duncan's multiple range test, and a pairwise comparison test was performed. A p < 0.05 denoted statistical significance in all tests (SPSS software, version 22.0, Chicago, IL, United States).

In addition, 10 µl of inoculant solution, pure B. amyloliquefaciens KC-1, and inactivated Pcc E1 suspension were inoculated onto the surface of petioles of healthy Chinese cabbage in the greenhouse, as described above. The symptoms of soft rot were observed and evaluated after inoculation and growth in the greenhouse at 22–28◦C for 3 days.

#### In vitro Co-culture Assay

Overnight culture of B. amyloliquefaciens KC-1 (10<sup>8</sup> CFU/ml) was used to initiate co-culture with Pcc E1 (10<sup>8</sup> CFU/ml) in LB medium and incubated at 28◦C at 150 rpm for 12 h. Similarly, Pcc E1 was inoculated and cultured alone as control to monitor the growth rate. The CFU of the co-cultured strains were recorded separately based on their distinct colony morphologies on the LB plates. Each treatment was performed in triplicate. The independent t-test (p < 0.05) was used to compare the CFU values of B. amyloliquefaciens KC-1 and Pcc E1 between strains grown alone and co-cultured.

#### PCR Amplification of Antibiotic Genes of Nonribosomal Peptide Synthetase and Polyketide Synthetase

The amplification and sequencing of the genes involving in synthesis of the polyketides difficidin, bacillaene, and macrolactin and the dipeptide bacilysin were performed using the primers listed in **Table 1**. The PCR sample mix was prepared as follows. Initially, 2 µl of EasyTaq buffer (10×) (TransGen, Beijing, China) was vortexed with 1.6 µl of dNTPs (2.5 mM), 1 µl of forward and reverse primers (10 µM), and 0.5 µl of EasyTaq DNA polymerase (TransGen, Beijing, China). Extracted DNA (50 ng) from the bacterium (HiPure Bacterial DNA Kit, Magen, China) sample and about 14.0 µl SDW were added to the PCR reaction mixture in a total volume of 20 µl. The PCR was performed in a WD-9402A Thermal Cycler (Applied Biosystems, Beijing, China) as follows: denaturation step (5 min at 94◦C), followed by 30 cycles of 40 s at 94◦C, 40 s at 57◦C, and 45 s at 72◦C, and a final extension of 10 min at 72◦C. PCR products were sequenced and analyzed as above.

#### Construction of Green Fluorescent Protein (GFP)-Tagged B. amyloliquefaciens KC-1

The GFP-tagged B. amyloliquefaciens KC-1 (KC-1-gfp) strain was obtained through conjugal transfer of the pHT01-P43GFPmut3a plasmid – carrying a GFP gene – into the KC-1 cytoplasm (He, 2014; Zhang et al., 2015). To assess the stability of the KC-1-gfp strain without selection, the KC-1-gfp was cultured overnight in LB medium (10 µg/ml chloramphenicol), followed by adjustment to a suspension (optical density [OD]<sup>600</sup> = 1.0) in LB broth. Subsequently, the stability was evaluated through continuous culturing in fresh LB broth (0.1% w/w, per 5 h) without chloramphenicol for 60 h at 37◦C under shaking (160 rpm). Furthermore, the suspension of KC-1-gfp and wild-type KC-1 were prepared as described above. Subsequently, 1.0 ml of suspension was introduced to 100 ml of fresh LB broth and cultured under the aforementioned conditions. During the initial 12 h post-inoculation, the OD<sup>600</sup> of the culture was measured every 2 h at OD600. Between 12 and 28 h, the OD<sup>600</sup> was measured every 4 h. Moreover, between 28 and 60 h, the OD<sup>600</sup> was measured every 8 h.

### Preparation of Bio-Organic Substrate

For the following experiments in the greenhouse, B. amyloliquefaciens KC-1-gfp strain was used instead of the wild-type strain. KC-1-gfp was incubated and resuspended in SDW as described above. The seedling substrate was initially moistened using the KC-1-gfp suspension. The bacteria concentration was 10<sup>7</sup> CFU/g. SDW at an equal volume was used as a negative control.

#### Biocontrol of Chinese Cabbage Soft Rot in Greenhouse Assays

Four independent greenhouse assays were performed to assess the effectiveness of biocontrol against soft rot in cabbages over a period of 2 years (2015–2016). Two different greenhouses (25–30◦C) located in Kunming city were used – one located at the NLJERCSAMS-YCFCL experiment site (Qinglong, Anning, Kunming, China) and the other at the FPP-YAU experiment site (Panlong, Kunming, China). The assays were termed according to their location and year as follows: A1 and A2 (2015 and 2016 in Anning, respectively) and P1 and P2 (2015 and 2016 in Panlong district, respectively).

A floating-seedling system was employed to cultivate the Chinese cabbage seedlings. The bio-organic substrate was gently spread on the foam trays (80 cm [L] × 50 cm [W] × 15 cm [H]). After sterilization of the surface (i.e., soaked in 70% ethanol for 2 min, immersed in 10% sodium hypochlorite for 5 min, and rinsed four times with SDW to remove residues), Chinese cabbage (Qingdao 83-1) seeds were sown in the commercial substrate, and trays were placed into a floating tank for 25 days in the greenhouse. The shoot height, fresh weight, and dry weight of 30 seedlings randomly selected were measured in triplicates prior to transplantation. The independent t-test (p < 0.05) was used to compare the values of shoot height, fresh weight, and dry weight between seedlings grown with bio-organic substrate and organic substrate.

A1 and A2 contained 80 plants per condition, while P1 and P2 contained 72 plants per condition. All experiments were set in a completely randomized design with six replicates under the following six conditions: seeds were directly sown in uninfected soil; seeds were directly sown in infected soil (10<sup>7</sup> CFU/ml of E1 suspension, 100 ml/plot); seedlings were grown in organic substrate or bio-organic substrate using the


TABLE 1 | The PCR detection of the polyketides and the dipeptide biosynthesis genes from B. amyloliquefaciens KC-1.

floating-seedling system; and after transplantation, plants were infected with Pcc E1 alone and in combination with biocontrol strain KC-1-gfp. The biocontrol and pathogenic bacteria were inoculated separately. Two days after the transplantation of seedlings, the pathogen Pcc E1 was inoculated at 10<sup>9</sup> CFU/plant onto transplanted plants through drenching with 100 ml cell suspension (10<sup>7</sup> CFU/ml). Four days after the transplantation of seedlings, 100 ml KC-1-gfp suspension (10<sup>7</sup> CFU/ml) was introduced to the plant petioles through watering. For each treatment – except for the uninfected plants – the plants were inoculated a total of four times at 7-day intervals after transplantation. Fertilizer, insecticides, and fungicides were applied according to the instructions provided by the manufacturers. At 70 days after transplantation, the incidence rate (IR), relative disease severity index (DSI), and protection value (PV) under each condition were used to determine the relative severity of the disease as follows (Tsuda et al., 2016): 0 = no symptoms; 1 = very small lesions on the outer leaves; 2 = rot on the outer leaves; 3 = rot on the outer leaves and part of the head; and 4 = rot on most parts of the head. IR = 6(the number of diseased plants) × 100/total number of plants), DSI = 6(scale × the number of diseased plants) × 100/(4 × total number of plants), PV = (DSI in control – DSI in treatment) × 100/DSI in control. For each treatment in the A1 and P1 experiments, the marketable yield of Chinese cabbage was recorded as grams of fresh weight per plant. The data for the IR, DSI, PV, and shoot fresh weight were compared according to Duncan's multiple range test, and a pairwise comparison test was performed. A p < 0.05 denoted statistical significance in all tests.

#### The Population of Pcc E1 Pathogen on the Leaves of Chinese Cabbage in the Greenhouse Assays

The transplanted plants were selected and collected to determine the population of the Pcc pathogen on the leaves of Chinese cabbage. On harvest day (ca. 70 days) in the A2 and P2 experiments, three asymptomatic leaves of Chinese cabbage were collected from each plant with a symptomatic score ≤3. Petioles of asymptomatic leaves from each transplanted plant were collected and pooled for the pathogenic bacteria enrichment process as follows (Des Essarts et al., 2016): the petioles were washed with sterile water thrice, 10 ml of phosphate buffer (2.7 g/l of Na2HPO<sup>4</sup> · 12H2O, 0.4 g/l of NaH2PO<sup>4</sup> · 2H2O, pH 7.2) was added, the petioles were ground to a homogenate suspension in a shaking incubator (200 rpm, 2 h) at 25◦C. Subsequently, 200 µl of homogenate suspension was transferred into 1,800 µl PEB (0.32 g/l of MgSO4, 1.08 g/l of (NH4)2SO4, 1.08 g/l of K2HPO4, and 1.7 g/l of sodium polypectate) liquid medium (Perombelon and Van Der Wolf, 2002; Czajkowski et al., 2010) and incubation was continued under the same conditions for 48 h. The bacterial culture was centrifuged to collect bacteria for total DNA extraction using the HiPure Bacterial DNA Kit (Magen, China). The presence of the Pcc pathogen was detected through polymerase chain reaction (PCR) using the following two Pcc-specific primers: EXPCCF (5<sup>0</sup> -GAA CTT CGC ACC GCC GAC CTT CTA-3<sup>0</sup> ) and EXPCCR (5<sup>0</sup> -GCC GTA ATT GCC TAC CTG CTT AAG-3<sup>0</sup> ) as previously described (Kang et al., 2003; Humphris et al., 2015). Data were analyzed using the Chi-squared test (p < 0.05).

#### Monitoring the Pathogenic and Biocontrol Population in the Greenhouse Assays

In the course of the A2 and P2 greenhouse assays, the cabbage leaves and rhizosphere soil samples (n = 3, each) were collected at the following seven time points: prior to the transplantation of the seedlings (ca. 0 day), inoculation of pathogen Pcc E1 alone (ca. 4 days, only soil samples were collected), 4 days after introduction of the biocontrol agent (ca. 8, 15, 22, and 29 days), and harvest day (ca. 70 days). All samples were used to monitor the population of B. amyloliquefaciens KC-1. In contrast, only four soil samples (ca. 0, 4, 29, and 70 days) were used to monitor the population of the Pcc E1 pathogen. Subsequently, the plant samples were cut and frozen using liquid nitrogen, and were ground to a powder for DNA extraction using the HiPure Bacterial DNA Kit (Magen, China). Soil samples were air-dried for DNA extraction using the HiPure Soil DNA Mini Kit (Magen, China). The concentration and the quality of the extracted DNA were determined using an Ultrospec 2100 pro UV/visible spectrophotometer (Amersham Biosciences, Pittsburgh, PA, United States) and through electrophoresis on agarose gels (1%, w/vol). Each sample was evaluated in triplicate.

Strain-specific primers were designed to monitor the B. amyloliquefaciens KC-1 and pathogen Pcc strain E1 from

the extracted bacterial and soil DNA using quantitative PCR (qPCR). These primers were designed using the software of the CLC Genomics Workbench 5.5 (CLC Bio, Denmark) and Primer premier 5.0 (Premier, Canada). Their specificity was tested using PCR on bacterial strains (Des Essarts et al., 2016). The primers of B. amyloliquefaciens KC-1-gfp and Pcc E1 were designed to amplify a DNA fragment of ca. 137 and 119 bp, respectively. Finally, the primers GFP-F: (5<sup>0</sup> -CGA CTT TCG GTT ATG GTG TTC A-3<sup>0</sup> ) and GFP-R: (5<sup>0</sup> -CGT GTA GTT CCC GTC ATC TTT G-3<sup>0</sup> ) were selected for the detection of B. amyloliquefaciens KC-1, while the primers E1-F: (5<sup>0</sup> -AGG TGC AAG CGT TAA TCG GA-3<sup>0</sup> ) and E1-R: (5<sup>0</sup> -GCC TCT AGC CTG TCA GTT TTG A-3<sup>0</sup> ) were used for the detection of Pcc E1.

obtained from the National Center for Biotechnology Information database are indicated in parentheses.

The qPCR sample mix was prepared as follows. Initially, 12.5 µl of master mix (Takara, Dalian, China) were vortexed with 0.5 µl of the specific forward and reverse primers (10 µM), and 0.5 µl of the ROX reference Dye (Takara, Dalian, China). Extracted DNA (5 µl) from the soil (HiPure Soil DNA Mini Kit, Magen, China) or plant (HiPure Bacterial DNA Kit, Magen, China) samples and 6.0 µl of SDW were added to the qPCR mix. The qPCR was performed as follows: denaturation step (2 min at 95◦C), followed by 40 cycles of 15 s at 95◦C, 15 s at 59◦C, and 30 s at 72◦C. The fluorescence was measured after each cycle. A melting curve analysis was performed at the end of the PCR run (15 s at 95◦C, 60 s at 60◦C, and 15 s at 95◦C) to ensure the amplification of only one PCR product. Data analysis was performed using the StepOne software v2.3 (ABI, PE Applied Biosystems, United States). The relative number of DNA copies per grams/sample was calculated as previously described (Des Essarts et al., 2016).

#### FIGURE 2 | Bacillus amyloliquefaciens KC-1 inhibiting the growth of Pcc E1.

### RESULTS

#### Characterization and Identification of B. amyloliquefaciens KC-1

KC-1 was found to be a typical Bacillus (Logan and De Vos, 2009). When grown on LB agar medium, it formed mucous wrinkles. Under the microscope, the KC-1 cells were rod-shaped and formed ellipsoidal endospores. Furthermore, physiological and biochemical analyses revealed that the KC-1 strain was motile, Gram-positive, and oxidase- and catalase-positive. Meanwhile, the gyrB gene sequence of KC-1 (1173 bp) showed 99% identity with B. amyloliquefaciens (**Figure 1**). The sequence was deposited in the GenBank database NCBI (Accession No. MH973156).

#### In vitro Antibacterial Activity

Clear growth inhibition zones were observed, indicating that B. amyloliquefaciens KC-1 exerted a strong effect against Pcc E1 on agar plates (**Figure 2**) previously inoculated with 200 µl Pcc E1 cell suspension (10<sup>6</sup> CFU/ml).

#### Evaluation of the Biocontrol Potential of B. amyloliquefaciens KC-1

SDW and Bacillus amyloliquefaciens KC-1 did not negatively affect the Chinese cabbage, following inoculation onto the sterilized surface of healthy tissues. There were no change observed on the tested tissues of petioles (**Figures 3B,C**). The pathogen Pcc E1 caused symptoms of soft rot (strong tissue maceration) on the cabbage (**Figure 3A** and **Table 2**). In comparison with pure Pcc E1 alone, only slight maceration was observed (**Figure 3D** and **Table 2**) following the co-inoculation of petiole slices with a mixed inoculant solution (i.e., Bacillus and pathogen at a 1:1 ratio). Furthermore, 3 days after inoculation with the aforementioned mixed inoculation solution, there were no symptoms or only slight symptoms of soft rot observed on the petioles of Chinese cabbage in the greenhouse, compared with control (**Figure 4**).

#### In vitro Influence of B. amyloliquefaciens KC-1 on the Growth of Pcc E1

The distinct colony morphologies of B. amyloliquefaciens KC-1 (i.e., opaque and large) and Pcc E1 (i.e., translucent and small) were used to quantify their growth in the co-culture



<sup>A</sup>Percentage maceration = the weight of macerated tissue after inoculation × 100/the weight of tissue before inoculation. Different letters above the bars indicate that the differences are significant (P < 0.05) using Duncan's multiple range test. Values represent the mean of four replications by checking 12 tissues.

FIGURE 4 | Biocontrol activity assay for soft rot caused by pathogen Pcc E1. (A) Chinese cabbage inoculated with Pcc E1 alone. (B) Chinese cabbage inoculated with inactivated B. amyloliquefaciens KC-1 alone. (C) Chinese cabbage co-inoculated with a mixed inoculation solution (B. amyloliquefaciens KC-1 and Pcc E1, 1:1 ratio).

TABLE 3 | Evaluation of antibacterial traits of B. amyloliquefaciens KC-1 in vitro.


experiments. Pcc E1 showed ∼10<sup>9</sup> CFU/ml when cultured alone and ∼10<sup>7</sup> CFU/ml in co-culture, indicating that its growth was inhibited by B. amyloliquefaciens KC-1 (**Figure 5**). Meanwhile, B. amyloliquefaciens KC-1 showed ∼10<sup>9</sup> CFU/ml under both culture conditions.

### Characterization of Potential Antibacterial Traits

Several traits potentially involved in antimicrobial activities were detected in vitro. Tested antibiosis traits included presence of antibiotic genes (**Table 3**). Grounded on the results of PCR analysis, strain KC-1 showed the presence of genes involved in the biosynthesis of the polyketides difficidin (dfnA), bacillaene (baeA), and macrolactin (mlnA) and the dipeptide

bacilysin (bacA). The results indicated that strain KC-1 possess antibacterial activities.

### GFP Tagging of the B. amyloliquefaciens KC-1

Validation of the KC-1-gfp transformants was performed by measuring their GFP fluorescence using photoelectric refractometer (data not shown). The growth characteristics in LB broth (**Figure 6A**) and its antagonistic ability against the Pcc E1 strain on LB agar plates (data not shown) did not exhibit a significant difference between the KC-1-gfp and KC-1 wild type, suggesting that the normal metabolism of KC-1 was not disrupted by the presence of the plasmid. The stability of the pHT01- P43GFPmut3a plasmid in KC-1-gfp was evaluated through continuous culture in LB broth without antibiotic treatment. After a 60-h incubation on LB agar plates only ∼18% of the KC-1-gfp culture lost the plasmid (**Figure 6B**), indicating that the KC-1-gfp strain could be used in the subsequent prolonged colonization experiments.

### Promotion of Seedling Growth by the Bio-Organic Substrate

The shoot fresh and dry weight, as well as the height, of the Chinese cabbage seedlings grown in a bio-organic substrate for 25 days was significantly increased by 52.8, 32.4, and 17.45%, respectively, compared with the control (**Figure 7** and **Table 4**). In contrast, the values of the root fresh and dry weight did not exhibit a significant difference between the two treatments.

#### Suppression of the Severity of Soft Rot by B. amyloliquefaciens KC-1 in the Greenhouse Assays

Four independent experiments (i.e., A1, A2, P1, and P2) were performed in two greenhouses over a period of 2 years

(**Table 5**). Chinese cabbage seedlings were grown in greenhouse soil or bio-organic substrate using the floating-seedling system for 25 days. Subsequently, they were transplanted and used in the greenhouse under different conditions. In all these experiments, there were no symptoms observed under the water control condition in plants treated without pathogen (data not shown). Similarly, there was no significant difference in soft rot symptoms between the plants grown using the floatingseedling system in the absence of B. amyloliquefaciens KC-1 and those directly sown with Pcc alone (**Table 5**). Following treatment with B. amyloliquefaciens KC-1, there was a tendency toward reduction in the severity of disease compared with the control. In addition, plants grown in the presence of the biocontrol agent both in the growth stage of the floatingseedling system and the greenhouse significantly reduced the severity of disease compared with application in one of the two stages (**Table 5**).

TABLE 4 | The efficacy of different treatments on promoting seedling growth in floating-seedling system.


<sup>a</sup>OS, organic substrate; BIO, bio-organic substrate containing B. amyloliquefaciens KC-1. The data were analyzed using the independent t-test. Values represent the mean of three replications by checking 30 plants. Asterisk indicates values that are significantly different (p < 0.05).

### Limitation of Spread of the Pathogen Pcc on the Leaves of Chinese Cabbage

On harvest day (ca. 70 days) of the A2 and P2 experiments, asymptomatic outer leaves were collected from the plant with a symptomatic score ≤3 and screened for Pcc through PCR (**Figure 8**). In experiment A2, all the combinations treated with B. amyloliquefaciens KC-1 exhibited reduced dissemination of the pathogen in the asymptomatic Chinese cabbage leaves. In experiment P2, a significant decrease in the pathogen was recorded following the application of B. amyloliquefaciens KC-1 in the growth stage in the greenhouse only, and combination of both floating-seedling system and greenhouse.

### Limitation Survival and Multiplication of the Pathogen Pcc in the Rhizosphere Soil of Chinese Cabbage

During the course of the A2 and P2 experiments, the population of Pcc E1 in the rhizosphere of Chinese cabbage was measured through qPCR using specific primers (**Figure 9**). The population size of Pcc E1 was attained at a level of 104–10<sup>6</sup> DNA copies/g of dry soil in all conditions after introduction of the pathogen (green lines in **Figure 9**). The B. amyloliquefaciens KC-1 applied in the greenhouse reduced one or two orders of magnitude of Pcc E1 compared with the control.

#### Population Dynamics of B. amyloliquefaciens KC-1 in the Greenhouse Assays

During the course of the A2 and P2 experiments, the population dynamics of B. amyloliquefaciens KC-1 in the cabbage leaves and rhizosphere was detected, exhibiting similar patterns (**Figure 9**). Notably, there was no DNA copy detected in the soil samples collected for all treatments prior to transplantation (i.e., 0 day) in the greenhouse. This finding confirmed the absence of B. amyloliquefaciens KC-1 in the soil prior to the application of the strain (red lines in **Figure 9**). Cabbage grown in the presence of KC-1 by floating-seedling system and without KC-1 in greenhouse (bio-organic substrate plus water), we detected a consistent cell concentration approximately 10<sup>4</sup> DNA copies/g of petiole. In the other treatments of bio-organic substrate plus B. amyloliquefaciens KC-1 and organic substrate plus B. amyloliquefaciens KC-1 was decreased by two orders of magnitude after its application (blue lines in **Figure 9**). Meanwhile, while biocontrol agent was introduced by spraying in the greenhouse, the population size of KC-1-gfp was attained at a level of 10<sup>6</sup> DNA copies/g of dry soil, three orders of magnitude higher than the treatment with biocontrol agent only inoculated in the floating-seedling system (bio-organic substrate plus water, red lines in **Figure 9**).

## DISCUSSION

Studies have reported the use of Bacillus sp. as biocontrol strains against pathogens associated with bacterial soft rot (Yang et al., 2011; Czajkowski et al., 2012; Des Essarts et al., 2016; Garge and Nerurkar, 2017; Gerayeli et al., 2018). However, certain Bacillus sp. have also been shown to cause bacterial soft rot in various vegetables, including the onion (Hwang et al., 2012), arrowhead (Zhong et al., 2015), and potato (Wang et al., 2017). Moreover, a study reported that Bacillus sp. cause severe decay in the potato, and may effectively control bacterial soft rot in potatoes, green peppers, and Chinese cabbages (Zhao et al., 2013). In the present study, B. amyloliquefaciens KC-1 was successfully isolated and identified. We revealed that this strain did not induce symptoms of soft rot on the tested plant tissues. However, we also demonstrated their in vitro and in vivo activity against the Pcc E1 strain. This result was also validated using PCR analysis of KC-1 showing the presence of genes involved in the biosynthesis of the polyketides difficidin, bacillaene, and macrolactin and the dipeptide bacilysin.

Real-time PCR is able to detect and quantify microorganisms in plant tissues and soil samples (Des Essarts et al., 2016). Considering the possible high abundance of native bacteria belonging to B. amyloliquefaciens in the rhizosphere soil of plants, the B. amyloliquefaciens KC-1-gfp was used in this study instead of its wild type. The qPCR results confirmed that B. amyloliquefaciens KC-1 can colonize the rhizosphere soil and leaves of Chinese cabbage. High densities of B. amyloliquefaciens KC-1-gfp were recovered in the rhizosphere soil prior to harvest. Colonization of the rhizosphere by the biocontrol agent is critical for the effective control of phytopathogens (Compant et al., 2005; Cao et al., 2011; Diallo et al., 2011). This suggests that the presence of high densities of B. amyloliquefaciens KC-1 strain in the rhizosphere soil is a prerequisite for suppressing the infection caused by the soil-borne pathogen Pcc. Biocontrol agents control soil-borne diseases through interference with their biological antibiotic responses, competition for niches and nutrients in the root and location of the lesion, and induction of systemic resistance in host plants (Bais et al., 2004; Compant et al., 2005; Haas and Défago, 2005; Cao et al., 2011). Recently, a study showed that the population of the pathogen and symptoms in the rhizosphere soil of Ralstonia solanacearum were significantly decreased following the introduction of a Bacillus


sp. (Tan et al., 2013). Of note, the introduction of Pseudomonas sp. reduced the blackleg and symptoms of soft rot symptoms; however, it did not succeed in limiting the concentration of the pathogen in the rhizosphere soil (Des Essarts et al., 2016). In comparison, the results of the present study clearly demonstrated that B. amyloliquefaciens KC-1 competes with the pathogen in the rhizosphere by significantly decreasing the pathogen copies in the rhizosphere soil and leaves of inoculated Chinese cabbage. This may be attributed to the adaptability of the Bacillus sp. in the rhizosphere and host plant. The biocontrol strains examined in this study may play the role of a "defender" against infections in plant by directly limiting the survival and transmission of the pathogen in the plant rhizosphere and tissues.

substrate; BIO, bio-organic substrate containing B. amyloliquefaciens KC-1. Significant reductions (Chi-squared test; p < 0.05) in the presence of Pcc E1

are indicated by asterisks.

The biocontrol agents associated with the roots may also prevent Pcc from invading the roots through antibiosis (Dong et al., 2004; Des Essarts et al., 2016; Tsuda et al., 2016; Garge and Nerurkar, 2017). In this study, B. amyloliquefaciens KC-1 induced a significant suppression of growth and reduction in the number of colonies of Pcc in plate assay and co-culture assay, respectively. These findings indicate the production of diffusible antibacterial compounds. Meanwhile, the percentage of Pcc-carrying asymptomatic leaves in Chinese cabbage inoculated with B. amyloliquefaciens KC-1 was reduced both directly and indirectly, suggesting that the transmission of Pcc to inner leaves was limited. The Bacillus sp. may act against pectinolytic bacteria

TABLE 5 | Disease

suppression

 of different treatments

 against bacterial soft rot of Chinese cabbage in greenhouse

 assays.

amyloliquefaciens

DProtection

END, not determined.

 value = (DSI in control – DSI in treatment) ×

 KC-1 after translating.

+

BIncidence

 rate = 6(the number of diseased plant) ×

100/DSI in control. Different letters above the bars indicate that the differences

100/total plant number).

CDisease severity

 are

seedlings grown in bio-organic

index = 6(scale ×

significant (P < 0.05) using Duncan's multiple range test.

 substrate and infected with Pcc E1 and B.

diseased plant number) × 100/(4 × total plant number).

fmicb-10-01471 July 2, 2019 Time: 17:45 # 9

by competing for nutrients and by producing signal molecules to disrupt the pectinolytic bacteria (Molina et al., 2003; Dong et al., 2004; Uroz et al., 2009; Garge and Nerurkar, 2017), and various antimicrobial molecules (Diallo et al., 2011) (e.g., siderophores, volatile compounds, antibiotics, and hydrogen peroxide).

The floating-seedling system – widely used in agriculture – assists in the growth of plant seedlings under non-pathogenic conditions (Yan et al., 2003; Sung and Hong, 2010). The organic substrate provides nutrients to ensure the survival and persistence of growth in the floating-seedling system. Recently, we reported that clubroot – one of the most destructive diseases of Chinese cabbage – may be controlled through seedling growth using the floating-seedling system in the absence of a biocontrol agent (He et al., 2019). However, there was no significant reduction observed in the symptoms of soft rot following the application of this method compared with direct seeding. In this study, we introduced a new biocontrol product – termed bio-organic substrate – by mixing the traditional organic substrate with B. amyloliquefaciens KC-1. Use of the bio-organic substrate in the floating-seedling system promoted growth of seedlings compared with the conventional organic substrate. In the greenhouse, three times introduction of B. amyloliquefaciens KC-1 culture to the soil reduced the incidence and severity index of soft rot in Chinese cabbage. The bio-organic substrate in the floating-seedling system alone reduced the density of B. amyloliquefaciens KC-1-gfp (104–10<sup>5</sup> copy/g) in plants. The disease incidence was not reduced but could decrease the DSI, which may be due to the plant niches occupied by the strain

KC-1 before Pcc E1 when the seedlings were grown in bioorganic substrate. Chinese cabbage seedlings grown in bioorganic substrate for 25 days, transplanted to the greenhouse, and treated with bacterial culture, exhibited a reduced incidence and severity index of soft rot, as well as impaired transmission of the Pcc strain in the cabbage leaves. Thus, colonization of Chinese cabbage by biocontrol bacteria prior to the introduction of Pcc may protect the plant from soft rot and suppress pathogen survival in the rhizoplane soil. Moreover, these results showed that B. amyloliquefaciens KC-1 – applied in an appropriate manner – may effectively manage soft rot in Chinese cabbage. However, field experiments are warranted to verify the competition and persistence of B. amyloliquefaciens KC-1 under natural conditions and its contribution to the reduction of disease symptoms and transmission of Pcc.

#### CONCLUSION

In conclusion, B. amyloliquefaciens KC-1 effectively reduced the symptoms of bacterial soft rot in Chinese cabbage and transmission of the Pcc pathogen under greenhouse conditions. Suitable carrier materials and methods are necessary for the function of this antagonist in the soil and its application in the field, providing a theoretical basis for the development of a bio-organic substrate. Use of a new bio-organic substrate at the initiation of the floating-seedling system and introduction of the B. amyloliquefaciens KC-1 strain to plants three times after transplantation resulted in good control of soft rot. This was

#### REFERENCES


achieved through successful colonization of the plant rhizosphere and tissues, consequently leading to a marked decrease in the population of the pathogen in the rhizosphere and transmission to the leaves. Furthermore, research investigating the bioantibacterial mechanisms of B. amyloliquefaciens KC-1 (i.e., quorum quenching, competition for nutrients, and induction of resistance in host plants) is currently ongoing in our laboratory.

#### 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

WC, PJH, PFH, YH, and YW designed the experiments and revised the manuscript. WC, PJH, PBH, PFH, XL, LY, BW, and SM performed the experiments and analyzed the data. WC, PJH, PFH, SM, and YH wrote the draft. All authors viewed the draft of manuscript.

#### FUNDING

This work was supported by grants from the National Key Research and Development Program (2018YFD0201202).


subsp. carotovorum causing soft rot. Biocatalysis Agric. Biotechnol. 9, 48–57. doi: 10.1016/j.bcab.2016.11.004


carotovorum subsp. atrosepticum) on Potatoes: a Laboratory Manual. Dundee: Scottish Crop Research Institute Occasional Publication.


**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 Cui, He, Munir, He, He, Li, Yang, Wang, Wu and He. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(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.

fmicb-10-01471 July 2, 2019 Time: 17:45 # 12

# Decoding Wheat Endosphere– Rhizosphere Microbiomes in *Rhizoctonia solani*–Infested Soils Challenged by *Streptomyces* Biocontrol Agents

*Ricardo Araujo1,2\*, Christopher Dunlap3, Steve Barnett1,4 and Christopher M.M. Franco1*

*1 Department of Medical Biotechnology, Flinders University, Adelaide, SA, Australia, 2 i3S, University of Porto, Porto, Portugal, 3 Crop Bioprotection Research, The United States Department of Agriculture, Peoria, IL, United States, 4 South Australian Research & Development Institute (SARDI), Adelaide, SA, Australia*

#### *Edited by:*

*Kalliope K. Papadopoulou, University of Thessaly, Greece*

#### *Reviewed by:*

*Sotirios Vasileiadis, University of Thessaly, Greece Rita Grosch, Leibniz Institute of Vegetable and Ornamental Crops, Germany*

> *\*Correspondence: Ricardo Araujo ricjparaujo@yahoo.com*

#### *Specialty section:*

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

*Received: 03 April 2019 Accepted: 24 July 2019 Published: 26 August 2019*

#### *Citation:*

*Araujo R, Dunlap C, Barnett S and Franco CMM (2019) Decoding Wheat Endosphere– Rhizosphere Microbiomes in Rhizoctonia solani*–*Infested Soils Challenged by Streptomyces Biocontrol Agents. Front. Plant Sci. 10:1038. doi: 10.3389/fpls.2019.01038*

The endosphere and the rhizosphere are pertinent milieus with microbial communities that perturb the agronomic traits of crop plants through beneficial or detrimental interactions. In this study, we challenged these communities by adding *Streptomyces* biocontrol strains to wheat seeds in soils with severe *Rhizoctonia solani* infestation. Wheat plants were grown in a glasshouse standardized system, and the bacterial and fungal microbiomes of 233 samples of wheat roots (endosphere) and rhizosphere soils were monitored for 20 weeks, from seed to mature plant stage. The results showed highly dynamic and diverse microbial communities that changed over time, with *Sphingomonas* bacteria and *Aspergillus*, *Dipodascus*, and *Trichoderma* fungi increasing over time. Application of biocontrol *Streptomyces* strains promoted plant growth and maturation of wheat heads and modulated the root microbiome, decreasing *Paenibacillus* and increasing other bacterial and fungal OTUs. The soils with the highest levels of *R. solani* had increased reads of *Thanatephorus* (*Rhizoctonia* anamorph) and increased root disease levels and increased *Balneimonas*, *Massilia*, *Pseudomonas*, and unclassified *Micrococcaceae*. As we enter the era of biologically sustainable agriculture, it may be possible to reduce and limit the effects of serious fungal infestations by promoting a beneficial microbiome through the application of biocontrol agents during different periods of plant development.

Keywords: 16S biodiversity, biocontrol agent, cereal microbiology, endophyte, ITS1 biodiversity, *Paenibacillus*, plant microbiome, *Streptomyces*

#### INTRODUCTION

Cereals represent the main carbohydrate food source in the world, particularly wheat that accounts for near 40% of the cereal supply worldwide (Charmet, 2011). For millennia, wheat has played a major role in the development of healthy societies and has supported economic and social stability (Würtenberger et al., 2006; Charmet, 2011). Intensive cereal cropping has shown an exponential increase in productivity and yield since the nineteenth century with the introduction of machinery and technology, but in the last few decades, the spread of soil infestations, soil degradation, and environmentally adverse conditions have been responsible for occasional decreases and instability in this cropping system (Lobell, 2009; Murray and Brennan, 2009). Understanding wheat crop system dynamics is critical, and several studies describe the rhizosphere as a pertinent milieu with microbial communities that perturb the agronomic traits through beneficial or detrimental interactions. The endosphere is the region inside the plant with microorganisms, namely, *Actinobacteria, Bacteroidetes,* and *Proteobacteria*  (Mendes et al., 2013; Turner et al., 2013, Schlaeppi and Bulgarelli, 2015; Vandenkoornhuyse et al., 2015), which influence root health and plant growth. From a sustainable perspective, microbiome management is important to predict the profitability of agricultural production systems, avoid soil degradation, understand plant responses to environmental challenges, and identify which microbes are more sensitive to each cropping practice (Hirsch and Mauchline, 2012; Mendes et al., 2013; Massart et al., 2015; Schlaeppi and Bulgarelli, 2015; van der Heijden and Hartmann, 2016).

*Rhizoctonia* root rot caused by *Rhizoctonia solani* AG8 (Kühn, teleomorph *Thanatephorus cucumeris*) is a major root infestation of cereals and other crops in dryland cropping systems, causing stunted seedlings and resulting in reduced access to water and nutrients by the plant (Paulitz et al., 2003; Schillinger and Paulitz, 2006; Anees et al., 2010; Jaffar et al., 2016). This can result in areas of high infestation levels with noticeable reduction of plant growth, or "bare patches," up to several square meters or up to 20% of the crop area (Schillinger and Paulitz, 2006; Anees et al., 2010). Infestation is increased in low rainfall areas resulting in low grain fill that exacerbates yield losses (Okubara et al., 2014; Sánchez-Cañizares et al., 2017). In Australia, *Rhizoctonia* infestation is most prevalent in the southern and western cropping regions, with registered and potential annual yield losses estimated at \$59 million and \$166 million, respectively (Murray and Brennan, 2009). *Rhizoctonia* root rot is difficult to control due to its wide host range (Cook et al., 2002), lack of commercially available resistant cereal cultivars, and increased prevalence in direct drill or minimal tillage practices (Schillinger and Paulitz, 2006). Current options for partial control include strategic tillage below seeds (Roget et al., 1996), removal of the green bridge with herbicide timing (Roget et al., 1987; Babiker et al., 2011), rotation with non-cereal crops (Angus et al., 2015), and in-furrow chemical fungicide treatments (Roget et al., 1987), and more recently, by using biocontrol-coated seeds (Franco et al., 2007; Barnett et al., 2019). Currently, the estimated cost of control measures is A\$106 million annually (Murray and Brennan, 2009).

*Rhizoctonia* root rot can be influenced by root-associated microorganisms, and biocontrol agent–coated seeds represent a biologically sustainable alternative with increasing potential in agriculture (Barnett et al., 2006; Dua and Sindhu, 2012; Yin et al., 2013; Mavrodi et al., 2014; Barnett et al., 2017). Endophytic *Streptomyces* species have been tested for biocontrol of phytopathogens in broad-acre cropping systems because of their ability to produce secondary metabolites, including antibiotics, and induce systemic resistance in the plant (Franco et al., 2016; Conn et al., 2008; Barnawal et al., 2017). Biocontrol agents can enhance root and shoot lengths, plant weight, higher tiller numbers, and/or induction of early flowering (Yang et al., 2012; Bokati et al., 2016; Franco et al., 2016; Araujo et al., 2017; Franco et al., 2017; Wemheuer et al., 2017). In addition, these *Actinobacteria*  produce spores for long-term viability and stability during storage (Emmert and Handelsman, 1999) and have the ability to produce siderophores (Wang et al., 2014), indole acetic acid (Khamna et al., 2009), and enzymes such as cellulases, chitinases, glucanases, and ACC deaminase (El-Tarabily, 2006; El-Tarabily, 2008). The enrichment of the root microbiome is a highly dynamic process that alters from the seed stage to the harvesting period. In order to understand and manage the microbiome, it is important to monitor the changes in microbial populations at each stage of plant growth (Turner et al., 2013; van der Heijden and Hartmann, 2016).

In the present study, we detail the dynamics of endosphere and rhizosphere microbiomes (both bacterial and fungal populations) in wheat plants for a period of 5 months. Wheat plants were grown in a standard glasshouse system in order to test the following hypotheses: 1) the endosphere and rhizosphere microbiomes of wheat crops change over time in a predictable manner, even in soils with severe *Rhizoctonia*  infestation; 2) the addition of biocontrol *Streptomyces* strains (e.g., F11, EN16, or F5) impacts endophytic and rhizosphere microbial populations; and 3) specific microorganisms existing in the plant roots and rhizosphere soils respond to high levels of *Rhizoctonia* infestation, especially during the first weeks.

#### MATERIALS AND METHODS

#### Biocontrol Cultures and Seed Coating

The strains F11, EN16, and F5 (all identified as *Streptomyces* sp.) (Franco et al., 2016) described as biocontrol agents (BCA) 1, 2, and 3, respectively, were used in this study. BCA1, BCA2, and BCA3 had previously reduced *Rhizoctonia* root rot in both pot bioassays and in the field and have demonstrated in vitro inhibition against *R. solani*  (Barnett et al., 2017; Barnett et al., 2019). The strains were identified by 16S rRNA gene sequencing and stored in culture collections of endophytic bacteria kept at Flinders University. A concentrated suspension of each strain was prepared in 0.3% (w/v) xanthan gum sticker solution and applied to 20g wheat seeds to a final count of ≈105 cfu/seed, as described in Barnett et al. (2017). Seeds were stored at room temperature for no more than 1 week before being used in pot bioassays. Seed cfu was assessed immediately and at 1, 2, and 7 d after application for confirmation of bacterial viability and concentration per seed (≈105 cfu/seed) as in Barnett et al. (2017).

#### Pot Bioassays in Glasshouse

Pot bioassays were prepared using field cropping soil collected at Waikerie, South Australia (34°14'32.91"S, 140° 5'44.31"E; details for soil features in **Supplement 1**). The bulk soil (150kg) was collected from the top 10cm of a 100-m2 section of the field, avoiding the collection of plants material larger than 2mm. This soil had a continuing *Rhizoctonia*

problem with background levels of R. solani AG8 of 492 pg DNA/g soil, determined by PreDictaB™ (SARDI, Urrbrae, SA, Australia; http://www.pir.sa.gov.au/research/services/ molecular\_diagnostics/predicta\_b), considered to carry a high risk of *Rhizoctonia* root rot (Ophel-Keller et al., 2008; Poole et al., 2015). There were no or low detectable levels of other root pathogens, such as *Pythium* sp., *Gaeumannomyces graminis* var. *tritici* (Ggt), *Fusarium pseudograminearum*, and *Fusarium culmorum*. Soil was air dried and sieved to <2 mm prior to use in pot bioassays. Soil chemical and physical soil properties were analyzed by CSBP Laboratories (https://www. csbp.com.au/CSBP-Lab, Perth, Western Australia, details in **Supplemental Material 1**). Water holding capacity was determined by the pressure plate method with a 1-m column (Marshall and Holmes, 1979) and the soil then adjusted to 60% water holding capacity for use. The pot experiments were prepared with the amount of soil per pot depending on the time of harvest: 600g for 4 weeks, 1,000g for 8 weeks, 1,125g for 12 weeks, 2,000g for 16 weeks, and 4,800g for 20 weeks. For half of the bioassays, three *R. solani* AG8 strain W19 infested millet seeds (https://www.keelangrainandfodder. com.au/) were placed in the center of each pot, for tests with higher levels of *R. solani* infestation, and the pots allowed to incubate for 1 week at 15°C in a controlled temperature room to allow for *R. solani* to colonize the soil. Then, five wheat seeds (*Triticum aestivum*), cultivar Yitpi, were planted per pot, covered with 50g of soil and 50g of coarse sand to reduce evaporation. Plants were grown in a 15°C room for the first 4 weeks and then were moved to a glasshouse under natural lightning and temperature conditions (mean temperature of 10 to 24°C during the autumn and early winter periods). The pots were watered twice a week to their original starting weight. Each pot used for testing time points, BCA treatments and soil infestation levels were run with four independent replicates arranged in a randomized complete block design. At 4, 8, 12, 16, and 20 weeks, plants were carefully washed and assessed for root rot disease on seminal and nodal roots using a 0–5 disease scale (0 = healthy roots with nodal and seminal roots, several secondary thin and long roots, no signs of disease; 5 = highly diseased and damaged nodal roots without seminal roots) (Roget, 1995). The number of plants per plot was also assessed at each time point. Pot bioassays were run from February to July 2016. Wheat plants were collected; the nodal and seminal roots were cut using sterilized scissors and washed at each time point to remove all the soil and organic matter; then, the surface of the roots was sterilized with sodium hypochlorite 2% (for 3 min) and ethanol 70% (for 3 min) and washed three times with sterile water. Rhizosphere soils were collected by recovering the small layer of soil on the surface of the roots; roots were initially collected, gently shaken to discard loosely adhering soil and the adjacent rhizosphere soil in the root surface collected by shaking the roots vigorously into a sterilized envelope (a sterile spatula was occasionally used on this procedure without damaging the roots—5g to 30g of rhizosphere soil was collected per independent pot/replicate).

#### DNA Extraction, Sequencing, and Bioinformatics

Root and soil samples were randomized (random numbers were attributed to the packages before storing to blind sample processing), stored at −80°C and processed for DNA extraction (groups of 16 random samples were processed simultaneously without any specific order). A fixed amount of 5 seeds, 1g of seminal roots (nodal roots were used for 4-week roots with serious disease and less than 1g seminal root material), or 2g of rhizosphere soil was used per replicate and subjected to a CTAB DNA extraction strategy (Zhang et al., 2010). The final DNA obtained was suspended in TE buffer. Polymerase chain reaction (PCR) was performed using KAPA HiFi PCR master mix (KAPA Biosystems Willington, MA, USA) using the following parameters: 95°C, 10 min, and 35 cycles of 95°C, 30 s; 58°C, 30 s; and 72°C, 60s. PCR primers for the bacterial community targeted the V3–V4 regions of the 16S rRNA genes with 341F and 806R primers (Muyzer et al., 1993; Caporaso et al., 2011), while for the fungal community that targeted the ITS1 region was targeted with ITS1F and ITS2 primers (Gardes and Bruns, 1993). The primers were incorporated into fusion primers for Illumina dual indexing and incorporation of Illumina adapters (Caporaso et al., 2012). After PCR, the amplicons were cleaned and normalized using a SequalPrep™ normalization plate (Thermo Fisher Inc., Waltham, MA, USA). The samples were pooled and the library quantified with a KAPA Library Quantification Kit (KAPA Biosystems Willington, MA, USA). The samples were sequenced using an Illumina MiSeq System with a MiSeq V3 2 x 300 bp sequencing kit. QIIME 1.9 (Caporaso et al., 2010) and USEARCH 9.2.64 (Edgar and Flyvbjerg, 2015) workflows were used for read merging, chimera removal (uchime2), operational taxonomic unit (OTU) picking, and taxonomic assigning (Ribosomal Database Project v11.4). Sequences with ≥97% identity defined the OTUs following sequence alignment in accordance to the model organism priors Escherichia coli; the clustering was produced in two passes of the swarm algorithm v2.1.6 (the first pass with an aggregation distance equal to 1 and the second pass with an aggregation distance equal to 3). Amplicon sequence variants (ASVs) were identified using a previously suggested R pipeline and DADA2 method (Callahan et al., 2016); Greengenes database (gg\_13\_8\_train\_set\_97) was used for the 16S rRNA amplicon classification and UNITE database (UNITE\_ public\_28.06.2017) for the ITS amplicon classification. The cutoff of more than or equal to 10 reads was considered for OTUs and ASVs included in this study.

#### Statistical Analysis

Plant and disease data from pot bioassays were analyzed as three-way factorial (five sampling times x two disease levels x four seed treatments) randomized complete block design with time fitted as a whole plot using GenStat version 14 (VSN International Ltd., Hemel Hempstead, England, UK). Fisher's least significant difference (lsd) was used to compare treatment means as the data was near normally distributed with homogeneity of variance between factors; *Rhizoctonia*-disease severity was analyzed by Kendall's coefficient of concordance (a non-parametric method). Data and statistical analyses were performed on Microsoft Office Excel 2013 (Microsoft Corporation, Santa Rosa, California, USA), STAMP 2.1.3 (Parks et al., 2014), PRIMER-6 (PRIMER-e, Auckland, New Zealand), and IBM SPSS Statistics 22 (IBM, New York, USA). Community diversity and distribution analyses were conducted by running analysis of similarities (ANOSIM) one-way analysis (calculating the resemblance and using similarity data type), non-metric multidimensional scaling (nMDS), clustering analysis (complete linkage), canonical analysis of principal components (CAP), homogeneity of dispersions (PermDISP; calculating the resemblance, similarity data type, using squared root of relative abundance (Legendre and Gallagher, 2001), Bray-Curtis similarities, and 999 permutations), and permutational multivariate analysis of variance (PERMANOVA) to reveal the effects of each factor (sample type, sampling time, biocontrol treatment or *Rhizoctonia* soil level) on the community composition (using squared root transformed data, Bray-Curtis similarities, and 4,999 permutations of residuals under a reduced model), and similarity percentages (SIMPER) analysis (using Bray-Curtis similarities and 90% cutoff for low contributions) (Anderson, 2001; Anderson et al., 2006; Clarke et al., 2008). Network analysis was conducted using the molecular ecological network analyses platform (http://ieg4.rccc.ou.edu/ MENA/) (Deng et al., 2012) to generate the networks, Cytoscape (Shannon et al., 2003) to visualize it, and cytoHubba (Chin et al., 2014) using maximal clique centrality (MCC) scores to select the top taxonomic groups with links in roots and rhizosphere soil samples. Random matrix theory (RMT)–based modeling was used for network analysis as this approach is powerful in delineating phylogenetic molecular ecological networks in microbial communities (following some steps microbial sequence collection, data standardization, Pearson correlation estimation, adjacency matrix determination by an RMT-based approach, network characterization, and module detection) and building an adjacency matrix that represents interactions in a network graph (Zhou et al., 2011). The reads in each sample were converted into percentage values according to the total number of OTUs or ASVs in the sample to eliminate the effect of the final number of reads (Zaura et al., 2009). These values were then transformed using the Hellinger approach (Legendre and Gallagher, 2001) to reduce the effects of overestimation among the most common taxa and the values compared on dissimilarity matrices that could be used for multiple population analyses. Post hoc analyses were done for multiple groups using one-way analysis of variance (ANOVA), Tukey-Kramer (0.95), and Eta-squared for effect size, while two-group analysis used Welch's t-test (two-sided, Welch's inverted for confidence interval method). The other data were compared at a significance level of 0.05 by the ANOVA test using the Bonferroni correction and by Student's t test (when the population could be assumed to be normally distributed) or Wilcoxon signed-rank test (when the population could not be assumed to be normally distributed) for paired samples.

#### RESULTS

#### Microbial Diversity and the Effect of Biocontrol Strains in Wheat Plants

Differences were observed on the wheat plants considering the studied factors: 1) *R. solani* level in the soil, 2) treatment with biocontrol strains (mainly F11 and EN16), and 3) sampling time. The agent F5 did not affect the plant growth, and the F5 plants were similar to the control wheat plants (in both soils with low and high *R. solani* levels). Wheat plants were obtained from control- and biocontrol-coated seeds grown in the glasshouse with the biocontrol-treated plants (F11 and EN16) having a higher biomass at later stages, earlier formation of wheat heads, and lower root disease indexes (more evident with EN16-coated seeds) (**Supplemental Material 2**). The roots and rhizosphere soils of each of these plants were then used for microbiome studies to compare untreated control versus biocontrol-treated plants (with F11, EN16, or F5 strains) in the presence of low and high levels of *R. solani* infestation. A total of 1,216,983 bacterial and 793,412 high-quality fungal sequences were organized into 6,880 bacterial and 861 fungal OTUs, or 16,248 bacterial and 969 fungal ASVs (details of ASVs in **Supplemental Material 3**). These sequences consisted of 628 bacterial and 204 fungal taxa (assigned at the genus or higher taxonomic levels) from the analyzed 233 samples. **Figure 1** shows the bacteria and fungi found in seed and root samples across the entire study (a set of 137 taxonomic groups were found in more than 75% of the samples, but only 13% of these taxa showed ASVs transversal to most of the collected samples); **Supplementary Material 4** shows the most frequent bacterial and fungal genera found in rhizosphere soils and wheat roots from the 20-week crop cycle. A set of 16 genera of bacteria and 7 fungi were found in all seed and root samples (**Figure 1**), being *Agrobacterium, Pseudomonas, Streptomyces,* and *Fusarium* the genera with the highest relative abundance in wheat roots or seeds.

The Shannon diversity indices for microbial communities of root and rhizosphere are shown in **Table 1**. This index was systematically higher in the rhizosphere soils compared to root or seed samples. The diversity was slightly decreased from the initial seeds or soils to 4-week sampled roots or rhizosphere soils, but then the diversity index increased in the following weeks. **Table 2** shows the statistical differences found in the microbial populations considering the multiple factors; ANOSIM showed a clear distinction between seeds, roots, and rhizosphere soil samples (sample types). Then, it also revealed the sampling time as the strongest factor (P < 0.001) responsible for the richness and composition of the microbial communities found in roots and rhizosphere samples, in comparison with the other factors biocontrol treatments and *Rhizoctonia* soil levels (**Tables 2** and **3**). *Streptomyces* biocontrol agents tested in this study showed a significant effect on the root microbial populations resulting in distinct microbiomes in the endosphere and rhizosphere (**Table 3** and **Supplemental Materials 5** and **6**); the effect of *R. solani* levels on root and rhizosphere microbiome was low


FIGURE 1 | (A) Clustering analysis of all samples included in the study; (B) relative abundance of the core bacteria and fungi found in seeds and roots during the entire study (in the table the taxa with at least 2% in any sampling time; the complete list of core bacteria and fungi classified at genus level is below the table and includes the taxa with low relative abundance—\* marks the taxonomic groups with core ASVs found in all sampling time points); the data for control and biocontrol treatments was pooled for these analyses. Details of *post hoc* plots can be seen in Supplemental Materials 6, 7, and 10.

**54**

TABLE 1 | Shannon diversity index and Margalef richness (at genus or higher taxonomic level) for wheat root and rhizosphere soil samples; average (minimum and maximum values).


*\*Values for seeds (root column) or initial soil (rhizosphere column).*

TABLE 2 | P values for analysis of similarities (ANOSIM) in roots and rhizosphere soil samples. The results represent the same samples according to sampling time, the biocontrol treatment, and finally according to pathogen infection level groupings.


*Bold mean the P value is significant (P<0.05). Analyses done in Primer v6 using squared root transformation of the data and resemblance Bray–Curtis similarity (with dummy variable).*

TABLE 3 | P values for analysis of similarities (ANOSIM) for each group of samples (values per week considering the effect of biocontrol agents and different levels of *Rhizoctonia* disease).


*Bold mean the P value is significant (P<0.05).*

and only significantly different by ANOSIM analysis for the root microbiome (**Table 3**). High levels of *R. solani* showed a significant alteration of the rhizosphere soil communities from 4 to 12 weeks, not in the subsequent weeks, suggesting some rhizosphere microorganism may respond to root disease.

#### Endosphere and Rhizosphere Microbiomes Over a 5-Month Period

A succession of microorganisms was observed in the wheat roots from 4 weeks to the mature plant after 20 weeks (**Figure 1** and **Supplemental Material 3**), with bacterial biodiversity being more prominent in the initial stages and fungal biodiversity increasing after the 12th week. Some bacterial and fungal OTUs were maintained in the plant root for several weeks, while others were only identified occasionally (**Figure 1** and **Supplemental Materials 3** and **4**). Although *Streptomyces* and *Paenibacillus* were predominantly found in the roots at week 4 (**Figure 1** and **Supplemental Material 3**), the OTUs of these bacterial genera were not mainly found in the same set of roots and rhizosphere samples. *Streptomyces* dominated samples obtained from F11- and EN16-treated plants, while Paenibacillus were abundant on control and F5-treated plants (**Figures 2A**, **B** and **Supplemental Material 5**). Similarity percentage (SIMPER) analyses also showed an abundance of *Arthrobacter, Bacillus,* and *Paenibacillus* OTUs in the control roots soon after 4 weeks. Other OTUs increased in F11- and EN16-treated roots, mainly fungal OTUs, were classified as *Exophiala, Phaeoacremonium,* and unclassified *Xylariaceae* (see post hoc plots in **Supplement 6**). The comparison of the OTUs in the control roots versus the biocontrol-treated roots showed a similarity in less than 10% at single time points (specific weeks), increasing to nearly 20% when the total 5-month period was considered (data not shown). By comparing the taxonomic groups (at genus level) of the OTUs for the control versus biocontrol-treated roots at the same time point (week), the similarity ranged from 43 to 88% (**Table 4**); the similarity was maximum at the 8th week of the wheat growth cycle for

reads (within fungal population) ± SEM in root samples (\*P < 0.05) (C).



both bacteria and fungi found in roots. The relative abundance of the major and most common taxa found on wheat roots also showed differences over time (**Figure 1** and **Supplemental Information 7**). Among the rhizosphere soil samples, the taxonomic similarity was consistent over time (around 70% in all samples), and multiple ASVs were found in common over the weeks (**Supplement 3**).

Distinct dynamics were found among bacterial and fungal OTUs: 1) *Pseudomonas* OTUs were high in the 4- and 8-week sampled roots and reduced after the 12th week (**Figure 1** and **Supplemental Information 7**); 2) *Sphingomonas* OTUs were particularly high after the 12th and 20th weeks (**Figure 1**); 3) Podospora OTUs increased after the 8th week (**Figure 1** and **Supplemental Information 7**); 4) *Bacillus*, *Curvularia*, and Rubrobacter OTUs reduced over time in the rhizosphere soils (**Supplemental Material 8**); 5) *Devosia* was abundant at 8th and 12th weeks (**Figure 1**); and 6) *Aspergillus*, *Dipodascus*, *Rhodoplanes*, and *Trichoderma* OTUs (P < 0.05 using post hoc two-group analysis) were more abundant in later periods (**Supplemental Material 8**). Notably, molecular ecological network analyses showed Streptomyces as a lateral genus in the population analyses, directly interacting with Ralstonia, and barely interacting with other groups (**Figure 3A**); *Afifella*, *Luteibacter*, *Methylibium*, and *Shinella* were found in the center of the bacterial network analyses with the major number of links within the microbiome (**Supplemental Material 9**). Additionally, *Thanatephorus* (anamorph of *Rhizoctonia*) was found in the core of the network interacting with multiple bacteria and fungi (see **Figure 3B** for details).

#### Biocontrol Strains and Other Taxa in Rhizoctonia Conducive Soils

**Figure 2C** shows the differences observed in reads of *Thanatephorus* OTUs (anamorph of *Rhizoctonia*) during the study. The highest values were seen at 4 weeks followed by a decrease of *Thanatephorus* OTUs in roots and rhizosphere soils in subsequent weeks. These values correlated with the disease incidence rate measured in the plant roots (**Supplemental Material 2**) showing higher root disease index at 4 weeks and decreasing in the weeks after. The relative abundance of *Thanatephorus* reads detected in F11- and EN16-treated roots at 4 weeks was significantly lower (P < 0.05) compared with the values for control roots; in the remaining sampling times, the relative abundance of *Thanatephorus* reads was less than 4%, and the differences were not significant (P > 0.05). Besides this direct effect observed by *Streptomyces* on the roots of plants exposed to *R. solani*–induced infestation, SIMPER, and ANOVA (STAMP) comparative analyses agreed that some other OTUs, classified as *Balneimonas* (*Bradyrhizobiaceae*), *Massilia (Oxalobacteraceae), Pseudomonas,* and unclassified *Micrococcaceae, Rhizophlyctidaceae,* and *Gemmatimonadaceae* were particularly dominant in the soils with highest levels of *R. solani* (**Supplemental Materials 10** and **11**). OTUs of *Bradyrhizobiaceae* (e.g., *Balneimonas*) and *Micrococcaceae* were mainly present in the rhizosphere soils, while some OTUs of *Pseudomonas* increased with higher levels of root disease.

### DISCUSSION

The microbiome of root endosphere and rhizosphere soils was monitored through the growth cycle (5 months) of the wheat crop growing in *Rhizoctonia*-infested soils with and without the influence of *Streptomyces* biocontrol strains tested at each step of plant development. This study confirmed that *Streptomyces* isolates (F11 and EN16) could modulate endosphere and rhizosphere microbiomes resulting in increased plant growth, reduced root disease, and increased number of wheat heads over the weeks. The third strain, *Streptomyces* F5, was less effective on plant physiology and produced a distinctively different microbiome compared with the first two endophytes. The addition of Streptomyces strains F11 and EN16 affected mostly *Paenibacillus* populations, commonly found in seeds (Yang et al., 2017), reducing their relative abundance drastically over the first weeks. In addition, the abundance of the OTUs of *Streptomyces* and the fungal OTUs classified as *Exophiala* (also found in seeds), *Phaeoacremonium*, and unclassified *Xylariaceae* increased during the same period. It is known that the enrichment of endosphere and rhizosphere microbiomes benefits the wheat plants (Yang et al., 2012; Bulgarelli et al., 2013; Reinhold-Hurek et al., 2015), and we have demonstrated that individual strains can shift the microbiome (not affecting the richness or evenness) and benefit the plants. It is possible that the efficacy of *Streptomyces* strains reported in glasshouse systems against *Rhizoctonia* in wheat (Goudjal et al., 2013) and tomato (Sabaratnam and Traquair, 2002) can produce predictable changes in the endosphere microbiome. The inoculum concentration of ≈105 cfu/seed has shown benefits for wheat plants in field trials (Barnett et al., 2017). A higher inoculum in the initial seeds might result in 3–7% rise of the endophyte population in wheat roots (Conn and Franco, 2004). Increasing numbers of reports suggest that a high biodiversity at the endosphere and rhizosphere levels may give extra protective "tools" to plants to respond to environmental constraints and infestation stresses (Ratnadass et al., 2012; Reinhold-Hurek et al., 2015; van der Heijden and Hartmann, 2016; Barnawal et al., 2017).

It is now known that the OTUs and taxonomic groups detected at each stage of wheat growth change over time (Lauber et al., 2013; Reinhold-Hurek et al., 2015; Rascovan et al., 2016; Mahoney et al., 2017). Such patterns were also observed in this study with *R. solani*–infested soils, independently of the disease levels. Bacterial OTUs tend to be dominant in the initial stages of plant growth, colonize the wheat roots, and reach successive peaks of biodiversity during the first 12 weeks. While some bacterial OTUs are dominant during the initial weeks, other bacteria, such as *Sphingomonas*, were found mostly in the later stages of wheat root maturation. *Sphingomonas* are usually found in multiple parts (roots, leafs, flowers) of mature plants and also have a plant protective role against infestations (Kim et al., 1998; Khan et al., 2014). These bacteria are common plant endophytes and are known to benefit plants by producing phytohormones and support plant maturation processes (Khan et al., 2014; Asaf et al., 2017). Later, fungi gain more relevance

and intensively colonize wheat roots; from the 12th to the 16th week, fungal biodiversity increased greatly, and some genera, such as *Aspergillus, Dipodascus,* and *Trichoderma*, described as protective for wheat and other plants (Nicolaisen et al., 2014; Bokati et al., 2016; Hertz et al., 2016; Barnett et al., 2017; Liu et al., 2018) were particularly abundant. In fact, it was clear that the microbial population changed over time, and distinct microbiomes can actually be considered for each stage of the wheat crop (**Supplement 3**) with only a fraction of OTUs/ASVs persisting during the entire period of this study (**Figure 1**).

Microbial communities were primarily affected by the sampling time of wheat roots, similar to findings for *Arabidopsis thaliana*  (Bulgarelli et al., 2012), but at a second level, it was possible to observe some differences in the endosphere and rhizosphere microbiomes in the presence of effective biocontrol agents. Notably, *Streptomyces* did not show the highest degree of connectivity in the network analyses, being the positions with more interactions taken by other taxonomic groups such as *Afifella, Luteibacter, Methylibium, Shinella,* and *Chitinophaga*, that are frequent colonizers of the rhizosphere and previously described as relevant endophytes for nuts, sugarcane, or potato plants (Manter et al., 2010; Chakraborty et al., 2016; Taulé et al., 2016; Su et al., 2017; Lo et al., 2018). These may represent keystone microbes in the rhizosphere microbiome (Berry and Widder, 2014). These bacteria networked to several others certainly represent interesting targets for microbial interaction studies, but further analysis of the soils with increased levels of *R. solani* showed similar or lower relative abundance of these bacteria compared with the soils with low levels of *Rhizoctonia*. Instead, a higher contribution of other lateral taxonomic groups (*Streptomyces* or *Paenibacillus*) was found to interfere with the wheat root endosphere communities from *Rhizoctonia*-infested soils.

Soils with low and high levels of *R. solani* were challenged by biocontrol strains and its effects on endosphere and rhizosphere microbiomes compared. The number of reads of *Thanatephorus*  (anamorph of *R. solani*) was not constant throughout the study, being peaks of reads collected in the initial weeks changing similarly to the values of disease severity seen in the wheat roots—higher values (roots were rated 4 and 5) in the first weeks and then decreasing to much lower levels (roots were rated 1 and 2) after the 12th week. The interaction between *R. solani* within the endosphere and rhizosphere microbiomes was complex with interactions between multiple taxonomic groups, some of these representing well-known endophytes or plant benefiting microbes (e.g., *Trichoderma, Gibberella,* and *Burkholderia*) with mechanisms of action against soil infestations (Panea et al., 2013; Gouda et al., 2016) and others being central taxa in the network analyses (e.g., *Luteibacter*). Nevertheless, the presence of F11 and EN16 strains on roots resulted in lower number of reads of *Thanatephorus* at 4 weeks, in agreement with lower disease severity reported in the plants. This suggests the percentage of each OTUs in the microbiome profiles may provide a semiquantitative perception of the infestation levels in the soils. Besides the impact of biocontrol *Streptomyces* on *Rhizoctonia*  root rot, some *Pseudomonas* OTUs in wheat roots and OTUs of *Bradyrhizobiaceae, Micrococcaceae,* and *Oxalobacteraceae* (e.g., *Massilia*) in rhizosphere soils were positively responsive to the higher levels of *R. solani* inoculation. Some strains of these groups may also play a protective role against this specific infestation, and occasional isolates of *Massilia* were described with such properties (Yin et al., 2013). Antifungal activity was reported for multiple members of *Oxalobacteraceae*, including *Collimonas* spp., against the ectomycorrhizal fungus *Laccaria bicolor*, also affecting the growth and hyphae branching of the fungus and potentially modulating the fungal gene in stress response (Leveau et al., 2010). It remains to be clarified if the "manipulation" of all these taxonomic groups simultaneously in standardized experiments can in fact result in a much higher physiological effect in wheat plants than the ones we observed with *Streptomyces* strains alone. The proliferation of these microbes at the right stage of the wheat crop, both in the plant and rhizosphere soil, may be a highly efficient barrier against the spread of fungal infestations, keeping the plant and the biodiversity of the soils stable. The addition of selective treatments to seeds, either as inoculants or materials to increase specific taxa, may add even more productivity to crops in coming years (Li et al., 2018), being important to integrate and complement these approaches.

In conclusion, the microbial populations of both endosphere and rhizosphere soils experience major changes from the early stages to the flowering phase with distinct groups of microbes dominating each stage. The addition of effective biocontrol *Streptomyces* strains impact the microbiome as these strains take over the dominant place of other bacteria, e.g., *Paenibacillus*, in the wheat root. *Paenibacillus* had higher relative abundance within the endophytic communities of some fruits, such as apple (Liu et al., 2018), suggesting that some observations of this study may be extended to these plants. Disease variation levels in the soil may be monitored by routine comparison of the endosphere microbiome profiles, which may also reveal OTUs directly responding to major levels of *R. solani* in the soils. We are entering the era of biological sustainability strategies for consolidation and promotion of plant productivity by acting at multiple levels to reduce and limit the consequences of serious infestations. By promoting, monitoring, and controlling the microbiome and the biocontrol agents within the plant at each period of development, we may effectively achieve exceptional nutritional and environmental standards.

### CONTRIBUTION TO THE FIELD STATEMENT

*Rhizoctonia* root rot is a major root infestation affecting cereals and other crops that cases considerable damage to the plant reducing access to water and nutrients. The estimated cost of control measures is of several hundreds of million dollars annually in Australia, being the options for partial control restricted to tillage, use of herbicides and other chemicals and rotation with non-cereal crops. The use of biocontrol strains, mainly *Streptomyces*, as seed coats promote wheat growth and plant maturation resulting in changes in the endosphere and rhizosphere soil microbiomes. These changes may impact Rhizoctonia infestation at root level and limit the damages. In this study, specific bacterial and fungal OTUs responded to crop age, addition of biocontrol strains (the effects of three strains were compared) and increased levels of *Rhizoctonia solani* infestation in the soils. Some OTUs of *Balneimonas, Massilia, Pseudomonas* and unclassified Micrococcaceae responded essentially in the soils representing potential protectors against *Rhizoctonia* infestation advance, without damaging the soils or affecting its bacterial and fungal biodiversity. The application of biologically sustainable approaches in agriculture may limit the damaging effects of serious fungal infestations and preserve high levels of microbial biodiversity in the soils.

#### DATA AVAILABILITY

The dataset supporting the conclusions of this article is available in the NCBI repository, under the ID SUB4129508, biosample SUB4129509 and bioproject PRJNA471385 (SRA study SRP149964; accessible with the link ftp://ftp-trace.ncbi.nlm.nih. gov/sra/review/SRP149964\_20190318\_110809\_27e795eb0f314e df0479737480ab0f2a).

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

All authors participated in the design, conception and implementation of the study, data analyses, manuscript writing and elaboration.

#### FUNDING

RA was supported by an Endeavour Postdoctoral Fellowship. This study was financed by Grains Research and Development Corporation (GRDC) project no. UF00008.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.01038/ full#supplementary-material


and environmental parameters. *Phytopathol.* 105, 1069–1079. doi: 10.1094/ PHYTO-07-14-0203-R


**Disclaimer:** The biocontrol agents included in this study are protected under the Australian Provisional Application No. 2017901523, filed April 27, 2017, and U.S. Provisional Application No. 62/568,763, filed October 5, 2017. The strains are stored in Flinders University and the South Australian Research and Development Institute (SARDI) and are available for research studies. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture. USDA is an equal opportunity provider and employer.

**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 Araujo, Dunlap, Barnett and Franco. 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.*

# Antifungal, Plant Growth-Promoting, and Genomic Properties of an Endophytic Actinobacterium Streptomyces sp. NEAU-S7GS2

Dongli Liu<sup>1</sup>† , Rui Yan<sup>1</sup>† , Yansong Fu<sup>1</sup> , Xiangjing Wang<sup>1</sup> , Ji Zhang<sup>1</sup> \* and Wensheng Xiang1,2 \*

<sup>1</sup> Heilongjiang Provinical Key Laboratory of Agricultural Microbiology, Northeast Agricultural University, Harbin, China, <sup>2</sup> State Key Laboratory for Biology of Plant Diseases and Insect Pests, Institute of Plant Protection, Chinese Academy of Agricultural Sciences, Beijing, China

#### Edited by:

Donald L. Hopkins, University of Florida, United States

#### Reviewed by:

Artemio Mendoza-Mendoza, Lincoln University, New Zealand Tanya Arseneault, Agriculture and Agri-Food Canada (AAFC), Canada

#### \*Correspondence:

Ji Zhang zhangji@neau.edu.cn Wensheng Xiang xiangwensheng@neau.edu.cn †These authors have contributed equally to this work

#### Specialty section:

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

Received: 12 May 2019 Accepted: 22 August 2019 Published: 10 September 2019

#### Citation:

Liu D, Yan R, Fu Y, Wang X, Zhang J and Xiang W (2019) Antifungal, Plant Growth-Promoting, and Genomic Properties of an Endophytic Actinobacterium Streptomyces sp. NEAU-S7GS2. Front. Microbiol. 10:2077. doi: 10.3389/fmicb.2019.02077 Diseases caused by Sclerotinia sclerotiorum have caused severe losses of many economically important crops worldwide. Due to the long-term persistence of sclerotia in soil and the production of air-borne ascospores, synthetic fungicides play limited roles in controlling the diseases. The application of antagonistic microorganisms can effectively reduce the number of sclerotia and eventually eradicate S. sclerotiorum from soil, and therefore considerable interest has been focused on biological control. Streptomyces sp. NEAU-S7GS2 was isolated from the root of Glycine max and its rhizosphere soil. It showed significant inhibitory activity against the mycelial growth of S. sclerotiorum (99.1%) and completely inhibited sclerotia germination. Compared to the control, in the pot experiment the application of NEAU-S7GS2 not only demonstrated excellent potential to control sclerotinia stem rot of soybean with 77 and 38% decrease in disease incidence and disease index, respectively, but could promote the growth of soybean. The light microscopy and scanning electron microscopy showed that co-culture of NEAU-S7GS2 with S. sclerotiorum on potato dextrose agar could lead to contorted and fragmented mycelia of S. sclerotiorum, which was associated with the secretion of hydrolytic glucanase and cellulase and the production of active secondary metabolites by NEAU-S7GS2. The plant growth promoting activity of NEAU-S7GS2 was related to the solubilization of inorganic phosphate, and production of 1-aminocyclopropane-1-carboxylate (ACC) deaminase and indole acetic acid (IAA). To further explore the plant growth promoting and antifungal mechanisms, the complete genome of strain NEAU-S7GS2 was sequenced. Several genes associated with ammonia assimilation, phosphate solubilization and IAA synthesis, together with genes encoding ACC deaminase, glucanase and α-amylase, were identified. AntiSMASH analysis led to the identification of four gene clusters responsible for the biosynthesis of siderophores including desferrioxamine B and enterobactin. Moreover, the biosynthetic gene clusters of lydicamycins, phenazines, and a glycosylated polyol macrolide showing 88% gene similarity to PM100117/PM100118 were identified. These results suggested that strain NEAU-S7GS2 may be a potential biocontrol agent and biofertilizer used in agriculture.

Keywords: antifungal activity, plant growth promotion, Streptomyces sp. NEAU-S7GS2, Sclerotinia sclerotiorum, genome

## INTRODUCTION

fmicb-10-02077 September 6, 2019 Time: 18:0 # 2

Stem rot caused by the phytopathogenic fungus Sclerotinia sclerotiorum is a very detrimental disease in many economically important crops including soybean, rapeseed oil, and sunflower (Wu et al., 2013; Arfaoui et al., 2018; Sabaté et al., 2018). S. sclerotiorum can infect more than 400 plant species belonging to 75 families and lead to typical stem rot symptoms, such as soft watery lesions or areas of light brown discoloration on leaves, main stems and branches (Kamal et al., 2015; Zhang et al., 2018). S. sclerotiorum commonly spreads by spores and in the forms of sclerotia that can infect stems, leaves and flowers, and eventually spread to adjacent plants and lead to devastating economic losses. It produces a long-lived melanized resting structure named as sclerotia, which can reside in soil for several years and germinate to form infectious hyphae or apothecia that release millions of airborne ascospores under appropriate environmental conditions (Smoliñska and Kowalska, 2018). Due to the unique life cycle pattern, it is difficult to control the disease caused by S. sclerotiorum. Although various disease management strategies such as crop rotation, fungicide treatments and the use of resistant varieties have been employed, none of these strategies can completely control the disease (Arfaoui et al., 2018). With the increasing concerns on the environmental pollution, food safety and chemical pesticide resistance, the use of beneficial microorganisms is considered as an environmentally friendly alternative way to combat crop disease.

Many beneficial microorganisms have been investigated to control sclerotinia diseases, and some of them are developed as commercial products for biological control, such as Coniothyrium minitans CON/M/91-08 (Contans <sup>R</sup> WG), Trichoderma harzianum T-22 (Plantshield <sup>R</sup> HC), Gliocladium virens GL-21 (SoilGard <sup>R</sup> ), Bacillus subtilis QST 713 (Serenade <sup>R</sup> MAX), and Streptomyces lydicus WYEC 108 (Actinovate <sup>R</sup> AG) (Budge and Whipps, 2001; Chitrampalam et al., 2008; Zeng et al., 2012). Streptomyces species are a diverse group of Gram-positive, filamentous, and spore-producing bacteria with relatively large genomes of approximately 8 to 9 Mbp in size and a high G + C content of more than 70%. They are well-known for the ability to produce various active compounds with agricultural applications (Antoraz et al., 2015). Additionally, Streptomyces strains also have beneficial effects on plant growth by providing nutrients from degradation of complex biological polymers in soil or producing plant growth factors (Khamna et al., 2009). Meanwhile, some species of Streptomyces exhibit biological control potential against phytopathogens, particularly phytopathogenic fungi such as Verticillium dahliae, Fusarium oxysporum, Pythium ultimum, Phytophthora sp., and Scleritium rolfsii through different mechanisms including the production of antibiotics, hyperparasitism, and induction of plant resistance response (Errakhi et al., 2009; Bubici et al., 2013; Tamreihao et al., 2016). Many Streptomyces species have been successfully developed as commercial biofungicides such as Actinovate based on S. lydicus, Mycostop based on Streptomyces griseoviridis and Rhizovit based on Streptomyces sp. DSMZ 12424 (Minuto et al., 2006; Berg et al., 2010; Zeng et al., 2012). Therefore, Streptomyces species represent an important resource of biofungicides or biofertilizers for agricultural use. In this study, the strain Streptomyces sp. NEAU-S7GS2 was isolated with plant growth promoting activity and broad antifungal activity from the root of Glycine max and its rhizosphere soil. It showed strong antagonistic activity against S. sclerotiorum and good biocontrol potential against sclerotinia stem rot on soybean. In order to fully understand the antifungal and plant growth promoting mechanisms, the genome of strain NEAU-S7GS2 was sequenced and analyzed.

### MATERIALS AND METHODS

#### Sample Collection and Bacterial Isolation

Soybean (Glycine max) along with the rhizosphere soil were collected from a saline-alkaline field located in Durbert Mongolian Autonomous County, Daqing, China (46◦ 30<sup>0</sup> N 124◦ 10<sup>0</sup> E). The root was processed as described by Bonaldi et al. (2015), except that the suspension of root tissues was spread on HV agar medium supplemented with cycloheximide (50 mg/l) and nalidixic acid (20 mg/l). Briefly, the root was surface sterilized with propylene oxide for 1 h, and then washed in washing solution containing sterilized 0.9% NaCl and 0.02% Silwet L-77. Subsequently, the root was finely homogenized in 3 ml washing solution and the suspension was plated in serial dilutions on HV agar medium. The soil sample (5 g) was suspended in distilled water (2 ml) followed by an ultrasonic treatment (160 W) for 3 min. After the addition of distilled water (43 ml), 3% yeast extract (w/v, 1 ml) and 2.5% sodium dodecyl sulfate (w/v, 1 ml), the soil suspension was incubated at 28◦C and 250 rpm on a rotary shaker for 20 min. Then, 200 µl of the suspension was spread on CMKA medium (Pan et al., 2016) supplemented with cycloheximide (50 mg/l) and nalidixic acid (20 mg/l). After 21 days of aerobic incubation at 28◦C, colonies were transferred and purified on International Streptomyces Project (ISP) 3 medium (Shirling and Gottlieb, 1966), incubated at 28◦C for 7 to 14 days and maintained as glycerol suspensions (20%, v/v) at −80◦C.

#### Strain Characterization

The utilization of sole carbon and nitrogen sources was examined as described previously (Williams et al., 1983; Kämpfer et al., 1991). Growth at different temperatures (5 to 50◦C, steps of 5 ◦C) was determined on ISP2 (Shirling and Gottlieb, 1966) medium after incubation for 14 days. Growth tests for pH range (3.0 to 12.0 in 1.0 pH unit intervals) were performed in ISP2 liquid medium at 28◦C for 7 days on a rotary shaker. The buffer systems were employed as described by Pan et al. (2016). The salinity resistance of strain NEAU-S7GS2 was tested in ISP2 liquid medium with the addition of 0 to 20% NaCl. The genomic DNA of NEAU-S7GS2 was extracted from cells grown in GY medium (Pan et al., 2016) at 28◦C for 7 days on a rotary shaker. Amplification of the 16S rRNA gene was performed by using the universal bacterial primers 27F and 1541R under conditions described previously (Pan et al., 2017). The PCR product was purified and cloned into the vector pMD19-T (Takara) and sequenced by using an Applied Biosystems DNA sequencer (model 3730XL)

and software provided by the manufacturer. Identification of phylogenetic neighbors and calculation of pairwise 16S rRNA gene sequence similarity was achieved using EzBioCloud (Yoon et al., 2017). The almost full-length 16S rRNA gene sequence of strain NEAU-S7GS2 was obtained and aligned with multiple sequences obtained from the GenBank/EMBL/DDBJ databases using CLUSTAL X 1.83 software. Phylogenetic trees were generated with the neighbor-joining algorithms using Molecular Evolutionary Genetics Analysis (MEGA) software version 7.0. The stability of the clades in the trees was appraised using a bootstrap value with 1000 repeats. All positions containing gaps and missing data were eliminated from the dataset (complete deletion option). The root position of the trees was inferred by using Nocardia carnea DSM 43397<sup>T</sup> (GenBank Accession No. X80607) as an outgroup.

#### Antagonistic Effects of NEAU-S7GS2 on Phytopathogenic Fungi

The antifungal activity of NEAU-S7GS2 against phytopathogenic fungi S. sclerotiorum, Exserohilum turcicum, Corynespora cassiicola, and Rhizoctonia solani was evaluated using Petri dish assays. Briefly, a colony of NEAU-S7GS2 was placed 1 to 3 cm from the margin of the potato dextrose agar (PDA) plates (8 cm in diameter) and incubated at 28◦C for 2 days. Then the fungal disk (5 mm) was taken from the test pathogen and also placed in a similar manner but directly opposite the NEAU-S7GS2. Plates inoculated with the tested pathogen in the absence of NEAU-S7GS2 served as negative controls. The plates were incubated at 20◦C until the control mycelium of pathogen reached the edge of the plates. Then, the antagonistic belt (inhibition zone) was recorded by measuring the distance between the edge of the fungal mycelium and strain NEAU-S7GS2. The percentage inhibition (I%) of radial growth was calculated using the formula: I% = [(r1-r2)/r1] × 100, where r1 was the radius of fungal mycelial growth in the control, and r2 was the radius of fungal mycelial growth that occurred toward NEAU-S7GS2. The assay was replicated three times.

### Evalution of Antagonism via Production of Volatile and Diffusible Materials

The inhibition of S. sclerotiorum mycelial growth via production of volatile compounds by NEAU-S7GS2 was evaluated using a divided plate method to prevent direct contact between the assayed organisms (Kamal et al., 2015). Petri plates split into two compartments were used, and each compartment was filled with PDA medium. Strain NEAU-S7GS2 was placed on one side and the tested fungus were placed on the opposite side. The plate was immediately wrapped in Parafilm M to trap the volatiles. The tightly sealed plates were incubated at 20◦C. A Petri plate containing only S. sclerotiorum was used as a control. Strain NEAU-S7GS2 was grown in 50 ml of liquid medium (5% glycerol, 2.5% corn flour and 0.5% yeast extract, pH 7.2) in a 250 ml Erlenmeyer flask and incubated at 28◦C for 7 days on a rotary shaker at 250 rpm. The culture was centrifuged at 8000 g for 10 min, and the supernatant of the fermentation broth was filtered with a 0.22 µm membrane filter. The mycelium was extracted with 50 ml of methanol, and the extract was also filtered with a 0.22 µm membrane filter. The antifungal activity of the supernatant and methanol extract was evaluated using the filter paper disk diffusion assay described previously (Fu et al., 2019b). Briefly, sterile filter paper disks (0.5 cm diameter), which were placed on the direct opposite of S. sclerotiorum disk on PDA plates, were saturated with the cellfree supernatant (20 µl) or methanol extract (20 µl). The filter paper disks saturated with supernatant of fermentation medium or methanol were used as negative controls. All the experiments were conducted in triplicate.

#### Sclerotia Germination Test

Mycelial plugs of S. sclerotiorum were inoculated on PDA plates, and then incubated at 20◦C. After 2 weeks, sclerotia of S. sclerotiorum formed in PDA plates were harvested. The sclerotia were sterilized by treating with 70% ethanol followed by 0.5% sodium hypochlorite (NaClO) for 5 min. The surface sterilized sclerotia were washed three times with sterile distilled water to remove all traces of NaClO and blotted dry on sterilized paper. A colony of NEAU-S7GS2 was placed in the center of the PDA plate and three sclerotia were placed around NEAU-S7GS2. The plates were incubated at 20◦C for 5 days to test whether the sclerotia could germinate. The antagonistic effects of NEAU-S7GS2 on sclerotia in soil were evaluated as described by Kamal et al. (2015) with the exception that the spore concentration of NEAU-S7GS2 was 10<sup>7</sup> colony forming units (cfu)/ml. Briefly, twenty uniform sclerotia grown on PDA plates were dipped into the spore suspensions of NEAU-S7GS2 and placed into nylon mesh bags with 2 mm<sup>2</sup> holes in the mesh. Pots containing clay loam field soil were watered to 80% field capacity prior to burying the nylon bags containing the sclerotia under 2 cm of soil. The pots were maintained at 20◦C for 30 days, and then the sclerotia were recovered and washed in sterile distilled water. The recovered sclerotia were sterilized as described above and the germination test was conducted. Sclerotia that produced mycelia were considered as viable sclerotia, and the percentage of germinating sclerotia was recorded. All the experiments were conducted in triplicate.

### Hyphal Morphology of S. sclerotiorum by Microscopic Observation

A colony of NEAU-S7GS2 was also placed on the margin of the PDA and incubated at 28◦C for 2 days. Then a fungal disk (5 mm) of S. sclerotiorum was placed on the direct opposite of NEAU-S7GS2. After incubation at 20◦C for 5 days, the hyphae of S. sclerotiorum adjacent to NEAU-S7GS2 were observed by light microscopy (ECLIPSE E200, Nikon) and scanning electron microscopy (SEM, S-3400N, Hitachi). For SEM, the samples were fixed in 2% glutaraldehyde for 24 h at 4◦C, rinsed three times with phosphate buffer (0.02M) and subsequently fixed with 2% osmium tetroxide for 2 h at 20◦C. The hyphae were then dehydrated in a graded series of ethanol concentrations (50, 70, and 90%) for 15 min each. After dehydration, the samples were gold-palladium sputtered and photographed with SEM.

### Biological Control of Sclerotinia Stem Rot Disease in Greenhouse Experiment

Strain NEAU-S7GS2 was assessed for its efficiency in suppressing Sclerotinia stem rot disease caused by S. sclerotiorum under greenhouse conditions. Soybean seeds [Glycine max (L.) Merr.] were surface sterilized with 5% NaOCl for 1 min and rinsed three times in sterile distilled water. Culture pots (17 cm high × 14 cm diameter) were filled with sterilized soil and 20 ml of different concentrations (105∼10<sup>9</sup> cfu/ml) of NEAU-S7GS2 spores suspended in water were added to the soil of each pot. Soil treated with 20 ml water was used as control. Three soybean seeds were used per pot, and all the pots were arranged in a completely randomized block design. At V2 stage of soybean seedling, 20 ml of ascospores of S. sclerotiorum (1 × 10<sup>6</sup> spores/ml) was added into the soil, and the disease severity was assessed after 7 days. Disease incidence was calculated as the percent of infected plants over the total number of plants. The degree of disease severity was determined from a score of 0 to 5 as described by Syed-Ab-Rahman et al. (2018) and calculated as follows: Disease index = [6(rating × number of plants rated)/(total number of plants × highest rating)] × 100. The experiment was performed in triplicate, and each treatment within an experiment consisted of 10 replicates.

### Plant Growth Promoting Effect of NEAU-S7GS2 on Wheat and Maize

#### Germination Bioassay

Strain NEAU-S7GS2 was grown on GY medium at 28◦C for 2 days. The spores were then collected and suspended in sterile water to various concentrations of 10<sup>5</sup> , 10<sup>6</sup> , 10<sup>7</sup> , 10<sup>8</sup> , and 10<sup>9</sup> cfu/ml. Seeds of wheat (Triticum aestivum L.) variety Longmai 30 or maize (Zea mays L.) variety Demeiya No. 1 were surface sterilized according to the method described previously (Fu et al., 2019a). The sterilized seeds were soaked in the spore suspension for 30 min. For control, seeds were dropped in distilled water only. The treated seeds were subsequently placed in sterilized Petri dishes covered with two sheets of filter papers, moistened with 5 ml of sterile distilled water. Each Petri dish contained 10 and 4 seeds for wheat and maize, respectively, and five replicates per treatment were conducted. All the Petri dishes were incubated at 25◦C in a dark incubator. After 1 week, the lengths of shoot and root were measured. The germination bioassay was performed in triplicate.

#### Pot Experiments

The treated seeds (10 wheat seeds/pot or 2 maize seeds/pot) were sown to a depth of approximately 1.5 cm in plastic pots (17 cm high × 14 cm diameter) filled with sterile soil. The seeds inoculated with sterile water were used as negative controls. All pots were placed in the greenhouse with natural light, 28◦C and 70% humidity. The plants were irrigated two times per day. After 30 days, the plants were collected carefully from the pots and washed with tap water to remove soil residues. The heights and fresh weights (FWs) of shoot and root were measured. Then, the root and the shoot of each plant were placed in paper bags in a drying oven with forced ventilation at 65◦C until they achieved a constant weight. Then, dry weights (DWs) of root and shoot were determined. Three replications were maintained with five pots per replication, and all the treatments were performed in a completely randomized block design.

#### Plant Growth-Promoting Traits of NEAU-S7GS2

#### Indole-3-Acetic Acid (IAA) Production

Strain NEAU-S7GS2 was grown in GY medium with the supplementation of 0.1% (w/v) <sup>L</sup>-tryptophan at 28◦C and agitated at 250 rpm. The cultures were sampled every 12 h and centrifuged (12,000 g, 15 min). The supernatant was mixed with equal volume of ethyl acetate, and then the obtained extracts were evaporated in vacuo at 37◦C and redissolved in methanol. Finally, the methanol samples were analyzed on high-performance liquid chromatography (Agilent 1100 series, United States) using a XDB-C18 column (250 mm × 4.6 mm i.d.; 5 µm particle size). The elution consisted of a mixture of methanol: water containing 0.05% acetic acid (methanol gradient: 10 to 35% in 40 min; 35% maintained for 20 min) at a flow rate of 0.5 ml/min. The column temperature was kept at 25◦C and IAA was detected by ultraviolet absorbance at 254 nm. The production of IAA was measured every 24 h interval. The quality of IAA was calculated according to the calibration curve of standard IAA (Sigma Aldrich, United States). Two independent experiments were performed in triplicate.

#### Phosphate Solubilization, Siderophore, and ACC Deaminase Production

Strain NEAU-S7GS2 was inoculated onto M9 minimal medium (Kandel et al., 2017) supplemented with Chrome Azurol S (CAS) and incubated at 28◦C for 7 days. The appearance of the orange halo zone around strain NEAU-S7GS2 was indicative of production of siderophores. The presence of 1-aminocyclopropane-1-carboxylate (ACC) deaminase was analyzed based on the ability of strain to grow on Dworkin-Foster (DF) salts agar medium (Dworkin and Foster, 1958) supplemented with 3 mM ACC instead of (NH4)2SO<sup>4</sup> as sole nitrogen source. Phosphate solubilizing properties of NEAU-S7GS2 were determined using National Botanical Research Institute's Phosphate (NBRIP) agar medium (Nautiyal, 1999), and the clear halo zones around bacterial growth for phosphate solubilization were checked. The plates were incubated at 28◦C for 7 days, and the experiment was repeated twice with three replicates each.

#### Genome Sequencing, Assembly, and Bioinformatics Analyses

The genomic DNA of NEAU-S7GS2 was extracted using a QIAamp DNA mini Kit (Qiagen) according to the manufacturer's protocols. The whole genome was sequenced with a 10-Kb SMRTbellTM template library using Single Molecular, Real-Time (SMRT) technology with the PacBio RS II sequencer. Sequencing was performed at the Beijing Novogene Bioinformatics Technology, Co., Ltd., China. The reads were de novo assembled using the RS Hierarchical Genome Assembly Process (HGAP Assembly Protocol version 3) in SMRT analysis version

2.3.0 software (Pacific Biosciences<sup>1</sup> ). GeneMarks version 4.17<sup>2</sup> was used to predict the open reading frames (Besemer et al., 2001). Software packages tRNAscan-SE v1.31 (Lowe and Eddy, 1997) and rRNAmmer v1.2 (Lagesen et al., 2007) were used to predict tRNA and rRNA, respectively. The gene functions were predicted by BLASTp using the non-redundant GenBank protein database<sup>3</sup> , the clusters of orthologous groups database (COG)<sup>4</sup> , GO database<sup>5</sup> , and KEGG database<sup>6</sup> . Genome mining for bioactive secondary metabolites was performed using "antibiotics and secondary metabolite analysis shell" (antiSMASH) version 4.0 (Blin et al., 2017).

#### GenBank Accession Number

The complete genome sequence of Streptomyces sp. NEAU-S7GS2 has been deposited in NCBI under the GenBank accession numbers NZ\_CP029541 (chromosome) and CP029542 (plasmid).

#### Statistical Analysis

The data were analyzed using analysis of variance (ANOVA) followed by Duncan's multiple-range test (p ≤ 0.05) using statistical software SPSS version 17.0 (SPSS, Inc., Chicago, IL, United States). The results were expressed as mean ± SD.

#### RESULTS

#### Bacterial Isolation and Identification

A total of 25 and 9 bacterial isolates were isolated from the root of Glycine max and its rhizosphere soil, respectively. Based on 16S rRNA gene analysis, these strains were classified into Streptomyces, Micromonospora, and Glycomyces. Most of the strains are Streptomyces spp., which accounted for 87% of the total isolates. Among the isolated strains, three strains including NEAU-S7GS2 that showed high antifungal activity were found both in the root and rhizosphere soil. The 16S rRNA gene sequence of NEAU-S7GS2 was submitted to NCBI under the accession number of MH675481. The phylogentic tree based on 16S rRNA sequence, constructed using MEGA 7.0 indicated that strain NEAU-S7GS2 was a member of the genus Streptomyces. It formed a subclade with the nearest neighbor Streptomyces libani subsp. libani NBRC 13452<sup>T</sup> , Streptomyces nigrescens NBRC 12894<sup>T</sup> , Streptomyces tubercidicus DSM 40261<sup>T</sup> , and Streptomyces angustmyceticus NRRL B-2347<sup>T</sup> with 99.86, 99.86, 99.79, and 99.72% 16S rRNA gene sequence similarity, respectively (**Supplementary Figure S1**). Strain NEAU-S7GS2 could utilize glucose, xylose, mannose, arabinose, raffinose, maltose, rhamnose, lactose, ribose, galactose, fructose, mannitol, galactitol, sorbitol, and inositol as sole carbon source. It could utilize tyrosine, glycine, asparagic acid, alanine, glutamic acid, serine, arginine, asparagine, threonine, proline, glutamine as sole nitrogen source. NEAU-S7GS2 was able to grow at 18 to 45◦C and pH 5.0 to 10.0, the optimal temperature and pH were 28 to 32◦C and 7.0 to 9.0, respectively. It could grow in ISP2 liquid medium with the addition of 0 to 14% NaCl. Although NaCl is not required for the growth of NEAU-S7GS2, the addition of NaCl even at a concentration of 10% has no obvious effect on its growth.

#### Antifungal Activity of Strain NEAU-S7GS2 Against Phytopathogenic Fungi

Strain NEAU-S7GS2 showed significantly inhibitory effects on the mycelial growth of the tested phytopathogenic fungi S. sclerotiorum, E. turcicum, R. solani, and C. cassiicola using Petri dish assays (**Figures 1A–D**). Compared to the control, the mycelia growth of S. sclerotiorum, E. turcicum, R. solani, and C. cassiicola was reduced by 99.1, 67.6, 65.3, and 57.9%, respectively, implying a broad antifungal spectrum (**Supplementary Table S1**). The effect of volatile organic compounds (VOCs) produced by NEAU-S7GS2 was evaluated, and the results suggested that the VOCs of NEAU-S7GS2 only slightly inhibited mycelial growth of S. sclerotiorum (**Figures 1I,J**). After fermentation, the cell-free supernatant of fermentation broth and methanol extract of mycelia showed obvious antifungal activity against S. sclerotiorum (**Figures 1E,F**); however, the supernatant of fermentation medium and methanol as negative controls exhibited no inhibitory effects on the growth of S. sclerotiorum (**Figures 1G,H**), indicating that NEAU-S7GS2 could produce diffusible and antifungal materials.

#### Effects of NEAU-S7GS2 on Hyphal Morphology and Sclerotial Germination

The hyphae morphological abnormalities of S. sclerotiorum, growing in the zone of interaction between S. sclerotiorum and NEAU-S7GS2, was observed by light microscopy. Compared to the control (**Figures 2a,b**), the mycelia of S. sclerotiorum was severely contorted and more pointed with increasing offshoots when S. sclerotiorum was co-cultured with NEAU-S7GS2 (**Figures 2c,d**). The front of normal mycelia was straight and apex obtuse (**Figures 2e,f,i,j**), while the hyphal front of treated S. sclerotiorum was atrophied and twisting, and many vacuoles appeared in the mycelia interior (**Figures 2g,h,k,l**). SEM revealed that the mycelia of S. sclerotiorum, adjacent to NEAU-S7GS2, were shriveled and irregular with a rough surface, and some hyphal fragmentation and perforations were also observed in cell walls (**Figures 3c,d**). In contrast untreated mycelia exhibited normal, regular, and homogenous morphology (**Figures 3a,b**). The sclerotia germination test showed that the sclerotia were able to germinate and ultimately formed white and cottony mycelia on the plate in the absence of NEAU-S7GS2 (**Figure 1L**), however no mycelia were observed on the plate in the presence of NEAU-S7GS2 (**Figure 1K**). When sclerotia were dipped into the spore suspension of NEAU-S7GS2 and placed in potted field soil for 30 days, the viability of these sclerotia was reduced to 9 ± 2% in comparison with 82 ± 3% of the untreated sclerotia.

<sup>1</sup>https://github.com/PacificBiosciences/SMRT-Analysis

<sup>2</sup>http://topaz.gatech.edu/GeneMark/

<sup>3</sup>www.ncbi.nlm.nih.gov/protein

<sup>4</sup>http://www.ncbi.nlm.nih.gov/COG/

<sup>5</sup>http://www.geneontology.org/

<sup>6</sup>http://www.genome.jp/kegg/

### Biological Control of Sclerotinia Stem Rot in Greenhouse Experiment

Given the notably inhibitory effect on S. sclerotiorum, the biocontrol potential of strain NEAU-S7GS2 in sclerotinia stem rot of soybean was further determined in a pot experiment under greenhouse conditions. Compared to the control that only inoculated S. sclerotiorum, the addition of NEAU-S7GS2 significantly (p < 0.05) reduced the disease incidence and disease index (**Figure 4A**). However, no significant difference (p > 0.05) was observed at high concentrations of NEAU-S7GS2. The remarkable effect was observed at a concentration of 10<sup>7</sup> cfu/ml, and disease incidence and disease index were 12 and 29%, respectively, which were 77 and 38% lower than the control. The sclerotinia stem rot disease symptoms on soybean, such as cottony growth of white mold in stems and appearance of yellow leaves, became visible at 7 days after inoculation of S. sclerotiorum (**Figure 4B**). In contrast, the pre-treatment of soils with the addition of strain NEAU-S7GS2 could alleviate sclerotinia stem rot disease symptom development.

### Plant Growth Promoting Activity of NEAU-S7GS2

Significant increases in the height of soybean seedling and numbers of leaf were observed when strain NEAU-S7GS2 was inoculated into the soil (**Figure 4B**). Compared to the un-treated control (10.8 ± 0.9 cm), the height of soybean seedling reached 12.4 ± 0.5, 12.5 ± 0.7, 14.4 ± 0.7, 13.7 ± 0.4, 12.6 ± 0.6 cm when strain NEAU-S7GS2 was added with a concentration of 10<sup>5</sup> , 10<sup>6</sup> , 10<sup>7</sup> , 10<sup>8</sup> , and 10<sup>9</sup> cfu/ml, respectively. The significantly increased height (p < 0.05) implied the plant growth promoting activity of strain NEAU-S7GS2. The plant growth promoting effect was enhanced with the increased concentration of strain NEAU-S7GS2 and then declined at high concentration. The optimal effect was observed at a concentration of 10<sup>7</sup> cfu/ml with the height of soybean increasing 33.3%. In order to fully investigate the plant growth promoting traits, the effects of strain NEAU-S7GS2 on wheat and maize were evaluated. In the germination test, NEAU-S7GS2 exhibited positive effects on the germination of wheat and maize (**Supplementary Figure S2**). Especially, strain NEAU-S7GS2 significantly increased the height of wheat shoot and the length of maize root. The similar plant growth promoting effects on wheat and maize were observed in the pot experiments (**Figure 5**). Compared to the effects on root of wheat, the effects on shoot were more significant, and the optimal inoculated concentration of NEAU-S7GS2 was 10<sup>7</sup> cfu/ml, which led to 18.3, 97.6, and 40% increases in the height, FW and DW of shoot, respectively (**Figures 5A–C**). In the case of maize, strain NEAU-S7GS2 exhibited notably beneficial effects on seedling growth (**Figures 5D–F**). With the optimal inoculation concentration of strain NEAU-S7GS2 (10<sup>7</sup> cfu/ml), the height, FW and DW of shoot increased 32.2, 33.7, and 111.8%,

respectively. By contrast, the promoting effect on root was more significant, and the length, FW and DW of root increased 55.9, 107.9, and 151.7%, respectively.

### General Genomic Features

In order to fully understand the molecular mechanisms of plant growth promotion and antagonism, the complete genome of NEAU-S7GS2 was sequenced. In total, 66,011 high-quality reads with 522,946,897 nucleotides were generated and subsequently de novo assembled. The complete genome of strain NEAU-S7GS2 contained one circular chromosome (9,641,634 bp) and one circular plasmid (45,805 bp) (**Figure 6**). The G + C contents of the chromosome and plasmid were 70.79 and 69.41%, respectively. The chromosome contained 7,864 protein coding genes (CDS), 21 rRNAs and 67 tRNAs genes. The plasmid harbored 51 protein coding genes (**Table 1**). Among these CDSs, 6,744 (85.76%) genes were classified into 24 clusters of orthologous groups of proteins. Most of the genes were associated with functions such as transcription, amino acid transport and metabolism, signal transduction mechanisms, carbohydrate transport and metabolism, lipid transport and metabolism, energy production and conversion, coenzyme transport and metabolism, inorganic ion transport and metabolism, and secondary metabolites biosynthesis, transport and catabolism (**Supplementary Table S2**). These functions are essentials for nutritional/spatial competition and antagonism against microorganisms to compete in various ecosystems.

#### Identification of Genes Associated With Plant Growth Promotion

Functional annotation of the genome revealed the presence of various genes that associated to plant growth-promoting traits. Strain NEAU-S7GS2 harbored the genes responsible for ammonia assimilation via both the GDH pathway using glutamate dehydrogenase (locus tag: DKG71\_41010, DKG71\_28945, DKG71\_16590) and glutamine synthetase (GS)-glutamate synthase (GOGAT) pathway using glutamine synthetase (DKG71\_09525, DKG71\_12715, DKG71\_12875, DKG71\_31225, DKG71\_33620, DKG71\_37070) and glutamate synthase (DKG71\_11785, DKG71\_11790, DKG71\_15225, DKG71\_15225). Furthermore, a gene encoding for 1 aminocyclopropane-1-carboxylate (ACC) deaminase (DKG71\_27775) was found. The ability of NEAU-S7GS2 to produce ACC deaminase was then confirmed by the fact that it could grow on minimal salt medium supplemented with ACC (**Supplementary Figure S3A**). Additionally, strain NEAU-S7GS2 has the potential to produce the major phytohormone indole acetic acid (IAA) with a maximal yield of 17.54 ± 0.61 µg/ml.

Several genes related to IAA biosynthesis were found in the genome of NEAU-S7GS2, such as the genes encoding for indole-3-glycerol phosphate synthase (DKG71\_11855, DKG71\_32585), indole acetimide hydrolase (DKG71\_30355), phosphoribosylanthranilate isomerase (DKG71\_11935), and anthranilate phosphoribosyltransferase (DKG71\_12405). Strain NEAU-S7GS2 also showed a clear solubilization zone on NBRIP plate (**Supplementary Figure S3B**), indicating the phosphate solubilizing potential. Several genes related to the solubilization of inorganic phosphate, phosphate metabolism and transporter were identified, such as the genes encoding for phosphohydrolase (DKG71\_28200, DKG71\_27640, DKG71\_27645), phosphate transporter (DKG71\_11035), polyphosphate kinase (DKG71\_05485).

#### Identification of Genes Responsible for Antifungal Property of NEAU-S7GS2

Strain NEAU-S7GS2 exhibited the potential to degrade glucan and cellulose on agar plates (**Supplementary Figures S4C,D**), and two glucanase (β-1,3-glucanase: DKG71\_10585; endoglucanase: DKD71\_15870), one α-amylase (DKG71\_32885) and four glucoamylase (DKG71\_10575, DKG71\_10960, DKD71\_31240, DKG71\_31870) genes were identified in the genome of NEAU-S7GS2. Although there are 48 protease (such as DKG71\_41140, DKG71\_39880, DKG71\_37495) and five chitinase (DKG71\_41330, DKG71\_38830, DKG71\_26705, DKG71\_15675, DKG71\_03385) genes distributed in the genome, NEAU-S7GS2 showed no protease activity and chitinase activity (**Supplementary Figures S4A,B**).

AntiSMASH analysis led to identification of 33 putative gene clusters in the genome of Streptomyces sp. NEAU-S7GS2 and some of them are responsible for the biosynthesis of siderophores, phenazines, and other polyketides (**Figure 7**). Strain NEAU-S7GS2 showed the potential to produce siderophore (**Supplementary Figure S3C**), and three siderophore gene clusters (clusters 11, 14, and 22) were obviously identified. Cluster 11 contains three siderophore biosynthesis protein genes (DKG71\_06530, DKG71\_06535 and DKG71\_06555) and four genes related to iron transport (DKG71\_06590, DKG71\_06520, DKG71\_06525, and DKG71\_06545). Cluster 22 contains one siderophore biosynthesis protein gene (DKG71\_30315) and one iron transport gene (DKG71\_30310). Cluster 14 shows 100 and 66% similarity to the biosynthetic gene cluster of desferrioxamine B in pseudouridimycin-producing Streptomyces sp. ID38640 and S. coelicolor A3(2), respectively. It contains four genes (DKG71\_10680, DKG71\_10685, DKG71\_10690, and DKG71\_10695) homologous to desA, desB, desC, and desD that responsible for the biosynthesis of desferrioxamine B and desferrioxamine E in S. coelicolor A3(2). Detailed survey and analysis of the predicted gene clusters in strain NEAU-S7GS2, a biosynthetic gene cluster (cluster 10-1) that contains entire intact genes necessary for the biosynthesis of catechol siderophore enterobactin was found to locate in cluster 10. DKG71\_06640, DKG71\_06625/DKG71\_06630, and

DKG71\_06645 encode isochorismate synthase, isochorismatase and 2,3-dihydro-2,3-dihydroxybenzoate dehydrogenase, respectively. These genes are responsible for the biosynthesis of 2,3-dihydroxybenzoate (2,3-DHB), which is the key moiety of catechol siderophore. DKG71\_06635 encodes 2,3-dihydroxybenzoyl adenylate synthase, which acitivates 2,3-DHB as a starter unit for non-ribosomal peptide synthase (NRPS) encoding by DKG71\_06620 to produce enterobactin. DKG71\_06610 encodes a 4-phosphopantetheinyl transferase that activates the thiolation (T) domains of 2,3-dihydroxybenzoyl adenylate synthase and NRPS. Additionally, there are some genes (DKG71\_06650–DKG71\_06655) encoding ABC transporter permease and ABC transporter ATP-binding protein in cluster 10-1 and one gene (DKG71\_06370) encoding siderophoreinteracting protein adjacent to cluster 10-1. These genes are responsible for siderophore export and uptake.

Cluster 10-2 shows 96% similarity to the biosynthetic gene cluster of lydicamycins. It is a hybrid PKS/NRPS gene cluster consisting of eight PKS (DKG71\_06775, DKG71\_06795, DKG71\_06800, DKG71\_06805, DKG71\_06810, DKG71\_06815, DKG71\_06820, DKG71\_06825) and one NRPS (DKG71\_06785) open reading frames to generate the backbone of lydicamycins. Three genes DKG71\_06750, DKG71\_06870 and DKG71\_06835, encoding amine oxidase, acyl-CoA ligase, and transaylase, respectively, are responsible for the starter unit 4-guanidinobutyryl CoA. DKG71\_06770 encodes a cytochrome P450, which is related to post-PKS modification of lydicamycin.

A phenazine gene cluster (cluster 23) was found to demonstrate 43% similarity to lomofungin biosynthetic gene cluster (lomo) in Streptomyces lomondensis S015. Cluster 23 contains five phenazine core biosynthetic genes phzGFEDB (DKG71\_31645, DKG71\_31650, DKG71\_31660, DKG71\_31665, and DKG71\_31670). The missing phzA (DKG71\_22560) was detected to be far away from phzGFEDB, and phzC (DKG71\_31600) were detected to be adjacent to phzGFEDB by BLASTp analysis. In addition, there are some genes encoding peptide synthases (DKG71\_31585, DKG71\_31630, DKG71\_31620), the genes related to shikimate pathway (DKG71\_08835, DKG71\_08840, DKG73\_08845), and some tailoring enzyme genes (DKG71\_31740, DKG71\_31700, DKG71\_31685, DKG71\_31680, DKG71\_31675, DKG71\_31640, DKG71\_31635) adjacent to phzGFEDB.

FIGURE 5 | The effects of NEAU-S7GS2 with various concentrations (105∼10<sup>9</sup> cfu/ml) on the growth of wheat and maize in the pot experiment. (A) The effects on the shoot height and root length of wheat; (B) the effect on the fresh weight (FW) of wheat shoot and root; (C) the effect on the dry weight (DW) of wheat shoot and root; (D) the effects on the shoot height and root length of maize; (E) the effect on the FW of maize shoot and root; (F) the effect on the DW of maize shoot and root. Different lowercase letters rows indicate significant difference (p < 0.05).

Cluster 6 was identified as a type I PKS gene cluster, in which 88% genes showed high similarity to that of PM100117/PM100118, which are glycosylated polyketide compounds consisting of a 36-membered macrocyclic lactone, three deoxysugars and a 1,4-naphthoquinone chromophore. Cluster 6 contains the intact genes that associated with the biosynthesis of naphthoquinone and deoxysugars. MqnA (encoding by DKG71\_04495), MqnC (DKG71\_04490), MqnD (DKG71\_04475) and a futalosine hydrolase (DKG71\_25050, which is far away from cluster 6) catalyze the biosynthesis of 1,4-dihydroxy 6-napthoic acid (DH6N), which is subsequently methylated by a putative S-adenosylmethionine-dependent methyltransferase (DKG71\_04545) to from 3-methyl-DH6N. Then, a putative synthetase-ligase (DKG71\_04565) catalyzes 3-methyl-DH6N and ATP to form 3-methyl-DH6N-AMP adduct, and the 3-methyl-DH6N is latter transferred to 11 PKS modules (DKG71\_04460, DKG71\_04535, DKG71\_04605, DKG71\_04610, DKG71\_04615, DKG71\_04625, DKG71\_04635, DKG71\_04645, DKG71\_04660, DKG71\_04665, DKG71\_04670) for elongation. Nine genes (DKG71\_04470, DKG71\_04510, DKG71\_04515, DKG71\_04520, DKG71\_04570, DKG71\_04580, DKG71\_04590, DKG71\_04675) dispersedly locate in cluster and these genes are responsible for the bioynthesis of deoxysugars. Four glycosyltransferase-coding genes (DKG71-04455, DKG71- 04550, DKG71-04575, DKG71-04595) are involved in the transfer of deoxysugars for tailoring modification.



–, not detected.

#### DISCUSSION

Stem rot caused by S. sclerotiorum is globally one of the most destructive soil borne diseases of many economically important crops. Fungicides have played important roles in controlling sclerotinia stem rot, however their intensive use can lead to microbial pathogen resistance and cause serious problems for human health and the quality of environment (Chen et al., 2016). Indeed, the development of benzimidazoles and dicarboxamides resistance in S. sclerotiorum has globally appeared after the introduction of these fungicides more than a decade, leading to control failure (Derbyshire and Denton-Giles, 2016). Thus, the use of biological control agents (BCAs) is considered as an alternative and sustainable strategy to control S. sclerotiorum. Although many efforts have been taken to investigate the antagonistic microorganisms, especially the mycoparasitic fungi Coniothyrium and Trichodema together with the antagonistic bacteria Bacillus and Pseudomonas, only C. minitans, T. harzianum, and B. subtilis have been commercialized as BCAs for use against S. sclerotiorum (Zeng et al., 2012; Derbyshire and Denton-Giles, 2016). Streptomyces spp. are well-known for the tremendous capacity to produce active secondary metabolites, rendering them a significant source of pharmaceutical leads and therapeutic agents. In comparison with the exploitation in pharmaceutical industry, there is only limited application of Streptomyces as BCAs in agriculture.

Frontiers in Microbiology | www.frontiersin.org 11

In the present study, Streptomyces sp. NEAU-S7GS2 that was distributed both in the root of Glycine max and its rhizosphere soil was isolated and exhibited a broad antifungal activity against phytopathogenic fungi. Especially, it inhibited mycelial growth of S. sclerotiorum by 99.1% in dual culture assays. In the pot experiments, the application of NEAU-S7GS2 could alleviate sclerotinia stem rot disease symptom development, and significantly decreased the disease incidence by 77%. The antagonistic effect of NEAU-S7GS2 against S. sclerotiorum was stronger than those of other Streptomyces spp. and the extensively studied Bacillus and Pseudomonas (Fernando et al., 2007; Wan et al., 2008; Baniasadi et al., 2009; Li et al., 2011; Zeng et al., 2012; Hu et al., 2013; Chen et al., 2014; Cheng et al., 2014; Wu et al., 2014). The excellent protection performance of NEAU-S7GS2 in the pot experiments may be mostly related with its ability to inhibit the germination of sclerotia and reduce the sclerotia viability. Sclerotia are hard, asexual, resting structures that are produced by many phytopathogenic fungi in Ascomycota and Basidiomycota. They are important dormant bodies that can survive for several years in soil and play a key role in the disease cycle (Zhang et al., 2019). Although chemical fungicides have played important roles in the control of disease caused by S. sclerotiorum, they could not continuously prevent the disease in future crop seasons due to the presence of sclerotia in the soils (Sabaté et al., 2018). The regulation of sclerotial germination is considered to be a major strategy for sustainable management of S. sclerotiorum, because it could break the disease life cycle by killing overwintering sclerotia in soils and prevent the formation of apothecia and myceliogenic germination (Zeng et al., 2012; Kamal et al., 2015). Sabaté et al. (2018) reported that the treatment of common bean seeds with chemical fungicide have no protective effect on the seed emergence in S. sclerotioruminfested soil, however the inoculation of BCA containing two Bacillusspp. led to 75% decrease in disease incidence compared to the control. Zeng et al. (2012) have reported that the application of fungal (C. minitans and T. harzianum) and bacterial (B. subtilis and S. lydicus) BCAs could significantly reduce sclerotia density of S. sclerotiorum in soil, resulting in reduction of the disease incidence in pot and field experiments. Therefore, NEAU-S7GS2 may be a promising BCA used to control the disease caused by S. sclerotiorum.

Many BCAs can produce hydrolytic enzymes, such as protease, glucanases, amylase and chitinase, to destroy the components of fungal cell wall, which is an important mechanism involved in the biocontrol of phytopathogenic fungi. Fernando et al. (2007) reported that the antifungal activity of Pseudomonas chlororaphis PA-23 against S. sclerotiorum is related to the secretion of chitinase and β-1,3-glucanase. The co-culture of Bacillus amyloliquefaciens and S. sclerotiorum could lead to fungal cell wall destruction and leakage of cell contents through the production of α-amylase, which can hydrolyze the glucans in S. sclerotiorum cell wall (Abdullah et al., 2008). The overexpression of endochitinase gene in Clonostachys rosea, a promising biocontrol fungus, significantly increased its biocontrol efficiency to soybean sclerotinia stem rot (Sun et al., 2017). Other reports also showed that Streptomyces spp. can inhibit the growth of S. sclerotiorum by producing chitinase (Tahtamouni et al., 2006; Baharlouei et al., 2011). Heterologous expression of chitinase from T. harzianum led to greatly increased biocontrol effect of S. lydicus A01 on fusarium disease (Wu et al., 2013). The culture filtrate of NEAU-S7GS2 inhibited the growth of S. sclerotiorum, however it lost the antifungal activity after the treatment of heat by boiling at 100◦C for 30 min. This suggested that the antifungal substances in the culture filtrate may be proteins. The SEM and light microscopy analysis revealed that NEAU-S7GS2 could lead to cell wall degradation in S. sclerotiorum, also indicating the production of extracellular lysis enzymes. Indeed, NEAU-S7GS2 exhibited the potential to degrade glucan and cellulose on agar plates, which was further confirmed by the identification of two glucanase and five amylase genes in the genome of NEAU-S7GS2. Although there are many protease genes distributed in the genome, NEAU-S7GS2 showed no protease activity due to the presence of genes encoding protease inhibitor, such as DKG71\_10355 and DKG71\_39690. Similarity, NEAU-S7GS2 contains five chitinase genes but no chitinase activity has been detected on agar plates. These chitinases show high sequence identity (95.36 to 99.21%) to those found in the commercial BCA S. lydicus WYEC 108. Chitinase plays important roles in the in vivo antifungal biocontrol activity of S. lydicus WYEC 108, however the production of chitinase is inducible with low constitutive levels and the fungal cell wall chitins especially that from the target fungi significantly enhance chitinase expression (Mahadevan and Crawford, 1997). Thus, we could not exclude the possibility that the expression of chitinase in NEAU-S7GS2 was not enough to be detected in agar plates.

Apart from hydrolytic enzymes, antagonistic microorganisms can produce VOCs or non-volatile organic compounds to control fungal diseases. Due to the low molecular weight, high vapor pressure, and ability to diffuse easily through the porous structure of soil and over great distances in the atmosphere, VOCs have recently received more attention (Yuan et al., 2012; Cho et al., 2017; Xing et al., 2018). It has been reported that VOCs from several species of Streptomyces possessed antifungal activity and caused severe morphological alterations on the hyphae and conidiophores of phytopathogenic fungi (Wan et al., 2008; Li et al., 2012; Boukaew et al., 2013; Marzieh et al., 2013; Wang et al., 2013; Wu et al., 2015; Cho et al., 2017; Xing et al., 2018). However, NEAU-S7GS2 only slightly inhibited the growth of S. sclerotiorum in the divided plate assay, indicating that VOCs produced by NEAU-S7GS2 are not the main antifungal components. It is well-known that Streptomyces are a prolific source of secondary metabolites with diverse structure, and some of which have been developed as fungicides for the control of fungal plant diseases, such as validamycin, blasticidin, kasugamycin, mildiomycin, and polyixins (Lee et al., 2003). Therefore, the antifungal activity of methanol extracts from mycelia of NEAU-S7GS2 against S. sclerotiorum was evaluated. The results showed that methanol extracts could inhibit the mycelial growth of S. sclerotiorum, indicating the production of antifungal secondary metabolites.

Subsequent genomic analysis led to identify three siderophore gene clusters designated as clusters 11, 14, and 22 in the genome of NEAU-S7GS2, and the production of siderophores was further confirmed by the appearance of yellow halo around the colony on CAS agar. Many gene clusters that

were homologous to cluster 11 and cluster 22 could be found in the sequenced Streptomyces, such as S. decoyicus NRRL 2666, S. lydicus A02 and S. chattanoogensis NRRL ISP-5002. All these strains demonstrate antifungal activity against plant pathogenic fungi, however no corresponding siderophore has been reported. Cluster 14 shows high similarity to the biosynthetic gene cluster of desferrioxamines in S. coelicolor A3(2) and contains four necessary genes responsible for the desferrioxamines biosynthesis. Detailed survey and analysis of the predicted gene clusters in NEAU-S7GS2 led to the identification of a biosynthetic gene cluster (cluster 10-1) that contains entire intact genes necessary for the biosynthesis of catechol siderophore enterobactin. Catechol-siderophores including coelichelin, enterobactin, griseobactin, streptobactin and qinichelins are featured with 2,3-dihydroxybenzoate (2,3- DHB) as key functional group and have been identified from the Streptomyces species (Reitz et al., 2017). Recently Gubbens et al. (2017) demonstrated that 2,3-DHB is an interwined precursor during the biosynthesis of the catecholate siderophores qinichelins, griseobactin, and enterobactin in Streptomyces sp. MBT76. Similar functional crosstalk was also found during the biosynthesis of enterobactin and other secondary metabolites benzoxazoles and caboxamycin in other Streptomyces species (Cano-Prieto et al., 2015; Losada et al., 2017). The antiSMASH analysis showed a large number of PKS and NRPS gene cluster distributed in the genome of NEAU-S7GS2, whether the similar crosstalk exists in NEAU-S7GS2 merits further investigation.

Cluster 10-2 showed 96% similarity to the biosynthetic gene cluster of lydicamycins, structurally unique type polyketides bearing an amidinopyrrolidine ring and a tetramic acid (Komaki et al., 2015). Lydicamycin was firstly isolated from S. lydicus 2249-S3 and demonstrated potent antimicrobial activity against Gram-positive bacteria including methicillin-resistant Staphylococcus aureus and antifungal activity against human pathogen Cryptococcus neoformans (Hayakawa et al., 1991). Four other lydicamycin congeners were lately isolated from a marine actinomycete Streptomyces platensis TP-A0598 and showed equivalent antimicrobial activity against Gram-positive bacteria to lydicamycin but no activity against Gram-negative bacteria and yeast (Furumai et al., 2002). To date, the antifungal activity of lydicamycins against plant pathogenic fungi has not been reported yet. Although whether lydicamycin associates with the antifungal activity against S. sclerotiorum is unclear, NEAU-S7GS2 represents the third lydicamycin-producing bacterium.

Cluster 23 showed 43% similarity to the biosynthetic gene cluster of lomofungin, a phenazine antibiotic possessed broadspectrum antibacterial and antifungal activity (Zhang et al., 2015). Phenazines, a class of microbial secondary metabolites containing a phenazine nucleus have demonstrated broadly inhibitory effects on the growth of pythopathogenic fungi (Mavrodi et al., 2010; Zhang et al., 2015). The functional groups attached to the phenazine nucleus led to different phenazine derivatives with diverse structure of biological activity. For example, the esmeraldin biosynthetic gene cluster in Streptomyces antibioticus Tü2706 contains a PKS gene that responsible for incorporating acetyl group to phenazine-1-carboxylic acid to yield esmeraldin and sapenamycin (Rui et al., 2012). Therefore, the presence of some peptide synthetases (DKG71\_31585, DKG71\_31630, DKG71\_31620) together with other different tailoring enzymes in cluster 23 implied major structural differences between the product of cluster 23 and lomofungin.

Cluster 6 was identified as a type I PKS gene cluster, in which 88% genes showed high similarity to that of PM100117/PM100118 biosynthetic gene cluster in a marine symbiotic actinobacteria Streptomyces camiferus GUA-06- 05-006A (Salcedo et al., 2016). PM100117 and PM100118 are glycosylated polyketide compounds, which consist of a 36-membered macrocyclic lactone, three deoxysugars and a 1,4-naphthoquinone chromophore. Compared to the PM100117/PM100118 biosynthetic gene cluster, cluster 6 contains more PKS modules, indicating that the corresponding product may possess a larger macrocyclic lactone ring. PM100117/PM100118 have attracted the attention due to their strong antitumor activity and slight antifungal activity against the opportunistic pathogen Candida albicans, but remarkably low toxicity (Pérez et al., 2016). Recently, another structurally related compound cyphomycin was isolated from the ant-associated Streptomyces sp. ISID311 and demonstrated significant in vitro antifungal activity against the ecologically relevant fungus-growing ant pathogen Escovopsis sp. and the resistant human pathogens Aspergillus fumigatus, Candida glabrata, and Candida auris (Chevrette et al., 2019). Meanwhile, other microbial secondary metabolites with similar structure to PM100117/PM100118 have been reported to demonstrate antifungal activity against phytopathogenic fungi. For example, deplelide B displayed strong antifungal activity against plant pathogenic fungi Cochliobolus miyabeanus and Pyricularia oryzae (Takeuchi et al., 2017). Other structurally closest related compounds with strong antifungal activity comprise 36-membered macrolides including liposidolide A and polaramycins, 32-membered macrolides including brasilinolides, novonestmycins and copiamycins, but they all lack the napthtoquinone unit (Kihara et al., 1995; Shigemori et al., 1996; Meng and Jin, 1997; Fukai et al., 1999; Wan et al., 2015). Liposidolide A exhibited strongly antifungal activity against phytophathogenic fungi Pyricularia oryzae and Colletotrichum lagenarium with MIC of ≤0.2 µg/ml and showed a preventive value of 99.4% against cucumber anthracnose at a dose of 50 ppm (Kihara et al., 1995); novonestmycin A and B showed a broad antifungal activity, especially against phytophathogenic fungi C. cassiicola, R. solani, and Septoria nodorum with MIC values of <1 µg/ml (Wan et al., 2015). Although Salcedo et al. (2016) proposed that the antitumor activity stems from the napthtoquinone unit, the antifungal activity is a common property of this type of compound and unrelated to the size of macrocyclic lactone ring. Therefore, we speculate that the compound biosynthesized by cluster 6 may be the major antifungal component produced by NEAU-S7GS2 although it contains a larger macrocyclic lactone ring than PM100117/PM100118.

Furthermore, NEAU-S7GS2 exhibited plant growth promoting activity through solubilization of inorganic phosphate, production of ACC deaminase, siderophore and the typical phytohormone IAA. Although Streptomyces has

been used in the biocontrol of soil-borne fungal pathogens, the commercialized biofertilizer based on Streptomyces is seldomly reported. The exception is S. lydicus WYEC108, which was originally isolated from a rhizosphere soil of linseed, however it could colonize in the root nodules of peas leading to increase root nodulation frequency and enhance plant growth (Tokala et al., 2002). The unique trait was then used later in the formulation and the commercialization of the well-known biocontrol product Actinovate <sup>R</sup> and Actino-Iron <sup>R</sup> (Crawford et al., 2005). Similarly, strain NEAU-S7GS2 was found to colonize the root of Glycine max and its rhizosphere soil and demonstrated antagonistic activity against phytopathogenic fungi and plant growth promoting activity. Endophytic colonization is usually considered as an important ecological advantage to the bacteria, due to that endophytic environment is more resistant to abiotic stress and microbial competition compared with the rhizospheresoil system (Shimizu, 2011). Therefore, strain NEAU-S7GS2 merits the further investigation for the use as biofertilizer.

#### CONCLUSION

Streptomyces sp. NEAU-S7GS2 was isolated from the root of Glycine max and its rhizosphere soil. Strain NEAU-S7GS2 showed significantly inhibitory activity against the mycelial growth and sclerotia germination of S. sclerotiorum by production of hydrolytic enzymes and active secondary metabolites. In the pot experiment, it not only demonstrated excellent potential to control sclerotinia stem rot of soybean but could promote the growth of soybean through solubilization of inorganic phosphate, and production of ACC deaminase and IAA. The genome sequencing and bioinformatic analysis identified several gene clusters related to the biosynthesis of phenazines, lydicamycins, enterobactin and an unknown glycosylated polyol macrolide. The significant antifungal and plant growth promoting activity implied the potent use in agriculture as biocontrol agent and biofertilizer.

#### DATA AVAILABILITY

The complete genome sequence of Streptomyces sp. NEAU-S7GS2 has been deposited in NCBI under the GenBank

#### REFERENCES


accession numbers NZ\_CP029541 (chromosome) and CP029542 (plasmid), the 16S rRNA gene sequence of NEAU-S7GS2 was submitted to NCBI under the accession number of MH675481.

### AUTHOR CONTRIBUTIONS

DL and RY performed the experiments. YF analyzed the genomic data. XW provided technical assistance and revised the manuscript. JZ and WX designed the experiments. JZ wrote the manuscript. All authors read and approved the final manuscript.

#### FUNDING

This work was supported by grants from the National Key R&D Program of China (2017YFD0200502), the National Natural Science Foundation of China (31471801), and the University Nursing Program for Young Scholars with Creative Talents in Heilongjiang Province (UNPYSCT-2016009).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.02077/full#supplementary-material

FIGURE S1 | Neighbor-joining phylogenetic tree based on 16S rRNA gene sequences showing the phylogenetic position of strain NEAU-S7GS2 and the related strains of the genus Streptomyces. Nocardia carnea DSM 43397<sup>T</sup> was used as an out group. Bootstrap values > 50% (based on 1000 replications) are shown at branch points. Bar: 0.01 substitutions per nucleotide position.

FIGURE S2 | The effects of NEAU-S7GS2 with different concentrations (105∼10<sup>9</sup> cfu/ml) on seed germination of wheat (a) and maize (b).

FIGURE S3 | Plant growth promoting traits of NEAU-S7GS2. (A) ACC deaminase, (B) phosphate solubilization, (C) siderophore production.

FIGURE S4 | Potential of NEAU-S7GS2 to show (A) protease activity, (B) chitinase activity, (C) glucanse activity, and (D) cellulase activity.

TABLE S1 | The antifungal activity of strain NEAU-S7GS2 against phytopathogenic fungi.

TABLE S2 | COG functional categories of the complete genome sequence of Streptomyces sp. NEAU-S7GS2.



of a Streptomyces corchorusii strain UCR3-16 and preparation of powder formulation for application as biofertilizer agents for rice plant. Microbiol. Res. 192, 260–270. doi: 10.1016/j.micres.2016.08.005


**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 Liu, Yan, Fu, Wang, Zhang and Xiang. 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.

fmicb-10-02077 September 6, 2019 Time: 18:0 # 16

# Monitoring of Rice Transcriptional Responses to Contrasted Colonizing Patterns of Phytobeneficial *Burkholderia s.l.* Reveals a Temporal Shift in JA Systemic Response

*Eoghan King1, Adrian Wallner1, Isabelle Rimbault1, Célia Barrachina2, Agnieszka Klonowska1, Lionel Moulin1 and Pierre Czernic1\**

*1 IRD, CIRAD, University of Montpellier, IPME, Montpellier, France, 2 Montpellier GenomiX (MGX), c/o Institut de Génomique Fonctionnelle, Montpellier, France*

#### *Edited by:*

*Massimiliano Morelli, Italian National Research Council (IPSP-CNR), Italy*

#### *Reviewed by:*

*Stéphane Compant, Austrian Institute of Technology (AIT), Austria Lisa Sanchez, Université de Reims Champagne-Ardenne, France*

> *\*Correspondence: Pierre Czernic pierre.czernic@umontpellier.fr*

#### *Specialty section:*

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

*Received: 21 June 2019 Accepted: 21 August 2019 Published: 24 September 2019*

#### *Citation:*

*King E, Wallner A, Rimbault I, Barrachina C, Klonowska A, Moulin L and Czernic P (2019) Monitoring of Rice Transcriptional Responses to Contrasted Colonizing Patterns of Phytobeneficial Burkholderia s.l. Reveals a Temporal Shift in JA Systemic Response. Front. Plant Sci. 10:1141. doi: 10.3389/fpls.2019.01141*

In the context of plant–pathogen and plant–mutualist interactions, the underlying molecular bases associated with host colonization have been extensively studied. However, it is not the case for non-mutualistic beneficial interactions or associative symbiosis with plants. Particularly, little is known about the transcriptional regulations associated with the immune tolerance of plants towards beneficial microbes. In this context, the study of the *Burkholderia* rice model is very promising to describe the molecular mechanisms involved in associative symbiosis. Indeed, several species of the *Burkholderia sensu lato* (*s.l.*) genus can colonize rice tissues and have beneficial effects; particularly, two species have been thoroughly studied: *Burkholderia vietnamiensis* and *Paraburkholderia kururiensis*. This study aims to compare the interaction of these species with rice and especially to identify common or specific plant responses. Therefore, we analyzed root colonization of the rice cultivar Nipponbare using DsRed-tagged bacterial strains and produced the transcriptomes of both roots and leaves 7 days after root inoculation. This led us to the identification of a co-expression jasmonic acid (JA)-related network exhibiting opposite regulation in response to the two strains in the leaves of inoculated plants. We then monitored by quantitative polymerase chain reaction (qPCR) the expression of JA-related genes during time course colonization by each strain. Our results reveal a temporal shift in this JA systemic response, which can be related to different colonization strategies of both strains.

#### Keywords: RNAseq, endophyte, symbiosis, *Burkholderia*, rice, jasmonic acid

#### INTRODUCTION

Plant microbiome is nowadays extensively studied, as it represents a huge potential for agriculture. Numerous studies describe the importance of microbes for plant's nutrient supply and resistance to diseases and pests (Finkel et al., 2017). Therefore, microbes are potential solutions to do the transition to a sustainable agriculture while maintaining yield (Busby et al., 2017). Especially, some rhizobacteria have been shown to have tremendous beneficial effects on plant growth (Hayat et al., 2010)

**79**

King et al. Rice Sensing of Beneficial *Burkholderia*

and resistance to pathogens (Beneduzi et al., 2012). These beneficial effects are induced through hormonal modulations following the colonization of plant roots (Vacheron et al., 2013) and inner tissues for endophytes (Hardoim et al., 2008) as well as systemic regulations of immunity (Pieterse et al., 2014). However, the perception of microbe-associated molecular patterns (MAMPs) generally leads to an immune response called MAMPs-triggered immunity (MTI) characterized by the synthesis of antimicrobial compounds (Jones and Dangl, 2006). In the same way pathogens suppress plant immunity, and beneficial microbes are able to escape (Trdá et al., 2015) or modulate (Zamioudis and Pieterse, 2012) the immune response of plants cells. Interestingly, the suppression of MTI by both pathogenic bacteria (Millet et al., 2010) and beneficial fungi (Jacobs et al., 2011) is commonly mediated *via* the jasmonic acid (JA) signaling pathway. However, the global physiological response and especially the transcriptional regulations induced by plants during the interaction with beneficial bacteria are not well described.

The interaction between plants and rhizospheric or endophytic bacteria-forming associative symbioses has been studied in several species belonging to the genera *Azoarcus*, *Azospirillum*, *Herbaspirillum*, *Acetobacter*, *Gluconacetobacter*, *Bacillus*, *Phyllobacterium*, *Pseudomonas*, and (*Para*-) *Burkholderia* (Ahemad and Kibret, 2014). Most studies on these models focused on the bacterial response to the interaction with its host plant (Shidore et al., 2012; Coutinho et al., 2015; Sheibani-Tezerji et al., 2015), while relatively few studies analyzed the transcriptional response of plants. Nonetheless, several studies described plants' transcriptional regulations including lowering of defense (Bordiec et al., 2011; Rekha et al., 2018), hormonal signaling (Drogue et al., 2014; Rekha et al., 2018), developmental reprogramming (Paungfoo-Lonhienne et al., 2016), and iron homeostasis (Brusamarello-Santos et al., 2019; Stringlis et al., 2018).

Within the diversity of bacteria interacting with plants, the *Burkholderia sensu lato* (*s.l.*) genus of Betaproteobacteria stands out for several reasons. It contains plant pathogenic species (*B glumae*, *Burkholderia gladioli*, and *Burkholderia plantarii*) (Maeda et al., 2006), N2-fixing nodule-forming species in association with tropical legumes (Gyaneshwar et al., 2011) as well as species-forming associative symbiosis particularly with cereals such as rice (Coutinho et al., 2013; Govindarajan et al., 2008). Also, a phylogenetic separation discriminates the *Burkholderia sensu stricto* (*s.s.*) genus, which contains animal or plants pathogens as well as human opportunists, and the *Paraburkholderia* genus, which contains mainly plant-associated and environmental species (Sawana et al., 2014; Estrada-de los Santos et al., 2016). Also, most recent phylogenetic refinements of *Burkholderia s.l.* taxa defined other genera—*Caballeronia*, *Robbsia*, *Trinickia*, and *Mycetohabitans*—which are supported by both differences in genomic and ecological features (Estrada-de los Santos et al., 2018).

Interestingly, two strains of the *Burkholderia s.l.* genus able to fix N2 have been described as growth promoters in rice. *Paraburkholderia kururiensis* M130 (hereafter *Pk*) is a beneficial rice endophyte (Mattos et al., 2008) related to environmental and plant-beneficial strains (Kaur et al., 2017). *Burkholderia vietnamiensis* TVV75T (hereafter *Bv*) is a rice-associated species having positive effect on yield (Trân Van et al., 2000; Govindarajan et al., 2008), which belongs to the *Burkholderia cepacia* complex, a complex of species that can cause serious risks to cystic fibrosis patients (Vial et al., 2011). In order to decipher how rice perceives these two beneficial strains belonging to genera with contrasted ecologic backgrounds, we studied the transcriptional responses of rice during the establishment of the interaction. Our aim was to identify plant physiological processes and potentially key genes involved in beneficial rice-rhizobacteria interactions and also differentially regulated by each strain. We first analyzed the colonization patterns of *Pk* and *Bv* on the *Oryza sativa* Nipponbare genotype. We then analyzed the root and leaf transcriptional responses to the bacterial colonization by RNAseq. This led us to the identification of a co-expression JA-related network. Therefore, we monitored, throughout the establishment of the interaction, the expression of JA-related genes by reverse transcriptase–quantitative polymerase chain reaction (RT-qPCR).

### MATERIALS AND METHODS

#### Plants and Bacterial Cultivation

*O. sativa* L. ssp. *japonica* cv Nipponbare was used in this study. For all experiments, seeds were dehusked and sterilized as follows: 70% ethanol for 10 min and 9.6% NaClO supplemented with 1% Tween 20 for 30 min. Treated seeds were rinsed twice with sterile distilled water, twice with 2% thiosulfate solution, and finally four times with sterile distilled water. To confirm surface sterilization, 100 µl of the last rinsing solution was plated in tryptic soy agar (TSA) medium (Sigma-Aldrich). Seeds were then put in sterile distilled water at 28°C for 24 h and transferred on 8% H2O agar plate for 30 h. Homogeneously germinated seeds were transferred to sterile magenta boxes (SPL Lifesciences Co. Ltd) containing 150 ml of autoclaved perlite and 200 ml of sterile hydroponic medium (recipe in **Supplementary Table 1**). Plants were grown in a growth chamber (16 h light; 8 h dark; 28°C; 70% humidity).

All bacterial strains (listed in **Supplementary Table 2**) were cultured as follows: Glycerol stocks (20%) of bacterial cells conserved at −80°C were plated in low-salt lysogeny broth (LB) (Sigma-Aldrich) agar plates and incubated for 72 h at 28°C. Liquid low salt LB medium of 20 ml was then inoculated in 50-ml Falcon tubes and incubated for 16 h under agitation (180 rpm) at 28°C. Overnight culture of 500 µl was inoculated in fresh liquid medium for 2 h. Bacterial cells were then centrifuged for 5 min at 4,000 rpm and resuspended in sterile distilled water. Each plantlet was inoculated with 107 bacterial cells 4 days after sowing in hydroponic system.

#### Bacterial Transformation

*P. kururiensis* M130 and *B. vietnamiensis* TVV75 cells were transformed by electroporation with the pIN29 plasmid (Vergunst et al., 2010). The plasmid chosen to transform the strains, pIN29, comprises a chloramphenicol resistance gene as well as the DsRed gene under the control of a constitutive TAC promoter. After 24 h of incubation of selective medium low salt LB Cm (200 μg·ml−1) at 28°C, the most fluorescent colonies were selected.

#### Rice Root Colonization Assays

The roots of plants were harvested at 1, 7, and 14 days postinoculation (dpi), weighted, and grinded in sterile water with a sterile ceramic bead using a FastPrep-24™ 5G at 6 m·s−1 for 40 s. The solution was then diluted and inoculated in low-salt LB selective medium containing 200 μg·ml−1 of chloramphenicol and incubated at 28°C for 24 h. Colony-forming units were then enumerated. The size of the root-associated bacterial population was measured during two independent experiments, each comprising nine plants. In order to measure the size of the endophytic population, the inoculated rice roots were surface disinfected for 1 min using a solution of 1% chloramine T (Sigma-Aldrich) supplemented with 0.1% Tween 20. Roots were then rinsed six times with sterile water. Controls of disinfection were performed by plating rinsing water in TSA medium (Sigma-Aldrich) overnight. Surface-disinfected roots were then treated as previously described. The size of the endophytic population was measured on five plants.

#### Microscopy

All microscopic observations of the bacterial colonization were restricted to the primary root in order to compare the colonization patterns on roots that have been in contact with the bacterial population for the same amount of time. Primary roots were harvested at 7 and 14 dpi and mounted between slide and slips cover and directly examined with the microscopes. Epifluorescence observations were performed using a Nikon Eclipse Ni-E microscope. Confocal Laser Scanning observations were performed using a Zeiss LSM880 confocal microscope.

### RNA Extraction

For the analysis of root and leaf transcriptional profiles, both plants' tissues were harvested at 6 h postinoculation (hpi), 1 dpi, 7 dpi, or 14 dpi with live bacterial cells. Each biological replicate consisted of five pooled root system or five pooled last mature leaves harvested from a single hydroponic system. For each time point and each inoculated strain, three biological replicates were harvested. Roots and leaves of untreated plants were collected at the same time points. After harvest, samples were snap-frozen in liquid nitrogen and stored at −80°C.

Rice roots were homogenized in liquid nitrogen using cooled mortar and pestle. Rice leaves were grinded using a TissueLyser II (Retsch) set to 30 Hz for 30 s. Total RNA extraction using TRIreagent (Sigma) was performed according to manufacturer's instructions. All samples were treated with DNase I (Ambion) and purified using the RNA Clean & Concentrator kit (Zymo) according to manufacturer's instructions. The integrity and quality of the total RNA were confirmed using a NanoDrop™ 1000 spectrophotometer (Thermo Fisher) and a 2100 BioAnalyzer (Agilent).

### RNA Sequencing and Mapping of Reads

Quality of RNA was checked by determining the RNA Integrity Number (RIN) with a Fragment Analyzer (Agilent). For the library preparation, samples with a RIN value > 6 were used. Eighteen RNA libraries were prepared using an Illumina TruSeq stranded mRNA sample preparation kit by MGX-Montpellier GenomiX core facility (MGX) France (https://www.mgx.cnrs.fr/). Library construction and sequencing were performed as described in Karmakar et al. (2019) on an Illumina HiSeq 2500. The quantitative and qualitative validation of the library was performed by qPCR, Roche LightCycler 480, and a Fragment Analyzer (Agilent) using a Standard Sensitivity NGS kit. Quality control and assessment of raw Illumina reads in FASTQ format were done by FastQC software (version 0.11.5) to obtain per base quality, Guanine-Cytosine (GC) content, and sequence length distribution. Clean reads were obtained by removing the lowquality reads, adapters, and poly-N-containing reads by using Trimmomatic v0.36 software (Bolger et al., 2014). RNAseq reads were aligned to the IRGSP 1.0 version of the rice genome using HISAT2 v2.0.5.1 (Kim et al., 2015). The number of reads mapped to each gene locus was counted using HTSEq-count v0.6.0 (Anders et al., 2015).

#### Differential Gene Expression and Gene Ontology (GO) Term Enrichment Analysis

DESeq2 v3.7 (Love et al., 2014) was used to calculate differential gene expression between non-inoculated and inoculated conditions. All genes having an adjusted *p*-value inferior to 0.01 were considered as significantly differentially expressed. All functional enrichment analyses were performed using g:Profiler (version e95\_eg42\_p13\_f6e58b9) with g:SCS multiple testing correction method applying significance threshold of 0.05 (Reimand et al., 2007; Raudvere et al., 2019).

#### Quantification of mRNA Levels Using RT-qCPR

cDNA was produced from 350 ng of DNase-treated total RNA using the SuperScript III Reverse Transcriptase (Thermo Fisher Scientific). The cDNA reaction was diluted five times before qPCR using MESA BLUE qPCR Master Mix for SYBR® assay (Eurogentec) on an Mx3005P qPCR system (Agilent Technologies). The relative expression level was calculated according to Pfaffl (2001). Three independent samples were analyzed for each condition, and each sample was assayed in triplicate. Primers used are listed in **Supplementary Table 3**.

## RESULTS

### Analysis of Root Colonization

We used hydroponic culture of rice plants, grown in axenic condition and inoculated with DsRed-tagged strains, to monitor the bacterial colonization at 1, 7, and 14 dpi. First, the root's colonization was measured by counting the bacterial populations on the rhizoplan and endosphere (see Materials and Methods). The roots of rice plants were rapidly colonized by both bacterial strains (**Figure 1A**). The populations of *Pk* and *Bv* reach a median value of 4.16 × 106 and 1.73 × 106 cfu·g−1, respectively, at 1 dpi. At 7 dpi, the maximum measured size of the bacterial population is reached for both strains: 7.49 × 108 and 3.18 × 108 cfu·g−1 for *Pk* and *Bv*, respectively. Then, between 7 and 14 dpi, the size of the total root-associated population decreases for each strain, reaching a median value of 4.6 × 108 and 1.46 × 108 cfu·g−1 for *Pk*  and *Bv*, respectively. Between the same time points, the variation of the endophytic population size differs between the two strains. Indeed, the endophytic population size of *Pk* decreases from 1.25 × 106 to 1.17 × 104 cfu·g−1 between 7 and 14 dpi, while the median number of endophytic *Bv* cells remain stable with 7.34 × 104 and 7.5 × 104 cfu·g−1 at 7 and 14 dpi, respectively.

Microscopic observations of the primary roots of inoculated rice plants demonstrated that both bacterial strains tagged with the DsRed gene colonized the root surface after inoculation (**Figures 1B**, **C**). Moreover specific zones were more densely colonized by bacteria such as the surface of root hairs and the emergence of lateral roots (**Supplementary Figure 1**). By comparing the colonization of the two strains, differences can be observed in the way they colonized the surface of the primary root. Several epidermal plant cells seemed colonized intracellularly by the tagged *Bv* cells, while the phenomenon was observed less frequently in *Pk*-colonized roots (**Figure 1C**; **Supplementary Figure 1**). Observations by confocal microscopy revealed that for *Pk*, the highly colonized epidermal cells appear to be only colonized on their surface as the whole outline of epidermal cells as well as in intercellular spaces (**Figures 2A**, **C**), while for *Bv*, most of the observations show that bacterial cells were able to cross the cell wall and were observed in the cytoplasm of the cell (**Figures 2B**, **D**). Thus, both strains colonized the roots of the Nipponbare cultivar both externally and endophytically but through apparently different intensities and entry roads. We then wondered if the host plant induces contrasted transcriptional regulations associated with the differential colonization pattern of each strain.

#### Transcriptional Response of Rice to Bacterial Inoculation

In order to identify changes in *O. sativa* transcriptome in response to both strains, we performed RNAseq on leaves and roots of non-inoculated controls, *Pk*-inoculated and *Bv*-inoculated plants at 7 dpi. We chose this time point to avoid the initial plant defense burst due to a bacterial inoculation in hydroponic system (hours to 1 dpi) and allow an advanced colonization stage of the roots but without any visible developmental effect such as increased growth. To confirm that the inoculated and noninoculated control plants were at the same developmental stage, we measured the dry weight of the plants and could not detect

FIGURE 1 | Colonization of the roots of hydroponically grown rice plants by *Bv* and *Pk.* (A) Population dynamics of DsRed-tagged *Bv* and *Pk* associated with rice roots and inside the plant roots. The data reported are the median of bacterial population size from 18 plants and two independent experiments for rhizosphere compartment and five plants for the endophytic compartment. The letters indicate the significance groups in each compartment according to *post-hoc* tests on a generalized linear model. Epifluorescence microscopy pictures of the colonization of rice primary roots at 7 days postinoculation by *Pk* (B) and *Bv* (C) DsRed cells. White bars represent 200 µm. *Bv*, *Burkholderia vietnamiensis* TVV75T; *Pk*, *Paraburkholderia kururiensis* M130.

any significant impact of the inoculation on biomass production (**Supplementary Figure 2**) nor on plant height. Therefore, we assume that the differences between the transcriptomes of inoculated plants and non-inoculated controls should be related to bacterial colonization of the roots rather than a developmental impact of inoculation.

A total of 843 million reads were sequenced with an average of 47 million reads per sample (**Table 1**). An average of 68% of reads was uniquely mapped to the *O. sativa* genome per sample. A principal component analysis also discriminates the transcriptome of non-inoculated plants compared with the inoculated ones as well as the plant responses to each bacterial strain (**Figures 3A**, **B**). Differential expression analysis yields a total of 4,951 and 5,275 significantly (*p* < 0.01) differentially expressed genes (DEGs) in response to *Bv* and *Pk*, respectively. Comparing leaf and root transcriptomes reveals that there are five and eight times more DEGs detected in leaves in response to *Bv* and *Pk*, respectively, than in roots (**Figures 3C**, **D**). When comparing the response to *Pk* and *Bv*, a large proportion of DEGs are commonly regulated in leaves (around 50% in response to both strains) contrarily to roots in which the transcriptional regulation appears to be more specific to each inoculated strain. Indeed, only 33% of the DEGs in response to *Bv* are also differentially expressed in response to *Pk*. The proportions of commonly up-regulated DEGs in roots represent 20% and 27% of DEGs in response to *Bv* and *Pk*, respectively.


In order to identify the biological processes transcriptionally regulated by the colonization of both bacterial strains, a GO enrichment analysis was carried out on the commonly regulated DEGs (intersections in **Figures 3C**, **D**). **Supplementary Figure 3** provides a visualization of the enriched GO terms from the commonly up-regulated or down-regulated DEGs in leaves and roots (complete list of enriched GO terms available in **Supplementary Table 4**). Three main biological processes are commonly regulated during the interaction with both strains in leaves and roots: response to stimuli, and metabolic and also developmental processes. First, stress-related genes are enriched in the commonly DEGs. Indeed, in leaves, the "response to abiotic stimulus" as well as the "response to oxidative stress" terms are enriched in the up-regulated DEGs, whereas defense-related GO terms are enriched in the down-regulated DEGs in both leaves and roots. Also, several hormone-related GO terms are enriched in the DEGs in both leaves and roots. In leaves, GO terms related to auxins and abscisic acid (ABA) response are enriched in the up-regulated genes, while in roots, the up-regulated DEGs are enriched in cytokinines (CK), brassinosteroids (BR), and ethylene response terms. Finally, in roots, down-regulated DEGs are enriched in gibberellic acid (GA) and salicylic acid (SA) response-related terms. The analysis also revealed that metabolic processes are transcriptionally regulated in response to the interaction with both strains: In leaves, the up-regulated DEGs are enriched in the "photosynthesis" and "translation" terms, while the down-regulated DEGs are enriched in "starch metabolic process" and "membrane lipid catabolic process" terms. Furthermore, in leaves, the "iron ion homeostasis" term is enriched in up-regulated DEGs, while the "metal ion transport" term is enriched in the down-regulated DEGs of roots. Although the inoculation of both strains did not significantly impact the biomass of rice plants (**Supplementary Figure 2**), several development-related GO terms are enriched in the DEGs in both leaves and roots. First, in leaves, among others, the "anatomical

structure development" term is enriched in the up-regulated DEGs, and the "glucan biosynthetic process" term is enriched in the down-regulated DEGs, which correspond to genes implicated in cellulose and callose synthesis. Finally, root transcriptional response is also enriched in development-related GO terms such as the "xylem development" term enriched in the up-regulated DEGs and the "lignin biosynthetic process" term enriched in the down-regulated DEGs.

Additionally, in order to identify the biological processes specifically induced during the interaction with each strain, we performed a GO term enrichment analysis on the DEGs specifically regulated by each strain (**Supplementary Figures 4** and **5**; **Supplementary Tables 5** and **6**). This analysis revealed that leaves of plants inoculated with *Pk* up-regulate genes related to biosynthetic process and translation, while the interaction with *Bv* induces the up-regulation of genes coding for components of the photosystem II and also the downregulation of protein folding genes (**Supplementary Figure 4,**  genes listed in **Supplementary Table 7**). Interestingly, the up-regulated DEGs in response to each strain are enriched in one hormonal signaling pathway. Indeed, the interaction with *Pk* induced the up-regulation of JA-related genes while cytokininrelated genes are enriched in the up-regulated DEGs in the leaves of *Bv*-inoculated plants. The analysis of the specific root transcriptome also revealed processes specifically induced by each strain. Particularly, the response to *Pk* in roots encompasses the down-regulation of oxidative stress response-related genes and also chitin catabolic process, while the interaction with *Bv* induced the down-regulation of genes involved in defense, JA signaling, and response to stimuli (**Supplementary Figure 5**, corresponding genes in **Supplementary Table 8**).

In order to deepen the transcriptome analysis, a functional categorization of the top 200 up-regulated and down-regulated DEGs identified in response to each strain was carried out using all available databases for rice annotation (UniProt, Kyoto

Encyclopedia of Genes and Genomes (KEGG), RAP-DB, and Oryzabase). The proportion of genes related to each category is presented in **Figure 4** and exemplifies rice-specific responses to the two bacterial strains. In leaves, more stress-related genes as well as genes involved in secondary metabolism are up-regulated in response to *Pk* compared with the response to *Bv*. Also, in response to *Pk*, 16 genes related to chromatin remodeling are down-regulated compared with two in response to *Bv*. Inversely, the inoculation of *Bv* induced the down-regulation in leaves of 22 stress-related genes, whereas only three are down-regulated in response to *Pk*. In roots, twice as much transcription regulation and protein degradation-related genes are up-regulated in response to *Bv* than in response to *Pk*. Conversely, approximately twice as much "nutrition/transport" and signaling-related genes are down-regulated in response to *Pk* than in response to *Bv*.

In order to identify plant key genes that could be related to the observed patterns of root colonization, we focused our analysis on defense (leaves in **Table 2**, roots in **Table 3**) and hormone-related genes (leaves in **Table 4**, roots in **Table 5**). Indeed, these processes are enriched among the strain-specific DEGs (**Supplementary Figures 4** and **5**) and could pinpoint differences in the physiological response of rice following the perception of each strain.

In roots, 13 defense-related genes are commonly regulated, while 33 and 18 are specifically *Pk* and *Bv* regulated (**Table 3**). Additionally, 12 hormone-related genes (implicated in JA, CK, ethylene, GA, and ABA syntheses or signaling) are commonly regulated in response to both strains, while 14 and 19 hormonerelated genes are *Pk* and *Bv* regulated, respectively (**Table 5**). The latest encompasses genes implicated in auxins, BR, SA, and strigolactone synthesis, or signaling. In leaves, 41 defenserelated genes are regulated, six being commonly regulated, while 11 and 24 are specifically induced by *Pk* and *Bv*, respectively (**Table 2**). Also hormone-related DEGs are detected in leaves; six genes implicated in GA, JA, auxins, ethylene and ABA synthesis, transport, or signaling are commonly regulated.

Finally, in response to each strain, 20 hormone-related specific DEGs are detected. Interestingly, only the interaction with *Pk* induced the regulation of BR- and SL-related genes, while only *Pv* induced the regulation of a SA-related gene. Other hormonal pathways, such as CK, ABA, ethylene, JA, auxins, and GA, are modulated by each strain but induced the regulation of different genes (**Table 4**).

## Validation of RNAseq Data by qCPR

To validate the RNAseq data, we selected genes implicated in defense, hormone signaling, or development regulated at 7 dpi

by one or both strains (**Supplementary Table 9**). We measured the level of expression of the selected genes by qRT-PCR in an independently conducted experiment. According to RNAseq analysis, in leaves, two defense-related genes are specifically regulated by each strain: *ALD1* is specifically down-regulated in response to *Bv*, and *WRKY71* is specifically up-regulated in response to *Pk*. Also, two hormone-related genes were detected as specifically regulated by each strain; namely, *bHLH148*, a JA signaling component, is down-regulated in roots inoculated with *Bv*; while *RR9*, a CK signaling component, is up-regulated in the leaves of *Pk*-inoculated plants. Finally, in roots, *RSL9* is up-regulated and *SHR5* is down-regulated in response to both TABLE 2 | Defense-related differentially expressed genes (DEGs) in leaves in response to *Paraburkholderia kururiensis* and *Burkholderia vietnamiensis*. Presented genes are part of the top 200 up-regulated and down-regulated significantly DEGs [false discovery rate (FDR) < 0.01]; all results for leaf transcriptome can be found in Supplementary Table 6.


strains. Gene expression changes obtained through qPCR analysis demonstrated a pattern similar to RNAseq (**Figure 5**) except for the over-expression of *RR9* in response to *Pk*, which was not detected as significant by DESeq2.

#### Temporal Analysis of Strain-Specific Marker Genes

Following the confirmation of strain-specific transcriptional regulations, we wanted to identify genes that are differentially expressed in both conditions but with opposite regulations. Among all DEGs detected in roots and leaves, only three DEGs are potential differential markers: *RERJ1* (Os04g0301500), a JA-responsive gene; *ATL15* (Os01g0597600), a putative amino acid transporter; and *DREB1B* (Os09g0522000), a droughtresponsive gene. Noteworthy, these three genes are only detected as DEGs in leaves and follow the same pattern of expression, as they are up-regulated in response to *Pk* and down-regulated in response to *Bv*.

Among these three genes, two of them, *ATL15* and *RERJ1*, are part of a co-expression network recovered from RiceFREND database (Sato et al., 2013), which contains four JA-related genes (*JAZ6*, *10*, *12*, and *AOS1*) (**Figure 6A**). As JA is one of the main phytohormones implicated in defense (Pozo et al., 2004), we further wanted to know if the whole co-expression network could act as a differential marker of the response to each bacterial strain. In order to describe the transcriptional regulation of this co-expression network throughout the establishment of the TABLE 3 | Defense-related differentially expressed genes (DEGs) in roots in response to *Paraburkholderia kururiensis* and *Burkholderia vietnamiensis*. Presented genes are part of the top 200 up-regulated and down-regulated significantly DEGs [false discovery rate (FDR) < 0.01]; all results for root transcriptome can be found in Supplementary Table 7.


(*continued*)

#### TABLE 3 | Continued


interaction, we produced a transcriptional kinetic of rice leaves tissues at 6 hpi, 1 dpi, 7 dpi, and 14 dpi and analyzed all six genes by qRT-PCR.

First, at 7 dpi, we can confirm that depending on the inoculated strains, the regulation of the co-expression network is opposite (**Figure 6**; **Supplementary Table 10**). On the one hand, *JAZ6*, *JAZ10*, *JAZ12*, *ATL15*, and *AOS1* are up-regulated following the inoculation with *Pk*; and on the other hand, *JAZ6* and *JAZ10* are down-regulated in response to *Bv*. Interestingly, and in sharp contrast to the 7 dpi response, for the short-term response (6 hpi), we observed the up-regulation of five out of the six genes of the co-expression network in rice leaves after inoculation with *Bv*, whereas only *JAZ12* is up-regulated in response to *Pk*. Then, the whole network appears to decline for the rest of the kinetics in response to *Bv*. Eventually, at 14 dpi, the expression levels of the six genes are quite comparable between the two conditions being all down-regulated compared with non-inoculated controls.

#### DISCUSSION

The aim of this study was to describe the transcriptional regulations induced by rice following the perception of naturally associated beneficial bacteria. Additionally, we took advantage of the particular phylogenetic organization of the *Burkholderia s.l.* genus to compare the responses of plants with two closely related beneficial species with different phylogenetic backgrounds in terms of ecology.

#### *P. kururiensis* M130 and *B. vietnamiensis* TVV75 Differentially Colonize Roots of *O. sativa* Cultivar Nipponbare

We first observed that both bacteria were able to efficiently colonize the rice Nipponbare roots. The amount of culturable bacterial cells associated with rice roots appears to be coherent with the literature, as Compant et al. (2010) described that generally a range between 107 and 109 cfu·g−1 of root fresh weight is found colonizing roots both externally and internally. The same goes for the endophytic population, which generally ranges between 105 and 107 cfu·g−1. The decrease of the bacterial population size between 7 and 14 dpi may be due to the increase of root biomass by the formation of newly emerged roots, which are not importantly colonized at least in the time of our experiments. However, by comparing the colonization of the two strains, two main differences can be observed in the dynamic of root colonization. First, *Pk* forms a significantly bigger population while colonizing rice roots surface than does *Bv* at every time postinoculation. Second, the dynamic of the endophytic population of both strains is very different, as the number of endophytic *Pk* cells declines between 7 and 14 dpi which is not the case for *Bv*. This could be due to the fact that the endophytic colonization by *Pk* is restricted by the plant throughout time in contrast to *Bv*, which maintains its population size. From this observation, two hypotheses can be proposed: Either the plant is not able to control the colonization by *Bv*, or this strain is more efficiently colonizing the newly emerged roots.

Both the maturation zone and the area of lateral root emergence were identified as hotspots for *Pk* and *Bv*  colonization. A similar area of colonization was identified for *Paraburkholderia phytofirmans* strain PsJN in grape plants (*Vitis vinifera* L.) (Compant et al., 2005) as well as for *B. vietnamiensis* MGK3, which also intensively colonizes the same areas of the roots of rice plants (Govindarajan et al., 2008). Root exudates, which contain essential nutrients for microbes, are released in the lateral root emergence zones (Badri and Vivanco, 2009). This may aid colonization and allow the possible entry of bacteria *via* mechanisms such as "crack entry" into the internal tissues (Hardoim et al., 2008).

Both strains were observed massively colonizing the maturation zone, on both the outside and even the inside of some root hair cells (**Supplementary Figure 1**). The accumulation of bacterial cells in the root hair zone has already been described for *Pk* (Mattos et al., 2008), and it was proposed to be a common hotspot for rhizobacterial colonization (Compant et al., 2010), as it is correlated to a higher local exudate concentration (Gamalero et al., 2004). Apparent intracellular colonization of epidermal cells was observed for both species but more frequently for *Bv*. In the case of *Bv*, the TABLE 4 | Hormone-related differentially expressed genes (DEGs) in leaves in response to *Paraburkholderia kururiensis* and *Burkholderia vietnamiensis*. Presented genes are part of the top 200 up-regulated and down-regulated significantly DEGs (false discovery rate (FDR) < 0.01); all results for leaf transcriptome can be found in Supplementary Table 6.


(*continued*)

#### TABLE 4 | Continued


fact that the vacuole is still intact when observing through the colonized epidermic cell offers evidence that the cell is still living (**Figure 2D**). Also, as the bacterial cells seem to have passed through the cell wall, the fluorescence signal appears to be cytoplasmic. Intracellular colonization of rice root epidermal cells by bacteria was also observed during the interaction with *Azoarcus* BH72 (Reinhold-Hurek et al., 2006) as well as during the colonization of ryegrass roots by *Paraburkholderia bryophila* Ha185 (Hsu et al., 2018). Furthermore, intracellular colonization of root hair cells was also observed by Prieto et al. (2011). They showed by confocal microscopy that *Pseudomonas putida* PICP2 and *Pseudomonas fluorescens* PICF7 are able to colonize root hair cells of olive trees (*Olea euroapea* L.) and subsequently move into epidermal cells. This observation led them to propose a new route of entry in intern plant tissues for endophytic bacteria, which starts with the colonization of single root hairs.

#### Root Colonization Induces More Transcriptional Regulations in Aerial Parts Than in Roots

The analysis of root and leaf transcriptomes revealed that colonization of rice roots by both strains induced more transcriptional regulations in leaves than in roots (**Figure 3B**). This could be due to the fact that the inoculated plants were harvested at 7 dpi and therefore at an established state of the interaction between bacteria and rice. In contrast, previous studies revealed that the inoculation of beneficial rhizobacteria induced the regulation of at least 1,000 genes in rice roots at earlier time points after inoculation (Drogue et al., 2014; Brusamarello-Santos et al., 2019; Rekha et al., 2018). Moreover, compared with the only study that analyzed the leaf transcriptional response of rice to the inoculation by beneficial rhizobacteria that retrieved only 2,414 DEGs at early stages of the interaction (Wu et al., 2018), when our study identified data of at least 4,000 DEGs in leaves in response to both strains.

#### Roots Trigger Contrasted Transcriptional Response Depending on the Inoculated Strain

In response to the colonization by both strains, transcriptional reprogramming of defense-related genes occurs in rice roots (**Table 3**). Only 13 of them appeared commonly regulated in the same way. First, three *WRKY* (*62*, *76*, and *86*),genes are commonly down-regulated; interestingly, those three genes are negative regulator of defense (Peng et al., 2008; Yokotani et al., 2013)—putatively for *WRKY86* (Choi et al., 2017). Also, two negative regulators of *NH1*, the rice *NPR1* homolog of the major SA response regulator (Yuan et al., 2007; Chern et al., 2012), namely, *RH2* and *RH3*, are also down-regulated (Chern et al., 2014). Taken together, as these genes are described as negative regulators of defense, their down-regulation may reflect a common defense response towards root colonizing bacteria.

The remaining 51 genes were differentially regulated in response to each strain (33 responding to *Pk* and 18 to *Bv*, respectively). Indeed, roots colonized with *Pk* induce the down-regulation of 11 pathogenesis-related (PR) genes as well as five chitinases and two xylanase inhibitors. One of the down-regulated PR genes is *PBZ1* (Os12g0555200), and it is described as a defense marker in leaves, but it is also up-regulated in response to root invasion by *Magnaporthe oryzae* (Marcel et al., 2010). Moreover, *bZIP79*, which is described as a phytoalexin synthesis suppressor (Miyamoto et al., 2015), is up-regulated. On the other hand, roots colonization by *Bv* induce the down-regulation of only one chitinase, whereas two PR genes are up-regulated, of which one gene, *BetvI*, is targeted by parasitic nematode to suppress root defense (Chen et al., 2018). Furthermore, four *WRKY* genes of which two are thought to encode negative regulators (*WRKY28* (Chujo et al., 2013) and *WRKY79* (Choi et al., 2017)) and *WRKY40*, which is up-regulated in striga-resistant rice roots (Swarbrick et al., 2008), are here down-regulated. Taken together, these results show that the colonization of rice roots by *Pk* is associated with a downregulation of *PR* genes, while the colonization by *Bv* is associated with the down-regulation of defense-suppressing *WRKY* genes. This transcriptional regulation of defense-related genes could be the consequence of the more invasive colonization pattern of *Bv* cells described above.

As previously described (Vacheron et al., 2013; Drogue et al., 2014; Rekha et al., 2018), the inoculation of beneficial rhizobacteria induces important hormone-related transcriptional regulation in rice roots (**Table 5**). Similarly, the inoculation of *Pk* and *Bv* induced the regulation of genes encoding for signaling components of CK and ethylene. Particularly, both bacterial strains induced the up-regulation of two genes encoding for ACO, which are enzymes TABLE 5 | Hormone-related differentially expressed genes (DEGs) in roots in response to *Paraburkholderia kururiensis* and *Burkholderia vietnamiensis*. Presented genes are part of the top 200 up-regulated and down-regulated significantly DEGs [false discovery rate (FDR) < 0.01]; all results for root transcriptome can be found in Supplementary Table 7.


implicated in the synthesis of ethylene (Ravanbakhsh et al., 2018). Interestingly, in response to the inoculation of other beneficial bacteria, genes coding for ACOs were down-regulated in rice roots (Drogue et al., 2014; Brusamarello-Santos et al., 2019). Also, important transcriptional regulations of ethylene responsive factors (ERFs) occur in response to both strains. In the same way, cytokinine signaling is impacted as both conditions induce the up-regulation of at least two *RR* genes and one cytokinine oxidase encoding gene (Tsai et al., 2012); however, it is not the same genes that are regulated by each bacterium.

*Bv*, *Burkholderia vietnamiensis* TVV75T; *Pk*, *Paraburkholderia kururiensis* M130; DEGs, differentially expressed genes.

#### Systemic Responses Associated With Root Inoculation Suggest a Time Shift in Defense Response Between the Two Species

Interestingly, the RNAseq analysis revealed that the inoculation of rhizobacteria induced major transcriptional regulations in leaves (**Tables 2** and **4**; **Supplementary Table 5**). As previously described (Persello-Cartieaux et al., 2003; Verhagen et al., 2004; Campos-Soriano et al., 2012), the colonization of beneficial microbes induces transcriptional regulations of defense-related genes in the leaves of the inoculated plants. Both strains induced the up-regulation of *WRKY7* as well as the down-regulation of three putative R genes and two PR genes. *Pk* induced the up-regulation of *WRKY71*, which is known to confer enhanced disease resistance to *Xoo* (Liu et al., 2007) and the down-regulation of *TFX1*, which is known as a susceptibility gene for *Xoo* (Sugio et al., 2007). On the other hand, *Bv* induces the up-regulation of five R genes: three (*NB-ARC*, *MLO4*, and *STV11*) (Wang et al., 2014), *NPR4*, as well as the down-regulation of three *WRKY* genes (*9*, *25*, and *69*), of which two are SA responsive (Liu et al., 2007; Choi et al., 2017), and also the down-regulation of *ALD1*, a basal immunity regulator needed for the accumulation of SA (Jung et al., 2016).

FIGURE 6 | JA-related co-expression network transcriptional regulations induced by *Pk* and *Bv* root colonization. (A) Co-expression network of differential markers (circled in blue); edges' width is proportional to the co-expression ratio between each gene according to RiceFREND database. (B) Gene expression dynamics of the co-expression network in response to *Pk* and *Bv* quantified by qRT-PCR. Transcript levels were normalized to that of reference gene EF 1α (Os03g0177400). Values between 0.5 and −0.5 are colored in gray. Data presented are the mean of log2 fold change (*n* = 3). JA, jasmonic acid; *Bv*, *Burkholderia vietnamiensis* TVV75T; *Pk*, *Paraburkholderia kururiensis* M130.

It is also in this organ only that differential markers have been detected, some of which designated the JA signaling pathway as a putative marker of the interaction with the two bacterial species. Also, interestingly, in contrast with the regulation of JA-related genes, several GA-related genes seem to be down-regulated in response to *Pk* and up-regulated in response to *Bv*. This antagonism between JA and GA has been described (Yang et al., 2012; Yimer et al., 2018) and proposed as one of the ways for plants to fine-tune the balance between growth and defense (Huot et al., 2014). Nonetheless, this apparent contrast in terms of JA-related genes transcriptional regulation at 7 dpi may only be a consequence of a delayed JA systemic signaling in response to *Pk* compared with the response to *Bv*, which induced the up-regulation of the co-expression network as soon as 6 hpi. As JA-induced plant defenses have been proposed to contribute to the restriction of endophytic colonization in grasses (Miché et al., 2006), this temporal shift in the induction of JA-related genes between the responses to the two bacterial strains could be associated with a delayed JA-induced defense signal. We propose the following interpretation: the perception of *Bv* induces very early (6 hpi) the up-regulation of JA-related genes in leaves to restrict the colonization, whereas in response to *Pk*, this JA signal happens at 7 dpi and results in the decrease of the bacterial population (**Figure 1A**).

The comparison of the interactions between rice and *Pk* and between rice and *Bv* revealed important differences in the process of root colonization and rice transcriptional regulations induced by each strain, which we summarize in **Figure 7**. First, the numerous intracellular colonization of root epidermic cells by *Bv* resembles a pathogen infection compared with the apoplastic colonization observed for *Pk* and other beneficial endophytes (McCully, 2001; Reinhold-Hurek et al., 2007). However, such intracellular colonization by beneficial endophytes has already been observed (Reinhold-Hurek and Hurek, 1998). Also, the specific root response to *Pk* is characterized by the downregulation of gene coding for chitinase and PR proteins, while *Bv* colonization specifically induced the down-regulation of several *WRKY* genes (**Table 3**). Moreover, *Bv* specifically induced the down-regulation of JA signaling genes (**Supplementary Figure 5**; **Supplementary Table 6**) in addition to the common SA-related down-regulation in roots. Therefore, the strategies to circumvent the immune system of roots appear to be different between the two strains. Finally, the fact that the root inoculation of *Bv* induced the up-regulation of JA-related genes in leaves in only 6 hpi compared with the delayed similar signal in response to *Pk* colonization at 7 dpi also supports the fact that *Bv* induces a response similar to that of pathogens. Indeed, the rice root colonization by *M. oryzae* induced the up-regulation of JA-related genes in leaves at 3 and 4 dpi (Marcel et al., 2010). All these elements support the fact that *Bv* appears to have a much more invasive colonization strategy in terms of both patterns and modulation of the plant immune system. This statement is in accordance with the opportunistic and pathogenic background of the *Burkholderia s.s.* genus (Eberl and Vandamme, 2016). Consequently, it would be of interest to analyze the response of the cultivar *Nipponbare* to a larger diversity of plant-associated *Burkholderia* and *Paraburkholderia* species to investigate if the

differences observed between the two strains can be extrapolated to other species of the respective clades.

Another interesting conclusion from this study is the importance of the JA signaling in the interaction with beneficial rhizobacteria. As previously discussed, this major component of plant defense classically associated with the resistance to herbivores and necrotrophic pathogens appears to be involved in a larger diversity of biotic interactions (Thaler, 2004). We have demonstrated that there is a temporal delay in the induction of JA-related genes in leaves following the colonization by *Pk* comparatively to the response to *Bv*. It would be of interest to investigate the role of JA in the establishment of these interactions in terms of colonization level and plant defense status using JA-deficient mutants.

## DATA AVAILABILITY

These sequence data for this study can be found in the EMBL database under accession number PRJEB31936.

## AUTHOR CONTRIBUTIONS

EK, PC, and LM contributed to the conception and design of the study. EK, AW, IR, CB, and AK performed data collection and analysis. EK wrote the first draft. PC and LM contributed to manuscript revision. All authors read and approved the submitted version.

## FUNDING

The authors acknowledge the CGIAR Research Program on Rice Agri-food Systems (RICE) and the MIC-CERES Project (FC Project ID 2013-1888; AF Project ID 1301-003, jointly supported by Agropolis Fondation through the "Investissements d'avenir" programme with reference number ANR-10-LABX-0001-01, and Fondazione Cariplo) for funding. EK and AW were supported by a fellowship from the French Ministry of Higher Education, Research and Innovation.

## ACKNOWLEDGMENTS

We would like to thank Annette Vergunst for providing the pIN29 plasmid. We would like to thank Antony Champion and Trang Hieu Nguyen for providing the JAZ qPCR primers.

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.01141/ full#supplementary-material

### REFERENCES


resistance in *Arabidopsis*. *Mol. Plant-Microbe Interact.* 17, 895–908. doi: 10.1094/MPMI.2004.17.8.895


disease resistance and cold stress tolerance. *J. Exp. Bot.* 64, 5085–5097. doi: 10.1093/jxb/ert298


**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 King, Wallner, Rimbault, Barrachina, Klonowska, Moulin and Czernic. 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.*

# Selection of an Endophytic Streptomyces sp. Strain DEF09 From Wheat Roots as a Biocontrol Agent Against Fusarium graminearum

Elena Maria Colombo\*, Andrea Kunova, Cristina Pizzatti, Marco Saracchi, Paolo Cortesi and Matias Pasquali\*

Department of Food, Environmental and Nutritional Sciences, University of Milan, Milan, Italy

#### Edited by:

Massimiliano Morelli, Italian National Research Council (IPSP-CNR), Italy

#### Reviewed by:

Juan Manuel Palazzini, National University of Río Cuarto, Argentina Hassan Etesami, University of Tehran, Iran Leen De Gelder, Ghent University, Belgium

#### \*Correspondence:

Elena Maria Colombo elenamaria.colombo@unimi.it Matias Pasquali matias.pasquali@unimi.it; matias.pasquali@gmail.com

#### Specialty section:

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

Received: 19 April 2019 Accepted: 27 September 2019 Published: 11 October 2019

#### Citation:

Colombo EM, Kunova A, Pizzatti C, Saracchi M, Cortesi P and Pasquali M (2019) Selection of an Endophytic Streptomyces sp. Strain DEF09 From Wheat Roots as a Biocontrol Agent Against Fusarium graminearum. Front. Microbiol. 10:2356. doi: 10.3389/fmicb.2019.02356 Selection of biological control agents (BCA) profits from an integrated study of the tripartite interactions occurring among the BCA, the plant and the pathogen. The environment plays a crucial role in the efficacy of BCA, therefore, the selection process shall utmost mimic naturally occurring conditions. To identify effective biocontrol strains against Fusarium graminearum, the major cause of Fusarium head blight (FHB) in wheat and deoxynivalenol (DON) accumulation in grains, a workflow consisting of in vitro and in vivo assays was set up. Twenty-one Streptomyces strains, 16 of which were endophytes of different plants, were analyzed. In vitro and in vivo tests characterized their plant growth promoting (PGP) traits. Biocontrol activity against F. graminearum was firstly assessed with a dual culture assay. An in vivo germination blotter assay measured Fusarium foot rot and root rot symptoms (FFR-FRR) reduction as well as growth parameters of the plant treated with the Streptomyces strains. A selected subset of Streptomyces spp. strains was then assessed in a growth chamber measuring FFR symptoms and growth parameters of the wheat plant. The approach led to the identification of an effective Streptomyces sp. strain, DEF09, able to inhibit FHB on wheat in controlled conditions by blocking the spread of the pathogen at the infection site. The results were further confirmed in field conditions on both bread and durum wheat, where DEF09 decreased disease severity up to 60%. This work confirms that FRR and FFR pathosystems can be used to identify BCA effective against FHB.

Keywords: endophytes, cereal, BCA, toxigenic fungi, PGP, Fusarium head blight, Fusarium root rot, Fusarium foot rot

### INTRODUCTION

Fusarium graminearum is a major threat to wheat, leading to Fusarium foot rot (FFR) and Fusarium root rot (FRR) (Smiley and Patterson, 1996), as well as Fusarium head blight (FHB), the major cause of wheat losses (Goswami and Kistler, 2004). Losses are aggravated by the accumulation of deoxynivalenol (DON), an internationally regulated mycotoxin (Wegulo et al., 2015). The pathogenic behavior of the fungus has been widely studied at the spike level both from a molecular point of view (Ilgen et al., 2009; Lysøe et al., 2011) and from a physiopathological point of view

**99**

(Boenisch and Schäfer, 2011). The pathogen, similarly to other known foot and root rot pathogens of wheat, such as F. culmorum (Scherm et al., 2013) and F. pseudograminearum (Chakraborty et al., 2006), has a specific pathway of infection and spread via roots (Wang Q. et al., 2015). Surprisingly, head blight, root, and foot rot caused by F. graminearum share most of the developmental steps of pathogenicity (Wang et al., 2018), including the DON synthesis (Covarelli et al., 2012).

Streptomyces spp. are well known Gram-positive bacterial symbionts of living organisms (Seipke et al., 2012), and can establish tight interactions with inner plant tissues (Coombs and Franco, 2003). They can act as plant growth promoters by producing phytohormones, facilitating nutrient uptake and inhibiting plant pathogens (Viaene et al., 2016; Vurukonda et al., 2018). They have been extensively investigated as a source of bioactive molecules (Watve et al., 2001) and Streptomyces-derived commercial products have been successfully applied for crop protection (Newitt et al., 2019). Indeed, several Streptomyces strains have been proposed as potential biocontrol agents against toxigenic fungi, including numerous Fusarium spp. causing diseases and mycotoxin accumulation in cereals (Nourozian et al., 2006; Palazzini et al., 2007; Yekkour et al., 2012; Jung et al., 2013).

Previous studies of Streptomyces strains effective against F. graminearum (Palazzini et al., 2007, 2017, 2018; Jung et al., 2013) did not assess their effect on the plant, despite the large arsenal of metabolites they produce may affect plant development. Moreover, often only in vitro tests are used to assess plant growth promoting (PGP) traits for strain characterization, and rarely BCA and PGP traits are evaluated in the presence of the host plant (Anwar et al., 2016).

One of the main limitations of historical biocontrol studies is that often the selection of strains is solely performed in vitro, which can result in the lack of activity in field conditions (Burr et al., 1996; Milus and Rothrock, 1997).

In an effort to set up a solid selection procedure of Streptomyces strains active against Fusarium head blight pathogens, the goal of this work was to characterize Streptomyces strains for both their PGP associated traits and their biocontrol activity, considering also tripartite interactions (plant, microorganism, pathogen) under different environmental conditions. To achieve this, the laboratory amenable pathosystems of FRR and FFR were used. This procedure proved successful in the identification of a Streptomyces sp. able to significantly limit FHB losses in field conditions.

## MATERIALS AND METHODS

#### Streptomyces Used in the Study

The collection of Streptomyces spp. maintained in the laboratory of Plant Pathology at the Department of Food, Environmental and Nutritional Sciences (DeFENS), University of Milan (Italy), hosts endophytic isolates from roots of different plants (Sardi et al., 1992) as well as from different sources (**Table 1**).

The twelve most active strains able to significantly inhibit FFR or FRR caused by F. graminearum (activity above 40%), identified in a comparative work of in vitro screening methods (Colombo et al., 2019), were selected for this study together with new isolates of diverse origin identified in this work (**Table 1**). Overall, twentyone strains were used.

#### Streptomyces Identification

Bacterial isolates DEF07, DEF09, DEF13, DEF14, DEF15, DEF16, DEF19, DEF20, DEF39, DEF41, DEF47, and DEF48 were identified in Colombo et al. (2019).

DNA from isolates DEF06, DEF08, DEF17, DEF18, DEF21, DEF31, DEF33, DEF40, and DEF46 was extracted following the method described by Sun et al. (2014). Briefly, a single bacterial colony was transferred to a sterile 1.5 mL tube containing 27 µL Tris (10 mM)-EDTA (1 mM) (pH 7.6); then, 3 µL KOH (0.4 M) – EDTA (10 mM) were added and incubated at 70◦C for 5 min. Next, 3 µL Tris-HCl (10 mM) (pH 4.0) were added to adjust the pH of the lysate. The lysate was used directly as a DNA template for the PCR amplification. 16S rRNA primers (Turner et al., 1999) used were 27F (5<sup>0</sup> -AGAGTTTGATCCTGGCTCAG-3<sup>0</sup> ) and rP2 (50 -ACGGCTACCTTGTTACGACTT-3<sup>0</sup> ). PCR was performed in a total volume of 50 µL, which contained 0.3 µL of GoTaq <sup>R</sup> DNA Polymerase 5 U/µL (Promega, United States), 10 µL of Green GoTaq <sup>R</sup> Reaction Buffer 5X (Promega, United States), 1 µL of 10 mM dNTP (Promega, United States), 1 µL of 10 µM primer forward, 1 µL of 10 µM primer reverse, 1 µL of template DNA and nuclease free water. The reaction conditions were initial denaturation at 95◦C for 5 min, followed by 35 cycles of denaturation at 95◦C for 20 s, annealing at 56◦C for 30 s and extension at 72◦C for 90 s. A final extension was performed at 72◦C for 7 min. Reaction products were separated by electrophoresis on a 1.5% agarose gel containing ethidium bromide and visualized under UV light. The PCR products were sequenced in both directions (Eurofins Genomics, Germany) using 27F and rP2 primers and two internal primers 16s\_p692f (50 -AATTCCTGGTGTAGCGGT-3<sup>0</sup> ) and 16s\_p782r (5<sup>0</sup> - ACCAGGGTATCTAATCCTGT-3<sup>0</sup> ). Assembled sequences were obtained with Geneious Prime 2019 (Biomatters, United States). EzBioCloud database was used to identify the strains based on 16S rRNA sequences (Yoon et al., 2017).

### Preparation of Bacterial Inoculum

Spores were collected after 2 weeks of incubation at 24◦C on Czapek Yeast Extract medium (CZY: 35 g/L czapek dox broth, Difco Laboratories, United States; 2 g/L yeast extract, Difco Laboratories, United States; 15 g/L agar, Amresco, United States; pH 6.5) scraping the surface of the colonies with a sterile loop and 5 mL of 10% sterile glycerol (ICN Biomedicals, United States) + 0.01% Tween20 solution (Sigma-Aldrich, United States). The concentration was determined using a hemocytometer and adjusted to 10<sup>7</sup> spores/mL. Small aliquots were then stored at −20◦C.

#### Antibiosis Assay

The antibiosis assay was performed using one medium and 22 treatments (21 Streptomyces strains + one water control). Three replicates were prepared. Briefly, 10 µL of Streptomyces spp. agar-spore suspension (10<sup>6</sup> spores/mL) or sterile water were

#### TABLE 1 | Streptomyces strains used in the study.

fmicb-10-02356 October 11, 2019 Time: 10:59 # 3


<sup>∗</sup>Strains identified in Colombo et al. (2019). Underlined strains were not originally isolated as endophytes.

inoculated on a Petri plate containing Wheat Meal Agar (WMA; Colombo et al., 2019). After 3 days, a plug of agar-mycelium (6 mm diameter) was taken from the edge of an actively growing colony of F. graminearum Fg8/1 (Boenisch and Schäfer, 2011) and inoculated upside down in the center of the plate at 25 mm distance from the bacterial strain. After a period of incubation (3 days at 24◦C in the dark), the antagonistic activity was assessed measuring the mycelial radial growth of the pathogen in the control (R1) and in the presence of the antagonist (R2). The percentage of mycelium growth inhibition compared to the control was calculated according to the Equation (1):

$$(\mathbb{R}1 - \mathbb{R}2)/\mathbb{R}1 \times 100\tag{1}$$

#### Screening for PGP Traits in vitro

The twenty-one strains involved in the study were screened for PGP characteristics. Indole acetic acid (IAA) production, tricalcium phosphate solubilization, siderophore and chitinase production, starch hydrolysis, nitrate reduction and growth in presence of salt were determined following the procedures described below.

The IAA production was evaluated following the method described in Bano and Musarrat (2003). Ten µL of Streptomyces strain spore suspension (1 × 10<sup>6</sup> spores/mL) were inoculated in 5 mL of CZY broth (35 g/L czapek dox broth, Sigma-Aldrich, United States; 2 g/L yeast extract, Difco Laboratories, United States; pH 6.5) adding 500 µg/mL of tryptophan (Sigma-Aldrich, United States). Three replicates were prepared. After a period of incubation (24◦C for 10 days at 125 rpm), the liquid cultures were centrifuged (10,000 rpm, 10 min, 15◦C) and 2 mL of supernatant were collected and mixed with 100 µL of 10 mM orthophosphoric acid (Carlo Erba Reagents, Italy) and 4 mL of Salkowski reagent (1 mL 0.5 M FeCl3, Sigma-Aldrich, United States; 49 mL 35% HClO4, Sigma-Aldrich, United States). The samples were incubated at room temperature for 20 min in the dark. The development of pink color indicated the IAA production. The absorbance of the samples was measured with a spectrophotometer (Perkin-Elmer Lambda 20, United States) at 530 nm. The concentration of IAA produced was calculated based on a standard curve of IAA obtained in the range of 1–50 µg/mL.

The ability to solubilize tricalcium phosphate was assessed in Petri plates (90 mm diameter) containing National Botanical Research Institute's Phosphate growth medium (10 g/L glucose, Sigma-Aldrich, United States; 5 g/L Ca3(PO4)2, Sigma-Aldrich, United States; 5 g/L MgCl<sup>2</sup> × 6H2O, Carlo Erba Reagents, Italy; 0.25 g/L MgSO<sup>4</sup> × 7H2O, Carlo Erba Reagents, Italy; 0.2 g/L KCl, Merck, Germany; 0.1 g/L (NH4)2SO4, Carlo Erba Reagents, Italy; 15 g agar, Amresco, United States; pH 7) as described in Nautiyal (1999), inoculated with 10 µL of spore suspension (1 × 10<sup>6</sup> spores/mL). Three replicates were prepared. After an incubation period (24◦C for 14 days), the halo was visually assessed. The halo width of 1 mm was marked with +, lack of halo was marked with −.

Siderophore production was observed using Chrome Azurol S agar overlay method (O-CAS) as described by Pérez-Miranda et al. (2007). The strains were grown on a modified CZY

medium without iron (pH 6.5) for siderophore production. Ten µL of agar-spore suspension (10 µL of spore suspension in 90 µL of 0.2% water agar) were inoculated in the center of a Petri plate (90 mm diameter) and kept for 14 days at 24◦C. Subsequently, 15 mL of Chrome Azurol S agar were cast upon culture agar plates (CAS agar: 60.5 mg/L Chrome Azurol S, Sigma-Aldrich, United States; 72.9 mg/L hexadecyltrimethyl ammonium bromide, Honeywell Fluka, Buchs, Switzerland; 30.24 g/L piperazine-1,4-bis(2-ethanesulfonic acid), Sigma-Aldrich, United States; 10 mL 1 mM FeCl<sup>3</sup> × 6H2O in 10 mM HCl, Sigma-Aldrich, United States; 9 g/L agar, Amresco, United States). Three replicates were prepared. In addition, a negative control was prepared using normal CZY medium. After 1 day of incubation at room temperature in the dark, the change of color around the colony (from blue to orange) indicated the siderophore production. The orange halo (D) and colony (d) diameters were measured and the production of siderophores was calculated according to the Equation (2):

$$(\mathbf{D} - \mathbf{d})/2 \tag{2}$$

Chitinase production (Kuddus and Ahmad, 2013) was assessed using chitin medium (6 g/L Na2HPO4, Carlo Erba Reagents, Italy; 3 g/L KH2PO4, Carlo Erba Reagents, Italy; 1 g/L NH4Cl, Carlo Erba Reagents, Italy; 0.5 g/L NaCl, Carlo Erba Reagents, Italy; 0.05 g/L yeast extract, Difco Laboratories, United States; 1% (w/v) colloidal chitin; 15 g/L agar, Amresco, United States; pH 6.5). Colloidal chitin was prepared adding 20 g of chitin (Sigma-Aldrich, United States) to 300 mL of 37% HCl (Merck, Germany). The chitin-HCl solution was kept for 60 min at 30◦C in continuous stirring and then precipitated adding 1 L of cold water. In order to allow the precipitation of the colloidal particles, the material was kept at 4◦C overnight and then collected by filtration on filter paper, washing with deionized water to bring up the pH at 6. Petri dishes (45 mm diameter) containing chitin medium were inoculated with 10 µL of agar-Streptomyces spore suspension (1 × 10<sup>6</sup> spores/mL). Three replicates were prepared. After an incubation period (24◦C for 14 days), the halo (D) and the colony (d) diameters were measured and the capacity to degrade chitin was expressed using Equation (2).

The ability to hydrolyze starch (Shirling and Gottlieb, 1966) was evaluated streaking a single colony of each Streptomyces strain on Petri dishes (90 mm diameter) containing ISP Medium 4 (Difco Laboratories, United States, pH 7.2) added with 10 g/L of soluble starch (Difco Laboratories, United States) and 1 mL of trace salts solution (per 100 mL: 100 mg FeSO<sup>4</sup> × 7H2O Merck, Germany; 100 mg MnCl<sup>2</sup> × 4H2O Carlo Erba Reagents, Italy; 100 mg ZnSO<sup>4</sup> × 7H2O Carlo Erba Reagents, Italy). Three replicates were prepared. After a period of incubation (24◦C for 14 days), the presence of the hydrolysis halo around the colonies determined the amylase activity.

The nitrogen reduction capability (Shirling and Gottlieb, 1966) was assessed by inoculating glass tubes containing 5 mL Bacto-Nitrate medium (13 g/L nutrient broth, Oxoid, Italy; 2 g/L KNO3, Carlo Erba Reagents, Italy; 2 g/L bacto agar, Difco Laboratories, United States; pH 6.5) with a single colony of each strain. Three replicates were prepared. After an incubation period (24◦C for 14 days), 200 µL of nitrate reagent A (α-naphthylamine, Sigma-Aldrich, United States) and B (sulfanilic acid, Sigma-Aldrich, United States) were added in each tube. The presence of nitrite was confirmed by the development of a red color after the formation of a diazonium salt caused by the reaction between the reagents A and B.

High salt concentration growth was evaluated streaking single colonies on Bennet's agar medium (Jones, 1949) (1 g/L yeast extract, Difco Laboratories, United States; 0.8 g/L lab-lemco, Oxoid, Italy; 10 g/L glucose, Sigma-Aldrich, United States; 2 g/L casitone, Difco Laboratories, United States; 15 g/L agar, Amresco, United States; pH 6.5) added with 3.5% or 7% (w/v) of NaCl (Carlo Erba Reagents, Italy). Three replicates were prepared. After 14 days of incubation at 24◦C, the growth of the strains in Petri dishes (45 mm diameter) was evaluated in comparison with control plates (0% NaCl).

#### Seed Treatments and Blotter Assay Germination

The twenty-one strains were tested for their potential growth promoting and biocontrol activities against FRR and FFR. Seeds of Triticum aestivum L. cv. Bandera were surface-sterilized in 0.7% sodium hypochlorite for 5 min and then rinsed 3 times in sterile water. In sterile Petri dishes, seeds (N = 40) were inoculated with 1 mL of Streptomyces strain spore suspension (10<sup>7</sup> spores/mL) and dried under the laminar flow hood. Control seeds were treated with 1 mL of deionized sterile water. For biocontrol experiments, seeds (N = 40) were treated in the same way. After 4 days, the seedlings were inoculated with an agarmycelium plug (6 mm diameter) taken from the edge of an actively growing F. graminearum Fg8/1 colony and inoculated upside down on the roots at a 10 mm distance from the seed. The assay took place in sterile glass dishes as seed trays (diameter 150 mm). In each dish, a filter paper was placed and soaked with 10 mL deionized sterile water before sowing. For each condition, four glass dishes containing 10 seeds arranged in three rows were prepared. The germination of the seeds followed the conditions described in Covarelli et al. (2013). Briefly, the dishes were placed at 5◦C in the dark for 24 h simulating a period of vernalization and then moved at 20◦C in the dark. Three days after seed bacterization, dishes were placed in a growth chamber (21◦C, 16 h photoperiod using fluora lamp osram L36W/77). Seedlings were watered with sterile deionized water every 2 days.

#### Evaluation of PGP and Biocontrol Effects in Germination Blotter Assay

Germination was assessed after 2 days of incubation in the growth chamber, when seeds were still in the dark to simulate normal germination process, while root and seedling length as well as root number were assessed after 3 and 10 days. At the 10th day, seedlings were dried and the root and shoot dry weight was assessed.

The biocontrol potential of the Streptomyces spp. against Fusarium root rot (FRR) and foot rot (FFR) was evaluated using F. graminearum Fg8/1 infected seedlings. Four days after pathogen inoculation the FRR was measured on 20 roots as necrosis development. FRR data were reported as millimeters

of necrosis extension. Percentages of necrosis inhibition were calculated using measurements of necrosis on the control (CN) and on the treated seedlings (TN) using the Equation (3):

$$(\text{CN} - \text{TN})/\text{CN} \times 100\tag{3}$$

Six days after seed bacterization, FFR was evaluated by scoring the symptoms at the crown level on 20 seedlings (Covarelli et al., 2013) with a 0–4 scale (0 = symptomless; 1 = slightly necrotic; 2 = moderately necrotic; 3 = severely necrotic; 4 = completely necrotic) (Colombo et al., 2019). The FFR disease severity was calculated for each treatment using the Equation (4):


(total number of plants × the highest disease grade) × 100 (4)

The ability of the antagonists to reduce symptom development was assessed with the Equation (5):

$$(\text{DC} - \text{DT})/\text{DC} \times 100\tag{5}$$

DC and DT were the disease severity in the control and the treated seedlings, respectively.

In addition, shoots from infected and control seedlings were dried and their weight was measured.

#### Streptomyces Biocontrol Activity Against FFR in Soil Substrate

Seed bacterization with Streptomyces spp. spore suspension was carried out as described above with strains that showed promising BCA features in vitro (DEF07, DEF09, DEF19, DEF20, DEF39, DEF47, and DEF48). DEF08 was used as a negative control, as it showed no FFR inhibition in the previous test. Conidia of F. graminearum Fg8/1 were produced in CMC medium (15 g/L carboxymethyl-cellulose, Sigma-Aldrich, United States; 1 g/L NH4NO3, Carlo Erba Reagents, Italy; 1 g/L KH2PO4, Carlo Erba Reagents, Italy; 0.5 g/L MgSO<sup>4</sup> × 7H2O, Carlo Erba Reagents, Italy; 1 g/L yeast extract, Difco Laboratories, United States; pH 6.5). Conidia were collected as described in Breakspear et al. (2011) after 5 days of incubation by filtering cultures through one layer of Miracloth (Calbiochem, United States) and centrifuging the filtrate for 10 min at 3000 rpm. Supernatant was discarded and the pelleted conidia were washed twice with sterile water (centrifuge 10 min, 3000 rpm).

Twenty inoculated seeds for each treatment (water or Streptomyces spp. spore suspension) were placed in sterile glass dishes (diameter 150 mm) to allow their germination at room temperature. After 6 days, wheat seedlings were transplanted in polystyrene seed trays (32 × 52 × 5.5 cm) containing sterile substrate (1:1 ratio of Irish peat and sand, pH 6.5, EC 0.2 dS/m, density 340 kg/m<sup>3</sup> , porosity 89% v/v, Vigorplant, Italy) watered with tap water. After that, roots were inoculated with one agar-mycelium plug (6 mm diameter) taken from a colony of F. graminearum Fg8/1 grown on V8 medium (Spanu et al., 2012). In addition, 1 mL of F. graminearum Fg8/1 conidia (1 × 10<sup>6</sup> conidia/mL) or a mixture of F. graminearum Fg8/1 (1 × 10<sup>6</sup> conidia/mL) + Streptomyces (5 × 10<sup>6</sup> spores/mL) was added to control plants or already bacterized plants, respectively (Simpson et al., 2000). Seedlings inoculated only with Streptomyces strains, as well as a non-inoculated seeds (wateronly) to be used as controls were prepared.

Plants were grown in a growth chamber (Conviron, Winnipeg, Canada) at 24◦C, 55% relative humidity and 15 h photoperiod, watered with tap water every 2 days. After 20 days, FFR disease symptoms were visually evaluated using a 0–4 scale (0 = symptomless; 1 = slightly necrotic; 2 = moderately necrotic; 3 = severely necrotic; 4 = completely necrotic) (**Supplementary File 1**). The FFR disease severity and protection level were calculated for each treatment using the Equations (4) and (5), respectively. Dried shoot weight of the infected seedlings was also assessed.

#### Streptomyces spp. Re-isolation From Inner Root Tissues and Evaluation of PGP Effect in Soil Substrate Assay

Control plants and plants inoculated only with Streptomyces (no-Fusarium inoculation) from the previous test were harvested 20 days after transplant and washed in sterile water to remove the excess soil. Shoot length and dried weight of wheat plants were assessed for each treatment.

For inner root tissue analysis, 10 seedlings for each treatment were selected and cut at the base. The roots were washed and surface sterilized with propylene oxide (Sigma-Aldrich, United States) for 1 h (Sardi et al., 1992). Subsequently, 10 or 15 root pieces were cut in sterile conditions and placed in water agar medium (WA) containing 15 g/L agar (Amresco, Italia), 25 mg/L nalidixic acid (Sigma-Aldrich, United States), 50 mg/L nystatin (Sigma-Aldrich, United States), and 50 mg/L cycloheximide (Sigma-Aldrich, United States). Plates were incubated for 7 days at 24◦C. Growth of Streptomyces spp. colonies on the plate was visually observed using a microscope. Morphological examination was carried out to confirm the re-isolation. Roots not inoculated with Streptomyces strains were used as negative control and subjected to the same procedure to check the presence of Streptomyces spp.

#### Evaluation of Biocontrol Activity Against FHB in Growth Chamber

Spring wheat (Triticum aestivum L.) cv. Bandera was cultivated in growth chamber following the procedure described in Watson et al. (2018) to speed up the plant development to reach anthesis in approximately 2 months. Briefly, seeds were sterilized as described before and placed in sterile glass dishes (diameter 150 mm) to allow their germination. After 3 days at 4◦C they were placed at room temperature for another period before sowing them in pots (21 × 13 × 15.5 cm, five seeds per pot) containing non-sterilized Irish and Baltic peat-based growth substrate (pH 6, EC 0.25 dS/m, density 120 kg/m<sup>3</sup> , porosity 90% v/v, Vigorplant, Italy). The lighting was set to 12 h light/12 h dark cycle for 4 weeks and then increased to an 18 h light/6 h dark photoperiod using fluora lamp osram L36W/77 until complete spike development. The temperature of

the growth chamber was set at 18◦C. Fusarium graminearum strain PH1 (Seong et al., 2009) was used to inoculate wheat heads. Bacterial spores of DEF09, which showed consistent biocontrol efficacy under all tested conditions, were prepared in CZY as described previously and F. graminearum conidia were prepared in CMC medium. The day of the treatment spores and conidia were collected and mixed with 0.01% Tween 20 (Sigma-Aldrich, United States) immediately before head inoculation. The final concentration of the mixture was 1 × 10<sup>7</sup> spore/mL for DEF09 and 1 × 10<sup>6</sup> conidia/mL for PH1. Ten µL of this mixture was used to inoculate the fifth centrally located spikelet from the bottom at anthesis. Three replicates were prepared for each treatment and arranged in a randomized block design. Three head treatments were performed: (1) F. graminearum, (2) F. graminearum + Streptomyces sp. DEF09, (3) Control (sterile distilled water + 0.01% Tween 20). Each spike was sealed in a plastic bag for 3 days. The FHB severity was visually estimated using a 0–100% scale 7 days after the treatment (Stack and McMullen, 1998). The average of FHB infection level was scored and the protection level calculated using the Equation (5).

#### Evaluation of Biocontrol Activity Against FHB in Field Conditions

In order to further verify the biocontrol effect of DEF09 against FHB under complex environmental conditions, a field trial was performed. Field trial was conducted in Travacò Siccomario, Pavia (45◦ 080 50.100N 9◦ 090 20.000E, Italy), during the growth season 2019. The spring wheat (Triticum aestivum L.) cv. Bandera and the durum wheat (Triticum turgidum L. ssp. durum**)** cv. Claudio (both susceptible to F. graminearum) were sown with a 200 kg/ha density at the end of October 2018 on a loamy soil (Sand 31.2%, Silt 47.5%, Clay 21.3, cation exchange capacity 21.3 cmol<sup>+</sup> kg DM−<sup>1</sup> , total organic carbon 1.51% DM, soil organic matter 2.60% DM, total Kjeldal nitrogen 0.19% DM, C/N ratio 7.95, P2O<sup>5</sup> Olsen 87 mg kg DM−<sup>1</sup> , where DM stands for dry matter) with neutral pH (7.1). The field was previously cultivated with soybeans. Nitrogen fertilization was 30 kg/ha at the sowing and 50 kg/ha before booting. Weeding was carried out with Arianne II (Corteva, Italy) the 15/03/2019 at a dose of 3.5 l/ha. DEF09 spores and F. graminearum PH1 conidia were freshly produced in the laboratory as described above and collected at the day of field inoculation. Biocontrol assays started at wheat anthesis stage (beginning of May). Flowering period of the two cultivars differed by 7 days. Spores were kept on ice (max 2 h) until inoculation. Thirty plants at anthesis stage were selected for each treatment. Controls included: (a) conidia of F. graminearum PH1 2 × 10<sup>6</sup> conidia/mL, (b) spores of Streptomyces DEF09 2 × 10<sup>7</sup> spores/mL; (c) sterile distilled water + 0.01% Tween 20 (Sigma-Aldrich, United States). The treatment consisted of bacterial suspension and conidia + 0.01% of Tween 20 (Sigma-Aldrich, United States), mixed before head inoculation. The final concentration of the mixture was 2 × 10<sup>7</sup> spores/mL for DEF09 and 2 × 10<sup>6</sup> conidia/mL for PH1. Ten µL of spore suspension per treatment were used to inoculate a single, centrally located spikelet at anthesis. Inoculation was arranged in a randomized block design. Wheat heads were evaluated after 30 days. The infected spikelets were counted and FHB disease severity was visually estimated using a 0–100% scale (Stack and McMullen, 1998) for both wheat cultivars. Protection level of DEF09 treatment was assessed using the Equation (5).

#### Statistical Analysis

Statistical analyses were performed using R software, version 3.5.1 (R Core Team, 2018), unless stated otherwise. To understand the effect of Streptomyces treatments on plant development and on FRR a Kruskal–Wallis test was applied, followed by a Dunn's test with Bonferroni's correction of the P-values to control the experiment-wise error rate (R package "dunn.test," Dinno, 2017). Unless stated otherwise P < 0.05 was considered significant.

In order to identify treatments able to protect seedlings from FFR symptom development in comparison to the untreated control, a Fisher's test was performed pooling the FFR symptoms in two groups [asymptomatic (class 0) or symptomatic (classes 1–4)]. Moreover, to assess also differences within the range of symptomatic seedlings, an additional Fisher's test was carried out comparing the group of mild symptomatic (classes 1–2) with the severely diseased group (classes 3–4). P < 0.01 for both tests was considered significant.

CORREL function [CORREL(x, y)] in Microsoft Excel was used to determine the correlation coefficient between the results of the dual culture assay and chitinase activity, FRR protection and FFR protection. The Equation (6) for correlation coefficient is:

$$
\sum (\mathbf{x} - \bar{\mathbf{x}})(\mathbf{y} - \bar{\mathbf{y}})/\sqrt{\sum (\mathbf{x} - \bar{\mathbf{x}})^2} \sum (\mathbf{y} - \bar{\mathbf{y}})^2 \tag{6}
$$

For field trials, a two-group analysis (Mann–Whitney test) using Estimation stats (Ho et al., 2018) was conducted for each cultivar on the number of diseased spikelets of control (PH1) and treatment (PH1 + DEF09). The results are presented on a Gardner-Altman estimation plots.

### RESULTS

#### Screening for Streptomyces Biocontrol and PGP Activities in vitro

Identification of the nine isolates not identified in a previous study (Colombo et al., 2019) by 16S rRNA confirmed that all 21 strains belong to Streptomyces spp. (**Table 1**).

Results of in vitro tests for physiological and biochemical features directly or indirectly involved in plant growth promotion are reported in **Table 2**. Chitinase activity is widespread among all strains, but it is not correlated with the ability to reduce F. graminearum mycelium development (r = 0.22). Low amount of IAA production was recorded at the tested conditions, except for DEF09 and DEF33 that produced 2.50 ± 0.04 and 7.51 ± 0.00 µg/mL of IAA, respectively. Siderophore production was observed for DEF06, DEF17, DEF18, and DEF46. The radius of the halo ranged from 3 to 36 mm and DEF46 showed the widest halo of siderophore production on CAS agar. Only DEF06, DEF17, and DEF21 were able to solubilize


TABLE 2 | The screening of plant growth promotion traitsin vitro andin planta(germination blotter or soil substrate assays).

 Underlined are the strains that were not originally isolated as endophytes. ∗Indicates significant difference (P < 0.05) from the control assessed with Dunn's test and Bonferroni correction for multiple comparison. nt Nottested. SD Standard Deviation.+, halo width 1 mm;−, not active. The growth at high salt concentration were compared to control plates (+grown like control plates; – not grown).


TABLE 3 | Screening tests for biocontrol activity against F. graminearum in vitro and in planta (germination blotter and soil substrate assays).

Underlined strains were not originally isolated as endophytes. <sup>∧</sup>Experimental data obtained from Colombo et al. (2019). nt Not tested. FRR Fusarium root rot. FFR Fusarium foot rot. The average of mycelium growth inhibitions recorded in dual culture assay against F. graminearum strain Fg8/1 is reported. In planta results of germination blotter assay or using soil are displayed as FRR protection (∗P < 0.05 is considered significant) and FFR protection (Fisher's test analysis: <sup>∗</sup>P < 0.01 treatment considered significantly able to maintain the seedling asymptomatic, ∗∗P < 0.01 treatment considered significantly able to increase disease severity). The average of infected shoot weight (mg) for each treatment assessed after 10 or 20 days after seed bacterization or transplant is also reported.

tricalcium phosphate on NBRIP medium. Starch hydrolysis was common among the strains except for DEF09, DEF13, DEF20, and DEF41. Eleven strains reduced nitrate at the tested conditions (**Table 2**). All strains except DEF33 were able to grow at 3.5% salt in the medium and 71% grew even at 7% salt concentration. The antifungal activity of the Streptomyces strains against F. graminearum Fg8/1 in dual culture assay varied from 41% inhibition for DEF31 to 70% inhibition for DEF07, DEF19, DEF20, and DEF48 (**Table 3**).

#### Evaluation of PGP Effects in Germination Blotter Assay

Under soilless conditions, none of the tested Streptomyces strains significantly altered the germination rate compared to the control plants, which had a germination percentage of around 99 (**Table 2**). A slight but significant reduction of the germination after seed bacterization was observed only for DEF17. Some strains inhibited the shoot and seminal root length 3 days after the seed bacterization (**Supplementary File 2**). After 10 days of incubation, an overall attenuation of these negative effects was observed (**Table 2**) except for DEF41, DEF46, DEF47, and DEF48, which still negatively affected both shoot and seminal root elongation (**Table 2**).

Ten days after seed bacterization, root number was not significantly different from the control with the exception of DEF41 and DEF09, which showed significantly lower number of roots compared to the control (4 versus 5) (**Supplementary File 2**).

To assess the potential gain/loss in biomass, root and shoot dry weight were also assessed after 10 days of growth (**Table 2**). Overall, the effect was minimal and a significantly lower weight was obtained only for shoots in plants treated with DEF17, while root dry weight was not significantly affected.

### Evaluation of Biocontrol Activity in Germination Blotter Assay

FRR was assessed 8 days after the antagonist inoculation. The results confirmed the biocontrol activity observed in vitro (**Table 3** and **Supplementary File 3**) with a correlation coefficient of r = 0.5. DEF07, DEF09, DEF16, DEF19, DEF20, DEF21, DEF31, DEF39, DEF41, and DEF48 significantly reduced the necrosis development on wheat roots in comparison with the untreated control (P < 0.05), showing up to 46% inhibition of necrosis.

The Fisher's test analysis of the FFR scores grouped in asymptomatic (0) and symptomatic (1-2-3-4) showed the ability

of DEF09 and DEF47 to maintain the seedlings healthy in comparison to the untreated control (P = 2.57e-08, P = 3.24e-05). Interestingly, seedlings treated with DEF08 showed more severe necrosis (45.45%) at the crown in comparison to the untreated control (P = 7.00e-04) (**Table 3** and **Supplementary File 4**). All P-values of Fisher's test analyses are reported in **Supplementary File 5**.

The strains with a capacity to reduce FRR did not reduce FFR symptoms in the same manner. The best performing strain against FRR was DEF19 (46.74%), while the best performing strain against FFR was DEF47 with protection percentages of 87.50%. Only DEF09 was able to control both symptoms of F. graminearum infection with high level of efficiency, resulting in approximately 80% inhibition of FFR development and > 40% in FRR development.

None of the non-endophytic strains showed the ability to effectively reduce the disease severity in planta with the exception of DEF31, which showed a partial efficacy against FRR only (29.14% reduction of necrosis extension).

In order to analyze if the BCA treatments were able to counteract the biomass loss following the infection, the shoots from infected seedlings were dried and weighed. Only DEF06 and DEF07 increased significantly the shoot weight compared to the Fusarium-treated control (**Table 3**).

#### Biocontrol and PGP Activities in Soil Substrate

To further verify whether the biocontrol and the PGP activities were consistent in a more complex environment – soil, and over a longer period of cultivation, 26 days-, FFR, stem shoot length and dried weight were evaluated for strains showing interesting biocontrol activities: DEF07, DEF09, DEF19, DEF20, DEF39, DEF47, and DEF48. DEF08 was used as negative control.

First, the colonization of inner root tissues by selected Streptomyces strains was verified. All root pieces (10/10) of wheat seedlings were extensively colonized by the tested Streptomyces strains on WA plates. They showed the ability to move in soil and internally colonize the plant, including DEF08 that was not originally isolated as endophyte.

The use of soil and the longer cultivation period until disease symptom evaluation and PGP analysis led to decreased BCA activity of most of the strains (**Table 3** and **Supplementary File 6**), with the exception of DEF07 and DEF09 which were able to significantly reduce FFR (61 and 46% level of protection, respectively) (**Supplementary File 7** for Fisher's test P-values). Plant growth promotion of non-infected plants colonized by the Streptomyces spp. strains was not significant for the two parameters analyzed (**Table 2**). Shoot dried weight of Fusariuminfected plants was, however, affected by some strains: DEF07, DEF09, DEF39, and DEF47 lowered the dried biomass in comparison with the Fusarium-treated control (**Table 3**).

#### Biocontrol Effects Against FHB Severity in Growth Chamber and Field Conditions

The strain DEF09 showed the best performance against both FRR and FFR diseases, with consistent results in all the

FIGURE 1 | Example of Fusarium head blight symptoms on wheat spikes grown in growth chamber. The red arrow indicates the spikelet of infection. Examples of water inoculated control (A), Fusarium inoculated control (B) and Fusarium + DEF09 treatment (C) are shown.

assays. Being an endophyte obtained from wheat, its efficacy against FHB disease was first assessed in fully controlled environment (growth chamber). The strain, co-inoculated with the pathogenic strain PH1, stopped the spreading of the disease at the first infected spikelet in all plants (**Figure 1**). High level of protection (75%) was reached under controlled conditions (**Supplementary File 8**).

In order to assess whether the strain could be effective also in field conditions, where different biotic and abiotic interactions occur, a field trial was carried out on bread and durum wheat. The P-value of the Mann–Whitney test was 2.47e-05 for Bandera and 1.35e-08 for Claudio. The presence of DEF09 reduced the number of diseased spikelets in comparison to the untreated control (**Figure 2**), decreasing FHB severity up to 60 and 45% on cv. Bandera and Claudio, respectively (**Supplementary File 9**).

### DISCUSSION

Comprehensive observation of different parameters, including the physiological characteristics of Streptomyces strains and their interaction with the plant (Colombo et al., 2019), is essential for successful selection and characterization of bioactive strains able to adapt to complex environmental conditions and microbiomes (Winter et al., 2019).

In this study, Streptomyces strains were extensively characterized for their plant growth associated features, together with detailed examination of their activity on germinating wheat and on wheat infected with F. graminearum.

The combination of in vitro and in vivo laboratory assays led to the identification of an effective strain, DEF09, which also showed promising results in field trials on both durum and bread wheat. The use of FRR and FFR pathosystems for selecting a strain effective against FHB proved successful. This study is in accordance with the observation by Wang L.-Y. et al. (2015), who showed a good correlation between FFR and FHB biocontrol activities for a diverse set of bacterial

FIGURE 2 | The mean difference between PH1 and PH1 + DEF09 for diseased spikelets (n◦ ) of cultivars "Bandera" (A) and "Claudio" (B) is shown in the Gardner-Altman estimation plot. The unpaired mean difference of data obtained between PH1 and PH1 + DEF09 is −3.73 (95.0% CI −5.13, −2.17) and −4.07 (95.0% CI −4.97, −3.03) for cv. Bandera and cv. Claudio, respectively. Both groups are plotted on the left axes; the mean difference is plotted on floating axes on the right as a bootstrap sampling distribution. The mean difference is depicted as a dot; the 95% confidence interval is indicated by the vertical error bar.

strains. It also confirms functional analyses of genes from wheatinfecting Fusarium species. Different genes were reported to be equally involved in the pathogenic mechanisms of both FHB and FFR (Spanu et al., 2012, 2018; Pasquali et al., 2013). From a physiopathological point of view, F. graminearum shows a common infection process during both root- and head infection (Wang Q. et al., 2015; Wang et al., 2018). Our work therefore supports the idea that the use of FRR and FFR pathosystems, being more manageable laboratory models than the FHB pathosystem, is suitable for selection of BCA strains effective against FHB.

In this work it was not possible to include strains previously selected as BCA in other scientific works therefore it is not possible to have a direct comparison of the activity of the strain DEF09 with other Streptomyces strains, given that results depend on the complex interactions occurring in the environment (Vurukonda et al., 2018). Nonetheless, based on the reported efficacy of the different microorganisms, the level of protection achieved by the strain DEF09 was comparable to that obtained in field trials using Bacillus sp. and Cryptococcus sp. (Schisler et al., 2002) and slightly higher than those achieved with other Streptomyces strains in field trials on bread wheat (Jung et al., 2013; Palazzini et al., 2017) and durum wheat (Palazzini et al., 2018). It is plausible that the inoculation method may affect the level of protection. Interestingly, Jung et al. (2013) reported significant protection against FHB by the BN1 Streptomycesstrain only when the strain was sprayed on spikes but not when it was co-inoculated. In our case, the high level of protection, comparable with fungicide treatments (Giraud et al., 2011), was obtained with co-inoculation. Other inoculation methods will need to be tested to better compare the level of protection obtained by DEF09 in different environmental conditions with that of previously studied strains. Novel approaches are also needed to explore the efficacy of the strains in large scale field trials.

The combination of the methods used to assess the bioactivity of the strains examined in our study allowed us to gain insight into their possible mechanisms of activity. For example, the in vitro assays carried out on DEF09 suggest that this strain blocks the growth of the fungus with specific antifungal molecules, as shown by the dual culture inhibition assay. Chitinase production has been identified as the main biocontrol mechanism in some studies (Herrera-Estrella and Chet, 1999). DEF09 is a chitinase producer, but the lack of correlation (r = 0.22) between the chitinase production in different strains and the growth inhibition of F. graminearum indicates that chitin degradation may not be the unique factor responsible for the observed bioactivity of the strains. Likely, the inhibition of fungal growth might be the result of a synergistic effect of different lytic enzymes and metabolites (Zhang and Yuen, 2000; Zhao et al., 2013). DEF09 directly affects wheat plant growth, modifying root development by way of seminal root elongation, as seen in the germination blotter assay after 10 days, and impacting overall plant growth (shoot dried weight) after pathogen infection. Interestingly, morphological changes of roots have been associated with the induction of systemic resistance (Zamioudis et al., 2013). Moreover, DEF09 was among the best IAA producers in the pool. Indeed, IAA is known to play a role in plant morphology as well as in disease modulation, stimulating plant defense (Pieterse et al., 2009) and therefore may contribute to protection against the pathogen. From all these data we may infer that DEF09 possesses multiple mechanisms leading to limitation of FHB on wheat. Metabolic profiling coupled with functional genomics of DEF09 will likely allow for the delineation of the mechanisms of action for the strain (Chen et al., 2018).

A large set of potentially bioactive strains was identified in this study. Strains able to significantly interfere with pathogen development also transiently affected plant growth, suggesting that a complex set of molecules is produced during the tripartite interaction (Mayo-Prieto et al., 2019). Our future goal will be to identify the determinants of these

specific interactions occurring among the BCA, the fungus, and the host, as detailed knowledge of their interaction is essential for developing novel plant protection strategies (Berendsen et al., 2012).

#### DATA AVAILABILITY STATEMENT

The datasets generated for this study can be found in NCBI GenBank, reported in **Table 1**. The raw data and additional figures are available in **Supplementary Material**.

#### AUTHOR CONTRIBUTIONS

MP, EC, MS, and AK contributed to the conception and design of the study. EC, MS, AK, and CP performed the experiments. CP and EC performed the statistical analysis. EC and MP wrote the first draft of the manuscript. AK, MS, PC, and CP wrote sections of the manuscript. All authors contributed to manuscript revision, read and approved the submitted version.

### REFERENCES


#### FUNDING

The work was supported by the BioHIT project, University of Milan. The Department of Food, Environmental and Nutritional Sciences, Università degli Studi di Milano, partially covered the Open Access APC.

#### ACKNOWLEDGMENTS

We acknowledge Elio Burrone for technical help, Alessandro Infantino for providing wheat seed, and Wilhelm Schäfer for providing Fg8/1 strain. Corby Kistler and Karen Broz are thanked for the precious revision of the discussion.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.02356/full#supplementary-material



**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 © 2019 Colombo, Kunova, Pizzatti, Saracchi, Cortesi and Pasquali. 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.

# Enhancement of Pathogen Resistance in Common Bean Plants by Inoculation With *Rhizobium etli*

*Armando Díaz-Valle, Alberto Cristian López-Calleja and Raúl Alvarez-Venegas\**

*Centro de Investigación y de Estudios Avanzados del Instituto Politécnico Nacional, Unidad Irapuato, Guanajuato, Mexico*

Symbiotic *Rhizobium*-legume associations are mediated by exchange of chemical signals that eventually result in the development of a nitrogen-fixing nodule. Such signal interactions are thought to be at the center of the plants' capacity either to activate a defense response or to suppress the defense response to allow colonization by symbiotic bacteria. In addition, the colonization of plant roots by rhizobacteria activates an induced condition of improved defensive capacity in plants known as induced systemic resistance, based on "defense priming," which protects unexposed plant tissues from biotic stress. Here, we demonstrate that inoculation of common bean plants with *Rhizobium etli* resulted in a robust resistance against *Pseudomonas syringae* pv. *phaseolicola*. Indeed, inoculation with *R. etli* was associated with a reduction in the lesion size caused by the pathogen and lower colony forming units compared to mock-inoculated plants. Activation of the induced resistance was associated with an accumulation of the reactive oxygen species superoxide anion (O2 −) and a faster and stronger callose deposition. Transcription of defense related genes in plants treated with *R. etli* exhibit a pattern that is typical of the priming response. In addition, *R. etli*–primed plants developed a transgenerational defense memory and could produce offspring that were more resistant to halo blight disease. *R. etli* is a rhizobacteria that could reduce the proliferation of the virulent strain *P. syringae* pv. *phaseolicola* in common bean plants and should be considered as a potentially beneficial and eco-friendly tool in plant disease management.

Keywords: induced systemic resistance, priming, nodule, *Rhizobium etli*, *Pseudomonas*

## INTRODUCTION

Over the last few decades, particularly after the Green Revolution, the global agricultural production increased considerably. However, due to the sustained increase in the global human population, the demand for higher crop production has also increased substantially. Accordingly, to achieve higher agriculture yields, farmers have adopted the extensive application of chemical fertilizers and pesticides, resulting in soil degradation and decrease in soil fertility (Gouda et al., 2018). In addition, nitrous oxide (N2O), which is a by-product of excess nitrogen (N) fertilization, is a chemical pollutant and contributes to global warming as it is the single most important ozone-depleting substance (Ravishankara et al., 2009; Vejan et al., 2016). Also, the excess application of ammonium nitrate in soils leads to a decline in symbiotic interactions established between microbes and legume plants (Vejan et al., 2016). Therefore, reducing the use of synthetic chemical pesticides and fertilizers is one of the challenges in modern society.

#### *Edited by:*

*Brigitte Mauch-Mani, Université de Neuchâtel, Switzerland*

#### *Reviewed by:*

*Andrés Gárriz, CONICET Institute of Biotechnological Research (IIB-INTECH), Argentina Kalliope K. Papadopoulou, University of Thessaly, Greece*

*\*Correspondence:*

*Raúl Alvarez-Venegas raul.alvarez@cinvestav.mx*

#### *Specialty section:*

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

*Received: 18 April 2019 Accepted: 23 September 2019 Published: 22 October 2019*

#### *Citation:*

*Díaz-Valle A, López-Calleja AC and Alvarez-Venegas R (2019) Enhancement of Pathogen Resistance in Common Bean Plants by Inoculation With Rhizobium etli. Front. Plant Sci. 10:1317. doi: 10.3389/fpls.2019.01317*

Some of the strategies involved in sustainable crop production and integrated pest management include development of novel cultivars exhibiting disease resistance, abiotic stress tolerance, and high nutritional value, application of genetic engineering tools to develop non-legume crops participating in N-fixing symbioses, and the use of biological control agents and plant growth promoting rhizobacteria (PGPR) (Bruce, 2010; Ubertino et al., 2016; Gouda et al., 2018). The use of PGPR is one of the most promising tools for disease management in contemporary agriculture and an alternative that could minimize reliance on agrochemicals (Timmermann et al., 2017). PGPR can be used to promote plant growth and development and enhance plant health without negatively affecting natural ecosystems and biodiversity and without environmental contamination (Calvo et al., 2014; Vejan et al., 2016). With regard to their association with root cells, PGPR could be classified into either intracellular PGPR (iPGPR, symbiotics) or extracellular PGPR (ePGPR, free living) (Martínez-Viveros et al., 2010). iPGPR may inhabit specialized root structures such as root nodules, while ePGPR may be found on root surfaces, in the rhizosphere, or in between root cortex cells (Martínez-Viveros et al., 2010; Gouda et al., 2018).

PGPR have numerous direct or indirect mechanisms for preventing growth and development of pathogens, such as antagonism, competitive exclusion, production of antibiotics, or extracellular secondary metabolites that inhibit pathogen growth, signal interference, competition for nutrients/resources (e.g., fixed N, ferric ion by siderophore production), hormone production, and the activation of the induced systemic resistance (ISR) plant defense system (Haas and Défago, 2005; Glick, 2012; Timmermann et al., 2017).

ISR is defined as an induced condition of enhanced defensive capacity in plants, which is activated in response to biological or chemical stimuli (e.g., colonization of the roots by PGPR or beneficial fungi or treatment with chemical compounds), and protects unexposed plant tissues against pathogenic invasion and insect herbivory (Pieterse et al., 2014; Timmermann et al., 2017; Gouda et al., 2018). ISR aids plants to fight numerous diseases and is effective against a wide range of pathogenic bacteria and fungi, and the activation of the defense mechanisms acts systemically in plant sections that are spatially separated from the inducer (Pieterse et al., 2014).

Activation of ISR by advantageous microorganisms is often based on priming (Pieterse et al., 2014), a physiological state that prepares the plant for a more rapid and/or greater activation of cellular defenses when exposed to biotic or abiotic stress, resulting in an enhanced level of resistance (Conrath, 2009; Conrath, 2011). In 1991, Van Peer and colleagues revealed that priming of defense responses is involved in ISR. In their study, the roots of carnation plants (*Dianthus caryophyllus*) were "bacterized" (primed) with *Pseudomonas* sp. strain WCS417r prior to inoculation with the pathogen *Fusarium oxysporum* f. sp. *dianthi*. Primed plants (bacterized) showing ISR exhibited enhanced synthesis and accumulation of phytoalexins following inoculation with *F. oxysporum*, when compared with noninduced control plants (Van Peer et al., 1991). Furthermore, induced resistance against pathogens has been associated with a faster and stronger accumulation of callose (Ton et al., 2005). For example, β-amino-butyric-acid (BABA) induced resistance in *Arabidopsis* correlates with primed deposition of callose-rich papillae (or enhanced accumulation of callose; Ton and Mauch-Mani, 2004).

Callose-containing cell-wall appositions, named papillae (where antimicrobial compounds can be deposited; Luna et al., 2011), are induce at the sites of pathogen attack, in primed plants, at early stages of pathogen invasion, and could also contribute to disease resistance by strengthening the plant cell wall at the site of pathogen invasion (Kohler et al., 2002).

In *Arabidopsis* plants primed by ISR (following colonization by PGPR), activation of the induced response involves the activation of the salycilic acid (SA)–, jasmonate (JA)-, and ethylene (ET)-signaling pathways. For example, priming with the PGPR *Paraburkholderia phytofirmans* PsJN reduces plant susceptibility to disease and the proliferation of the virulent strain *Pseudomonas syringae* pv. tomato DC3000 (PstDC3000) (Timmermann et al., 2017). In addition, *P. phytofirmans* PsJN induces transcriptional changes in key defense-related genes of the SA-, JA-, and ET-signaling pathways (e.g., *PATHOGENESIS RELATED 1*, *PLANT DEFENSIN 1.2*) in plants infected with PstDC3000. Also, the plant hormone abscisic acid (ABA), which regulates numerous plant developmental processes and adaptive responses to stress, has been involved in priming by beneficial microbes for enhanced callose deposition (Ton et al., 2005; Asselbergh et al., 2008; Pieterse et al., 2014). Furthermore, brassinosteroids modulate plant defense responses against pathogens and protect plants from environmental stresses, independently of SA-mediated defense signaling and PR gene expression (Krishna, 2003; Bari and Jones, 2009).

However, the activation of the different hormone-signaling pathway(s) following the application of the ISR-inducing PGPR seems to be specific to the plant species, as well as the pathogen and PGPR involved (De Vleesschauwer and Höfte, 2009; Timmermann et al., 2017). Basic research on the application of PGPR and in the exploitation of existing tools and techniques for safer and productive agricultural practices remains limited (Gouda et al., 2018).

Legume plants, particularly the common bean (*Phaseolus vulgaris* L.), are great model systems for studying ISR, as well as defense-priming and plant-pathogen interactions. For instance, common bean interacts with symbiotic diazotrophic bacteria (rhizobia). The symbiotic interaction results in the formation of new plant organs, N-fixing nodules, which fix atmospheric N (Taté et al., 1994). Therefore, symbiotic N-fixing rhizobacteria have positive effects on the host plants, not only facilitating nutrient availability but also priming defense against biotic and abiotic stresses *via* activation of ISR (Franche et al., 2009; Gouda et al., 2018).

To determine if rhizobacteria elicits the ISR in common bean against *P. syringae* pv. *phaseolicola* (PspNPS3121), the causal agent of the halo blight disease, we evaluated the effect of using *R. etli* to induce resistance based on pathogen accumulation, symptom emergence, nodule number, N fixation, callose deposition, and changes in levels of expression of defense related genes, within and across generations.

### MATERIALS AND METHODS

#### Plant Material

*P. vulgaris* cultivar "Rosa de Castilla" (Jiménez-Hernández et al., 2012), a type of bean produced under rainfed conditions and susceptible to the halo blight bacterial disease, was used in the present study. Seeds were surface sterilized (15 min in 2% NaOCl), rinsed three times with sterile deionized water, and germinated under sterile conditions, in moist paper towels, at 28°C. Three days after germination (dag), the seedlings were transferred to 2.3-L pots containing vermiculite, inoculated with 4 ml of the *Rhizobium etli* CE3 suspension, and placed in a greenhouse (Guanajuato, México, 101°09′01″ W, 20°30′09″ N; 1730 masl; as described by Martínez-Aguilar et al., 2016). The experiments were performed under daylight conditions, with maximum and minimum temperatures of 30 and 18°C, respectively. All inoculated plants were fertilized once a week with a B&D nutrient solution (Broughton and Dilworth, 1971) without N. Non-inoculated plants were fertilized with a B&D solution supplemented with N (8 mM KNO3; Estrada-Navarrete et al., 2007).

#### Rhizobacteria Growth Conditions and Inoculation

*R. etli* strain CE3 was grown at 28°C on PY media (5 g/L of peptone, 3 g/L of yeast extract, 0.7 g/L of calcium chloride) supplemented with 20 μg/ml nalidixic acid, 100 mg/ml streptomycin (Cárdenas et al., 2006), and 7 mM CaCl2 until the bacterial culture reached an OD600 of 0.5–0.6. The bacterial culture was centrifuged (5,000 rpm; Sorvall GSA Rotor), the pellets washed in 10 mM MgSO4, and the cells resuspended in 10 mM MgSO4 (Cárdenas et al., 2006). Subsequently, 3-day-old roots from "Rosa de Castilla" cultivar were inoculated, by drench, with 4 ml of the bacterial resuspension.

### Pathogen Infection

Pathogen infection in F0 plants was performed as described by Martínez-Aguilar et al., 2016. Briefly, *P. syringae* pv. *phaseolicola* NPS3121 (PspNPS3121) was cultivated for 36 h at 28°C on KB media containing 50 μg/ml rifampicin; then, the cells were resuspended in MgCl2 (10 mM) at a final concentration of 5 × 108 colony-forming units per milliliter (CFU/ml). Four milliliters of the bacterial solution were placed in a syringe without a needle, and the abaxial surfaces of the second trifoliate leaves were infiltrated with the pathogen 17 dag. Six infiltration points per leaf were employed. Negative control plants were neither treated nor infiltrated, while positive control plants were not infiltrated but treated with *R. etli*. To determine the systemic effect, samples were obtained from distal leaves not exposed to the pathogen, 1 day before infection, and 1 and 5 days after infection. **Table 1** presents an outline of the experimental design.

### Disease Evaluation

Bacterial growth was assessed 10 days after infection as follows: three leaf discs adjacent to the infection sites were excised using TABLE 1 | Experimental design.


*dag, days after germination.*

a 1-cm diameter stainless steel cork borer, then rinsed and homogenized with sterile deionized water, and plated in serial dilutions on KB media containing 50 μg/ml rifampicin and 6 mM MgSO4. The plates were incubated for 36 h at 28°C, and the total numbers of CFUs from three plates for each dilution were counted. The percentage of leaf damage (total chlorotic leaf area/total leaf area) and lesion size (total chlorotic and necrotic leaf area/total leaf area), on leaves that had been exposed to the pathogen, was determined using Fiji software (Schindelin et al., 2012).

### Callose Quantification

To determine callose deposition in leaves of common bean plants, we followed the protocol described by Luna et al. (2011) with slight modifications. Seventeen days old *P. vulgaris* plants (co-cultivated with the symbiont) were immersed, for 30 s, in a bacterial solution (MgCl2 10 mM, Silwet L-77 0.05%, and PspNPS3121 at a concentration of 5 × 108 CFU/ml). After 24 h of pathogen infection, the plants were destained for 24 h in 95% ethanol, washed in 0.07 M phosphate buffer (pH 9), and incubated for 24 h in 0.07 M phosphate buffer containing 0.01% aniline blue (Sigma–Aldrich, catalog #415049). Imaging of callose deposition in three plants per treatment, from two leaves per plant (and four discs per leaf, with a diameter of 1 cm each), was performed with an Olympus BX50 microscope (UV filter BP 340–380 nm, LP 425 nm), and pictures were taken with a Lumenera INFINITY 3 camera. Quantification and callose intensity were determined by using the Fiji software (https://imagej.net) and following the protocol described by Jin and Mackey (2017).

### Histochemical Detection of Superoxide Radical

To detect the accumulation of the superoxide anion (O2 −), a reactive oxygen species, we followed the protocol described by Timmermann et al. (2017) with some modifications. Aerial parts of 17 days old *P. vulgaris* plants (co-cultivated with the symbiont) were immersed, for 30 s, in a bacterial solution (MgCl2 10 mM, silwett L-77 0.05%, and PspNPS3121 at a concentration of 5 × 108 CFU/ml). After pathogen infection (24 h), leaves were removed from plants at 1.5, 3, 6, and 9 h and incubated for 24 h at room temperature in a phosphate buffered saline solution (137 mM NaCl; 2.7 mM KCl; 10 mM Na2HPO4; 2 mM KH2PO4) containing 0.06 mM nitroblue tetrazolium (Sigma–Aldrich, catalog # N6639). The leaves were then cleared in 80% ethanol (50 ml) at 60°C to extract chlorophylls (Timmermann et al., 2017). Imaging of purple formazan deposits (visualized as a blue precipitate), which result from the reaction of nitroblue tetrazolium with O2 − (Carvalho et al., 2008), identified the areas where O2 − was accumulated. Images were taken using a digital camera Nikon 5500 (Nikon Corp., Tokyo, Japan). Quantification of O2 − was performed using the protocol described by Juszczak and Baier, 2014. Each experiment was independently performed in triplicate.

#### Acetylene Reduction Assay

Acetylene reduction assays were performed using whole nodulated roots 24 days post-inoculation (dpi) according to Barraza et al. (2018). Each experiment was performed in triplicate. Nitrogenase-specific activity was expressed as nmol−1 of C2H2 g−1 h−1 of nodule fresh weight (Vessey, 1994).

#### Candidate Defense Genes With Enhanced Transcription in Response to Colonization of the Roots by *R. etli*

To identify genes in common bean that were potentially activated by ISR and that confer resistance to plant pathogens, we carried out searches in studies published on the subject of ISR and defense priming for the genes (according to Mayo et al., 2016). The following *P. vulgaris* genes were selected (orthologues of *Arabidopsis* genes): *ERF6* (*Phvul.002G055800*), *PR10* (*Phvul.002G209500*), *MYC2* (*Phvul.003G285700*), *NPR1* (*Phvul.006G131400*), *PR1* (*Phvul.006G196900*), *WRKY29* (*Phvul.002G293200*), *WRKY70* (*Phvul.009G043100*), *PAL2* (*Phvul.007G150500*), and *WRKY33* (*Phvul.008G251300*).

In addition, we analyzed the transcript levels of some of the genes coding for enzymes directly involved in plant hormone synthesis. Specifically, we determined the transcript levels of the gene orthologs to the *Arabidopsis AMI1* (*Phvul.007G180900*) and *NIT1* (*Phvul*.*011G096700*), both involved in auxin biosynthesis (Barraza et al., 2015b); *AAO3* (*Phvul.008G210000*), an aldehyde oxidase that catalyzes the last step of ABA biosynthesis and a marker for the sites of ABA production (Koiwai et al., 2004) and ABA biosynthesis in leaves (Seo et al., 2004); and *BR6OX2.2* (*Phvul.004G041700*), which catalyzes the last reaction in brassinosteroid (BR) biosynthesis (Kim et al., 2005). The importance of the genes was evaluated by the quantification of their transcripts *via* qPCR in response to *R. etli* and pathogen infection.

#### Quantitative PCR (qPCR) Analysis

qPCR was performed according to Martínez-Aguilar et al. (2016). Data from qRT-PCR experiments were analyzed based on the relative quantification method, or the 2−ΔΔCT method (Livak and Schmittgen, 2001). Fold changes in the target genes were normalized to *PvAct11* (Borges et al., 2012) and *PvTUB* (Livak and Schmittgen, 2001; Barraza et al., 2013, Barraza et al., 2015a), as endogenous controls, and relativized to the expression in control samples (relative gene expression in control plants was defined as 1). All experiments were run in triplicate (technical replicates) from three biological replicates. For a list of all primers used, see **Supplementary Table ST1**. Statistical significance was determined as described by Martínez-Aguilar et al. (2016).

#### Transgenerational Inheritance of Priming

Additional experiments were conducted in the F1 progeny to explore the transgenerational priming effect. First, all common bean plants (F0 generation) from all the different treatments (rhizobacteria only, rhizobacteria plus pathogen, pathogen only, rhizobacteria plus water, and without treatment) were self-pollinated and grown to seed set to generate the F1 progeny. Subsequently, for the transgenerational priming analysis, seeds from the F1 progenies were germinated and exposed to the pathogen only, without rhizobacteria treatment. Plants were inoculated with the pathogen at the age at which the parental lines were infected (17 dag). Samples from inoculated plants were obtained from distal leaves that had not been exposed to the pathogen 24 h before infection, and 24 h and 120 h after infection (or 16, 18, and 22 dag, respectively). All samples were obtained in triplicate and stored at −80°C.

#### RESULTS

#### Inoculation With *R. etli Protects P. vulgaris* Against PspNPS3121

As shown in **Figure 1A**, inoculation of the symbiont *R. etli* in cultivar "Rosa de Castilla" before pathogenic bacteria inoculation resulted in a considerable increase in resistance to halo blight disease caused by PspNPS3121. Treatment with *R. etli* induced a 75% reduction in lesion size (Re/Ps), compared to plants treated with the pathogen only (-/Ps; **Figure 1B**). Also, bacterial populations in leaves from plants treated with *R. etli* and exposed to the pathogen (Re/Ps) were significantly lower, approximately 50% less, than in plants treated with the pathogen only (-/Ps; **Figure 1C**). We inferred that common bean plants treated with the *R. etli* were efficiently protected against PspNPS3121, in contrast to plants that were not primed.

Given that infection development induces oxidative stress, we performed a fast and efficient histochemical assay to examine activation of the induced resistance and to detect accumulation of the reactive oxygen species (ROS) superoxide anion (O2 −) in common bean plants at 1.5, 3, 6, and 9 h after PspNPS3121 infection. Plants inoculated with *R. etli* and exposed to the pathogen accumulated 3–4 times more O2 − than control plants (**Figure 2**), and a high accumulation of O2 − was observed 3 h after infection at the whole-leaf level. The above indicates that treatment with *R. etli* primes common bean plants for improved induction of certain cellular defense responses.

#### Effect of *R. etli* Inoculation on Plant Height, Nodule Number and Size, and Acetylene Reduction After Pathogen Infection

Plant height in all treatments was quantified before pathogen inoculation and 13 days after infection (17 and 30 dag,

respectively; **Figure 3**) by measuring the length of the shoot systems from 24 plants per treatment. Plants inoculated with *R. etli* and exposed to PsNPS3121 were similar in size to plants that were only inoculated with the symbiont, or to the control plants. There were no significant differences in plant height after 13 dpi.

Subsequently, at 13 dpi with PsNPS3121 (or 30 dag), the plants were extracted, the soil removed, and the lengths of the root systems measured. Root system lengths were not influenced by the different treatments (**Figure 4A–B**). In addition, we counted the total number of nodules from six plants from each of the different treatments: (i) *R. etli* plus PsNPS3121 (Re/Ps), (ii) *R. etli* plus water (Re/H2O), and (iii) *R. etli* alone (Re/-). This experiment showed that the numbers of nodules among different treatments are not significantly different (**Figure 4C-D**). We further analyzed the effect of PsNPS3121 on nodule formation by measuring nodule diameter in six plants in each of the different treatments. Nodule diameters from PsNPS3121 infected plants were not significantly different compared to control nodules (Re/ H2O and Re/-; **Figure 4E**).

Subsequently, we evaluated nitrogen fixing activity in nodules. By measuring nitrogen fixation (nitrogenase activity), *via* an acetylene reduction assay, we observed that all plants from the Re/Ps treatment had capacity for nitrogen fixation similar to the control nodulated roots (**Figure 4F**).

leaves from plants co-cultivated or not with *Rhizobium etli*, after exposure to PspNPS3121. (B) Quantification of O2 − was performed using the protocol described by Juszczak and Baier, 2014. Images were taken using a digital camera Nikon 5500 (Nikon Corp., Tokyo, Japan). Statistical significance was determined using one-way ANOVA followed by Tukey's test, p < 0.05. Means with the same letter are not significantly different.

### *R. etli* Prime Plants for Enhanced Expression of Defense Related Genes

Next, we performed a methodical analysis to select illustrative common bean genes, based on their involvement in plant defense and priming, for their expression analysis by qPCR (according to Mayo et al., 2016). After conducting an exhaustive literature review, we selected and identified the corresponding genes at the NCBI and Phytozome databases. Based on transcriptomic data available at the Phytozome, several genes were selected according to their expression patterns (genes with low or no expression in leaves were eliminated). After we confirmed their expression in common bean leaves, three types of genes were selected: (a) genes highly expressed in leaves, (b) genes exhibiting medium expression levels in leaves, and (c) genes exhibiting low expression levels in leaves (**Table 2**).

Accordingly, we examined the effect of *R. etli* inoculation (priming) on the transcription levels of nine genes related to ISR and defense and potentially involved in priming. Samples

significantly different.

were obtained from plants treated with *R. etli* (Re/-) only, with the pathogen (Ps/-) only, or during the interaction with *R. etli* and exposed to the pathogen (Re/Ps). After qPCR analysis, the transcript levels of only three genes from plants that had been primed with *R. etli* and inoculated with PspNPS3121 showed a transcriptional pattern characteristic of the priming response (e.g., biphasic curve). In systemically resistant leaves of *R. etli*–treated plants, priming alone did not enhance transcription of *PvWRKY33, PvERF6,* and *PvPAL2*. However, 120 h after pathogen inoculation, there was high accumulation of transcripts, compared to the control plants (**Figure 5** and **Supplementary Figure SF1**). WRKY33 is a transcription factor (TF) involved in the activation of the salicylic acid (SA)– related host response (Birkenbihl et al., 2012); ERF6 is a TF that activates expression of PR genes (Huang et al., 2016), whereas PAL2 functions in plant secondary metabolites synthesis and callose deposition (Kohler et al., 2002). Consequently, *R. etli* primed *P*. *vulgaris* plants for potentiated gene activation, which was subsequently induced by PspNPS3121 infiltration.

On the contrary, there was enhanced transcription of *PvPR1* in the late stages of priming (120 h after bacterial inoculation; **Supplementary Figure SF2**). Such a transcriptional pattern, however, was also induced by the pathogen alone. Therefore, the rest of the genes could not be considered as having been primed, under the present experimental conditions.

#### *R. Etli* Activates the Expression of Genes Involved Plant Hormone Biosynthesis

Plant hormones regulate numerous developmental processes and signaling networks involved in plant responses to different types of biotic and abiotic stresses (Bari and Jones, 2009). Furthermore, the study of key genes and enzymes involved in

plant hormone biosynthesis provides important information concerning the regulation of plant hormones pathways and can offer new information regarding the function of plant hormones during plant-microbe interactions (Seo et al., 2004). Thus, we determined the transcript levels of some key genes involved in the synthesis of auxins (IAA; involved in the modulation of defense and development responses; Bari and Jones, 2009), ABA (involved in biotic stress; Ton and Mauch-Mani, 2004; Ton et al., 2009), and BR (implicated in induced systemic tolerance to biotic stress and in the modulation of plant defense responses; Bari and Jones, 2009; Li et al., 2013) (**Figure 6** and **Supplementary Table ST1**). Specifically, we determined the transcript levels of the *AMI1* (*INDOLE-3-ACETAMIDE HYDROLASE 1*), *NIT1* (*NITRILASE 1*), *AAO3* (*ABSCISIC ALDEHYDE OXIDASE 3*), and *BR6OX2.2* (*BRASSINOSTEROID-6-OXIDASE 2 ISOFORM 2*). Compared to control plants (**Figure 6** and **Supplementary Figure SF3**), the transcripts of *PvAAO3* (which codes for an enzyme that catalyzes the last step in the ABA biosynthesis pathway, from abscisic aldehyde to ABA; Koiwai et al., 2004; Seo et al., 2004) and *PvBR6OX2.2* (enzyme that catalyzes the last step in the production of brassinolide; Kim et al., 2005) increased in


TABLE 2 | Genes selected for defense response and priming, and their experimental expression during ISR in *Phaseolus vulgaris* leaves.

all treatments where the plants were inoculated with *R. etli*. As shown in **Figure 6**, there was an induction in *PvAAO3* and *PvBR6OX2.2* transcripts in all three samples that were inoculated with *R. etli* (Re/-, Re/H2O, Re/Ps). In control plants (-/-) or in plants treated with the pathogen only (-/Ps), however, the transcript levels of *PvAAO3* and *PvBR6OX2.2* remained very low. Intriguingly, infiltration with water (Re/H2O), after *R. etli* inoculation, resulted in similar *PvAAO3* and *PvBR6OX2.2*  transcript levels to the Re/Ps treatment (at 24 h and 120 h after PspNPS3121 infection, respectively), indicating that their induction was potentiated by treatment with water. This result suggests that *R. etli* inoculation enhances transcription of *PvAAO3* and *PvBR6OX2.2*, which in turn could increase ABA and brassinolide synthesis, independently of PspNPS3121 infection. The rest of the genes analyzed (*PvNIT1*, *PvAMI1*), involved in auxin biosynthesis, did not show significant changes in their transcripts as a result of the different treatments.

#### *R. etli*–Induced Callose Deposition in *P. vulgaris*

To quantify activity of plant defense response, we determined callose deposition (a quick cellular defense outcome; Kohler et al., 2002), in response to *R. etli* inoculation and to PspNPS3121 infection. After bacterial treatment, common bean leaves were stained with aniline blue analyzed by epifluorescence microscopy, and callose was quantified and documented as the "relative number of callose-corresponding pixels (callose intensity) or the relative number of callose depositions" (Luna et al., 2011). As shown in **Figure 7**, plants primed with *R. etli* and inoculated with PspNPS3121 (Re/Ps) showed an increase in the number of callose depositions and callose intensity compared to the rest of the different treatments (Re/H2O; -/Ps; Re/-), depicting the primed callose response. Furthermore, the size of the individual callose depositions was larger in Re/Ps treated plants, indicating that *R. etli* enhanced the production of callose and generated greater amounts of callose per deposition, which was induced by PspNPS3121 infection.

The number of depositions and callose intensity between control plants (-/-) and plants inoculated only with the symbiont (Re/-; Re/H2O) did not differ statistically. Whereas the relative number of depositions and intensity did not differ statistically between plants inoculated only with the symbiont (Re/-; Re/H2O) and plants infected only with the pathogen (-/Ps).

#### Transgenerational Inheritance of Priming in the Common Bean

To evaluate if *R. etli*–primed plants had the potential to generate an transgenerational memory and produce offspring that are more resistant to the halo blight disease than progeny of unprimed parents, all plants (F0 generation), from all the different treatments (-/-, Re/-, Re/H2O, -/Ps, and Re/Ps), were self-pollinated and grown to seed set to generate the F1 progeny.

As shown in **Figure 8A**, there was no significant difference in the number of seeds produced by plants treated with *R. etli* (Re/-, Re/H2O, Re/Ps) and the number of seeds produced by control plants (-/-) or plants treated with the pathogen only (-/ Ps). However, there was a marginal difference between plants that were nitrogen-supplemented (-/- and -/Ps) and plants that were fertilized with a nitrogen-free solution (Re/-, Re/H2O, Re/ Ps). The results suggest that *R. etli* directly or indirectly protected its hosts against pathogen attack, and its presence facilitates the maintenance of proper crop yield.

Next, we germinated F1 seeds from all the different F0 treatments (-/-, Re/-, Re/H2O, -/Ps, and Re/Ps) in germination trays under sterile conditions, and the lengths of the radicles were measured 3 dag, before transferring the seedlings to pots. There were significant differences in radicle length in F1 seedlings when compared with F0 control seedlings (**Figure 8B**, **C**). Particularly, seedlings from plants that were treated with the pathogen only in the F0 generation were 29% shorter than the untreated control plants. Notably, seedlings from plants that were treated with the symbiont in the F0 [(F0 Re/Ps): F1; (F0 Re/-): F1; **Figure 8C**] had radicles that were 17% longer than the radicles in the control plants (F0 -/-).

At 17 dag, two trifoliate leaves from all F1 plants were exposed to PspNPS3121 [(F0 Re/Ps): F1 -/Ps; (F0 Re/H2O): F1 -/Ps; (F0 Re/-): F1 -/Ps; (F0 -/Ps): F1 -/Ps]. Treatment with the pathogen resulted in a 5–8% lesion size (of total leaf area),

among all plants—that is, an 85% reduction in lesion size when compared to the control plants [(F0 -/-): F1 -/Ps] (**Figure 9A**). In addition, PspNPS3121 bacterial populations (CFU) in leaves from plants treated with *R. etli* in the F0 generation and exposed to the pathogen in the F1 generation [(F0 Re/Ps): F1 -/Ps; (F0 Re/ H2O): F1 -/Ps; (F0 Re/-): F1 -/Ps] were significantly lower than in the control plants treated with the pathogen only ([F0 -/-]: F1 -/Ps) (**Figure 9B**). Therefore, common bean plants treated with *R. etli* in the F0 generation were effectively protected in the F1 generation against PspNPS3121, in contrast to plants that were not primed [(F0 -/-): F1 -/Ps]. Moreover, F0 plants that were treated with the pathogen only ("sensitized"; Conrath, 2009) and exposed to the pathogen in the F1 generation [(F0 -/Ps): F1 -/ Ps] also have shown lower CFUs compared to the control plants (**Figure 9B**). We inferred that infection with PspNPS3121, in the F0 generation, induced enhanced pathogen defense (priming

etli, followed by inoculation with PspNPS3121 (Re/Ps), inoculated only with the symbiont (Re/- ; and water, Re/H2O), infected only (-/Ps), or neither primed nor inoculated (control, -/-). Data were normalized to the Actin11 (PvActin11) and Tubulin (PvTUB, see Supplementary Figure SF3) reference genes. Data represent mean ± SD from three independent experiments (n = 3). Statistical significance was determined using two-way ANOVA followed by Tukey's multiple comparison test, p < 0.05. Dag: days after germination. Means with the same letter are not significantly different.

effect) in the F1 generation against the halo blight bacterial disease. However, continuous exposure to the pathogen [(F0 -/ Ps): F1 -/Ps] influenced plant height when compared to plants that were primed with *R. etli* in the F0 generation (**Figure 9C**). That is, plants exposed only to the pathogen in F0 and F1 [(F0 -/ Ps); (F0 -/Ps): F1 -/Ps; **Figure 9C**] were 17% shorter 30 dag than plants inoculated with the symbiont [(F0 Re/H2O): F1 -/Ps; (F0 Re/-): F1 -/Ps; Figure 9C]. However, there was no significant difference in the number of seeds produced between F1 plants treated PspNPS3121 and the control plants (-/-) (**Figure 9D**).

We next wished to examine the activation of the induced resistance in F1 plants by detecting accumulation of the O2 − at 1.5, 3, 6, and 9 h after PspNPS3121 infection (**Figure 10**). All plants from the F0 generation, when exposed to the pathogen in the F1 generation, accumulated more O2 − than control plants [(F0 -/-): F1 -/Ps], at 1.5 h after infection. However, higher accumulation of O2 − was observed 3 h after infection at the whole-leaf level, in leaves from plants treated with *R. etli* and PspNPS3121 in the F0 generation and exposed to the pathogen in the F1 generation [(F0 Re/Ps): F1 -/Ps], as well as in plants treated with *R. etli* and water [(F0 Re/H2O): F1 -/Ps], when compared to control plants (**Figures 7A**, **B**). Such O2 − accumulation pattern suggests a transgenerational effect in plants primed in the F0 generation with *R. etli* ("bacterizing the roots," or root colonized by bacteria; Van Peer et al., 1991) and induced in the F1 by PspNPS3121 on cellular defense responses.

In addition, to determine whether the primed callose response has a transgenerational effect against PspNPS3121, we analyzed callose depositions in the F1 generation of all the different treatments, after exposure to the pathogen [(F0 -/-): F1 -/Ps; (F0 -/Ps): F1 -/Ps; (F0 Re/H2O): F1 -/Ps; (F0 Re/-): F1 -/Ps; (F0 Re/Ps): F1 -/Ps]. As shown in **Figure 11**, plants treated with *R. etli* and PspNPS3121 in the F0 generation and exposed to the pathogen in the F1 generation [(F0 Re/Ps): F1 -/Ps] showed a faster and stronger accumulation of callose depositions after pathogen infection, even though the relative number of depositions was 2.6 times lower than in the F0 generation. This suggests that the ability to produce callose in F1 is caused by the priming effect in the F0. Plants from the rest of the treatments, however, failed to accumulate enhanced levels of callose after pathogen treatment.

We then selected the primed-responsive genes (*PvWRKY33, PvPAL2,* and *PvERF6*) highly induced against PspNPS3121 in the

parental generation (F0) and analyzed their expression patterns in the F1 generation before and after pathogen attacks (**Figure 12** and **Supplementary Figure SF4**). As shown in **Figure 12A**, the genes that exhibited an enhanced transcription pattern following pathogen attack were *PvERF6* (2.5 to three-fold induction, 24 h after infection) and *PvPAL2* (0.5-fold induction, 24 h after infection) when compared to control plants. Consistent with previous results, the finding suggests a transgenerational effect in common bean plants that were primed in the F0 generation with *R. etli* and induced by PspNPS3121 infiltration. In addition, F0 plants treated with *R. etli* were sensitized and displayed enhanced *PvPAL2* and *PvERF6* transcript levels in the F1 generation. However, expression levels of the ABA-synthesis-related gene *PvAAO3*, in the F1 generation (without *R. etli* inoculation), were not significantly different before and after PspNPS3121 infiltrations and also with respect to the control {[(F0 -/-): F1 -/-]; **Figure 12B**}. This suggests that inoculation with *R. etli* is required for *PvAAO3* expression and ABA synthesis and that *PvAAO3* expression is not directly related to the transgenerational induced resistance, under the present experimental conditions.

#### DISCUSSION

Common bean is a valuable crop worldwide and an important grain legume in the human diet. Common bean production, however, is affected by numerous pathogens. Some of the major approaches for controlling common bean infectious diseases include crop rotation, sanitation, using certified seeds, growing resistant cultivars, stress and wound avoidance, chemical control or application of pesticides, and to a lesser extent, the use of bacterial biocontrol agents and PGPR. In legumes, the use of

PGPR has been mostly limited to rhizobia manipulation in studies aimed at improving legume growth and development, particularly *via* nodulation and nitrogen fixation (Pérez-Montaño et al., 2014). In addition, several studies have examined the hypothesis that PGPR might enable plants to maintain their yield with reduced fertilizer application rates (Yang et al., 2009).

PGPR also suppress disease-causing microbes by activating the ISR plant defense system. In addition, legume-*Rhizobium* interactions induce PGPR "like-responses," improving tolerance/resistance to different types of stress (Fernandez-Göbel et al., 2019). However, few reports have been published on *Rhizobium* as elicitors of resistance to bacterial diseases (Osdaghi et al., 2011). Therefore, the goals of the present study were, first, to determine whether the symbiont *R. etli* elicits the ISR in *P. vulgaris* against PspNPS3121. Second is to determine if *R. etli* has a role in decreasing or inhibiting the harmful effects of PspNPS3121 on plant growth, development, and nitrogen fixation in *R. etli* treated bean plants. Achievement of the above goals could facilitate the use *R. etli* (and rhizobacteria in general) as a defense-priming agent to induce ISR in *P. vulgaris* to minimize susceptibility to pathogens and enhance plant breeding activities and overall crop productivity. Therefore, we studied the effect of priming on gene activation and the generational effect of the primed state using *R. etli* as an elicitor.

Inoculation of common bean with *R. etli* reduced halo blight severity when compared to plants not treated with the symbiont. For example, foliar pathogen lesion size was 75% lower in the *R. etli* inoculated plants, than in the non-inoculated plants, while PspNPS3121 CFU was 50% lower in *R. etli* inoculated plants. Therefore, the symbiotic relationship between *R. etli* and common bean reduced disease incidence. The finding is notable since any method that can reduce the foliar symptoms of halo blight disease at low costs and in an environmentally safe manner (e.g., without the application of commercial pesticides) is of considerable importance in agriculture.

In our experiments, there were no significant changes in root length, diameter and number of nodules, and nitrogen fixation (nitrogenase activity) among the different treatments. The finding is also critical, since the effects of co-inoculation of *R. etli* and PspNPS3121 on common bean suggest that *R. etli* improves tolerance against halo blight disease without affecting crop productivity. Although further studies are required to corroborate the effectiveness of distinct strains of *Rhizobium* spp. in reducing the incidence of halo blight disease, the utilization of *R. etli* in common bean planting areas is recommended and

could be an alternative to the application of synthetic pesticides and fertilizers.

Accordingly, inoculation with *R. etli* elicited ISR and primed the whole plant for enhanced defense against PspNPS3121. Aboveground parts of the plant (foliar tissue) acquired an enhanced level of resistance against infection by the PspNPS3121 pathogen following inoculation with *R. etli*. In addition, *R. etli*– treated plants exhibited considerably higher ROS superoxide anion accumulation at non-toxic levels at the site infected by the pathogen. Moreover, during the different stages of the symbiotic interaction, ROS can be translocated extensively. Therefore, the systemic redox signaling plays an important role in the regulation of systemic acclimatory mechanisms under stress conditions (Fernandez-Göbel et al., 2019). Consequently, ROS signaling networks control a wide range of biological processes, including responses to biotic stimuli, and functions as a general priming signal in plants (Baxter et al., 2014). Moreover, plant defense hormones modulate plants' ROS status (Sewelam et al., 2013), which suggests that inoculation of *R. etli* to the root system sensitized distal plant parts for enhanced pathogen defense (priming effect). Therefore, long-term responses (Mittler et al., 2011) enhanced the oxidative stress defense capacity in common bean leaves when exposed to the PspNPS3121 pathogen.

Interestingly, our study demonstrates that plants treated with *R. etli* responded more efficiently to pathogen attack *via*  an augmented deposition of callose. Thus, *R. etli* sensitizes or primes *P. vulgaris* plants for stronger accumulation of callose biosynthesis at the site of infection, under the present experimental conditions. Hence, we propose that *R.* 

*etli*–induced superoxide anion accumulation promotes callose deposition.

*R. etli*–treated plants exhibited a potentiated defense-related gene expression. The transcript levels of *PvWRKY33, PvERF6,*  and *PvPAL2* exhibited a typical priming response. Although *R. etli* did not trigger their expression, after PspNPS3121 inoculation, transcripts accumulated at levels higher than in the unprimed, inoculated controls. In plants, a large number of *WRKY* genes code for TF induced by pathogen infection are involved in plant defense responses and regulate crosstalk between JA- and SA-regulated disease response pathways (Zheng et al., 2006). In *Arabidopsis*—for example, WRKY33 TF interacts with MAP kinase 4 (MPK4) and MKS1 within the nucleus and upon exposure to PsDC3000 or following elicitation with flg22 (a 22-amino acid sequence of flagellin), WRKY33 is released from the trimeric complex and targets the promoter of

*PHYTOALEXIN DEFICIENT3* (*PAD3*), encoding an enzyme necessary for the synthesis of antimicrobial camalexin, a phytoalexin with antimicrobial and antioxidative properties (Qiu et al., 2008). In addition, inoculation with *R. etli* greatly enhanced *Phenylalanine AMMONIA-LYASE2* (*PAL2*) expression induced by PspNPS3121 infection. *PAL* genes catalyze the first step in the phenylpropanoid pathway, where L-phenylalanine undergoes deamination to produce trans-cinnamate and ammonia. *PAL* is also involved in cellular defense responses and the formation of lignin, phytoalexins, coumarins, and other flavonoids (Zhang et al., 2017). Furthermore, it has been shown that PAL2 functions in callose deposition (Kohler et al., 2002). *ETHYLENE RESPONSE FACTOR 6* (*ERF6*), a defense activator involved in *Arabidopsis* immunity, is also induced upon pathogen attack *via* the activation of mitogen-activated protein kinases (MAPKs/MPKs) (Huang et al., 2016). *ERF6* also plays a key role in oxidative stress signaling and is necessary for the expression of antioxidant genes in response to biotic and abiotic stresses (Sewelam et al., 2013).

Primed plants responded more efficiently to PspNPS3121 attack by a higher level of superoxide anion production, a faster and stronger accumulation of callose, and a faster activation of plant defense gene expression. In addition, treatment with *R. etli* increased the transcript levels of the ABA-synthesis-related gene *PvAAO3*, after pathogen attack. Suggesting the activation of an ABA-dependent defense mechanism in primed plants

(C) *PvERF6*. F0 plants were self-pollinated and grown to seed set to generate the F1 progeny. Seeds from the F1 progenies were germinated and exposed to the pathogen only, without rhizobacteria treatment. Samples were obtained at 16, 18, and 22 dag. Data were normalized to the actin11 (*PvActin11*) and tubulin (*PvTUB*, see Figure SF4) reference genes. (B) Transcript levels of the *PvAAO3* gene involved in ABA synthesis. Data were normalized to the actin11 (*PvActin11*) and tubulin (*PvTUB*) reference genes. Data are mean ± SD from three independent experiments (n = 3). Statistical significance was determined using two-way ANOVA followed by Tukey's multiple comparison test, p < 0.05. Dag, days after germination.

(Ton and Mauch-Mani, 2004). Actually, it has been shown that ABA functions as a positive regulator of disease resistance through—for example, potentiation of callose deposition (Ton and Mauch-Mani, 2004; Ton et al., 2005; Flors et al., 2008). Furthermore, environmental conditions such as light, drought, and salt stress have been suggested to regulate ABA biosynthesis (Xiong and Zhu, 2003). This could explain the higher expression levels of the *PvAAO3* gene in *R. etli*–treated plants that were inoculated with water. Thus, ABA displays an important function in plant defense and its task depends on distinctive plantmicroorganism interactions (Bari and Jones, 2009).

Therefore, *R. etli* facilitated greater and more rapid activation of defense-related genes following infection with the pathogen, which suggests that priming is an important cellular mechanism in ISR in common bean plants. In addition to the critical role of rhizobacteria in plant growth and development, and in maintaining soil fertility, the indirect biotic stress tolerance effect induced by *R. etli* inoculation should be considered as a major factor influencing the activities of phytopathogenic microorganisms.

Plants exposed to the pathogen in the F1 generation exhibited reduced lesion sizes, low numbers of pathogenic bacterial populations (CFU), and higher transcript levels of *PvERF6* and *PvPAL2*, when compared with the control plants. The results indicate that *R. etli*–primed plants could develop a transgenerational defense memory and could produce offspring that are more resistant to the halo blight disease. However, the accumulations of superoxide anion and callose deposition, in the F1 generation, were much lower than those detected in the F0, and the expression of the *PvAAO3* gene remained without change, after pathogen attack. Notably, up-regulation of *PvERF6* (an ET-independent TF that activates expression of PR genes; Huang et al., 2016) suggests that the transgenerational defense mechanisms implicated in common bean protection mediated by *R. etli* could encompass ET-independent signaling pathway.

#### CONCLUSIONS

The use of rhizobacteria can stimulate ISR in plants, with the potential to transform modern agriculture; however, research on the exploitation of PGPR remains limited. Considering the positive effects of *R. etli* on crop productivity in the form of biotic stress tolerance and generational and transgenerational inheritances of priming effects, in addition to N-fixation and reduced pesticide application rates, adoption of ISR and PGPR in agriculture should be encouraged as a tool for managing plant stress. Nevertheless, more studies are required to explore the role of root nodule symbiosis and plant innate immunity under stressful conditions such as disease infestation.

#### DATA AVAILABILITY STATEMENT

All datasets for this study are included in the manuscript/ **Supplementary Files**.

### AUTHOR CONTRIBUTIONS

RA-V provided the idea of the work. RA-V and AD-V designed the experiments. AD-V and AL-C conducted the experiments and performed the statistical analysis. RA and AD-V participated in the interpretation of results and critically reviewed the manuscript. RA-V wrote the paper. All authors read and approved the final manuscript.

#### FUNDING

This work was supported by Consejo Nacional de Ciencia y Tecnología, grant CB-2015/257129 to RA-V.

#### ACKNOWLEDGMENTS

The authors thank José Antonio Vera-Núñez for assistance with the acetylene reduction analysis.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.01317/ full#supplementary-material

SUPPLEMENTARY FIGURE S1 | Transcript levels of genes involved in plant defense as determined by q-PCR at various days after germination (dag). (A) PvWRKY33; (B) PvPAL2; (C) PvERF6. Plants were primed with R. etli, followed by inoculation with P. syringae pv. phaseolicola (Re/Ps), inoculated with the symbiont only (Re/- ; and water, Re/H2O), infected only (-/Ps), or neither primed nor inoculated (control, -/-). Data were normalized to the Tubulin (PvTUB) reference gene. Data are mean ± SD from three independent experiments (n = 3). Statistical significance was determined using two-way ANOVA followed by Tukey's multiple comparison test, p<0.05. Dag: days after germination. Means with the same letter are not significantly different.

SUPPLEMENTARY FIGURE S2 | Transcript levels of genes involved in plant defense. q-PCR analysis was performed at 16, 18, and 22 days after germination (dag). (A) PvPR1; (B) PvPR10; (C) PvWRKY29, (D) PvWRKY70; (E) PvNPR1; (F) PvMYC2. Plants were primed with R. etli, followed by inoculation with PsNPS3121 (Re/Ps), inoculated only with the symbiont (Re/- ; and water, Re/ H2O), infected only (-/Ps), or neither primed nor inoculated (control, -/-). Data were normalized to the PvActin11 reference gene. Data are mean ± SD from three independent experiments (n = 3). Statistical significance was determined using two-way ANOVA followed by Tukey's multiple comparison test, p<0.05. Dag: days after germination. Means with the same letter are not significantly different.

SUPPLEMENTARY FIGURE S3 | Transcript levels of genes involved in plant defense. q-PCR analysis was performed at 16, 18, and 22 days after germination (dag). (A) PvAAO3; (B) PvAMI1; (C) PvNIT1; (D) PvBROX2.2. F0 plants were self-pollinated and grown to seed set to generate the F1 progeny. Seeds from the F1 progeny were germinated and exposed to the pathogen only, without rhizobacteria treatment. Samples were obtained at 16, 18, and 22 dag. Data were normalized to the Tubulin (PvTUB) reference gene. Data are mean ± SD from three independent experiments (n = 3). Statistical significance was determined using two-way ANOVA followed by Tukey's multiple comparison test, p<0.05. Dag: days after germination. Means with the same letter are not significantly different.

SUPPLEMENTARY FIGURE S4 | Transgenerational transcript levels of genes involved in plant defense. (A) PvWRKY33; (B) PvPAL2; (C) PvERF6. F0 plants were self-pollinated and grown to seed set to generate the F1 progeny. Seeds from the F1 progeny were germinated and exposed to the pathogen only, without rhizobacteria treatment. Samples were obtained at 16, 18, and 22 dag. Data were normalized to the Tubulin (PvTUB) reference gene. Data are mean ± SD from three independent experiments (n = 3). Statistical significance was determined using two-way ANOVA followed by Tukey's multiple comparison test, p<0.05. Dag: days after germination. Means with the same letter are not significantly different.

#### REFERENCES


resistance against Pseudomonas syringae and Alternaria brassicicola. *Plant J.* 54, 81–92. doi: 10.1111/j.1365-313X.2007.03397.x


promotion by rhizobacteria. *J. Soil Sci. Plant Nutr.* 10, 293–319. doi: 10.4067/ S0718-95162010000100006


Induced Resistance. *Mol. Plant Microbe. Interact.* 30, 215–230. doi: 10.1094/ MPMI-09-16-0192-R


**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 © 2019 Díaz-Valle, López-Calleja and Alvarez-Venegas. 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.*

# Endophytes and Epiphytes From the Grapevine Leaf Microbiome as Potential Biocontrol Agents Against Phytopathogens

#### Edited by:

Kalliope K. Papadopoulou, University of Thessaly, Greece

#### Reviewed by:

Stéphane Compant, Austrian Institute of Technology (AIT), Austria Katerina Karamanoli, Aristotle University of Thessaloniki, Greece

#### \*Correspondence:

Laure Weisskopf laure.weisskopf@unifr.ch †These authors have contributed equally to this work

#### Specialty section:

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

Received: 13 August 2019 Accepted: 08 November 2019 Published: 29 November 2019

#### Citation:

Bruisson S, Zufferey M, L'Haridon F, Trutmann E, Anand A, Dutartre A, De Vrieze M and Weisskopf L (2019) Endophytes and Epiphytes From the Grapevine Leaf Microbiome as Potential Biocontrol Agents Against Phytopathogens. Front. Microbiol. 10:2726. doi: 10.3389/fmicb.2019.02726 Sébastien Bruisson† , Mónica Zufferey† , Floriane L'Haridon, Eva Trutmann, Abhishek Anand, Agnès Dutartre, Mout De Vrieze and Laure Weisskopf\*

Department of Biology, University of Fribourg, Fribourg, Switzerland

Plants harbor diverse microbial communities that colonize both below-ground and above-ground organs. Some bacterial members of these rhizosphere and phyllosphere microbial communities have been shown to contribute to plant defenses against pathogens. In this study, we characterize the pathogen-inhibiting potential of 78 bacterial isolates retrieved from endophytic and epiphytic communities living in the leaves of three grapevine cultivars. We selected two economically relevant pathogens, the fungus Botrytis cinerea causing gray mold and the oomycete Phytophthora infestans, which we used as a surrogate for Plasmopara viticola causing downy mildew. Our results showed that epiphytic isolates were phylogenetically more diverse than endophytic isolates, the latter mostly consisting of Bacillus and Staphylococcus strains, but that mycelial inhibition of both pathogens through bacterial diffusible metabolites was more widespread among endophytes than among epiphytes. Six closely related Bacillus strains induced strong inhibition (>60%) of Botrytis cinerea mycelial growth. Among these, five led to significant perturbation in spore germination, ranging from full inhibition to reduction in germination rate and germ tube length. Different types of spore developmental anomalies were observed for different strains, suggesting multiple active compounds with different modes of action on this pathogen. Compared with B. cinerea, the oomycete P. infestans was inhibited in its mycelial growth by a higher number and more diverse group of isolates, including many Bacillus but also Variovorax, Pantoea, Staphylococcus, Herbaspirillum, or Sphingomonas strains. Beyond mycelial growth, both zoospore and sporangia germination were strongly perturbed upon exposure to cells or cell-free filtrates of selected isolates. Moreover, three strains (all epiphytes) inhibited the pathogen's growth via the emission of volatile compounds. The comparison of the volatile profiles of two of these active strains with those of two phylogenetically closely related, inactive strains led to the identification

of molecules possibly involved in the observed volatile-mediated pathogen growth inhibition, including trimethylpyrazine, dihydrochalcone, and L-dihydroxanthurenic acid. This work demonstrates that grapevine leaves are a rich source of bacterial antagonists with strong inhibition potential against two pathogens of high economical relevance. It further suggests that combining diffusible metabolite-secreting endophytes with volatileemitting epiphytes might be a promising multi-layer strategy for biological control of above-ground pathogens.

Keywords: endophytes, epiphytes, Bacillus, Botrytis cinerea, Phytophthora infestans, volatiles

#### INTRODUCTION

Plants are densely colonized by a variety of microbes (Compant et al., 2019), some of which – the epiphytes – stay on the surface of plant organs, while others are able to penetrate further inside the plants and are called endophytes (Hardoim et al., 2015). Early studies on the structure and function of plant microbiotas have focused on the roots and shown the interplay between soil and host plant in shaping the root and rhizosphere microbiotas (Berg and Smalla, 2009; Philippot et al., 2013). Later investigations demonstrated that the leaves also offer habitats to complex, albeit less diverse microbial communities (Vorholt, 2012). The importance of rhizosphere microbes for plant health is well documented, e.g., through studies on suppressive soils (Mendes et al., 2011; Schlatter et al., 2017), or on the ability of root-colonizing bacteria to stimulate plant defenses, through the so-called "induced systemic resistance" (Pieterse et al., 2014). While the functional roles of phyllosphere inhabitants are still less well understood than those of rhizosphere microbes, few reports have shown that they also significantly contribute to plant health (Innerebner et al., 2011; Ritpitakphong et al., 2016). Indeed, it would seem useful for the plant to be able to count on local defenses provided by phyllosphere microbes to fend off pathogens attacking leaves, in addition to the systemic plant-encoded resistance triggered by root-colonizing microbes. The main aim of this study was therefore to harness the potential of phyllosphere microbiota for protection against above-ground pathogens.

We selected grapevine as a model plant for two main reasons. First, we expected this perennial plant to host a better adapted leaf endophytic microbiome than annual plants, since strong overlap had been reported between communities living in the soil and those colonizing the above-ground plant organs (Zarraonaindia et al., 2015). This suggests that the vineyard soil acts as a reservoir from which microbes colonize the phyllosphere. Second, grapevine cultivation is a heavy consumer of fungicides (Sabatier et al., 2014) due to its high sensitivity to multiple pathogens, among which the oomycete Plasmopara viticola causing downy mildew (Gessler et al., 2011) and the ascomycete Botrytis cinerea causing gray mold (Pertot et al., 2017). Current disease management practices strongly rely on synthetic fungicides (or copper-based products), but the increasing awareness of the deleterious side-effects of these compounds on environmental and human health urges us to find efficient, less toxic alternative strategies for disease management. Biological control using microbial antagonists of the diseasecausing agents represents one such alternative strategy (Pal and Mc Spadden Gardener, 2006; Syed Ab Rahman et al., 2018). In grapevine, the search for biocontrol agents has been ongoing for some years to find alternative solutions against gray mold as well as against downy mildew. Several fungal and bacterial biological control agents are available for the control of gray mold (recently reviewed in Abbey et al., 2019), while only few have been reported for downy mildew (Dagostin et al., 2011; Puopolo et al., 2014b). In contrast to synthetic fungicides, biological control agents offer the advantage of having various modes of action (Haidar et al., 2016), which can be direct (e.g., through antibiosis or niche competition) or indirect, through induction of resistance (Pertot et al., 2016). The latter has been object of many studies dealing with grapevine protection against diseases (Delaunois et al., 2014; Aziz et al., 2016).

Independently of the mode of action of biological control agents, one major difficulty associated with their use is their inconsistent performances in the field. Combination of few biological control agents has been suggested as a mean to increase robustness by providing functional redundancy as well as complementarity in the modes of action (De Vrieze et al., 2018), yet some studies have shown that single strains can perform as well as mixed consortia (Pertot et al., 2017). Moreover, most commercially available biocontrol strains have not been isolated from the plant/organ they are intended to protect, which might at least partially account for their inconsistent survival and efficacy in the field. Moreover, most available bacterial biocontrol agents originate from the rhizosphere but are used as leaf sprays, although both plant habitats greatly differ, as do their native microbial colonizers (Müller et al., 2016). For this reason, the major aim of this study was to isolate bacteria from grapevine leaves and to test their potential suitability for the biological control of important disease-causing agents. Different cultivars differ in their sensitivity to diseases, but the relative contribution of the host plants and their microbiotas in mounting this resistance has not yet been elucidated. We therefore compared isolates obtained from two sensitive varieties, Pinot Noir and Chasselas, with those from the disease-resistant variety Solaris. Moreover, we isolated both epiphytic and endophytic bacteria in order to discriminate between tightly associated bacteria and those more loosely connected to the leaves. After taxonomic identification, we assembled a collection of 78 non-redundant

strains, which we tested for in vitro growth inhibition of Botrytis cinerea and Phytophthora infestans. The latter was used as a surrogate oomycete pathogen for Plasmopara viticola, which cannot be cultivated in vitro. In addition to classical dual assays, we also tested the potential of the grapevine isolates to inhibit the growth of phytopathogens through the emission of volatile compounds, which have gained attention in recent years as promising antifungal agents emitted by plantassociated bacteria (Bailly and Weisskopf, 2017). Finally, we mined our grapevine leaf strain collection for phylogenetically closely related strains of differing volatile-mediated activity to identify candidate molecules underlying the observed pathogeninhibiting effects.

#### MATERIALS AND METHODS

#### Microbial Strains and Culture Media

In total, 194 bacteria were isolated as described earlier (Vionnet et al., 2018) from both the epiphytic and endophytic compartments of three grapevine cultivars, one resistant to fungal diseases (Solaris) and two sensitive to fungal diseases (Pinot noir and Chasselas). Bacterial strains were routinely grown on LB (Luria-Bertani) medium and incubated in the dark at room temperature. LB medium was prepared by dissolving 12.5 g/L of LB Broth Miller and 10 g/L of LB Broth Lennox (Fisher Bioreagents)<sup>1</sup> in distilled water, to which 15 g/L of Agar-Agar Kobe I (Roth)<sup>2</sup> were added for solid LB medium. Botrytis cinerea strain BMM was provided by Brigitte Mauch-Mani (University of Neuchâtel, Switzerland) and grown on Potato dextrose Agar (PDA). This medium was prepared by dissolving 39 g of PDA Powder (Sigma-Aldrich) in one liter of distilled water. Botrytis cinerea plates were incubated at 20◦C with 12 h light/12 h dark cycle. Phytophthora infestans strain Rec01 (Hunziker et al., 2015) was grown on V8 agar medium and incubated at 18◦C in the dark. V8 100% hot spicy vegetable juice<sup>3</sup> was prepared at 100 mL/L in distilled water and 1 g/L of CaCO<sup>3</sup> was added. Agar (15 g/L) was added for solidified V8. Each medium was sterilized by autoclaving at 120◦C during 20 min. For long-term storage, bacterial strains were kept at -80◦C in 25% glycerol (Reactolab SA) in cryogenic tubes (Sarstedt). B. cinerea and P. infestans were kept as mycelial plugs in 10% glycerol in a nitrogen tank after a few hours at −20◦C and after one night at −80◦C.

#### Taxonomic Identification of Grapevine Epiphytes and Endophytes

For taxonomic identification, amplification of the fulllength 16S rRNA gene was carried out. To this end, two to three colonies of each bacterial strain were lysed in 50 µL of distilled water by boiling at 100◦C during 10 min. The polymerase chain reaction (PCR) was performed in a total volume of 25 µL with 5 µL of lysed bacterial solution as template. The primers BactF (5<sup>0</sup> -AGA GTT TGA TYM TGG CTC-3<sup>0</sup> ) and BactR (5<sup>0</sup> -CAK AAA GGA GGT GAT CC-3<sup>0</sup> ) were used at a final concentration of 0.5 µM, while the loading gel track and the polymerase Accustart II PCR toughmix (VWR) were added according to the manufacturer's protocol. The PCR program was performed as follows: an initial denaturation step at 94◦C for 3 min, then 35 cycles of three steps made up of a denaturation step at 94◦C during 30 s, an annealing step at 56◦C during 30 s and an extension step at 72◦C during 1.5 min, followed by a final extension step at 72◦C during 10 min. Three µL of PCR product were run on a 1% agarose gel to verify the correct size of the amplified product. The rest of the PCR product was purified by QIAquick PCR purification Kit (Qiagen) following the protocol of the manufacturer. Twelve µL of sample were mixed with 3 µL of either the BactF or the BactR primer at 10 µM and were sent to an external company for sequencing (Microsynth). The obtained chromatograms were visually inspected and sequences were manually corrected when necessary.

#### Assembly of a Collection of Non-redundant Strains and Construction of a Phylogenetic Tree

In order to identify redundant isolates (=same bacterial strain isolated twice), all sequences retrieved from bacteria isolated from the same cultivar and compartment (endovs. epiphytes) were aligned using the Geneious software. In each of these subsets of sequences, strains having a distance matrix inferior to 0.2 were considered redundant and only one representative of these redundant strains was selected for further analysis. In total, 78 strains were kept in the collection of non-redundant strains. These sequences have been submitted to the NCBI database (accession numbers MN555571-MN555648). The 78 sequences were then blasted using the nucleotide database from NCBI<sup>4</sup> to determine the genus and species affiliation for each strain. The blast results are listed in the supplement (**Supplementary Table S1**). A phylogenetic tree based on the sequences of these 78 non-redundant isolates was constructed using the following parameters in the Geneious software: a global alignment with free end gaps was performed, with a similarity index of 65%. A neighbor-joining tree was then constructed using the Tamura-Nei genetic distance model. In addition to the grapevine isolates, sequences corresponding to known bacterial species of the following genera were added prior to the alignment: Variovorax (NR\_113736.1), Cupriavidus (NR\_074704.1), Erwinia (NR\_148650.1), Sphingomonas (NR\_104893.1), Rhodopseudomonas (NR\_114302.1), Methylobacterium (NR\_115219.1), Microbacterium (NR\_042480.1), Micrococcus (NR\_134088.1), Paenibacillus (NR\_044524.1), and Sporosarcina

<sup>1</sup>www.fisher.co.uk

<sup>2</sup>www.carlroth.com

<sup>3</sup>www.junkfood.ch

<sup>4</sup>https://blast.ncbi.nlm.nih.gov/Blast.cgi

(NR\_025049.1). For the Bacillus and Staphylococcus genera, the sequences from the following species were added: B. circulans (FJ581445.1), B. aryabhattai (NR\_115953.1), B. butanolivorans (MN235850.1), B. zhangzhouensis (NR\_148786.1), B. stratosphericus (MH973204.1), B. licheniformis (NR\_118996.1), B. subtilis (NR\_113265.1), B. halotolerans (NR\_115063.1), B. cereus (NR\_074540.1), S. saprophyticus (L37596.1), S. hominis (L37601.1), S. epidermidis (NR\_113957.1), S. pasteuri (NR\_114435.1) and S. warneri (NR\_025922.1). A Flavobacterium sequence (MK246909.1) was used as an outgroup.

#### In vitro Dual Assays

fmicb-10-02726 November 27, 2019 Time: 17:28 # 4

The antagonistic activity of the 78 non-redundant isolates was tested against both pathogens (B. cinerea and P. infestans) on standard (Greiner bio-one) and (Sarstedt) Petri dishes. V8 medium was used for standard plates and a combination of V8 (fungus/oomycete) and LB (bacteria) media were used for two-compartment plates. The assays on standard plates allowed both volatile and diffusible compounds to be exchanged between the two partners, while only volatiles could be exchanged in the split plate assays (**Supplementary Figure S1**). Liquid cultures of the bacterial strains were prepared by suspending two to three colonies from fresh LB agar cultures in 3– 5 mL of sterile LB broth and incubating overnight at 28◦C under shaking at 180 rpm. After a first centrifugation step at 5000 rpm during 5 min, the supernatant was removed. The bacterial cell pellet was washed one time in 0.9% (w:v) NaCl with the same centrifugation conditions and the bacterial pellet was then resuspended in 0.9% NaCl. Optical density (OD) was measured at 595 nm to evaluate cell density, and was adjusted to 1 with 0.9% NaCl. Three drops of 10 µL of OD<sup>595</sup> = 1 bacterial suspension were then inoculated at the border of standard Petri dishes, and at the border of one compartment for the split Petri dishes (**Supplementary Figure S1**). The pathogens (B. cinerea and P. infestans) were inoculated by placing a 5 mm plug of a 3-day-old B. cinerea culture grown on PDA medium plate or of a 2-week-old P. infestans culture grown on V8 agar medium plate at the center of the standard Petri dishes and at the center of one compartment of the split Petri dishes (**Supplementary Figure S1**). Plates were then sealed with Parafilm M. For dual assays with P. infestans, both partners (bacterium and oomycete) were inoculated on the same day. For dual assays with B. cinerea, the bacteria were given a head start of 3 days before fungus inoculation to take into account the very fast growth of the pathogen. Control plates without bacteria were supplemented with three droplets of 10 µL of 0.9% NaCl and one plug of the respective pathogen mycelium. Dual assay plates were incubated in the dark at 23◦C and pictures were taken at several time points during incubation. Pictures corresponding to the time point when B. cinerea and P. infestans reached the border of the Petri dishes in the control plates were used to measure mycelial growth with ImageJ and to calculate the percentage of growth inhibition caused by bacterial exposure. For dual assays on full Petri dishes, the inhibition area (Ai) was calculated as follow: A<sup>i</sup> = A<sup>c</sup> - A<sup>b</sup> - A<sup>t</sup> , where A<sup>c</sup> is the mean of the area

colonized by the pathogen in the control plates, A<sup>b</sup> is the area colonized by the bacteria in the test plate and A<sup>t</sup> is the area colonized by the pathogen in the test plate. The percentage of growth inhibition was given by the formula: inhibition percentage = 100 × Ai/(A<sup>c</sup> – Ab), where A<sup>c</sup> – A<sup>b</sup> is the area available for the pathogen growth in test plates (**Supplementary Figure S2a**). For dual assays on two-compartment Petri dishes, the inhibition area was calculated as: A<sup>i</sup> = A<sup>c</sup> - A<sup>t</sup> and the percentage of growth inhibition was calculated as: 100 × Ai/A<sup>c</sup> (**Supplementary Figure S2b**). Statistical analysis was performed using a two-tailed Student's t-test, <sup>∗</sup>p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

#### Effects of Selected Bacterial Strains on B. cinerea Spore Germination

The spores of B. cinerea were harvested from a PDA culture in sterile water and filtered through glass wool to remove hyphae. After a centrifugation step at 800 rpm for 10 min, the spore pellet was suspended in 100 µL of sterile water. The spore concentration was determined and diluted at 2 × 10<sup>5</sup> spores/mL in clarified V8 medium [i.e., medium filtered through a PVDF 0.2 µm syringe filter (Fisher Scientific)]. Six Bacillus strains were selected according to their inhibitory effect on the mycelial growth of B. cinerea: CHD4, CHP14, PID5, SOD5, SOD20, and SOP51. Moreover, two strains showing no inhibitory effect on the mycelium growth of B. cinerea were chosen as controls (PIP26, SOP5). The effect of the bacterial strains on the germination of B. cinerea spores was tested using two modalities, one with the bacterial cells and one with their spent medium (cell-free filtrate). Two to three bacterial colonies grown on LB agar medium were inoculated in filtered V8 broth and incubated overnight at 28◦C under 180 rpm shaking. The overnight bacterial liquid culture was diluted to OD<sup>595</sup> = 1 in filtered V8 broth. To obtain cellfree filtrates, bacterial cultures with OD<sup>595</sup> = 1 were filtered with sterile syringe filters PVDF 0.2 µm (Fisher Scientific). Nine µL of bacterial culture or filtered bacterial culture were mixed with 3 µL of B. cinerea spores at a final concentration of 5 × 10<sup>4</sup> spores/mL on a glass slide. Controls contained 9 µl of filtered V8 medium instead of bacterial culture or cell-free filtrate. The glass slides were placed in a humid box and incubated for 8 and 24 h, respectively. The germination of spores was observed using a Leica DMR microscope with bright-field settings. At least 30 spores of B. cinerea were analyzed. The analysis criteria were the percentage of germinated spores, the percentage of spores with multiple germ tubes as well as the percentage and type of anomalies observed. Three different anomalies were distinguished: hyphal swelling, swelling at the germ tube tip and multiple swellings in a row. In addition, germ tube length was measured using ImageJ. All above-mentioned parameters were scored after 8 h of incubation. Statistical analysis was performed using a Chi-square test with Yate's correction for continuity (p < 0.05) to reveal differences in percentages of germination and in percentages of abnormal germination between the bacterial treatments and the control. For the germ tube lengths, a two-tailed Student's

t-test was used (∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001). Furthermore, the presence of bacterial accumulation around spores, the occurrence of spore or germ tube vacuolization and the production of conidiophores were assessed at a later time point (24 h after inoculation).

#### Effects of Selected Bacterial Strains on P. infestans Zoospore and Sporangia Germination

Zoospores of P. infestans were collected by adding 10 mL of ice-cold sterile water to 12–14 days old plates of P. infestans grown on V8 medium. The plates were incubated for 2 h at 4◦C. After incubation, the plates were left at room temperature in the dark for 20 min. Swimming zoospores were collected by pipetting the upper layer of liquid from the Petri dishes. Zoospore concentration was 1.8 × 10<sup>5</sup> spores/mL. Sporangia of P. infestans were harvested by scraping off the mycelium of 12–14 days old cultures grown on V8 medium and by suspending the mycelium in sterile water. After vigorous shaking, the suspension was filtered using a mesh to remove the hyphae. Sporangia concentration was adjusted to 2 × 10<sup>5</sup> spores/mL. As with B. cinerea, bacterial isolates CHD4, CHP14, PID5, SOD5, SOD20, and SOP51 showed strong inhibitory potential on the mycelium growth of P. infestans and were therefore selected, with non-active bacterial strains PIP26 and SOD5 added as controls. Bacterial strains were prepared as described above for the germination of B. cinerea spores and the effects on zoospores and sporangia germination of both the bacterial cells and their spent medium (cell-free filtrate) were assessed. Nine µL of bacterial culture or filtered bacterial culture were mixed with 3 µL of zoospores or sporangia suspensions at a final concentration of 4.5 × 10<sup>4</sup> and 5 × 10<sup>4</sup> spores/mL respectively. Controls contained 9 µl of filtered V8 medium instead of bacterial culture or cell-free filtrate. For the zoospores, the mixtures were pipetted on the inner side of the lid of 96-well plates, in order to provide a solid surface for the zoospores to adhere to prior to germination. For the sporangia, the mixtures were pipetted onto water agar disks (0.8%, Ø 8 mm) in Petri dishes. The 96-well plates and Petri dishes were placed in a humid box and incubated at 18◦C for 4 and 24 h respectively. Germination of both type of spores was observed using a Leica DMR microscope with bright-field settings. Pictures of the mixtures of bacteria with zoospores and sporangia were taken at a 10-fold and 5-fold magnification respectively and analyzed. Counts of germinated and non-germinated zoospores were performed. On average, the pictures contained 60 zoospores or 75 sporangia. Additionally, for the germinating zoospores, normal, delayed and abnormal germination of zoospores were distinguished in the counting. Percentage of germination and the percentages of the different germinated zoospore phenotypes were computed. Differences between bacterial treatments and the control were assessed by using a two-tailed Student's t-test (∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001) for zoospore and sporangia germination. A Chi-square test with Yate's correction for continuity (p < 0.05) was performed to compare the detailed germination behavior of the zoospores.

#### Collection and Analysis of Volatiles Emitted by Selected Strains

Two strains exhibiting volatile-mediated inhibition of P. infestans were selected for volatile analysis: CHP14 and CHP20. Furthermore, two closely related strains with no volatilemediated activity were selected for comparison: CHD4 (closely related to CHP14) and SOD6 (closely related to CHP20). For each of these four strains, two or three colonies from a LB agar plate were resuspended in 3 mL of LB broth, incubated 24 h at 28◦C and shaken at 190 rpm. The bacterial culture was adjusted to a density of OD<sup>595</sup> = 1.0 in LB broth and 100 µL of this cell suspension were spread on LB agar medium poured into 5 cm glass Petri dishes. LB broth inoculated on LB-agar glass plates was used as control. The strains were grown overnight at room temperature before collecting the volatiles during 48h using closed-loop-stripping analysis (CLSA) as described by Hunziker et al. (2015). Trapped volatiles were extracted from the charcoal filter by rinsing the filter three times with 25 µL dichloromethane (≥99.8%, VWR). The experiment was repeated four times. The volatiles were analyzed by gas chromatography-mass spectrometry (GC/MS). The samples were injected in a HP6890 gas chromatography connected to a HP5973 mass selective detector fitted with an HP-5 ms fused silica capillary column (30 m; 0.25-mm inside diameter; 0.25–µm film; Agilent Technologies). Conditions were as follows: inlet pressure, 67 kPa; He, 15 mL/min; injection volume, 2 µL; transfer line, 300◦C; injector, 250◦C; electron energy, 70 eV. The gas chromatograph was programed as follows: 5 min at 50◦C, then increasing 5◦C/min to 320◦C and hold for 1 min. MZmine-2.20 was used to process the chromatograms. Statistical analysis of the processed data was performed using Metaboanalyst 3.0 (Xia et al., 2015). The differentially enriched compounds were identified using NIST 17 database using OpenChrom software.

#### RESULTS

#### Diversity of Cultivable Bacteria in Grapevine Leaves

The 16S rRNA sequencing of the 194 bacterial strains isolated from grapevine leaf wash off (= epiphytes) and from surface-sterilized leaves (= endophytes) revealed 78 non-redundant strains. A large majority of these were Gram-positive bacteria, with a strong representation of two genera: Bacillus and Staphylococcus. As expected, the diversity on genus level was much lower among endophytes than among epiphytes (**Figures 1**, **2**): except for three Micrococcus, two Paenibacillus strains and one Erwinia strain, most endophytes belonged to the Bacillus or Staphylococcus genera. The proportions between these two

FIGURE 1 | Phylogenetic tree of 78 non-redundant isolates from the microbiome of grapevine leaves. The tree was constructed using the Genious software (see Materials and Methods for details). Isolates from Pinot Noir are shown in blue, with epiphytes (PIP) in dark blue and endophytes (PID) in light blue. Isolates from Solaris are shown in red/orange, with epiphytes (SOP) in red and endophytes (SOD) in orange. Isolates from Chasselas are shown in green, with epiphytes (CHP) in dark green and endophytes (CHD) in light green. A Flavobacterium sequence (MK246909.1) was used as an outgroup and reference sequences from different genera and different species within the Bacillus (B.) and Staphylococcus (S.) genera corresponding to the following accession numbers were included: Variovorax (NR\_113736.1), Cupriavidus (NR\_074704.1), Erwinia (NR\_148650.1), Sphingomonas (NR\_104893.1), Rhodopseudomonas (NR\_114302.1), Methylobacterium (NR\_115219.1), Microbacterium (NR\_042480.1), Micrococcus (NR\_134088.1), Paenibacillus (NR\_044524.1), Sporosarcina (NR\_025049.1), B. circulans (FJ581445.1), B. aryabhattai (NR\_115953.1), B. butanolivorans (MN235850.1), B. zhangzhouensis (NR\_148786.1), B. stratosphericus (MH973204.1), B. licheniformis (NR\_118996.1), B. subtilis (NR\_113265.1), B. halotolerans (NR\_115063.1), B. cereus (NR\_074540.1), S. saprophyticus (L37596.1), S. hominis (L37601.1), S. epidermidis (NR\_113957.1), S. pasteuri (NR\_114435.1), andS. warneri (NR\_025922.1).

Species names of the strains correspond to the closest blast hits (Supplementary Table S1).

genera among endophytes differed in the three cultivars: they were similarly frequent (38%) in Chasselas, while Staphylococcus were slightly more frequently isolated (42%) than Bacillus (33%) in Pinot Noir (**Supplementary Table S1**). In the disease-resistant Solaris cultivar, Bacillus isolates constituted 54% of total endophytes, while Staphylococcus represented 31%. Overall, only two Gram-negative isolates (one Erwinia and one Sphingomonas) were found among endophytes in our survey (**Supplementary Table S1**). Gramnegative bacteria were more frequently retrieved among epiphytes, including Cupriavidus, Methylobacterium, or Rhodopseudomonas strains (**Figure 1**). In addition to Bacillus and Staphylococcus strains, which were also isolated from epiphytic communities of all three cultivars, nine different genera were found among Pinot Noir epiphytes, seven among Chasselas and four among Solaris epiphytes. Bacillus isolates made up one third of the isolated epiphytes in Chasselas and Solaris, but only 5% of Pinot Noir epiphytes (**Figure 1** and **Supplementary Table S1**).

#### Screening for Antagonistic Activity of Grapevine Leaf Isolates Against Pathogens

In dual assays allowing exchange of both volatile and diffusible metabolites (full plates), mycelial growth inhibition of P. infestans was more widespread than the inhibition of B. cinerea for both epiphytes and endophytes. With one exception (Variovorax strain PIP3), only Bacillus strains were able to strongly reduce mycelial growth in B. cinerea, while significant inhibition of P. infestans was also observed for strains belonging to the genera Pantoea, Variovorax, Staphylococcus, Erwinia, Sphingomonas, Herbaspirillum or Micrococcus (**Figure 2** and **Supplementary Table S2**). The ability to inhibit the mycelial

growth of P. infestans was more frequently observed among endophytes (41%) than among epiphytes (21%). This tendency was also visible for B. cinerea, although less pronounced (**Figure 2**). When only volatile compounds were allowed to reach the target pathogens (two-compartment plates), the strains did not inhibit the growth of B. cinerea. However, P. infestans was significantly inhibited in its growth by a few strains, which were all epiphytes from Chasselas (**Figure 2** and **Supplementary Table S2**). Two of these emitters of growth-inhibiting volatiles belonged to the Bacillus genus: one clustered with the B. subtilis group (CHP14), and the second with the B. cereus group (CHP20). The third strain was also a Gram-positive bacterium belonging to the genus Frigoribacterium (CHP33) (**Figure 1**). Moderate P. infestans growth inhibition also occurred upon exposure to volatiles from endophytes, but was more frequently observed among epiphytes (**Figure 2** and **Supplementary Table S2**). It should be noted that in this two-compartment plate bioassay, bacteria were grown on LB medium, not on V8, since the volatiles emitted when the bacteria grew on V8 did not lead to significant pathogen growth reduction (data not shown).

#### Selected Bacillus Isolates Strongly Affected Spore Germination of Botrytis cinerea

The above-mentioned screen for antagonistic activity revealed six Bacillus isolates that reduced mycelial growth of B. cinerea by 57–74% (**Figure 2** and **Supplementary Table S2**). The dual assay was repeated for these six strains and led to the same results, which are shown with representative pictures in **Figure 3**. Interestingly, these six Bacillus strains (CHP14, CHD4, SOP51, PID5, SOD5, and SOD20) were phylogenetically closely related and clustered together in a particular subgroup of the tree with reference strains such as B. subtilis, B. zhangzhouensis, B. licheniformis, or B. stratosphericus (**Figure 1**). To characterize the antifungal potential of these strains beyond mycelial growth inhibition, their ability to interfere with spore germination was assessed by co-inoculating B. cinerea spores with either the bacterial cells themselves or their spent medium (cell-free filtrate). Two strains showing no activity on mycelial growth (PIP26 and SOP5) were included as negative control in this analysis. In contrast to their effect on mycelial growth, the six Bacillus strains strongly differed in their impact on spore germination when applied as cells (**Figures 4A,E**): SOD5, CHD4, PID5, and SOD20 had no or only very marginal effect on the pathogen's germination rate, while CHP14 reduced it to 20% and SOP51 to 0%. When applied as cell-free filtrates, the effects of the strains on germination rate were generally less strong, although PID5, SOP51, and to a lesser extent SOD20, still caused marked reductions (**Figures 4B,E**). When measuring the length of the germ tube, significant reduction was obtained with all except SOD5 with both cells (**Figure 4C**) and cell-free filtrates (**Figure 4D**). Not only did the strains cause drastic tube length reduction, but they also led to specific developmental anomalies (depicted in **Figure 5**) such as hyphal swelling, multiple swelling in a row or swelling at the end of the germ tube. Proportions of these different phenotypes were strain specific and differed between spores exposed to cells and those exposed to cell-free filtrates of the same isolate (**Figures 5a,b**). For instance, multiple swelling in a row only occurred when exposed to cell-free filtrates and hyphal swelling was much more frequently observed in filtrate-treated than in cell-treated spores (**Figures 5a,b**), which might be partially due to the lower proportion of non-germinating spores in filtratetreated compared with cell-treated samples (**Figure 5**). In order to see whether the active strains (or their filtrates) only delayed germination or completely arrested it, spores were analyzed for germination rate, conidiophore production, the above-described

anomalies, as well as vacuolization and degradation after 24 h of incubation. The results are summarized in **Table 1** and illustrated in **Supplementary Figure S3**. Spore germination was observed for all treatments after 24 h (**Table 1**), but it was severely reduced in samples exposed to cells and cell-free filtrates of all strains but the inactive strains PIP26, SOP5 as well as SOD5, which confirmed the earlier time point observations (**Table 1** and **Supplementary Figure S3**). CHD4 had a strong inhibitory effect on germination when applied as cell suspension but its filtrate did not prevent the formation of conidiophores nor did it lead to vacuolization or visible degradation of fungal structures, unlike all other active strains (**Table 1** and **Supplementary Figure S3**).

#### Both Zoospores and Sporangia From Phytophthora infestans Were Affected in Their Germination by the Selected Bacillus Strains

Mycelial growth of P. infestans upon direct exposure to the six selected Bacillus strains was reduced from 86.2 to 96% (**Figure 2** and **Supplementary Table S2**). The effects of direct exposure to the same bacterial strains and their spent media on zoospore and sporangia germination (**Figure 6**) were more variable. While zoospore germination was slightly reduced by the presence of cells of PIP26, it was seemingly unaffected by the presence of SOP5, SOD5, CHD4, and SOP51 (**Figure 6A**). Although this latter strain led to complete absence of spore germination in B. cinerea, only PID5 and to a lesser extent CHP14 distinctly reduced the germination of zoospores of P. infestans (**Figure 6A**). Moreover, the phenotypes of the germinated zoospores appeared to be affected by all strains except SOP51 (**Figure 7a**). When in the presence of PIP26, SOD5, CHD4, and CHP14, the proportions of zoospores showing retarded germination (i.e., germination without appressorium formation, **Figure 7e**) were higher (**Figure 7a**). For the latter three Bacillus strains (SOD5, CHD4, and CHP14), the proportion of zoospores showing anomalies (**Figures 7c–f**) was higher than in the control. For SOP5, used as a negative control in these experiments, the reverse effect was observed (less retarded germination) but the proportion of zoospores showing anomalies was still high (**Figure 7a**). Zoospore germination was strikingly affected by PID5, showing only 1.75% of normal germination on average. Though for B. cinerea only germ tube length and not germination percentage was strongly reduced by PID5, the cells and their spent medium both seemed to specifically block the germination process of P. infestans zoospores (**Figures 7a,b**). Interestingly, when exposed to the spent medium, percentage of zoospore germination dropped significantly for SOD5, SOD20, and SOP51, whereas for CHP14, it was higher when compared to the bacterial

cells (**Figure 6B**). However, the germination inhibition observed with bacterial cells (**Figure 6A**) might be underestimated since in some treatments (e.g., SOD20), very few zoospores could be found, which may be due to aggregation with bacterial cells and/or degradation of zoospores by the bacteria (**Figure 6E**). A closer look at the germination phenotypes revealed strong reductions in normally germinating zoospores for the spent medium of SOD5 and PID5, and a higher proportion of delayed germination for the SOD5 cell-free filtrate (**Figure 7b**). On the contrary, the effects of the Bacillus strains on sporangia germination were stronger for the bacterial cells when compared to the cell-free filtrates (**Figures 6C,D**). Despite an overall low germination rate of sporangia, all strains but SOP5 and SOD20 reduced sporangia germination. CHP14 and SOP51 were the strongest inhibitors of sporangia germination, as observed for B. cinerea spores (**Figure 4A**).

### Anti-oomycete Volatiles Emitted by Selected Epiphytic Grapevine Isolates

Using a two-compartment Petri dish assay in our screening, we observed significant reduction of mycelial growth when the oomycete pathogen P. infestans was exposed to volatiles from three epiphytic strains isolated from the Chasselas cultivar (**Figure 2**). In contrast to the phylogenetically closely related Botrytis-inhibiting Bacillus isolates described above, these three strains (CHP14, CHP20, and CHP33) belonged to three different subgroups: B. subtilis for CHP14, B. cereus for CHP20 and Frigoribacterium for CHP33 (**Figure 1**). We expected their volatile blends to differ too strongly to allow identification of common volatiles involved in anti-oomycete activity. In order to identify the volatiles responsible for the mycelial growth reduction, we therefore looked for closely related strains that lacked such volatile-mediated activity. While no close relative could be found for CHP33, which is why we did not include this strain in the analysis, we selected CHD4 as "inactive counterpart" for CHP14, and SOD6 for CHP20. The differing activity of these four strains is shown with representative pictures in **Figure 8**. To compare their overall volatile profiles, we performed a principal component analysis (PCA) on both couples of active vs. inactive strains (**Figure 9**). As shown in the PCA score plots, there was – as expected for closely related strains - partial overlap between the volatilomes of the active vs. non-active strains, but multivariate analysis revealed features (mass ions corresponding to specific compounds) that were significantly different between the active and the inactive strain of each couple (in pink in the Volcano plots). The features that are of interest in our case are those enriched in or specific to active strains and can be found in the top right of the plots (**Figure 9**). The corresponding molecules are listed in **Table 2**. Only one of the two compounds that were significantly more abundant in the volatilome of CHP14 compared to CHD4 could be identified with confidence: the tryptophan derivative L-dihydroxanthurenic acid (**Figure 9A** and **Table 2**). When comparing the volatiles emitted by the active CHP20 with those emitted by its closely related yet less active counterpart SOD6, a stronger overlap was observed in the PCA scores plot for this couple than for the CHP14/CHD4 couple (**Figure 9**). Nevertheless, three compounds were significantly enriched in CHP20: trimethylpyrazine, dihydrochalcone, and the L-dihydroxanthurenic acid already detected in CHP14 (**Figure 9B** and **Table 2**).

#### DISCUSSION

#### Diversity of Cultivable Bacteria in the Leaves of Three Grapevine Cultivars

The major aim of this study was to isolate and identify new potential candidates for biological control of fungal and oomycete grapevine pathogens rather than to provide a comprehensive view of the leaf microbiota. This is why we focused on a cultivation-based approach to compare the endophytic and epiphytic bacterial communities present on three different grapevine cultivars grown on the same organically managed experimental vineyard in Prangins (Switzerland) (Vionnet et al., 2018). One striking feature in our survey was the overrepresentation of Bacillus and Staphylococcus species in all three cultivars, and especially among endophytic communities (**Figure 1**). Bacteria belonging to these genera are readily cultivable on standard microbiology media such as the PCA (Plate Count Agar) used in our study (Vionnet et al., 2018), which might partially explain their dominance. However, members of both genera have also been identified in cultivation-independent surveys of phyllosphere grapevine microbiota (Campisano et al., 2014; Perazzolli et al., 2014;

TABLE 1 | Presence or absence of different phenotypes in spores exposed to bacterial cells or their spent medium (filtrate) for 24 h.


Ten pictures per treatment were examined for the presence of the described phenotypes.

cells (left) or cell-free filtrates (right) are shown in (E). The experiment was repeated twice with similar results.

Yousaf et al., 2014; Singh et al., 2018). Vineyard management was shown to greatly influence not only the soil and rhizosphere (Calleja-Cervantes et al., 2015; Vega-Avila et al., 2015) but also the phyllosphere microbiota (Campisano et al., 2014). Notably, higher proportions of Staphylococcus and other commensal and opportunistic pathogens associated with animals and humans were observed in the endosphere of grapevine from organically managed vineyards (Campisano et al., 2014; Yousaf et al., 2014), which might be linked to the use of organic manure rather than mineral fertilizer. Such organic amendments are applied to the soil, but since above-ground microbial communities in grapevine have been shown to share a substantial proportion of their taxa with the soil (Zarraonaindia et al., 2015), it seems likely that microbes contained in organic manure might also

technical replicates were analyzed. Per technical replicate, one picture was taken and analyzed. Representative pictures of zoospores co-incubated with bacterial

FIGURE 8 | Volatile-mediated growth inhibition of P. infestans by selected strains. Representative pictures of dual assays with P. infestans non-exposed to the bacteria (Control) or exposed to four bacterial strains isolated from the grapevine microbiome are shown in (A). The pictures were taken 2 weeks after P. infestans inoculation. The percentage of P. infestans mycelium inhibition induced by each strain compared to the non-exposed control that represents 100% is shown in (B). Bars represent average inhibitions of 3–5 replicates with standard deviation. The percentage of inhibition was determined 2 weeks after P. infestans inoculation. Asterisks represent statistically significant differences in comparison to the control (t-test; <sup>∗</sup>p < 0.05; ∗∗p < 0.01).

FIGURE 9 | Multivariate analysis of volatiles emitted by active vs. closely related non-active strains. Two dual strain comparisons are shown: (A) CHP14 vs. CHD4; (B) CHP20 vs. SOD6. Active strains are shown in red, inactive relatives in green. Principal component analysis (PCA) score plots are shown in the upper row. These PCA plots display data variance over two principle axes, with explanatory percentages shown in brackets. Volcano plots displaying significance over fold-change are shown in the lower row. Mass features varying with high significance (p < 0.05) and fold-change (>2) are displayed in pink and framed red for features associated with active strains vs. green for those associated with inactive strains.

TABLE 2 | Volatile compounds enriched in active strains, as revealed by multivariate analysis on log transformed, normalized GC/MS data using MetaboAnalyst 3.0.


Compounds listed were significantly more abundant (p < 0.05) by a factor of at least two in active than in non-active strains. Identification of the enriched compounds was carried out based on the NIST17 database using the OpenChrom software. RT, retention time; m/z, mass/charge; Match, probability of correct compound identification. Lines in italic represent compounds identified with low match quality and therefore classified as "unknown."

colonize above-ground plant parts. Similarly to our study, few recent reports focusing on cultivable bacteria from grapevine leaves also identified Bacillus isolates as major components of endophytic communities (Baldan et al., 2014; Andreolli et al., 2016). In addition, they detected other genera such as Methylobacterium and Pantoea (Andreolli et al., 2016) or Staphylococcus, Paenibacillus, Microbacterium, Micrococcus, and Variovorax (Baldan et al., 2014), which were also isolated in the present study, either among endophytes or epiphytes (**Figure 1**). This relatively strong overlap at genus level is surprising when considering that these studies investigated different cultivars, i.e., Corvina (Andreolli et al., 2016) and Glera (Baldan et al., 2014), and that plant genotype is known to play a major role in shaping rhizosphere and phyllosphere microbiotas (Cardinale et al., 2015; Wagner et al., 2016; Berlanas et al., 2019). In our study, differences were observed between the three cultivars, especially among epiphytes (**Figure 1**): higher cultivable diversity was found in Pinot Noir than in Chasselas and Solaris, which

was associated with a lower proportion of Bacillus strains than in the two other cultivars. Among endophytes, all three cultivars shared a dominance of the genera Bacillus and Staphylococcus, although the proportions between these two varied between the cultivars (**Figure 1** and **Supplementary Table S1**). Apart from these rather minor differences between cultivars, the main differences in cultivable diversity occurred between epiphytes and endophytes: while only few genera were represented among endophytes, the leaf surface contained higher diversity, and notably a higher proportion of Gram-negative bacteria, such as the two well-described leaf inhabitants Sphingomonas and Methylobacterium (Müller et al., 2016; Compant et al., 2019). Lower diversity among endophytes than epiphytes is expected (Bulgarelli et al., 2013) and can be linked to a stronger selection pressure by the plant and to the necessity for endophytes to be able to overcome additional barriers, such as the endodermis in roots or the cuticle in leaves. Moreover, the stronger overlap between the three cultivars among endophytes than among epiphytes (**Figure 1**) might relate to tight association between endophytes and their host. In contrast to endophytic communities, epiphytic communities might harbor both highly adapted strains selected for their ability to withstand harsh abiotic conditions (e.g., UV, desiccation), and "transient passengers" brought by wind and rain but lacking the specific equipment necessary to establish stable populations in the phyllosphere. Current isolation procedures do not allow distinguishing between these two types of epiphytes and identifying plant-beneficial strains in this particular compartment thus does not warrant their long-term survival in the plant phyllosphere. In contrast, endophytes may represent a more promising pool of isolates to search for host-adapted microbes with plant-beneficial functions, such as health protection through inhibition of plant pathogens (Brader et al., 2014; Hardoim et al., 2015).

### Antagonistic Activity of Grapevine Leaf Isolates Against Botrytis cinerea

When comparing the antagonistic potential of epiphytic vs. endophytic isolates on two phytopathogens, we observed that a high proportion of endophytes was able to inhibit mycelial growth of both pathogens in full plate assays, while this ability was less widespread among epiphytes (**Figure 2** and **Supplementary Table S2**). As discussed above, this might be due to lesser selection pressure on plant surface- than on inner tissuecolonizing bacteria, although further studies on a more extensive strain collection would be needed to confirm this hypothesis. In such full plate assays where both diffusible and volatile compounds can be exchanged between the two partners, only seven strains (three epiphytes and four endophytes) were able to strongly inhibit B. cinerea mycelial growth (>50%, red color in **Figures 2**, **3**). From these seven strains, six belonged to the genus Bacillus (**Figure 1**). Botrytis cinerea is one of the major fungal pathogens worldwide, causing great pre- and post-harvest losses in a wide range of agronomically relevant crops (Dean et al., 2012). One key element promoting fast disease dispersion of this pathogen is the massive production of conidia that are easily spread by wind (Williamson et al., 2007). An efficient biocontrol agent should therefore not only inhibit mycelial growth but also prevent spore formation or germination. We thus tested whether the six Bacillus grapevine isolates showing similar and high inhibition of B. cinerea mycelial growth would also inhibit spore germination. This was indeed the case for five of them, while the sixth (SOD5), which clustered in a different clade closely related to B. licheniformis (**Figure 1**), did not interfere at all with spore germination (**Figures 4**, **5**). From the five remaining strains, different phenotypes were observed, some strains (e.g., CHD4) showing no effect on germination rate but leading to smaller germ tubes, others strongly affecting both germination rate and germ tube length (e.g., CHP14 and SOP51). Specific anomalies in germ tube development were observed for some strains or groups of strains: CHD4 filtrate led to a high proportion of hyphal swelling at the germ tube tip, and the three closely related PID5, SOD20 and SOP51 induced similar disturbances when applied as cell-free filtrates, while they differed in their effects when applied as cell suspensions (**Figure 5**). This suggests that the spent medium of these three strains contained the same (or similar) active compounds. While we did not determine their identity in this study, one likely candidate class of compounds that could be responsible for the observed effects is the lipopeptides. Indeed, Bacillus strains are well known for their ability to produce a wide range of such surface-active molecules with broad effects on phytopathogens (Ongena and Jacques, 2008; Liu et al., 2014), some of which were shown to inhibit B. cinerea development (Touré et al., 2004; Haggag, 2008; Zhang et al., 2013). Beyond lipopeptides, other antimicrobial compounds produced by Bacillus strains, such as polyketides (e.g., macrolides), or lytic enzymes (e.g., chitinases), could also be involved in the observed effects (Caulier et al., 2019). Reports of volatile-mediated activity of Bacillus strains on B. cinerea also exist (Chen et al., 2008; Liu et al., 2008; Zhang et al., 2013), but this pathogen was not inhibited in its growth in our volatile assays, in contrast to the oomycete P. infestans.

#### Antagonistic Activity of Grapevine Leaf Isolates Against Phytophthora infestans

Downy mildew caused by the oomycete Plasmopara viticola is one of the most important diseases threatening grapevine health (Gessler et al., 2011; Kamoun et al., 2015). This pathogen is an obligate parasite, which makes in vitro assays such as the screen of grapevine isolates for antagonistic activity very challenging. This is why we have selected a closely related pathogen, the oomycete P. infestans as a surrogate for the early selection of strains with anti-oomycete activity. Although both pathogens might react differently to biological control agents, earlier work on Lysobacter strains revealed their efficacy against both pathogens (Puopolo et al., 2014a,b), which raises hope that some of the strains identified as inhibitors of P. infestans might also inhibit P. viticola. Our screening revealed that in contrast to B. cinerea, which was inhibited only by few Bacillus strains, P. infestans mycelial growth was significantly reduced by a number of grapevine isolates affiliated with different genera (Bacillus, Staphylococcus and Micrococcus among Gram-positive

and Variovorax, Pantoea, Herbaspirillum and Erwinia among Gram-negative) (**Figure 2**). This suggests higher sensitivity of the oomycete to the diffusible metabolites produced by the bacteria during full plate confrontation. A high proportion of endophytes severely inhibited P. infestans mycelial growth, and they mostly belonged to the Bacillus genus. As discussed above for B. cinerea, lipopeptides might be involved in this strong growth inhibition, as susceptibility of this pathogen (and of P. viticola) to different lipopeptides has been reported (Stanghellini and Miller, 1997; Tran et al., 2007; Zachow et al., 2015). In addition, small cyclic peptides such as those produced by Lysobacter capsici might also play a role in this inhibition (Nishanth Kumar et al., 2012; Puopolo et al., 2014a). Both types of compounds (lipopeptides and small cyclic peptides) should be contained in the cell-free filtrates and indeed, the filtrates of the three most active strains against B. cinerea (PID5, SOP21 and SOP51) also strongly perturbed P. infestans zoospore germination, indicating broad range activity of the secreted molecules. Beyond Bacillus strains and in view of the broad phylogenetic distribution of strains that reduced P. infestans mycelial growth through diffusible compounds, a similarly broad chemical diversity of responsible molecules (Stringlis et al., 2018) might underlie the growth inhibition and identifying these molecules shall be the scope of future studies. In contrast to diffusible compound-mediated growth inhibition, reduction of P. infestans mycelial growth by the emission of volatile compounds seemed largely restricted to Bacillus strains (with the exception of one Frigoribacterium, CHP33), and specifically to those isolated as epiphytes from the Chasselas cultivar (**Figure 2**). Although volatiles emitted on nutrient-rich media such as LB might only very partially resemble those emitted on leaf surfaces (Blom et al., 2011), they still provide an idea of the metabolic potential of strains showing volatile-mediated inhibition of pathogensin vitro. When comparing the volatilomes of these active isolates to those of phylogenetically closely related, but non-active isolates, we identified few volatiles enriched (or only detected) in active strains. These compounds, which might therefore be involved in the volatile-mediated growth inhibition of the oomycete pathogen (**Figure 9** and **Table 2**), included trimethylpyrazine, which is known to have antifungal activity (Méndez-Bravo et al., 2018). Interestingly, one compound appeared as enriched in both active strains compared with their inactive counterparts: L-dihydroxanthurenic acid. Very little is known about this molecule, which appears to be produced in insects by a kynurenine transaminase (Dolores Real and Ferré, 1991). Given the molecular structure of this tryptophan derivative, one might indeed speculate that it is an intermediate in the kynurenine pathway, which has been implicated in the biosynthesis of (non-volatile) antimicrobial and cell-cell communication signals such as quinolones (Gross and Loper, 2009) or the siderophore thioquinolobactin involved in anti-oomycete activity (Matthijs et al., 2007). Moreover, a recent study identified a genomic locus in a Pseudomonas strain involved in anti-oomycete activity, which contained several genes encoding different steps of the kynurenine pathway (Wagner et al., 2018). Although the biosynthesis and biological activity of L-dihydroxanthurenic acid is so far unknown, this newly detected volatile might be an interesting candidate for further studies of its potential implication in the inhibition of oomycetes such as P. infestans or P. viticola.

#### Harnessing the Potential of Grapevine Endophytes and Epiphytes for Sustainable Disease Control

Overall, our results show that grapevine leaves are a rich source of potential biocontrol agents of fungal and oomycete pathogens. There was no striking difference in the relative abundance of antagonistic strains between the different cultivars, although emitters of anti-oomycete volatiles were all isolated from the disease-sensitive Chasselas cultivar. More comprehensive surveys on larger pools of sensitive vs. resistant cultivars would be needed to draw conclusions about the role of the phyllosphere microbiome in disease resistance, which was not the aim of the present study. However, it seems from our results that leaf isolates producing diffusible compounds inhibiting pathogen growth were more frequently retrieved in the endophytic communities, while all emitters of anti-oomycete volatiles belonged to epiphytic communities. Indeed, volatile compounds are likely more efficient in fending off disease-causing agents when emitted on the surface of the plants, in contact with air, which allows their dispersion to places where spores of pathogens might land. In contrast, diffusible substances and contact-mediated inhibition of penetrating pathogens might constitute a further line of defense provided by bacteria living in the leaf endosphere. Accordingly, providing grapevine with particularly health-protective endophytes (e.g., by soil drenching) at the beginning of the season, could be combined with later leaf spraying treatments with epiphytic emitters of pathogen-inhibiting volatiles. Together with the use of disease-resistant cultivars and continuous monitoring of disease pressure to allow timely treatments, such application of biological agents adapted to their host plant might contribute to efficient disease control strategies of lower environmental impact than the traditionally used synthetic fungicides.

#### DATA AVAILABILITY STATEMENT

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

LW, FL'H, and MD designed the research. SB, MZ, FL'H, ET, MD, AA, and AD performed the experiments. SB, MZ, FL'H, ET, AA, and MD analyzed the data. LW, FL'H, SB, MZ, MD, and AA wrote the manuscript with help from ET.

#### FUNDING

Funding from the Swiss National Science Foundation (grant 179310 to LW) is gratefully acknowledged.

### ACKNOWLEDGMENTS

fmicb-10-02726 November 27, 2019 Time: 17:28 # 16

The authors are grateful to Dr. Aurélie Gfeller and Léo Vionnet for their help in the initial strain isolation efforts. They further

#### REFERENCES


thank Dr. Paolina Garbeva for her advice on the statistical analysis of volatile profiles and Lucas Caiubi Pereira for his help in dual assay experiments.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.02726/full#supplementary-material

hyphal growth of the plant pathogen, Botrytis cinerea. Biotechnol. Lett. 30, 919–923. doi: 10.1007/s10529-007-9626-9629



and carposphere: an NGS approach. Microorganisms 6:96. doi: 10.3390/ microorganisms6040096


**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 © 2019 Bruisson, Zufferey, L'Haridon, Trutmann, Anand, Dutartre, De Vrieze and Weisskopf. 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.

# Beneficial Endophytic Bacteria-*Serendipita indica* Interaction for Crop Enhancement and Resistance to Phytopathogens

Alejandro del Barrio-Duque, Johanna Ley, Abdul Samad, Livio Antonielli, Angela Sessitsch and Stéphane Compant\*

Bioresources Unit, Center for Health and Bioresources, AIT Austrian Institute of Technology, Tulln, Austria

Serendipita (=Piriformospora) indica is a fungal endophytic symbiont with the capabilities to enhance plant growth and confer resistance to different stresses. However, the application of this fungus in the field has led to inconsistent results, perhaps due to antagonism with other microbes. Here, we studied the impact of individual bacterial isolates from the endophytic bacterial community on the in vitro growth of S. indica. We further analyzed how combinations of bacteria and S. indica influence plant growth and protection against the phytopathogens Fusarium oxysporum and Rhizoctonia solani. Bacterial strains of the genera Bacillus, Enterobacter and Burkholderia negatively affected S. indica growth on plates, whereas Mycolicibacterium, Rhizobium, Paenibacillus strains and several other bacteria from different taxa stimulated fungal growth. To further explore the potential of bacteria positively interacting with S. indica, four of the most promising strains belonging to the genus Mycolicibacterium were selected for further experiments. Some dual inoculations of S. indica and Mycolicibacterium strains boosted the beneficial effects triggered by S. indica, further enhancing the growth of tomato plants, and alleviating the symptoms caused by the phytopathogens F. oxysporum and R. solani. However, some combinations of S. indica and bacteria were less effective than individual inoculations. By analyzing the genomes of the Mycolicibacterium strains, we revealed that these bacteria encode several genes predicted to be involved in the stimulation of S. indica growth, plant development and tolerance to abiotic and biotic stresses. Particularly, a high number of genes related to vitamin and nitrogen metabolism were detected. Taking into consideration multiple interactions on and inside plants, we showed in this study that some bacterial strains may induce beneficial effects on S. indica and could have an outstanding influence on the plant-fungus symbiosis.

Keywords: *Serendipita indica*, bacterial endophytes, symbiosis, *Mycolicibacterium*, tripartite interactions, biocontrol, fungal stimulation

### INTRODUCTION

In all terrestrial plants and natural ecosystems, multipartite interactions take place between plants and different kinds of microbes (Hardoim et al., 2015). While some bacteria can have a negative impact on the surrounding microflora (Barea et al., 2005; Deveau et al., 2018), other bacteria are known as boosting colonization and establishment of endophytic fungi such as in

#### *Edited by:*

Kalliope K. Papadopoulou, University of Thessaly, Greece

#### *Reviewed by:*

Gustavo Santoyo, Universidad Michoacana de San Nicolás de Hidalgo, Mexico Rong Li, Nanjing Agricultural University, China

> *\*Correspondence:* Stéphane Compant stephane.compant@ait.ac.at

#### *Specialty section:*

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

*Received:* 19 September 2019 *Accepted:* 02 December 2019 *Published:* 19 December 2019

#### *Citation:*

del Barrio-Duque A, Ley J, Samad A, Antonielli L, Sessitsch A and Compant S (2019) Beneficial Endophytic Bacteria-Serendipita indica Interaction for Crop Enhancement and Resistance to Phytopathogens. Front. Microbiol. 10:2888. doi: 10.3389/fmicb.2019.02888 the case of mycorrhiza (van Overbeek and Saikkonen, 2016), acting therefore as helper bacteria (Frey-Klett et al., 2007). Some of these bacteria are also plant growth-promoting rhizobacteria (PGPR), capable of increasing plant growth and resistance against fungal pathogens by mobilizing nutrients, production of siderophores, auxins, ACC-deaminase, polyamines as well as antibiosis and chitinolytic compounds (Whipps, 2001; Glick, 2014). Most of the helper bacteria were isolated from the rhizosphere. However, some could thrive as endophytes in their host plants, crossing from the root surfaces to the inner plant tissues (Brader et al., 2017). These bacteria are particularly important for the development of biostimulant products, as those microbes with the ability to colonize roots are more prepared to survive and exert beneficial effects on the plant (Compant et al., 2005). Bacteria can further penetrate fungal hyphae and establish symbiosis, playing sometimes critical roles for the survival of the fungal hosts (Bertaux et al., 2005; Bonfante and Desirò, 2017; Guo et al., 2017). Fungal hyphae provide a nutrientrich niche for bacterial growth (Linderman, 1992; Boer et al., 2005; Scherlach et al., 2013). The hyphae-associated bacteria may reciprocally promote hyphal development and root colonization through the supply of vitamins, nitrogen, phosphorus, sugars, and secondary metabolites to the fungal partner or by increasing ATP production (Hildebrandt et al., 2006; Ghignone et al., 2012), as exemplified by the mycorrhizal helper bacteria of arbuscular mycorrhizal fungi (AMF) (Frey-Klett et al., 2007). However, bacteria with suppressive effects on fungi (e.g. secretion of antifungal compounds, mycophagy) have also been widely reported (Bonfante and Anca, 2009; Kobayashi and Crouch, 2009). Understanding and exploiting the association between different kinds of beneficial microorganisms, together with the management and engineering of the plant microbiome, may lead to improving soil fertility, crop productivity, and biological control of plant pathogens (Berg, 2009; Collinge et al., 2019; Compant et al., 2019).

An example of beneficial fungi interacting with symbiotic bacteria is Serendipita indica, previously known as Piriformospora indica (Varma et al., 1999). It is a well-known root endophytic fungus that mimics the capabilities of AMF, but in contrast to AMF it can be cultured axenically (Varma et al., 1999). The fungus improves crop yield and confers resistance against biotic and abiotic stresses by triggering induced systemic resistance (ISR), boosting antioxidant capacity, mobilizing nutrients, and manipulating the hormone levels of the plant (Franken, 2012; Gill et al., 2016). Interestingly, it has been demonstrated that this fungus hosts an endobacterium, Rhizobium radiobacter (Sharma et al., 2008), which has endophytic as well as plant growthpromoting properties, although its functional role is not yet fully understood (Glaeser et al., 2016). Only few studies have demonstrated in vitro that particular bacterial strains can be detrimental for the growth of S. indica (Varma et al., 2013) or in contrast have stimulatory effects, like strain WR5 of Azotobacter chrococcum, which enhanced mycelial growth and sporulation of S. indica in vitro (Bhuyan et al., 2015). In the last few years, some researchers have developed co-inoculations (microbial consortia) of fungi and bacteria, searching synergisms between two beneficial microbes (Artursson et al., 2006; Collinge et al., 2019). Special focus has been laid on the combination of S. indica with PGPR. Kumar et al. (2012) detected increased plant growth by combining S. indica and pseudomonad R81. However, combinations of S. indica and bacteria have not always been successful (Sarma et al., 2011). Similarly, several combinations of different biocontrol agents have not improved the effect exerted by the most efficacious one, indicating no synergistic but more likely antagonistic interactions (Xu et al., 2011). This suggests that more research is needed in the exploration of compatible combinations. We hypothesized, nevertheless, that several bacterial taxa or strains can positively interact with the beneficial fungus S. indica and that these bacterial-fungal interactions could be exploited for crop enhancement and resistance against different phytopathogens.

The aim of this study was to identify bacterial taxa that stimulate the growth and effects of the beneficial fungus S. indica and to select the most promising "fungus-conducive" bacteria for dual-inoculation with S. indica. The effect of these combinations on plant growth and biocontrol activity against fungal pathogens was further assessed. In particular, we aimed at researching the tripartite interactions between S. indica, various bacterial strains and the plant pathogens F. oxysporum or R. solani. We further explored potential mechanisms involved using genome analysis of most promising bacterial endophytes showing positive interaction with S. indica.

### MATERIALS AND METHODS

#### *Serendipita indica* Cultivation

Serendipita (=Piriformospora) indica strain DSM 11,827 was provided by Pr. Philipp Franken and obtained from the "Deutsche Sammlung für Mikroorganismen und Zellkulturen," Braunschweig, Germany (Varma et al., 1999). The fungus was maintained at −80◦C in sterile Potato Dextrose Broth (PDB) (Carl Roth, Germany) amended with 25% glycerol and grown on Potato Dextrose Agar (PDA) plates or in liquid culture containing Aspergillus complete medium (Pontecorvo et al., 1953).

To produce inoculum, roughly fifty 2-mm agar plugs from a 2-week old culture of S. indica grown on PDA were transferred to 250 ml Erlenmeyer flasks containing 100 ml of Aspergillus CM and incubated for 3 weeks under constant shaking (150 rpm) at 26 ± 1 ◦C. Mycelium and spores were collected by centrifugation (4,500 rpm, 5 min) and the remaining pellet was washed 3–5 times with sterile phosphate-buffered saline (PBS) of pH 6.5. The mixture of mycelium and spores, resuspended in PBS, was ground with a homogenizer Ultra Turrax T25 (IKA <sup>R</sup> , Staufen, Germany) for 3 min in intervals of 30 s. The number of spores + mycelium fragments was estimated with a hemocytometer (NanoEnTek, Seoul, Korea) and the viability of the CFU confirmed by plating on PDA. Final concentrations were adjusted with PBS.

### Plant Assay for Isolation of Endophytic Bacteria

An agricultural soil was sampled in a field located in Meires, Lower Austria, Austria (48◦ 46′ 43.8′′N; 15◦ 17′ 23.4′′E) at a depth of 5–15 cm and stored at 4◦C. To create an isolated microsystem, one potato tuber (Solanum tuberosum L. cv. Romina; NÖ. Saatbaugenossenschaft, Austria) and four tomato seeds (Solanum lycopersicum L. cv. Moneymaker; Austrosaat, Vienna, Austria) were grown in closed Magenta boxes (Sigma-Aldrich) containing ∼200 g of the agricultural soil. The plants were kept in greenhouse with a Day/Night temperature of 22/21◦C, a relative humidity 50/35% and a 12 h light/dark photoperiod. Plants were watered weekly with 20 ml of water. After 4 weeks, potato and tomato plants were inoculated by drenching a mixture of spores + fragmented mycelium of S. indica at 10<sup>4</sup> CFU/g of soil. Control treatment was mock-inoculated with PBS. To mitigate a possible big shift in the bacterial community caused by S. indica, an additional treatment with low-concentrated (10<sup>2</sup> CFU/g). S. indica inoculum was included for potato plants. In total five treatments were prepared with three replicates (boxes) per treatment (**Supplementary Figure 1**).

#### Isolation of Endophytic Bacteria

Three weeks after inoculation, 1.5 g of potato and 0.2 g of tomato roots were harvested from each magenta box to isolate endophytic bacteria. For this, roots were rinsed abundantly with tap water, then surface-sterilized with 70% ethanol for 10 s followed by 2.5% sodium hypochlorite for 3 min and then rinsed 3 times with sterile water. For each sample, 100 µl of the final wash were plated in triplicates on Nutrient Agar No2 (NA) (Sigma-Aldrich, St. Louis, USA) until 6 days of incubation at 26◦C to confirm the surface sterilization. To isolate bacteria, sterile roots, macerated in 5 ml of sterile 0.85% NaCl, were smashed with a mortar and pestle and homogenized by vortexing for 30 s at maximum speed. Smashed roots from each sample were 10-fold diluted on PBS and 100 µl from each dilution were plated on NA and further incubated at 26◦C for 6 days. Based on visual differences in morphology, single colonies were randomly picked only from the most diluted cultures to avoid contaminations. Selected bacteria were further purified by repeating streaking on NA plates and all the recovered isolates were stored at −80◦C in Nutrient Broth (NB) (Difco, Detroit, MI) supplemented with 25% glycerol. At least 100 isolates were obtained from every treatment with potato plants, and 50 with tomato plants.

#### DNA Isolation and 16S rRNA Gene Sequencing

Bacterial DNA of each isolate was extracted using the UltraClean <sup>R</sup> Microbial DNA Isolation Kit (QIAGEN, Venlo, Netherlands) according to the manufacturer's instructions. The 16S rRNA genes were amplified using the primers 8F (5′ -AGAGTTTGATCCTGGCTCAG-3′ ) (Weisburg et al., 1991) and 1520R (5′ -AAGGAGGTGATCCAGCCGCA-3′ ) (Edwards et al., 1989). A conventional PCR amplification of 20 µl PCR reaction mix containing 2.5 mM MgCl2, 0.2 mM dNTPs, 0.3 mM of each primer, 1–2 µl of DNA template, 1 U HOT FIREPol <sup>R</sup> DNA polymerase (Solis BioDyne) and 1 × PCR reaction buffer (Invitrogen) was carried out in a thermocycler peqSTAR 96X HPL (PEQLAB Biotechnologie GmbH). An initial denaturation step at 95◦C for 5 min was followed by 30 cycles of denaturation at 95◦C for 45 s, annealing at 54◦C for 60 s and elongation at 72◦C for 90 s, plus a final extension at 72◦C for 10 min were performed. Sequencing of the PCR product was performed by LGC-Genomics (Teddington, UK) using the primers 8F and 1495r (5′ -CTACGGCTACCTTGTTACGA-3 ′ ) (Lane, 1991). To remove duplicate sequences from the library, sequences were de-replicated and clustered at 100% similarity with the Avalanche NextGen Workbench (http:// www.visualbioinformatics.com/bioinf/index.html). Strains showing the same partial 16S rRNA genes and the same phenotypic interaction with S. indica were removed. The identification of isolates was performed by BLAST search on a local installation of the complete NCBI's nt database (downloaded in October 2018), targeting the first more significant 50 hits. Taxonomic assignment of BLAST hits was then refined with the approach implemented in BlobTools (Laetsch and Blaxter, 2017). Sequence data are available at NCBI database and GenBank under the accession numbers MN180888–MN181366.

### Effect of Endophytic Bacteria on Mycelial Growth of *S. indica*

To determine the effect of bacteria on S. indica, the growth of the fungus in interaction with each endophytic bacterium was expressed in terms of hyphae expansion on agar plates. For this, bacteria were pre-cultured for 4 days on NA and the fungus for 2 weeks on PDA. To study the interaction between bacteria and S. indica, each bacterium was then streaked on 1 cm<sup>2</sup> in the center of a Petri dish (9 cm Ø, containing 15 ml PDA) and a 0.5-cm<sup>2</sup> agar plug of active S. indica mycelium was placed inverted over the streaked bacterium. As a control, S. indica was grown alone. All the co-cultures were replicated four times. Aiming a confirmation of results, some selected isolates were additionally co-cultured with S. indica under different nutrient conditions (a mixture 1:1 of PDA+NA), and under longer bacterial growth phase (9 days of bacterial preculture).

Plates were incubated at 26◦C in darkness. After 13 days of dual-culturing, the surface of the plate covered with S. indica mycelium was measured with ImageJ 1.48 software (https://imagej.nih.gov/ij/) and the average measurement of the four replicated plates was employed to determine the type of interaction. The difference in growth between S. indica cocultured with a bacterium (dual-cultured S. indica) and control was calculated as relative increase/decrease of dual-cultured S. indica growth respect to the control. To characterize the type of interaction between fungus and bacteria, an arbitrary scale was further established. When the growth of dual-cultured S. indica was reduced by more than 90% in respect to the control, the interaction was considered as a complete inhibition. A reduction between 90 and 20% was determined as a negative interaction. If the growth of dual-cultured S. indica was decreased or increased up to 20% respect to the control, a neutral interaction was established. An increment of dual-cultured S. indica growth larger than 20% defined a positive interaction (**Figure 1B**).

FIGURE 1 | Isolates of bacteria and interaction with S. indica. (A) Relative abundance of bacterial families isolated from potato and tomato roots, and from plants pre-inoculated or not with S. indica. (B) Growth of the fungus alone (a) and in combination with bacteria completely inhibiting fungal growth (b), interacting negatively (c), in a neutral way (d), and stimulating the growth of the fungus (e). Bar corresponds to 3 cm. (C) Number of bacterial clusters (strains) and their phylogeny per type of interaction.

### Combining Selected Beneficial Endophytic Bacteria and *S. indica* for Tomato Growth Promotion

Since most of the isolates from one genus (Mycolicibacterium) stimulated S. indica growth, four isolates (P1-5, P1-18, P9-22, and P9-64) of this genus were selected for further experiments. With the aim of studying the effect of dual inoculations of S. indica and bacteria on tomato plants, a pot experiment was conducted in the greenhouse.

For the pot experiment, S. indica inoculum was produced as described above. In the case of bacteria, they were grown for 2.5 days in 10 ml bottom-rounded Falcon tubes containing 5 ml NB at 26◦C and constant shaking (190 rpm). The cultures were centrifuged (4,600 rpm, 6 min, room T◦C) and washed 3 times with sterile PBS to remove traces of media. Cell growth was determined by measuring the OD and CFU were estimated by standard serial dilution on NA.

Tomato seeds cv. Moneymaker were germinated for 4 days in a Whatman <sup>R</sup> filter paper (110 mm Ø) at room temperature on Petri dishes. Germinated seeds were transferred to 50 ml falcon tubes containing 15 ml of PBS and either (i) bacterial cells (5 × 10<sup>7</sup> CFU/ml), (ii) spores + hyphae fragments of S. indica (5 × 10<sup>5</sup> CFU/ml), or (iii) a mixture of bacterial cells (5 × 10<sup>7</sup> CFU/ml), and S. indica (5 × 10<sup>5</sup> CFU/ml). Falcon tubes were maintained in a Tube Roller RS-TR5 (Phoenix Instrument GmbH, Garbsen, Germany) for 30 min. For control treatment, the seeds were immersed in 1 × PBS.

Two seeds per pot were then sown at 1 cm depth in pots (1 l capacity) containing the substrate "Fruhstorfer Erde Typ Nullerde" (Hawita Gruppe, Vechta, Germany). The experiment had 10 treatments, ± S. indica and ± individual bacterial strains with 14 replicates (7 pots × 2 plants per pot) for each treatment. The plants were grown in the greenhouse (with conditions described before) and watered twice a week with tap water. Plants were harvested 6 weeks after planting and shoot fresh weight and leaf area were measured (using ImageJ). A confirmatory experiment was repeated with a richer soil in which pots were filled with a mixture (1:1:1 v/v) of perlite, sand, and the substrate "Tonsubstrat ED63 Special" (Einheitserde, Germany). After harvesting, shoot fresh and dry weight (after oven-drying for 3 days at 70◦C) were measured.

#### *In vitro* Interaction Between Selected Bacteria and Plant Pathogens

In addition to determining the effect of selected bacteria on the growth of S. indica and tomato plants, they were also tested for their effects on pathogens such as Fusarium oxysporum f. sp. lycopersici Fol4287 (kindly provided by Maria E. Constantin, University of Amsterdam, Netherlands) (Di Pietro and Roncero, 1996) and Rhizoctonia solani AG-3 (kindly provided by Rosanna C. Hennessy, University of Copenhagen, Denmark). These fungi were maintained at −80◦C in PDB amended with 25% glycerol. In vitro dual culture assays between these fungal pathogens and selected bacterial isolates were performed as described above for the beneficial fungus S. indica, but due to different growth rates, fungal preculture and final measurements were shortened to 3 days in case of R. solani, and 6 days for F. oxysporum.

### Tomato Protection Against *Fusarium oxysporum* and *Rhizoctonia solani* Using Multipartite Interaction

Biocontrol of F. oxysporum was evaluated by pot experiment in which plants were infected with the pathogen and with single or dual inoculations of S. indica and selected bacteria. Inoculum production of S. indica and bacteria and seed inoculation were carried out as described above. The seeds were planted in 1 l pots containing a mixture (1:1:1 v/v) of perlite, sand and the substrate "Tonsubstrat ED63 Special" (Einheitserde, Germany) (2 seeds per pot, 5 pots per treatment).

For production of Fusarium, spores were obtained according to van der Does et al. (2019). Briefly, an agar plug from a 6-day old PDA culture of Fusarium was transferred to a 250-ml Erlenmayer flask containing 100 ml minimal media (3% sucrose, 0.17% yeast nitrogen base without amino acids or ammonia, and 100 mM KNO3), and incubated for 5 days at 26◦C, 190 rpm. Spores were filtered through a Miracloth filter (Millipore), washed twice with sterile 1×PBS and diluted to a concentration of 10<sup>7</sup> spores/ml.

Ten days after planting, tomato plants were infected with the solution of F. oxysporum described before according to the root dip method (Wellman, 1939). Seedlings were uprooted and trimmed leaving roughly 1 cm of root, to facilitate the penetration of Fusarium. Roots were placed for 30 min in the spore suspension of Fusarium, and directly repotted. Five weeks after inoculation, plant weight above the cotyledons was measured, and the extent of disease progression was scored according to de Lamo et al. (2018). Briefly, disease index was 0 = no symptoms, 1 = one brown vessel above the soil, 2 = one or two brown vascular bundles at the cotyledon level, 3 = at least three brown vessels and growth distortion, 4 = all vessels brown or the plant is small and wilted, 5 = dead plant.

The effect of dual-inoculation of S. indica and endophytic bacteria against the damping-off causative agent R. solani was analyzed, in parallel to F. oxysporum test, by a germination assay in closed boxes (Steri Vent Containers 107 × 94 × 96 mm, Duchefa Biochemie b.v, Haarlem, Netherlands). These boxes contained 120 g of a sterile (2 times, 121◦C, 20 min) 1:4 mixture (w/w) of vermiculite (2–3 mm, Sigma-Aldrich) and distilled water. Tomato seeds cv. Moneymaker were surface-sterilized with 2.5% sodium hypochlorite for 5 min and rinsed 8 times with sterile water. Seeds were inoculated with either (i) S. indica, (ii) (x4) bacteria, or (iii) (x4) combination of fungus and bacteria as earlier described.

For this, Rhizoctonia was cultured on 1/5 PDA for 3 weeks. Five agar plugs of Rhizoctonia mycelium were placed in a row in the center of each box, and 2 rows of tomato seeds (5 seeds per row) were sown on both sides of the pathogen row at 2 cm distance. Both phytopathogen and seeds remained at 0.5 cm depth. Control was prepared with plugs of 1/5 PDA. Three replicated boxes were prepared per treatment and maintained in the greenhouse (see above). The number of germinated seedlings per box was monitored regularly for 4 months by scoring as follows: 1 = Plant germinated and no disease symptoms, 0.5 = Plant germinated, alive, with necrotic areas in leaves and stems, 0 = plant dead.

#### Bacterial Genome Sequencing and Analysis

Bacterial genomic DNA from 4 selected Mycolicibacterium strains were isolated using a phenol-chloroform based protocol according to Samad et al. (2016). Concisely, cells were grown on NB for 3 days and collected by centrifugation. Each bacterial pellet was resuspended in lysis buffer (5 mM EDTA pH8, 50 mM Tris-Cl, 1% SDS, 0.5 M NaCl, 0.2 mg/ml Proteinase K) and incubated at 65◦C overnight, 400 rpm. DNA was extracted 2 times using 1 volume of phenol-chloroform-isoamylalcohol (25:24:1) and collected by centrifugation. Genomic DNA was further cleaned with Amicon Ultra 0.5 mL 30K Centrifugal Filter Units (Millipore, Cork, Ireland) and re-suspended in water. Whole-genome shotgun sequencing was performed on an Illumina HiSeq (GATC Biotech, Konstanz, Germany), producing 2 × 150 bp reads.

Illumina reads were checked for the presence of PhiX using Bowtie 2 (v2.3.4.3) (Langmead and Salzberg, 2012) and adapters were removed with fastp (v0.19.5) (Chen et al., 2018). Sequence quality and length distribution were checked via FastQC<sup>1</sup> (Andrews, 2010). Genome assembly was carried out with SPAdes v3.13.0 (Bankevich et al., 2012) and short (<500 bp), low-abundant (<2×) contigs filtered out. The presence of contaminant contigs was assessed using BlobTools and alien contigs were eventually removed. Genome assembly quality was then inferred using QualiMap v2.2 (Okonechnikov et al., 2015) and QUAST v5.0.0 (Gurevich et al., 2013) and genome completeness reconstruction was evaluated with BUSCO v3.0 (Waterhouse et al., 2017). Gene annotation was performed using Prokka v1.12 (Seemann, 2014) and NCBI Prokaryotic Genome Annotation Pipeline (PGAP). Contigs were further screened for the presence of antimicrobial resistance or virulence genes with ABRicate v0.8.10. The presence of plasmids was ascertained by using Mash v2.1 against the PLSDB database (Galata et al., 2018).

Functional annotation was performed using EggNOG 4.5 (Huerta-Cepas et al., 2015) and the ClassicRAST (Rapid Annotation using Subsystem Technology) web server (http:// rast.nmpdr.org) (Aziz et al., 2008). Prediction of biosynthetic gene clusters and secondary metabolites was additionally carried out using antiSMASH version 4.0.2 (Weber et al., 2015). CAZy families were identified with dbCAN2 according to the DIAMOND database. A cutoff of E-Value of 1e-102 was set for the output. When a gene contained a CBM with other CAZy classes, the gene was classified as CBM. Protein annotation was based on the CAZy database (Cantarel et al., 2008; Lombard et al., 2013).

The Average Nucleotide Identity (ANI) analysis was further used to determine the relatedness between the assembled genomes and affiliated genomes available in NCBI database classified as Mycobacterium or Mycolicibacterium. For this, 169 genomes were downloaded using the script available at https://github.com/kblin/ncbi-genome-download and ANI was calculated with the pyani Python module available at https://github.com/widdowquinn/pyani, using BLAST (ANIb) and TETRA methods. Based on the ANI pair-wise values, a distance matrix representing ANI-divergence (defined as 100% ANI data) (Chan et al., 2012) was compiled to display a heat map and compute a dendrogram using the hierarchical clustering adopting the complete linkage algorithm, with the software Morpheus (https://software.broadinstitute.org/morpheus/).

The draft genome sequences for the Mycolicibacterium strains P1-5, P1-18, P9-22, and P9-64 are available at NCBI, BioProject PRJNA393298, with the DDBJ/ENA/GenBank accession numbers NPKT00000000, NPKR00000000, NPKP00000000, and NPKO00000000, respectively.

#### Statistical Analysis

Statistical analysis of in vitro fungal growth, analysis of biomass and leaf area of samples from the growth enhancement experiments were performed in R 3.5.1 (R Core Team, 2019). Data distributions were checked using the fitdistrplus package (Delignette-Muller and Dutang, 2015) and linear or linear mixed-effects models (nlme R package) (Pinheiro et al., 2019), when applicable, were generated. After graphical verification of homogeneity assumption, ANOVA was applied on previously generated models, followed by pairwise comparisons (Tukey's method, P = 0.05) calculated using Estimated Marginal Means (emmeans R package) (Lenth, 2019). Quantitative data were processed with dplyr package (Wickham et al., 2019) and results visualized with boxplots using ggplot2 package (Wickham and Chang, 2016). The samples from the biocontrol experiment were analyzed with PRISM 8.0 (GraphPad). Concerning Fusarium, a non-parametric Mann-Whitney U-test was applied on the fresh weight and disease index data (de Lamo et al., 2018). The germination assay with Rhizoctonia was analyzed by one-way ANOVA and Tukey's test (P = 0.05).

#### RESULTS

#### Isolation and Identification of Bacteria

In total, 479 isolates were recovered and identified at the genus level (**Table 1**). The most abundant families of isolates were Bacillaceae (21.09% of the total isolates), Enterobacteriaceae (13.78%), Rhizobiaceae (11.48%), and Paenibacillaceae (10.02%) (**Table 1** and **Supplementary Figure 2A**). At the genus level, the overall top five genera were Bacillus (21.09% of isolates), Enterobacter (13.57%), Paenibacillus (8.56%), Burkholderia (7.52%), and Agrobacterium (6.05%). After de-replication with a 100% threshold, we obtained 260 different clusters (strains) with different 16S rRNA genes. The most abundant strains belonged to the families Bacillaceae (17.31%), Paenibacillaceae (15.0%), Enterobacteriaceae (11.15%), and Microbacteriaceae (9.23%). At the genus level, the most abundant strains belonged to Bacillus (17.31%), Paenibacillus (12.69%), Enterobacter (10.77%), and Mycolicibacterium (7.31%).

Tomato roots were thinner than potato roots, thus fewer isolates were recovered from tomato plants (25.47% of the total isolates) using the sterilization procedure. After de-replication,

<sup>1</sup>FastQC. https://www.bioinformatics.babraham.ac.uk/projects/fastqc/

#### TABLE 1 | Species identified by sequencing the 16S rRNA gene and number of isolates and clusters (strains) assigned to species.


(Continued)

#### TABLE 1 | Continued


(Continued)



Ctrl stands for control plants.

only 15 strains were shared by tomato and potato plants (**Supplementary Figure 2B**). The bacterial community isolated from potato plants differed considerably to the tomato plants (**Figure 1**). Interestingly, strains belonging to Enterobacteriaceae (overall top second family) were only found in potato roots. Abundances of bacterial taxa isolated from the same plant species, but under different treatment, were more similar. Nevertheless, there were some differences between control plants and plants inoculated with S. indica (**Table 1** and **Figure 1A**). Particularly, strains of the genus Burkholderia (9 strains, 36 isolates) were only found in control plants of tomato and potato, or in plants inoculated at 10<sup>2</sup> CFU/g, but never in the treatments of plants heavily (10<sup>4</sup> CFU/g) inoculated with S. indica (**Table 1** and **Supplementary Figure 2C**).

#### Interaction Between *S. indica* and Bacteria

From the whole assemblage of bacteria co-cultured in vitro with S. indica, similar number of strains were found in each type of interaction. Twenty percentage of the total strains were completely inhibitory, 26% negative, 28% neutral, and 26% positive for S. indica growth (**Figure 1C**), revealing that S. indica must coexist with antagonistic microbes, but also with synergistic ones, during the process of root colonization by the fungus.

Bacillaceae, Enterobacteraceae, and Burkholderiaceae were the most detrimental families for S. indica growth (**Figure 1C**). All the strains from Enterobacteriaceae displayed an inhibitory or negative interaction with the beneficial fungus. Similar results were obtained with members of Bacillaceae, especially strains of B. subtilis, B. pumilus, B. velezensis, and B. thuringiensis, although strains of B. simplex displayed a neutral effect to S. indica growth. In the Burkholderiaceae family, all the strains of Burkholderia completely inhibited S. indica growth or were strongly negative, while one strain of Paraburkholderia displayed positive interactions (**Supplementary Figure 2D**). Likewise, the great majority of Leifsonia, Rathayibacter, and Pseudomonas (95%) strains showed an inhibitory or negative interaction with S. indica.

The most abundant families showing a positive interaction were Mycobacteriaceae (22.67% of the total positive strains), Rhizobiaceae (13.33%), Xanthomonadaceae (10.67%), Paenibacillaceae (10.67%), and Rhodanobacteraceae (9.33%). Considering the type of interaction within a family, every strain of Xanthomonadaceae, and 17 out of 19 for Mycobacteriaceae stimulated S. indica growth. At the genus level, several strains classified as Achromobacter and Sphingomonas were further positive for S. indica growth, although some others were just neutral.

### Screening of Bacterial Strains for Further Experiments

Since Mycobacteriaceae was the family that contains more strains stimulating fungal growth, four isolates of the genus

Mycolicibacterium (P1-5, P1-18, P9-22, and P9-64) recovered from potato roots were selected for further experiments. These isolates strongly stimulated S. indica growth when co-cultured on PDA and after 4 days of bacterial pre-culture (**Figure 2A**). The four isolates further stimulated S. indica growth on different growing media (PDA+NA) and bacterial growth phase (9 days preculture) (**Supplementary Figure 3**), except for P9-22 that did not significantly increase fungal growth on PDA+NA. To rule out the hypothesis that the stimulating effect of bacteria on fungal growth is due to stressed hyphae running away from the bacterium, we further confronted S. indica with the selected isolates, but growing on different zones of the Petri dish. Concomitantly, S. indica growth was also stimulated when the bacteria were streaked few centimeters away from the fungus (**Figure 2B**).

### Effect of Combined Inoculation of Endophytic Bacteria and *S. indica* on Tomato Growth

To determine effects of selected Mycolicibacterium strains and S. indica on plants, tomato plants were inoculated with single or dual inoculations. The fresh weight in all the inoculated treatments was never lower than control (untreated) plants (**Figure 3**), therefore none of these microbes seemed detrimental or pathogenic for plant growth. Apart from the strain P1- 5, single inoculations of bacteria increased plant growth, but only P1-18 and P9-22 increased shoot fresh weight significantly. Inoculation of plants with S. indica increased shoot fresh weight (3.3-fold). Dual inoculations of S. indica+P1-5 and S. indica+P1- 18 further enhanced the beneficial effect triggered by S. indica, but it resulted significant only for leaf area measurements of plants inoculated with S. indica+P1-5. Contrarily, dual inoculations of S. indica+P9-22 and S. indica+P9-64 displayed lower performance than inoculation of S. indica alone. To confirm the plant growth promotion triggered by these microbes, this experiment was repeated in soil with high content of nutrients but in this experiment no differences were obtained between treatments (**Figure 4**).

#### *In vitro* Interaction Between Selected Bacterial Strains and Fungal Pathogens

By in vitro dual-culturing, the effect of selected Mycolicibacterium strains on Fusarium and Rhizoctonia growth was studied. In contrast to the beneficial interaction observed between S. indica and the selected Mycolicibacterium strains, none of these bacteria stimulated F. oxysporum growth. Interestingly, some strains significantly reduced hyphal growth respect to the control (**Supplementary Figure 4**) under certain growth conditions. In the case of Rhizoctonia solani, none of the four bacteria restrained fungal growth in vitro. Contrarily, the strains P9- 22 and P9-64 slightly stimulated fungal growth under certain growth conditions.

#### Tomato Protection Against *Fusarium oxysporum* and *Rhizoctonia solani* Using Dual Inoculations

Tomato plants infected with the pathogen F. oxysporum (Fol) showed typical symptoms as leaf yellowing, necrotized vessels, wilting, and death (as described in van der Does et al., 2019). On average, fresh weights of Fusarium-treated plants were always reduced in comparison with mock-inoculated plants, although it was statistically significant uniquely for plants single-inoculated with P9-64 (**Figure 5B**). The extent of disease progression (i.e., yellowing, brown bundles, wilting)

non-significantly different mean values, according to Tukey's test (P < 0.05) after ANOVA. n = 14. Bar in the picture corresponds to 6 cm.

seemed to be alleviated when the plants were treated with the beneficial fungus S. indica but it was not statistically significant. Only during combined treatments of S. indica+P1-18 and S. indica+P9-22, the level of disease progression was significantly reduced (**Figure 5A**). Furthermore, single inoculation of P9-22 significantly reduced Fol symptoms to the same extent as the combination S. indica+P9-22.

Concerning the experiments with Rhizoctonia, the effect of seedlings germinated and alive was regularly monitored up to 111 days. The effects of the damping-off caused by Rhizoctonia first appeared 7 days after planting (dap), affecting only treatments in which S. indica was not inoculated (**Figure 5C**). Seedlings inoculated with S. indica as well as combinations of S. indica + bacteria were not affected by the pathogen until 11 dap. In contrast, single inoculation of Mycolicibacterium P1- 5 accelerated the damping off caused by Rhizoctonia and the number of seedlings alive was significantly decreased 25 dap in comparison to Rhizoctonia control. In general, inoculation of seedlings with S. indica conferred resistance against the pathogen, but uniquely dual inoculations of S. indica + bacteria significantly maintained the number of seedlings alive for all the measurements from 18 to 111 dap, in comparison to Rhizoctonia control. The combinations of S. indica+P1-18 (at 73 dap) and S. indica+P9-64 (at 73 and 111 dap) further significantly increased the number of plants alive in comparisons to single inoculation of S. indica.

#### Genome Analysis

#### Genomic Features, ANI and Phylogeny of Sequenced Strains

The genomes of the Mycolicibacterium strains P1-5, P1-18, P9- 22, and P9-64 have a total of 5.47, 6.70, 6.79, and 7.34 Mb with an average G+C content of 65.95, 68.74, 66.89, and 66.27%, respectively. No evidence of plasmids was ascertained. The analysis of antimicrobial resistance or virulence genes detected few genes (**Supplementary Table 1**). In particular, strain P1-5 shows the presence of the rbpA gene, that can confer resistance to rifampin, and the strains P1-18 and P9-64 harbor the gene tet(V), possibly involved in tetracycline resistance. The genomic features of the four genomes are summarized in **Table 2**. To determine the relatedness of the sequenced strains to genomes publicly available at the NCBI database, ANI was calculated with the BLAST algorithm (ANIb), and with tetranucleotide frequency correlation coefficients (TETRA). The maximum values of ANIb for the four sequenced genomes were only in the range of 80–90% (**Supplementary Table 2**) and therefore the proposed threshold of ≈95% ANIb as the putative boundary for species circumscriptions was not reached (Konstantinidis and Tiedje, 2005; Richter and Rosselló-Móra, 2009). Contrarily, the TETRA values for the isolates P1-5 and P9-22 reached the 99% threshold (**Supplementary Table 2**) required to support the species circumscription (Richter and Rosselló-Móra, 2009). However, since TETRA values > 99% should agree with ANIb > 95–96% (Richter and Rosselló-Móra, 2009), we did not assign species names to these isolates.

The dendrogram computed with the distance matrix of pairwise ANI values showed two separated groups (**Figure 6**). One group represents the clade of slow-growing mycobacteria designated as "Tuberculosis-Simiae" and depicts the emended genus Mycobacterium (Gupta et al., 2018; Oren and Garrity, 2018). This group includes well-known human pathogens (Gupta et al., 2018), most notably Mycobacterium leprae and

Mycobacterium tuberculosis, causative agents of leprosy and tuberculosis, respectively (Magee and Ward, 2012; Lory, 2014). The second group, in which the four sequenced bacteria are included, encompasses species from the clade "Fortuitum-Vaccae," recently transferred to a new genus, Mycolicibacterium gen. nov. (Oren and Garrity, 2018).

#### Genes and Proteins Predicted to Stimulate S. Indica Growth

Analysis of the genomes revealed the four Mycolicibacterium strains contain numerous genes predicted to be involved in the stimulation of S. indica growth. Some vitamins and cofactors are indispensable for fungal growth (van Overbeek and Saikkonen, 2016), and since some vitamins like cobalamin (B12) can only be synthesized by bacteria (Ghignone et al., 2012; Danchin and Braham, 2017), we hypothesized that production of vitamins could be one of the key factors in the stimulation of fungal growth. We identified several genes involved in the synthesis of six vitamins of the vitamin B complex: cobalamin (B12), biotin (B7), thiamin (B1), riboflavin (B2), pyridoxin (B6), and folate (B9), as well as menaquinone and phylloquinone of the vitamin K complex in all the four genomes. Moreover, genes related to nitrogen metabolism such as those for nitrate and nitrite reductase, ammonification, ammonium transporters, and glutamine synthase were detected in all four genomes (**Table 3** and **Supplementary Table 3**).

#### Genes and Proteins Related to Plant Growth Promotion Traits

The RAST annotation and the functional annotation of proteins based on the eggNOG protein database detected various genes related to plant growth promotion (PGP) traits. In the genomes of the strains P1-18, P9-22, and P9-64, we identified key genes attributable to well-known plant growth-promoting compounds like siderophore synthesis and receptors as well as siderophore-interacting proteins, auxin biosynthesis and acetoin and butanediol metabolism (**Table 4** and **Supplementary Table 4**). Regarding strain P1- 5, the antiSMASH analysis detected a secondary metabolite cluster identified as mycobactin (**Supplementary Table 5**), a siderophore used by members of the genus Mycobacterium to shuttle free extracellular iron ions into the cytoplasm (McMahon et al., 2012). This cluster was also detected in the genomes of P9-22 and P9-64.

The four genomes, and especially P9-64, contain also genes involved in phosphate solubilization (**Table 4** and **Supplementary Table 4**). These genomes encode phosphatases and pyrroloquinoline quinone biosynthesis (pqqE genes) that catalyzes the synthesis of gluconic acid, considered as one of the major organic acids responsible for mineral phosphate solubilization (Wagh et al., 2014; Liu et al., 2016). Phosphate solubilization was additionally confirmed in vitro (data not shown). Furthermore, inorganic phosphate transport and uptake may be facilitated by low- and high-affinity phosphate transport systems, detected in these genomes. Similarly, iron is an essential nutrient for plant nutrition that is mainly absorbed by plants as ferrous iron (Morrissey and Guerinot, 2009). Genes coding for ferrous iron transporters are present in the four genomes analyzed, contributing to the provision of iron to the plants. Polyamines are phytohormone-like compounds with biological activity in processes like plant growth, development, and stress mitigation (Niemi et al., 2002; Kuznetsov et al., 2006). Furthermore, we identified several proteins involved in the transport and synthesis of the polyamines spermidine and putrescine in all genomes. Moreover, the strains P1-18 and P9-64 contain the gene acdS for 1-aminocyclopropane-1-carboxylate (ACC) deaminase (**Supplementary Table 6**), that enhances plant growth by lowering plant ethylene levels (Glick, 2014).

**159**


#### Genes and Proteins Related to Stress Tolerance

The four genomes encode genes implicated in protection against diverse stresses (**Table 4** and **Supplementary Table 4**), and this protection may indirectly lead to growth promotion (Liu et al., 2016). We further found genes involved in resistance to heavy metals and metalloids including cobalt, zinc, cadmium, copper, arsenic, mercury, chromium, and selenite. We also identified numerous proteins and compounds that protect the cell from oxidative stress: peroxidases, catalases, hydroperoxide reductases, superoxide dismutases, glutathione S-transferases, and mycothiol (Newton et al., 2008). The four genomes also encode domain proteins of rhodanese (**Supplementary Table 6**), an enzyme that detoxifies cyanide (Cipollone et al., 2008), and nitrilases and cyanide hydratases, enzymes with critical roles in plant-microbe interactions for defense, nitrogen utilization, detoxification, and synthesis of plant hormones (Howden and Preston, 2009). Strain P1-18 further contains genes involved in the synthesis of mycosporines (**Table 4**), secondary metabolites considered to be amongst the strongest natural absorbers of UV radiation and with antioxidative capacities (Oren and Gunde-Cimerman, 2007).

The four strains might potentially confer tolerance to salt stress as they possess several copies of the genes encoding choline dehydrogenase and betaine aldehyde dehydrogenase (**Table 4** and **Supplementary Table 4**), required to produce glycine betaine, one of the most important solutes to face osmolarity fluctuations (Nau-Wagner et al., 2012). The genomes contain also several genes involved in trehalose biosynthesis, a sugar with a protective effect under salt and drought stress (Garg et al., 2002). The strain P9-22 further harbors genes for ectoine (**Tables 4**, **5**), an osmolyte that helps organisms survive extreme osmotic stress (Bernard et al., 1993). Complementarily, all strains also contain various K<sup>+</sup> transport and Na+/H<sup>+</sup> antiporters that contribute to resist hyperosmotic stress (Liu et al., 2016).

#### Resistance to Antibiotics and Production of Antibiotic Compounds

Genes for antibiotic resistance may protect the plant against other pathogenic microbes. The four genomes encode the enzymes β-lactamases, that provide multi-resistance to βlactam antibiotics such as penicillins (Neu, 1969), genes involved in resistance to the bactericide fluoroquinolone, and genes encoding multidrug resistance proteins (**Table 4** and **Supplementary Table 6**). P1-18 and P9-64 have genes involved in the degradation of oxalate, a compound secreted by fungi to promote their growth and colonization of substrates (Dutton and Evans, 1996), which might contribute to plant defense against pathogenic fungi.

Similarly, we identified several genes involved in the production of antibiotics compounds (**Tables 4**, **5** and **Supplementary Tables 4–6**), including bacteriocins, clavulanic acid (Reading and Cole, 1977), aminoglycosides antibiotics (Davies and Wright, 1997) and type IV pili, a bacterial virulence mechanism that appears operational during pathogenesis of fungal hosts (Dörr et al., 1998). Polyketides are secondary metabolites that have antimicrobial properties, including the mycotoxins produced by fungi (Huffman et al., 2010). We identified in the four genomes several enzymes involved in the synthesis of polyketides (**Table 5** and **Supplementary Tables 5, 6**). Among the biosynthetic gene clusters (BGCs) characterized as polyketides synthases (PKS) by antiSMASH analysis, several have been identified as alkylresorcinols, phenolic lipids with the ability to inhibit bacterial and fungal growth (Stasiuk and Kozubek, 2010). Others BGCs characterized as PKS were identified as the antibiotics rifamycin, FK520 (ascomycin), ansamitocin and tetrocarcin A, and as the siderophore griseobactin (Patzer and Braun, 2010) (**Supplementary Table 5**). However, the percentage of gene match was very low in comparison to their homologs BGCs, suggesting that these BGCs might encode novel antibiotic and siderophore biosynthetic pathways (de Los Santos-Villalobos et al., 2018). Other antibiotic BGCs detected by antiSMASH with antifungal properties, although with low percentage of gene match, include galbonolides (Fauth et al., 1986) identified in all the genomes, bacillomycin (Gu et al., 2017) in the genomes of P9-22 and P9-64, angucycline Sch 47554 (Basnet et al., 2006), pimaricin (Aparicio et al., 2016) in P1-18, and fengycin (Vanittanakom et al., 1986) in P1-5. We further identified proteins involved in the synthesis of phenazines (**Supplementary Table 6**), heterocyclic compounds that have been shown to control a wide range of plant pathogenic fungi (Chin-A-Woeng et al., 2003) and to elicit ISR (Pierson and Pierson, 2010).

FIGURE 6 | Heat map and dendrogram of average nucleotide identity (ANI) values amongst different strains of Mycobacteriaceae showing separation of two groups. The group on the left include common human pathogens and are included in the genus Mycobacterium, clade "Tuberculosis-Simiae." The group on the right consists of species of the new genus Mycolicibacterium, clade "Fortuitum-Vaccae." The four sequenced strains P1-5, P1-18, P9-22, and P9-64 are included in the group of Mycolicibacterium (right side).



n ◦R, number of subsystem roles/proteins; n◦G, number of protein encoding genes (peg).

In addition, these bacteria contain genes encoding enzymes involved in the degradation of the fungal cell-wall, like βhexosaminidases and chitooligosacharide deacetylases, which have been proved to degrade chitin and chitooligosaccharides (Barber and Ride, 1989; Zhao et al., 2010). Furthermore, genes encoding chitinases are present in the genome of P1-5 (**Tables 4**, **6**) and these enzymes can potentially contribute to biocontrol of fungal pathogens by disruption of fungal cell walls (Whipps, 2001).

#### CAZy Analysis

We identified roughly 140 putative genes encoding carbohydrateactive enzymes (CAZy) in each genome analyzed (**Table 7**). CAZy were distributed unevenly among the six CAZy families and no enzymes belonged to the class polysaccharide lyases (PL). In the classes carbohydrate esterases (CE) and glycoside hydrolases (GH), we identified several plant cell-wall degrading enzymes (PCWDEs) related genes in all the genomes analyzed (**Table 6** and **Supplementary Table 7**). The PCWDEs identified have the potential to degrade many plant cell-wall polymers, including cellulase, hemicellulose, pectin, and cutin. In addition, P1-18 encodes proteins involved in lignin degradation (**Supplementary Table 6**). Many other genes related to enzymes involved in the degradation of different plant intracellular polysaccharides were detected (**Table 6**), suggesting that the strains harbor traits for endophytic colonization.

#### DISCUSSION

In our study, we confirm that the endophyte S. indica coexists with communities of deleterious, neutral and beneficial bacteria inside roots as revealed by in vitro assays, and there seems to be an ecological balance among these microbial communities (Varma et al., 2012). Bacillaceae, Enterobacteraceae, and Burkholderiaceae were the most detrimental families for S. indica growth. These results are congruent with the widely reported antifungal properties of several strains of Bacillus and Burkholderia (Compant et al., 2005; de Los Santos-Villalobos et al., 2018). In contrast, many Rhizobiaceae strains representing different species and genera stimulated S. indica growth in vitro. Intriguingly, it has been confirmed that S. indica hosts an endobacterium of Rhizobium radiobacter inside its hyphae and the bacterium increases host fitness (Sharma et al., 2008; Glaeser et al., 2016). Hence, it might be that this interaction is not specific and several Rhizobium relatives may promote S. indica growth. It is not known whether different rhizobia can colonize the fungus internally and it also remains unclear if the isolated Rhizobiaceae strains, rather than stimulating the fungus directly, stimulate the activity of the endobacterium of S. indica leading to improved growth. Likewise, several isolates identified as Paenibacillus exhibited a positive or neutral interaction with S. indica. These results are in line with Hildebrandt et al. (2006) who demonstrated that the bacterium Paenibacillus validus stimulates growth of the AMF Glomus intraradices. Moreover, one of the few endofungal bacteria detected in the Serendipita (=Sebacina) vermifera complex belongs to Paenibacillus (Sharma et al., 2008), suggesting a possible synergistic interaction between some strains of Paenibacillus and Serendipita. The fact that rhizobia and Paenibacillus inoculants are frequently applied as biofertilizers (Sessitsch et al., 2002; Grady et al., 2016), suggests that strains of these taxa could be further tested for application jointly with S. indica.

TABLE 4 | Protein encoding genes predicted to be involved in plant growth promotion and resistance of strains P1-5, P1-18, P9-22, and P9-64 determined by RAST.


n ◦R, number of subsystem roles/proteins; n◦G, number of protein encoding genes (peg).

Most striking were the effects of Mycolicibacterium strains, which highly stimulated S. indica growth. This genus is only poorly understood regarding its interaction with plants, although some strains have been tested as bioinoculants owing to plant growth promotion effects (Egamberdiyeva, 2007). The predicted traits from the genomes of the four Mycolicibacterium strains revealed the presence of many genes responsible for vitamin production, which are potentially relevant for supplying vitamins

to S. indica and thereby enhancing its growth. It has been reported that endosymbionts of mycorrhizal fungi are important for the provision of vitamin B12 (cobalamin) to their host (Ghignone et al., 2012), and vitamin B1 (thiamin) is implicated in the growth-promoting effect of Pseudomonas fluorescens on the ectomycorrhizal fungus L. bicolor (Deveau et al., 2010). Moreover, it has been described that S. indica possesses biotrophassociated genomic adaptations, such as lacking genes related



to nitrogen metabolism and therefore suffering from some metabolic deficiencies. Congruent with this, S. indica barely grows on nitrate, but shows good growth on ammonium and glutamine as N source (Zuccaro et al., 2011). In accordance with these studies, we found several genes coding for nitrate and nitrate reductase as well as amino acid and peptide ABC transporters (genes families of proteins that have undergone contraction in S. indica genome) in all four Mycolicibacterium genomes. The strains also encode ammonium transporters and glutamine synthase, an enzyme that plays an essential role in the metabolism of nitrogen by catalyzing the condensation of glutamate and ammonia to form glutamine. These genomic features might be involved in increasing supply of glutamine and ammonium to the fungus, complementing the predicted metabolic deficiencies. Furthermore, it has been identified that trehalose is involved in the stimulation of hyphal growth of mycorrhizal fungus (Duponnois and Kisa, 2006; Hildebrandt et al., 2006). The Mycolicibacterium strains tested in this study encode genes for synthesis of trehalose, a compound that can be present as a disaccharide in the cytoplasm, but it is also present in the cell-wall glycolipids of Mycobacteria (Argüelles, 2000). The secretion of this sugar and the degradation of cellwall glycolipids into oligosaccharides might also contribute to S. indica growth stimulation. To this point, there is no clear evidence, if only one, or a combination of the above-mentioned bacterial traits are responsible for the positive interactions with S. indica.

The four isolates contain many genes related to plant growth promotion traits. As predicted by the analysis of the genomes, strains P1-18, P9-22, and P9-64 enhanced plant growth, while strain P1-5 did not improve plant growth. A plausible reason is that, unlike the other strains, P1-5 lacks some of the most well-known genes involved in PGP, like those for biosynthesis of auxin, the volatile acetoin and ACC deaminase, besides harboring fewer number of genes involved in siderophore and phosphatases production as well as nitrilases, that have a potential role in the biosynthesis of indole-3-acetic acid (auxin) (Park et al., 2003). S. indica has been extensively shown to increase plant growth in different crops (reviewed in Franken, 2012). Consistent with this, in our study the inoculation of tomato plants with S. indica increased shoot fresh weight and leaf area.

Considering the PGP traits of the Mycolicibacterium strains tested in this study, a dual inoculation of S. indica with these bacteria can potentially increase plant growth. However, only dual inoculations of the strains P1-5 and P1-18 with S. indica moderately enhanced plant growth in comparison to single inoculations. These results are in agreement with Sarma et al. (2011) and Kumar et al. (2012), who observed plant growth promotion by combining pseudomonads and S. indica, and concomitantly with our results, the dual inoculation increased only slightly in comparison to single inoculations of each microbe. Interestingly, the strain P1-5 did not enhance plant growth when single-inoculated but was notably effective in plant growth promotion when applied in combination with S. indica. This synergistic effect can be ascribed to cooperation in the supply of phosphorus and nitrogen to the plant. These bacteria and S. indica encode genes involved in the provision of P to the plant, but it has been controversially discussed if, and how, S. indica supplies P to the plant. It has been shown that the fungus is able to solubilize phosphate from inorganic, but not from organic P sources (Ngwene et al., 2016), but also that S indica is not involved in the phosphate transfer to host plant (Achatz et al., 2010). The inconsistency of these results reveals how complex the interaction is, and the outcome might be dependent on abiotic factors, like adequate pH. In this regard, the combined effect of bacterial and fungal phosphatases and phosphate transport systems under certain pH levels triggered by the presence of both microbes might be the key factors to plant P uptake and growth promotion. Similarly, the fact that P1-5 and P1-18 possess genes for nitrate reductase, an enzyme which plays a key role in nitrate acquisition in plants (Gill et al., 2016), and for ferrous iron transport, might complement the nitrogen and iron supply to the host plant. Contrarily, co-inoculation of the strains P9-22 and P9-64 with S. indica displayed lower performance than single inoculations. These results coincide with Sarma et al. (2011), in which dual inoculation of S. indica and the pseudomonad R62 was more detrimental than R62 and S. indica inoculated singly. Some authors claim that the negative interaction might be due to niche competition for both space and nutrients as described by Whipps (2001). Similarly, this incompatibility might also be ascribed to alterations in the IAA (auxin) levels of the plants. Provision of low levels of IAA stimulate plant growth whereas high concentration of IAA may influence plant growth negatively (Sarwar and Frankenberger, 1994). S. indica can produce auxins, and these strains harbor the largest number of genes involved in auxin biosynthesis, therefore the dual inoculation might lead to auxin overproduction and the consequent imbalance of TABLE 6 | Plant and microbe cell-wall polysaccharide degrading enzymes (CE and GH classes) of strains P1-5, P1-18, P9-22, and P9-64 based on genome analysis.


The numbers refer to the number of genes found in each CAZy family.

IAA levels, causing less growth promotion than each microbe inoculated singly.

Furthermore, our study reveals that the beneficial effect of the endophytes on plant growth is dependent on the cultivation substrate. Dual and single inoculation under nutrient-rich conditions did not exhibit shoot growth promotion 6 weeks after planting while on low-nutrient soil it did. Concomitant with our results, it has been shown that under certain nutrient conditions and plant stage S. indica and PGPR do not exert beneficial effects (Egamberdiyeva, 2007; Fakhro et al., 2010; Gill et al., 2016) and the outcome of fungal-bacteria interactions can be different from mutualistic to antagonistic. For example, Gorka et al. (2019) recently reported that the interaction between ectomycorrhiza and soil bacteria responded negatively to soil nitrogen application.

S. indica has been shown to confer resistance against several fungal pathogens (Waller et al., 2005; Qiang et al., 2012) including Fusarium (Deshmukh and Kogel, 2007; Sarma et al., 2011; Rabiey et al., 2015) and Rhizoctonia (Knecht et al., 2010), through activation of the antioxidant system (Prasad et al., 2013), defense related genes (such as PR, LOX2, and ERF1) (Zarea et al., 2012) and ISR. In agreement, S. indica always alleviated to

TABLE 7 | Genes related to carbohydrate-active enzymes (CAZymes) in strains P1-5, P1-18, P9-22, and P9-64.


CBM, Carbohydrate-Binding Module; CE, Carbohydrate Esterase; GH, Glycoside Hydrolase; GT, Glycosyl Transferase; PL, Polysaccharide lyases; AA, Auxiliary Activity.

some extent the symptoms caused by the fungal pathogens in these experiments. Bacteria employ many different mechanisms involved in biocontrol of fungal pathogens (Compant et al., 2005). Two of the most important mechanisms of biocontrol are the production of siderophores, depriving pathogenic fungi from iron acquisition (Kobayashi and Crouch, 2009), and production of antibiotics and fungal cell-wall degrading enzymes (Whipps, 2001). The genomes of Mycolicibacterium strains tested in this study encode genes for siderophore synthesis and receptors, as well as several antimicrobial compounds, including antibiotics, polyketides, phenolic lipids (alkylresorcinols), phenazines, and chitinolytic enzymes, confirming the genetic potential of these strains to exhibit biocontrol effects against pathogenic fungi. However, single inoculations of these bacteria did not protect the plants from pathogen attack. This might not be surprising considering the bacteria did not show in vitro direct antagonism to Fusarium and Rhizoctonia. Furthermore, inoculation of tomato seedlings with the strain P1-5 increased the negative effect caused by Rhizoctonia. A possible explanation is that in the same way the bacterium stimulates S. indica growth, it stimulates also the growth of Rhizoctonia enhancing damping off. Moreover, the bacterium P1-5 might help to detoxify the plant material after pathogen attack (Howden and Preston, 2009), by production of cyanide hydratases, carotenoids and antioxidants. Only strain P9-22 alleviated the symptoms of Fusarium wilt disease when single-inoculated. In reference to its genome, this biocontrol effect might stem from activation of ISR and competition for iron, as this strain encodes the highest number of genes implicated in siderophore synthesis and receptors (including BGCs identified as mycobactin, coelichelin, scabichelin) and numerous genes involved in triggering ISR (phenolic lipids—alkylresorcinols, cell wall-degrading enzymes, polyketides, antibiotics), as well as from direct antagonisms, like secreting the antifungal bacillomycin.

Overall, in this study dual inoculations of S. indica and bacteria enhanced the protective effect conferred by S. indica against the two pathogens. The strains that better performed were P1-18 and P9-22 against Fusarium, and P9-22 and P9- 64 against Rhizoctonia. Similar results showing synergism by dual-inoculations have been earlier reported (Whipps, 2001). For instance, dual inoculation of S. indica and pseudomonad R81 improved biocontrol of Fusarium compared to single inoculations (Sarma et al., 2011). This could be explained by cooperation in triggering the plant ISR. According to its biotrophic lifestyle, S. indica lacks also genes potentially involved in biosynthesis of toxic secondary metabolites and cyclic peptides, and family proteins involved in PKS and NRPS are contracted (Zuccaro et al., 2011). This deficiency could be ameliorated by these bacterial helpers, as they possess several genes involved in secondary metabolite production, including PKS and NRPS. These secondary metabolites might supplement S. indica metabolism, bioenergetic capacity, activation of defense related genes and production of antibiosis compounds (Bonfante and Anca, 2009; Bhuyan et al., 2015; Salvioli et al., 2016), that ultimately raises the plant ISR. These strains possess also genes that might cooperate in restraining pathogen expansion, like strains P1-18 and P9-64 that might degrade oxalate produced by pathogens. Moreover, competition for niche and nutrients (carbon, nitrogen, and iron) has been shown to be a mechanism associated with biocontrol or suppression of Fusarium wilt in several systems (Whipps, 2001). All in all, perhaps the combined effect of bacterial and S. indica-mediated ISR, bacterial production of siderophores and antimicrobial compounds [e.g., polyketides, non-ribosomal peptides, phenazines, chitinolytic enzymes, bacillomycin (P9-22 and P9-64), angucycline and pimaricin (P1-18), galbonolides] and the nutrient and niche competition between the pathogen and beneficial microbes might explain the enhanced resistance of plants inoculated with S. indica + Mycolicibacterium.

These results demonstrate the potential of Mycolicibacterium-S. indica combinations for biocontrol of plant pathogens, but the safety of these bacteria should be carefully addressed. As the dendrogram shows, these strains are separated from the clade "Tuberculosis-Simiae," and included in the clade "Fortuitum-Vaccae," primarily comprised of environmental species (Gupta et al., 2018). Besides, the analysis of antimicrobial resistance or virulence genes detected very few antimicrobial resistance genes of concern and the most related strains were classified as risk 1 according to the German classification TRBA. Nevertheless, the fact that these isolates are not related to the well-known human pathogens does not imply that they are completely safe. Moreover, future studies must consider how the application of fungal and bacterial inoculants affect soil microbial communities. For example, several studies provided evidence that mycorrhizal fungi modify the bacterial communities in the rhizosphere (Nuccio et al., 2013). Similarly, Meena et al. (2010) showed that S. indica affected population dynamics of pseudomonads in chickpea and also Nautiyal et al. (2010) observed changes in the microbial community structure in soil inoculated with S. indica. Most probably, the previously reported shifts in the bacterial communities might be attributed to the modified plant physiology, altered composition of root exudates and changes in the pH (Linderman, 1992; Barea et al., 2005; Svenningsen et al., 2018). Future research is needed in culture independent analysis to study the effect of S. indica in the native soil populations, and in exploring the interaction between S. indica and other positive strains belonging to different taxa that were isolated in this study. This might help to better understand bacteria-fungal interactions, and the selection of compatible microbial strains for field application, aiming at crop enhancement and biocontrol of fungal pathogens.

#### DATA AVAILABILITY STATEMENT

The datasets generated for this study can be found in the Sequence data are available at NCBI database and GenBank under the accession numbers MN180888–MN181366. The draft genome sequences for the Mycolicibacterium strains P1-5, P1-18, P9-22, and P9-64 are available at NCBI, BioProject PRJNA393298, with the DDBJ/ENA/GenBank accession numbers NPKT00000000, NPKR00000000, NPKP00000000, and NPKO00000000, respectively.

#### AUTHOR CONTRIBUTIONS

AB-D carried out the experiments and wrote the manuscript. JL and ASa helped for analysis. LA was responsible of genome assemblage and statistics. AB-D, ASe, and SC designed the experiments. All authors gave intellectual input and critically revised the manuscript.

#### REFERENCES


#### FUNDING

This project has received funding from the European Union's Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement No. 676480.

#### ACKNOWLEDGMENTS

We are grateful to Maria E. Constantin for sharing expertise in scoring Fusarium wilt disease, Philipp Franken and Rosanna C. Hennessy for providing S. indica and R. solani, respectively, and Negar Ghezel Sefloo for laboratory support.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.02888/full#supplementary-material


Hodkinson, F. Doohan, M. Saunders, and B. Murphy (Cambridge: Cambridge University Press), 25–51.


**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 © 2019 del Barrio-Duque, Ley, Samad, Antonielli, Sessitsch and Compant. 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.

# Anatomical and Biochemical Changes Induced by Gluconacetobacter diazotrophicus Stand Up for Arabidopsis thaliana Seedlings From Ralstonia solanacearum Infection

*María V. Rodriguez1, Josefina Tano2, Nazarena Ansaldi1, Analía Carrau2, María S. Srebot 1, Virginia Ferreira3, María L. Martínez 1, Adriana A. Cortadi 1, María I. Siri <sup>3</sup> and Elena G. Orellano2,4\**

#### Edited by:

Massimiliano Morelli, Italian National Research Council (IPSP-CNR), Italy

#### Reviewed by:

Kei Hiruma, Nara Institute of Science and Technology (NAIST), Japan Carlos Henrique Meneses, State University of Paraíba, Brazil

> \*Correspondence: Elena G. Orellano orellano@gmail.com

#### Specialty section:

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

Received: 14 July 2019 Accepted: 18 November 2019 Published: 23 December 2019

#### Citation:

Rodriguez MV, Tano J, Ansaldi N, Carrau A, Srebot MS, Ferreira V, Martínez ML, Cortadi AA, Siri MI and Orellano EG (2019) Anatomical and Biochemical Changes Induced by Gluconacetobacter diazotrophicus Stand Up for Arabidopsis thaliana Seedlings From Ralstonia solanacearum Infection. Front. Plant Sci. 10:1618. doi: 10.3389/fpls.2019.01618

1 Área Biología Vegetal (CONICET), Facultad de Ciencias Bioquímicas y Farmacéuticas, Universidad Nacional de Rosario, Rosario, Argentina, 2 Facultad de Ciencias Bioquímicas y Farmacéuticas, Instituto de Biología Molecular y Celular de Rosario (CONICET-UNR), Universidad Nacional de Rosario, Rosario, Argentina, 3 Departamento de Biociencias, Facultad de Química, Universidad de la República, Montevideo, Uruguay, 4 Área Biología Molecular (CONICET), Facultad de Ciencias Bioquímicas y Farmacéuticas, Universidad Nacional de Rosario, Rosario, Argentina

Nowadays, fertilization and pest control are carried out using chemical compounds that contaminate soil and deteriorate human health. Plant growth promoting bacteria endophytes (PGPBEs), are a well-studied group of bacteria that offers benefits to the host plant, such as phytostimulation, biofertilization, and protection against other microorganisms. The study of Gluconacetobacter diazotrophicus–which belongs to PGPBEs-aids the development of alternative strategies of an integrated approach for crop management practices. Ralstonia solanacearum is responsible for bacterial wilt disease. This phytopathogen is of great interest worldwide due to the enormous economic losses it causes. In this study the action of G. diazotrophicus as a growth promoting bacterium in Arabidopsis thaliana seedlings is analyzed, evaluating the antagonistic mechanisms of this beneficial endophytic bacterium during biotic stress produced by R. solanacearum. Effective colonization of G. diazotrophicus was determined through bacterial counting assays, evaluation of anatomical and growth parameters, and pigments quantification. Biocontrol assays were carried out with Ralstonia pseudosolanacearum GMI1000 model strain and R. solanacearum A21 a recently isolated strain. Inoculation of A. thaliana (Col 0) with G. diazotrophicus Pal 5 triggers a set of biochemical and structural changes in roots, stems, and leaves of seedlings. Discrete callose deposits as papillae were observed at specific sites of root hairs, trichomes, and leaf tissue. Upon R. pseudosolanacearum GMI1000 infection, endophyte-treated plants demonstrated being induced for defense through an augmented callose deposition at root hairs and leaves compared with the non-endophyte-treated controls. The endophytic bacterium appears to be able to prime callose response. Roots and stems cross sections showed that integrity of all tissues was preserved in endophyte-treated plants infected with R. solanacearum A21. The mechanisms of resistance elicited by the plant after inoculation with the endophyte would be greater lignification and sclerosis in tissues and reinforcement of the cell wall through the deposition of callose. As a consequence of this priming in plant defense response, viable phytopathogenic bacteria counting were considerably fewer in endophyte-inoculated plants than in not-inoculated controls. Our results indicate that G. diazotrophicus colonizes A. thaliana plants performing a protective role against the phytopathogenic bacterium R. solanacearum promoting the activation of plant defense system.

Keywords: biocontrol, Gluconacetobacter diazotrophicus, induced systemic resistance, plant growth promoting bacteria endophyte, Ralstonia solanacearum

### INTRODUCTION

Global climate change and increase in human population generate large pressure over natural resources, including the demand of land and water resources available for food production. In addition, plant diseases represent a serious threat to agricultural crops (Velivelli et al., 2014). Chemical pesticides used for the control of phytopathogens are currently known for their adverse effects both on the environment and on the health of consumers. This issue has become a matter of growing concern among consumers and generates social pressure for food free of pesticide residues (Gerbore et al., 2014). Legislation limiting the use of certain agrichemicals, the high awareness, and the lack of acceptance by consumers of genetically modified crops, in addition to their strict regulation, leads the development of new sustainable practices for agriculture. The immediate task that stakeholders face is the search for a sustainable crop production system to solve the problems that threaten global food security. An example of such a sustainable crop-production system is the strategy that incorporates beneficial microorganisms to improve plant health.

The beneficial effects of bacterial endophytes on host plants appear to take place through two types of mechanisms: direct growth promotion activity or indirect mechanisms (Velivelli et al., 2014). Within the direct activity of growth promotion there are several ways in which different PGPB directly facilitate the proliferation of their host plants: they can fix atmospheric nitrogen and supply it to plants, synthesize different phytohormones to intensify the growth of the host plant; they also have solubilization mechanisms of minerals such as phosphorus, to improve their availability. A particular endophytic microorganism can affect the growth of the plant and its development through the use of one or more of these mechanisms. Endophytic bacteria reduce or prevent the deleterious effects of phytopathogenic organisms, and this ability can be considered as an indirect promotion of plant growth (Lodewyckx et al., 2002). The direct inhibition of pathogens carried out by endophytic bacteria is commonly mediated by the synthesis of inhibitory allelochemicals such as antibiotics, iron chelating siderophores, antifungal metabolites, and the degradation of signals produced by pathogens (quorum sensing quenching). Indirect biocontrol mechanisms of endophytic bacteria include the induction of systemic resistance in plants that inhibits a broad spectrum of phytopathogens (Liu et al., 2017). Disease elimination by the biocontrol agents is the prolonged manifestation of the interaction among the plant, the phytopathogen, the biocontrol agent, the microbial community surrounding the plant and the environment (Handelsman and Stabb, 1996).

The induced state of systemic resistance (ISR) is characterized by the activation of latent defense mechanisms that are subsequently expressed more rapidly and intensively in response to an infection by a pathogen at a low physiological cost for the plant. The ISR is characterized for being activated after the interaction between beneficial microorganisms and their host plants; this induction is signaled by the ethylene (ET) and jasmonic acid (JA) or salicylic acid (SA) pathways or a combination of both signaling pathways. When the induced resistance is demonstrated to be SA dependent, is referred to as systemic acquired resistance (SAR) (Pieterse et al., 2014). Several studies showed that relatively slight changes occurred in the transcriptome in systemic tissues upon a colonization of the roots by a beneficial microorganism, especially when compared with the massive transcriptional reprogramming that occurs during the attack of pathogens (Verhagen et al., 2004). Because the defense mechanisms remain dormant after interaction with beneficial microorganisms, it is sometimes difficult to recognize them in plants that have not been challenged by an interaction with a pathogen; therefore the combination in the interaction plant-beneficial microorganism-pathogen allows studying and visualizing the ISR changes easily (Pieterse et al., 2014).

*Gluconacetobacter diazotrophicus* is a Gram-negative bacterium, tolerant to acid, obligate aerobic and rod-shaped with rounded ends (0.7–0.9 μm x 1–2 μm) with lateral or peritrichous flagella (Cavalcante and Dobereiner, 1988; Gillis et al., 1989; Muthukumarasamy et al., 2002; Chawla et al., 2014). *Gluconacetobacter* belongs to the *Proteobacteria* phylum, in the α-proteobacteria section, *Rhodospirillales* order and *Acetobacteraceae* family (Kersters et al., 2006). *G. diazotrophicus* (Yamada et al., 1997) formerly named *Acetobacter diazotrophicus*, (Gillis et al., 1989). This bacterium was originally isolated from sugar cane (Cavalcante and Dobereiner, 1988) and has the ability to fix atmospheric nitrogen without forming nodules (Stephan et al., 1991; Alvarez and Martínez-Drets, 1995). Its endophytic nature was confirmed in Brazil by the counting of this bacterium in roots, stems, and aerial parts of sugarcane (Reis et al., 1994). The potentially beneficial effects promoted by this bacterium in plants Rodriguez et al. Structural/Physiological Features Primed by Endophytes

are nitrogen fixation, phytohormones production, inhibition/ suppression of pathogen, and solubilization of mineral nutrients (Fuentes-Ramirez et al., 1993; Fisher and Newton, 2005). In 2009, the genome of the Pal5 strain of *G. diazotrophicus* was completely sequenced and genes involved in nitrogen fixation, sugar metabolism, transport systems, polysaccharide biosynthesis, quorum sensing, and auxin biosynthesis were identified, confirming its importance (Bertalan et al., 2009). The potential use of *G. diazotrophicus* as antagonist against *Colletotrichum falcatum*, the pathogenic fungus of "red rot" in sugarcane was first demonstrated by Muthukumarasamy et al. (2000). When *G. diazotrophicus* and the fungal pathogen were cultured in the same medium, a clear zone of inhibition against the pathogen was visualized. Similarly, its potential as antagonist of *Xanthomonas albilineans*, the organism responsible for leaf scald disease in sugarcane, was demonstrated. *G. diazotrophicus* secretes certain proteins (bacteriocins) that prevent the growth of *X. albilineans* and imparts a lysozyme-like activity to the inner cell wall of the pathogen (Blanco et al., 2005). *G. diazotrophicus* presented antifungal activity against several species of *Fusarium spp*. when they were grown on potato-dextrose agar (PDA) (Logeshwarn et al., 2011). Regarding the induction of resistance in the host plant by *G*. *diazotrophicus*, there are reports that show an increase in a marker of the JA/ET defense pathway in rice plants when these were inoculated with this endophytic bacterium (Filgueiras et al., 2019). The activation of genes involved in the ET signaling pathway in sugarcane plants colonized by *G. diazotrophicus* was also demonstrated (De Nogueira et al., 2001; Cavalcante et al., 2007). In addition, the accumulation of polysaccharides and tannins in the parenchymal cells surrounding the metaxylem of sugarcane plants inoculated with *G. diazotrophicus* was reported, suggesting that the plant's defense system is activated during the interaction with the bacterium (Dong et al., 1997). The inoculation of *G. diazotrophicus* in plants of *A. thaliana*, *NahG*, mutated in the SAR, revealed that this route of signaling related to the immune system of the plant plays an important role during the stages of early association of the endophytic bacterium with the host plant (Rangel de Souza et al., 2015). The activation of these defense systems after *G. diazotrophicus* inoculation indicates their role in the biocontrol of pathogens by priming.

*Ralstonia solanacearum* is a phytopathogenic β-proteobacteria of great importance worldwide due to the enormous economic losses that it causes, since it attacks a wide variety of crops and wild plants (Hayward, 1991; Genin and Denny, 2012; Peeters et al., 2013). *R. solanacearum* attacks more than 200 species of plants belonging to more than 60 different botanical families, affecting not only solanaceous plants such as potato and tomato but also many agricultural crops, shrubs, trees, and weeds. This unusual wide host range expands continuously, and descriptions of new hosts are very common. This bacterium produces the disease known as bacterial wilt, which is characterized by the loss of leaf turgor and general decay of the whole plant, due to the obstruction of the conducting tissues that transport water and nutrients throughout the stem (Gabor and Wiebe, 1997). Although it is generally considered a plant pathogen, *R. solanacearum* mainly behaves like a saprophyte bacterium capable of surviving for long periods of time in various natural habitats, such as superficial water and different types of soil. As a consequence, it is able to use the wide variety of carbon sources and face the toxic compounds present in the soil. The bacterium has a large repertoire of catabolic genes, genes responsible for the detoxification of harmful compounds and adhesion, which allow efficient colonization and permanence in specific ecological niches (Genin and Boucher, 2004).

This work investigates the action of *G. diazotrophicus* as a growth promoting bacterium in *A. thaliana* seedlings, evaluating the antagonistic mechanisms of this beneficial endophytic bacterium during the biotic stress produced by *Ralstonia pseudosolanacearum* GMI1000 and *R. solanacearum* A21. These strains attack agronomic crops of interest worldwide. In addition, in this work we have genotyped the strains of *R. solanacearum* isolated from the Northeastern region of Argentina.

#### MATERIALS AND METHODS

#### Bacterial Strains and Growth Conditions

*G. diazotrophicus* Pal 5 was kindly ceded by Ing. Agr. Paola Delaporte Quintana, who works in the Instituto Superior de Investigaciones Biológicas (INSIBIO, Tucumán, Argentina). *R. solanacearum* A21, phylotype IIA-sequevar 50 strain isolated from tomato in Argentina, was obtained from Culture Collection of the Intituto Nacional de Tecnología Agropecuaria (INTA Bella Vista, Corrientes, Argentina) and typified by Dr. MI Siri and Bioq. V Ferreira (Departamento de Biociencias, Facultad de Química, Universidad de la República, Montevideo, Uruguay). In the present work, the strain *R. pseudosolanacearum* GMI1000, wild reference strain (Salanoubat et al., 2002; Prior et al., 2016; Safni et al., 2014) was also used. *R. pseudosolanacearum* reporter strain that express the green fluorescence protein (GFP) constitutively, GMI1000-GFP was constructed, validated previously, and kindly provided by Dr. Marc Valls (CRAG, Barcelona, España; Monteiro et al., 2012). The reporter system was introduced in a neutral genome region of *R. pseudosolanacearum* GMI1000 (Monteiro et al., 2012).

Cells of *G. diazotrophicus* strain Pal 5 were grown in LGI-P medium (g L−1): sucrose, 100.0; K2HPO4, 0.2; KH2PO4, 0.6; MgSO4·7H2O, 0.2; CaCl2·2H2O, 0.2; Na2MoO4·H2O, 0.002; FeCl3·6H2O, 0.01 and pH adjusted to 5.5 at 30°C and 200 rpm (Cavalcante and Dobereiner, 1988). Cells of *R. solanacearum* strains GMI1000, A21, and GMI1000-GFP were grown in bactoglucose (BG) medium (g L−1): casein peptone, 10.0; yeast extract, 1.0; casamino acids 1.0 at 28°C and 200 rpm (Boucher et al., 1985). Tetracycline (10 μg ml−1) was added to the medium for cultivation of the tetracycline-resistant *R. pseudosolanacearum* strain, GMI1000-GFP.

The bacterial strains and culture media used in this work are indicated in the **Supplementary Table 1**.

#### Molecular Typing of Isolated Ralstonia solanacearum Strains From Argentina

Phylotype affiliation of the *R. solanacearum* strains was performed by multiplex PCR on the internal transcribed spacer region as described by Fegan and Prior (2005). Identification of phylotypes I, II, III, and IV was accomplished with four forward primers: Nmult 21:1F, Nmult 21:2F, Nmult23:AF, and Nmult 22: InF, respectively, and a common reverse primer Nmult 22:RR. Amplified fragments for each phylotype had an expected specific length (I: 144 bp, II:372 bp, III: 91 bp, and IV: 213 bp). The multiplex PCR also included a pair of primers common to all phylotypes (759/760). Amplifications were carried out in a total volume of 25-μl containing 1X PCR buffer, 0.2 mM of each deoxynucleoside triphosphate (dNTP), 1.5 mMMgCl2, 6 pmol of each phylotype-specific primer, 4 pmol of species-specific primers 759/760, 2 U of Taq DNA polymerase (Promega), and 50 ng of DNA template. Amplifications were performed in an automated Corbett thermocycler with an initial denaturation step at 96°C for 5 min; followed by 30 cycles of denaturation at 94°C for 15 s, annealing at 59°C for 30 s, and extension at 72°C for 30 s; with a final extension step at 72°C for 10 min. PCR products were analyzed by electrophoresis through 2% agarose gels and revealed under UV light.

The phylogenetic assignment of *R. solanacearum* strains was also determined based on analysis of the partial nucleotide sequences of the endoglucanase (*egl*) gene. PCR amplification of a 750-bp region of the *egl* gene was performed using the Endo-F and Endo-R primers pair as previously described (Fegan et al., 1998). Reactions were performed in a total in a total volume of 25 μl containing 1× DNA polymerase buffer, 1.5 mM MgCl2, 0.2 mM of each dNTP, 10 pmol of each primer, 1 U of Taq DNA polymerase (Promega), and 50 ng of DNA template. Amplification cycling conditions included an initial denaturation step at 96°C for 9 min; followed by 30 cycles of denaturation at 95°C for 1 min, annealing at 70°C for 1 min, and extension at 72°C for 2 min; with a final extension step at 72°C for 10 min. PCR products were purified and sequenced by Macrogen Services (Kumchun-ku, Seoul, Korea) using Endo-F and Endo-R primers. Forward and reverse chromatograms were edited using the Geneious v.7 software package. The determination of sequevars was assumed by *egl* sequence divergence values less than or equal to 1% (Fegan and Prior, 2005). Approximately maximumlikelihood phylogenetic trees were built including also sequences from worldwide reference strains representing the whole diversity of the *R. solanacearum* species complex. Phylogenetic and molecular evolutionary analyses were conducted in MEGA version X (Kumar et al., 2018).

#### Plant Material, Growth Conditions, and Inoculation With Gluconacetobacter diazotrophicus

Seeds of wild-type *A. thaliana* Columbia-0 (Col-0) and *A. thaliana sid2* mutants were germinated on small plastic pots containing 33 g of soil: pearlite (90:10). Cultivation of plants occurred in a growth chamber at 22/24°C, 60% relative humidity with a photoperiod of 16 h light/8 h darkness for 14 days. For plant inoculation, *G. diazotrophicus* Pal5 cells were grown overnight (16 h) at 28°C in constant agitation at 200 rpm in liquid LGI-P medium. Cells were centrifuged at 12,000 ×g and the supernatant was discarded. Cells were washed twice in ultrapure sterile water and suspended in sterile distilled water to a final concentration of 106 CFU/g of soil. Plants were inoculated by soil drenching with the adjusted suspension of *G. diazotrophicus* Pal 5 which was poured onto the rooted soil of a plant. Sterile water was used as negative control (Morel et al., 2017). Three independent growth chamber assays were performed using fourteen plant replicates of each treatment arranged in a complete randomized design. A total of 42 plants of each treatment were used to compare growth parameters. The pigment determination and counting colony-forming unit (CFU) assays were realized on three to six seedling replicates randomly selected of the three independent experiments.

#### Colonization of Arabidopsis thaliana by Gluconacetobacter diazotrophicus

Colonization analyses were conducted at 28-day post-inoculation (dpi), using defined portions of roots and shoot soft six seedling replicates randomly selected of the three independent experiments. Each plant was carefully removed from the plastic cup and washed with sterile distilled water to take away the traces of soil. For tissue surface sterilization, the root and stem of each plant were immersed 70% v/v ethanol for 3 min under a laminar flow chamber. Then they were rinsed three times with sterile water and placed on a sterile absorbent paper. Extract of each organ was made under sterile conditions, cutting from each plant, 2 and 6 cm of stem and root, respectively, and were placed in a sterile Eppendorf tube containing 500 μl of NaCl 0.9% w/v. Then the tissue was mortared to obtain the corresponding extract. Twenty microliters of the extract and serial dilutions were plated in petri dishes with LGI-P agar 1.8% w/v supplemented with bromothymol blue (Cavalcante and Dobereiner, 1988; Reis et al., 1994; Baldani et al., 2014; Ferreira et al., 2017). The same samples were plated in petri dishes with LGI-P agar 1.8% w/v supplemented with antibiotics (chloramphenicol 20.0 mg/L, cycloheximide 150.0 mg/L). The plates were placed in an oven at 30°C for 7 days. After this time, colony forming units per g of plant tissues was determined (CFU/g).

To evaluate the effect of the colonization of the plants, the following growth parameters were measured: fresh weight of the different organs; length of the main root; length of the stem; number of rosette leaves; number of stem leaves; size of rosette leaves; size of stem basal leaves; size of leaves along the stem (Poupin et al., 2013; Poupin et al., 2016).

#### Biocontrol In Vivo Assays of Gluconacetobacter diazotrophicus Against Ralstonia pseudosolanacearum GMI1000 and Ralstonia solanacearum A21

Inoculation of *Arabidopsis thaliana* Plants Containing *Gluconacetobacter diazotrophicus* With Phytopathogenic Strains of *Ralstonia solanacearum* Inoculation with the phytopathogenic strains of the *R. solanacearum* species complex (GMI1000 or A21) was carried out in *A. thaliana* Col 0 and *sid2* mutant plants previously inoculated with the endophytic bacteria and in mock-inoculated plants. The inoculation was done by soil drenching, adding 1 mL of the bacterial suspension onto the rooted soil of plants. Previously, the roots were damaged with a tip to promote bacterial invasion of roots.

For plant inoculation, cells of the corresponding strains of *R. solanacearum* were grown overnight (16 h) at 28°C and under constant agitation at 200 rpm, in liquid BG medium. Bacterial cultures were diluted with sterile distilled water until a final concentration of 106 CFU/g of soil was obtained. Gentamicin (10 μg ml−1) was added to the medium for cultivation of the gentamicinresistant *R. pseudosolanacearum* strain, GMI1000-GFP.

A total of 144 plants in three independent experiments were used with a completely randomized design with six replicates per treatment. Three different phytopathogen strains were used. The following four treatments were applied: mock inoculated plants (*Gd*<sup>−</sup> *Rso* strain−): plants without bacteria; endophyte inoculated plants (*Gd*<sup>+</sup>*Rso* strain−): plants inoculated only with *G. diazotrophicus*; pathogenic bacterium inoculated plants (*Gd*<sup>−</sup>*Rso* strain+): plants inoculated only with *R. solanacearum*; endophyte and pathogenic bacterium inoculated plants (*Gd*<sup>+</sup>*Rso* strain+): plants inoculated with both bacteria. *Rso* strain: GMI1000/ GMI1000-GFP/A21.

#### Colony-Forming Unit Counting Assays

Determination of microbial population in plant organs was performed at 28 days post-inoculation with *G. diazotrophicus* and 12 dpi with the strains GMI1000 and A21 of *R. solanacearum*, using defined portions of roots and shoots. For tissue surface sterilization of plants and determination of *G. diazotrophicus* population, the procedure described in *Colonization of Arabidopsis thaliana by Gluconacetobacter diazotrophicus* was followed.

Determination of CFU of different strains of *R. solanacearum* was carried out by plating 20 μl of the extract and serial dilutions in petri dishes with modified sorbitol MacConkey agar (mSMSA) agar 1.8% w/v supplemented with 0.05 g/L of 2,3,5-triphenyltetrazolium chloride, TTC (French et al., 1995) containing the following (g L−1): casamino acids, 1.0; peptone 10.0; glucose 5.0. The following antibiotics were added to medium:5 mg/L chloramphenicol, 100 mg/L cycloheximide, 0.5 mg/L penicillin-G, 5 mg/L crystal violet, and 2,3,5-triphenyl tetrazolium chloride. The plates were placed in an oven at 28°C for 7 days. After this time, CFU/g was determined. Six seedling replicates of each treatment of three independent experiments were used (Engelbrecht, 1994 modificate by Elphinstone et al., 1996).

#### Anatomical Studies to Determine Structural and/or Physiological Changes Inclusion in Paraffin and Differential Staining

Different treatments samples from tissue *A. thaliana* Col 0 and *sid2* mutant plants were taken at 14, 20, and 28 dpi with *G. diazotrophicus* and 12 dpi with *R. solanacearum* strains and were fixed in FAA solution (50% ethanol, 5% glacial acetic acid, 30% formaldehyde, 15% water). Different organs of each treatment were separated and dehydrated with ethanol and ethanol/xylene of ascending concentration solutions. Then they were embedded in paraffin and roots, stems, and leaf blades were cut at 10 μm thickness with Minot microtome and stained with safraninfast green. Samples were mounted in Canada balsam natural (Biopack). For polychromatic dye with toluidine blue solution (0.05% w/v), cuts manually obtained were incubated in the dye during 5 min, washed with distilled water, and mounted with glycerol-water (50% v/v) (D´Ambrogio, 1986). Observations were made with a light microscope (Zeiss MC 80 Axiolab) equipped with a camera.

#### Confocal Microscopy

Cross and longitudinal sections of fresh root and stem samples (from plants of four treatments) obtained with the technique of freehand cutting were mounted on a slide, surrounded with solid Vaseline and covered with agarose (1% w/v) used as a mean of immersion and fixation. Samples were observed using a confocal microscope (Nikon C1SiR attached on a Nikon TE2000 inverted microscope). Three independent experiments were performed.

#### Callose Detection

The detection of callose deposits was carried out in *A. thaliana* Col 0 and *sid2* mutant plants of the four treatments described above with *R. solanacearum* GMI1000 strain. Three replicates of plants were randomly selected. Three independent experiments were performed. Roots and leaves of these plants were kept overnight in alcohol 96°; once the organs were completely decolorated, they were incubated in sodium phosphate buffer (0.07 M, pH 9) for 30 min and then in an aniline blue solution (0.05% w/v) for 60 min (Daurelio et al., 2009). Finally, the samples were mounted in a glycerol-water mixture (50% v/v) and observed immediately using an Epi-fluorescence-UV microscope (MIKOBA F320 with mercury lamp power box).

### Pigment Determination

To evaluate potential changes of pigments in *A. thaliana* plants, the content of chlorophyll *a* (Chl a), chlorophyll *b* (Chl b) and total chlorophyll (Chl a + b) was determined at 28 dpi with *G. diazotrophicus* and 12 dpi with *R. pseudosolanacearum* GMI1000 or *R. solanacearum* A21. *A. thaliana* seedlings were treated with *G. diazotrophicus* and/or *R. pseudosolanacearum* GMI1000, *G. diazotrophicus*, and/or *R. solanacearum* A21 12 dpi and mock inoculated plants. For this quantification, two leaves per plant of three plant replicates from each treatment were used. The experiments were performed independently by triplicate. For the quantification of chlorophylls, two leaf discs of 0.8 cm diameter were cut and incubated in 1 ml of N,N-dimethylformamide (DMF) for 72 hs at room temperature and in the dark. The absorbance of the samples was then measured at 664 and 647 nm. The content of total chlorophyll, chlorophyll a, and chlorophyll b were determined according to the equations described by Porra (2002).

### Data Analysis

For comparison between two treatments, Shapiro-Wilk normality test was performed, and Student's t or Mann Whitney test was applied. For comparison of a larger number of treatments an analysis of variance (ANOVA) and multiple comparisons test were performed. It was considered statistically significant for *p* < 0.05. The data obtained were processed with the InfoStat program (2014). Chloroplast length values were obtained from 10 fields/leaf cross sections; using 5 different sections. Chloroplast length values were obtained using the ImageJ processing program.

### RESULTS

#### Gluconacetobacter diazotrophicus Induces Anatomical Changes in Arabidopsis thaliana Seedlings

Endophytic bacterial population in roots and stems was analyzed at 28 dpi to corroborate the presence of *G. diazotrophicus* in *A. thaliana* seedlings. CFU counting assays revealed that the endophytic population within the root and stem of *A. thaliana* was (2.57 ± 0.21) x105 CFU/g and (2.60 ± 3.66) x105 CFU/g, respectively. The colonization of endophytic bacteria was tested in roots of *A. thaliana* Col 0 using a magnifying glass. The observation of inoculated seedlings showed some colonies associated with the root, mainly in the places where the lateral roots emerge; in the vicinity of the radical apex and in the radical hairs (**Supplementary Figures 4C, D**). The roots of the mock inoculated plants did not show associated bacterial colonies (**Supplementary Figures 4 A, B**). Radical hairs with associated colonies showed no visible damage (**Supplementary Figure 4D**).

Despite of not having exomorphological changes between 28 dpi inoculated plants and mock inoculated controls (**Figure 1** and **Supplementary Figure 1**), endomorphological analysis showed that *G. diazotrophicus* colonizes and promotes significant anatomical modifications in *A. thaliana* Col 0 plants inoculated with 106 CFU/g of soil. These were observed at 28 dpi in root (**Figures 2O–Q**) and stem (**Figures 2G, H**) and only in root at 14 dpi (**Figures 2M, N**). Roots inoculated with *G. diazotrophicus* 14 and 28 dpi, presented an increase in the diameter and a greater lignification of the xylem vessels. Epidermis, exodermis, endodermis, and pericycle of the main root sclerosis was also part of the observed structural features in plants inoculated with *G. diazotrophicus* after 28 days (**Figures 2P, Q**). Stems (**Figures 2G, H**) showed an increase in xylem tissue and in the amount of sclerosed cortical parenchyma tissue between the vascular bundles with respect to the mock inoculated controls (**Figures 2C, D**). Greater lignification of the xylem was also observed in inoculated plants. There were not significant structural changes between the leaves of inoculated 28 dpi and control plants (**Figures 3A, B**). Nevertheless, changes were detected in the chloroplasts between treatments. Although there was no increase in quantity, a greater size of chloroplasts was observed in the inoculated plants 28 dpi (6.56 ± 0.83 μm) with respect to the mock inoculated controls (5.16 ± 0.71 μm) (**Figure 3C**). Values of chlorophyll *a*, chlorophyll *b*, and chlorophyll *a + b* in plants inoculated with *G. diazotrophicus* at 28 dpi revealed significantly higher amount of pigments in plants inoculated with the endophytic bacteria than the control plants (**Figure 4**). We also tested the lignin presence in *A. thaliana sid2* mutants plants and the presence of lignin thickening caused by inoculation of endophytic bacteria was not observed in any tissue (stems and root) of these mutant plants (**Supplementary Figure 7**)*. A. thaliana sid2* mutant accumulate much less SA in comparison to wild-type plants since this plant mutant is deficient in the induction of SA accumulation having blockage the SA biosynthesis.

#### Gluconacetobacter diazotrophicus Activates Plant Defense Protecting Arabidopsis thaliana Seedlings From the Invasion of Ralstonia pseudosolanacearum GMI1000

To analyze the possible antagonistic effects of the *G. diazotrophicus* Pal5 strain in the presence of the *R. pseudosolanacearum* GMI1000

FIGURE 2 | Micrographs of cross sections of Arabidopsis thaliana plants inoculated with Gluconacetobacter diazotrophicus at different dpi stained with safraninfast green. (A–H) Stem; (I–Q) Root. (A–D, I–L) Mock inoculated plants; (E–H, M–Q) inoculated plants. (B–D), Stem vascular bundle detail of 14 and 28 dpi mock inoculated plants (A–C) showing less lignification and amount of xylematic tissue than in inoculated plants at 28 dpi. (F–H), Stem vascular bundle detail of 14 and 28 dpi inoculated plants (E–G). (H) shows a greater lignification and amount of xylematic tissue than in mock inoculated plants. (J), Root vascular cylinder detail of mock inoculated plants 14 dpi (I) showing smaller diameter and lignification of the xylematic vessels. (N), Root vascular cylinder detail of inoculated plants 14 dpi (M) showing grater diameter and lignification of the xylematic vessels. (L), Root vascular cylinder detail of mock inoculated plants 28 dpi (K) showing low lignification of the xylematic vessels. (O), Root vascular cylinder of inoculated plants 28 dpi showing greater lignification of the xylematic vessels. (Q), Root epidermis detail of inoculated plants 28 dpi (P) showing sclerosis of tissue. cp, cortical parenchyma; e, epidermis; p, pith; scp, sclerosed cortical parenchyma; vb, vascular bundle; vc, vascular cylinder; xv, xylematic vessels. Asterisk indicated more xylematic vessels lignification. Each image is a representative result of observation of at least 10 sections from five biological replicates.

strain, experiments were performed on *A. thaliana* plants during biotic stress produced by the phytopathogenic bacterium.

At 6 and 9 dpi, *A. thaliana* seedlings inoculated only with *R. pseudosolanacearum* GMI1000 (**Figures 5 G, H**) did not show exomorphological differences regarding to those plants inoculated with *G. diazotrophicus* and *R. pseudosolanacearum* GMI1000 (**Figures 5J, K**), inoculated only with *G. diazotrophicus* (**Figures 5D, E**), or mock inoculated controls (**Figures 5A, B**). At 12 dpi, exomorphological changes arose in those plants only inoculated with *R. pseudosolanacearum* GMI1000. The leaves of the rosette were chlorotic and dehydrated (**Figure 5I**).

CFU counting assays were performed to evaluate bacterial number (CFU/g) of *G. diazotrophicus* (28 dpi) and *R. pseudosolanacearum* GMI1000 (12 dpi) of respective *A. thaliana* plants treatments. The results are indicated in **Table 1**. In roots of *A. thaliana* plants *R. pseudosolanacearum* GMI1000 population decreased from (9.40 ± 0.50) x1010 CFU/g in the absence of *G. diazotrophicus* to (8.23 ± 1.90) x106 CFU/g, when is confronted to the endophyte. Stem extracts and dilutions of *A. thaliana* plants with *G. diazotrophicus* did not show *R. pseudosolanacearum* GMI1000 growth in the selective medium mSMSA (Elphinstone et al., 1996).

Although exomorphological alterations were not observed in inoculated seedlings at 6 and 9 dpi, stem longitudinal sections stained with toluidine blue (1% w/v) showed the presence of *R. pseudosolanacearum* GMI1000 along the stem, but only in those plants that were not previously inoculated with *G. diazotrophicus* (**Supplementary Figures 2A–H**). In plants previously inoculated with *G. diazotrophicus*, the presence of phytopathogenic bacteria was not observed (**Supplementary Figures 2I–L**).

using five different sections. The error bars represent the standard deviation. Significant differences between treatments are indicated by an asterisk

To corroborate the previous results and to confirm the presence or absence of the phytopathogenic bacterium, *A. thaliana* seedlings were inoculated with an *R. pseudosolanacearum* GMI1000*-*GFP strain. Stems and roots of plants previously inoculated with *G. diazotrophicus* and plants without *G. diazotrophicus* were cut in cross and longitudinal sections and confocal microscopy technique was applied. This experiment allowed locating the bacterium within the plant tissue. The results observed agree

treatments are indicated by an asterisk (Student's t test, p < 0.05).

with those previously obtained. Phytopathogenic bacterium colonization in seedlings *Gd*<sup>−</sup>*RsoGMI1000-*GFP+ was observed in the xylem tracheary elements as well as the parenchyma cells in the cortex at 12 dpi (**Figures 6A, B**). The presence of *R. pseudosolanacearum* GMI1000*-*GFP was not manifest in plant stems previously inoculated with *G. diazotrophicus* (**Figures 6C, D**). These results show that bacterial colonization is restricted in stems of seedlings that previously were inoculated with *G. diazotrophicus*. *R. pseudosolanacearum* GMI1000*-*GFP was observed in root cross section in both treatments showing correspondence with CFU counting assays (**Figure 7**, **Table 1**). Strikingly, in those plants without *G. diazotrophicus* (*Gd*<sup>−</sup> *RsoGMI1000-GFP*+), bacteria were localized with a broader colonization in cortical zone proximal to the vascular cylinder and several of xylem vessels were filled (**Figures 7A, B**). Contrary to this, in seedlings *Gd*<sup>+</sup>*RsoGMI1000-GFP*+, the *R. pseudosolanacearum* GMI1000*-*GFP distribution seems to be different. Bacteria seem to be surrounding the xylematic vessels probably in the xylem parenchyma cells. No large colonization of *R. pseudosolanacearum* GMI1000*-*GFP was observed in the cortical zone of the root. Thus, bacterial invasion of the vascular cylinder appears restricted and larger metaxylem elements are not significantly colonized (**Figures 7C, D**).

Effects of *G. diazotrophicus* on the cellular defense response mediated by callose deposition were investigated. Callose was localized using aniline blue solution leading to yellow fluorescence. Papillae are appositions that reinforce cell wall at sites of interaction with pathogenic microorganism. These papillae structures were observed at discrete sites of root hairs in plants inoculated with *G. diazotrophicus* or *R. Pseudosolanacearum* GMI1000, indicating that in both treatments callose deposition occurs (**Figures 8D–I**). The callose deposits were clearly increased in those plants infected with *R. pseudosolanacearum*

(Student's t test, p < 0.05).

pseudosolanacearum GMI1000. Three independent growth chamber assays were performed using six replicate plants of each treatment. (A–C) Mock inoculated plants; (D–F) plants inoculated with G. diazotrophicus Pal5; (G–I) plants inoculated with R. pseudosolanacearum GMI1000; (J–L) plants inoculated with G. diazotrophicus and with R. pseudosolanacearum GMI1000.

TABLE 1 | Counting values of Gluconacetobacter diazotrophicus and Ralstonia solanacearum in roots and stems of Arabidopsis thaliana plants.


n.d., not detected.

Data expressed as mean averages of six seedling replicates of three independent experiments ± standard error. Shapiro-Wilk normality test was performed, and Student's t or Mann Whitney test was applied. Different letters indicate significant differences in counts (p<0.05).

GMI1000 which had previously been inoculated with *G. diazotrophicus* (**Figures 8J–L**). Similarly, newly callose deposits were accumulated within leaf tissues and trichomes in those seedlings previously inoculated with *G. diazotrophicus* and subsequently with *R. pseudosolanacearum GMI1000* (**Figure 9**). This assay was also performed using *A. thaliana sid2* mutants. No callose deposition was observed in plants inoculated with *G. diazotrophicus* or *R. Pseudosolanacearum* GMI1000 or both bacteria in trichomes or portions of the epidermis of these seedlings (**Supplementary Figures 5B–K**). No deposits of callose were observed in the papillae of the radical hairs of any of the treatments tested in the roots of the *A. thaliana sid2* mutant seedling (**Supplementary Figures 6B–K**).

There were significant statistical differences in the amount of pigments between those plants only exposed to *R. pseudosolanacearum* GMI1000 and those pretreated with *G. diazotrophicus*. No variation in the plant pigments between *Gd*<sup>+</sup>*RsoGMI1000+* and *Gd*<sup>+</sup>*RsoGMI1000−* treatments was

observed. Pigments concentrations were higher in plants treated with *Gd*<sup>+</sup>*RsoGMI1000+* and *Gd*<sup>+</sup>*RsoGMI1000−* than in control (*Gd*<sup>−</sup>*RsoGMI1000*−) for both treatments. Chlorophyll *a*, chlorophyll *b*, and chlorophyll *a + b* concentrations decreased in those plants only inoculated with *R. pseudosolanacearum GMI1000* at 12 dpi (*Gd*<sup>−</sup>*RsoGMI1000*+) (**Figure 10**).

#### Gluconacetobacter diazotrophicus Successfully Controls the Infection of Arabidopsis thaliana Seedlings Inoculated With an Argentine Isolation of the Bacterium Ralstonia solanacearum A21

Three strains were isolated from infected tomato (A11, A21) and pepper (A31) plants from the Argentina north eastern region. These strains were characterized as *R. solanacearum* using microbiological and *in planta* tests. The genomic DNA from the three *R. solanacearum* strains was obtained and using for the genotypic identification. Phylotype-specific multiplex PCR analyses amplified the expected 280 and 372 bp fragments, indicating that all three isolates belonged to the *R. solanacearum* phylotype II. Based on phylogenetic analysis of partial *egl* sequences strains from Argentina were assigned to the phylotype IIA sequevar 50 (strains A11 and A21) and sequevar 38 (strain A31) (**Supplementary Figure 3**). Both sequevars are associated to strains isolated from solanaceous crops in South America. Since all the Argentinean strains presented a similar tomato symptom, we decided to work with the one of them, named *R. solanacearum* A21.

At 12 dpi, exomorphological changes began to be observed in plants inoculated only with *R. solanacearum* A21. The leaves of the rosette were chlorotic and dehydrated (**Figure 11I**). Stem and root cross sections showed significant anatomical changes at 12 dpi between the plants that were previously inoculated with *G. diazotrophicus* and those that did not possess the endophyte bacterium. Similarly, an increase in the amount of xylematic tissue, with vessels that present greater lignification and greater amount of sclerosed cortical parenchyma tissue between the vascular bundles were observed in those plants with *G. diazotrophicus* and *R. solanacearum* A21 (**Figures 12D–F**) as happened with

the pathogenic strain *R. pseudosolanacearum* GMI1000 (data not shown). This was already observed to a lesser extent in plants only inoculated with *G. diazotrophicus* at 28 dpi (**Figures 2G, H**). These structural differences result in colonization resistance of the stems to the pathogenic bacteria. Seedlings treated with *Gd*<sup>+</sup>*RsoA21*+ at 12 dpi presented integrity in all the tissues of the stem (**Figure 12D**) while seedlings treated with *Gd*<sup>−</sup>*RsoA21*+, showed breakage of the cortical parenchyma and areas of the vascular bundles (**Figures 12A–C**). *R. solanacearum* A21 provokes plasmolysis of epidermal, cortical, and endodermal root cells in those seedlings without *G. diazotrophicus* (**Figures 13A, B**). Plasmolysis zones appear in alignment with one of the xylem pole axes. Integrity of all root tissues was observed in plants previously inoculated with *G. diazotrophicus* (**Figure 13C**), in addition, an increase in lignification of xylematic vessels was observed (**Figures 13D, E**).

CFU counting assays were performed to evaluate bacterial number (CFU/g) of *G. diazotrophicus* (28 dpi) and *R. solanacearum* A21 (12 dpi) in *A. thaliana* plants under each treatment. Results are indicated in **Table 1**. In roots of *A. thaliana* plants with *G. diazotrophicus*, *R. solanacearum* A21 counting decreased from (1.40 ± 0.68) x107 CFU/g to (1.92 ± 2.31) x105 CFU/g. *R. solanacearum* A21 was not detected in stems that had

the endophytic bacteria. In stems of *A. thaliana* plants without *G. diazotrophicus,* CFU/g of *R. solanacearum* A21 was (3.79 ± 4.82) x105 . These results indicate that the bacterial population of *R. solanacearum* A21 decreases in the presence of *G. diazotrophicus*, and therefore this microorganism exerts some mechanism of biocontrol on *G. diazotrophicus*.

Chlorophyll *a*, chlorophyll *b*, and chlorophyll *a + b* concentrations increased in those plants that were previously inoculated with *G. diazotrophicus* (Gd+*RsoA21*− or Gd+*RsoA21*+). Plants only inoculated with *R. solanacearum* A21 presented lower concentrations of pigments. This treatment (Gd−*RsoA21*+) did not show significant statistical differences with the mock inoculated plants (**Figure 14**). The argentine *R. solanacearum* strain presented a deferential phenotype compared to GMI 1000 strain.

### DISCUSSION

Plants live in complex environments where they interact closely with a wide range of microorganisms (Verhage et al., 2010). *G. diazotrophicus* is an endophytic bacterium able to colonize many plant species, both monocotyledons and dicotyledons. This

FIGURE 8 | Observation of callose deposits with epifluorescence microscope in roots of Arabidopsis thaliana. In each column the treatment is specified. The white arrows indicate root hairs. The black arrows indicate root hair papillae. The red arrows indicate zones and cells of the root surface with callose. The experiments were performed at 12 dpi with pathogenic bacterium. (A–C) Mock inoculated plants; (D–F) plants inoculated with Gluconacetobacter diazotrophicus Pal5; (G–I) plants inoculated with Ralstonia pseudosolanacearum GMI1000; (J–L) plants inoculated with G. diazotrophicus and with R. pseudosolanacearum GMI1000. Each image is a representative result of observation of at least 10 sections from three biological replicates.

FIGURE 9 | Observation of callose with epifluorescence microscope in leaves of Arabidopsis thaliana. In each column the treatment is specified. The black arrows indicate trichomes. The experiments were performed at 12 dpi with pathogenic bacterium. (A–B) Mock inoculated plants; (C–D) plants inoculated with Gluconacetobacter diazotrophicus Pal5; (E–F) plants inoculated with Ralstonia pseudosolanacearum GMI1000; (G–H) plants inoculated with G. diazotrophicus and with R. pseudosolanacearum GMI1000. Each image is a representative result of observation of at least 10 sections from three biological replicates.

FIGURE 10 | Bar graph showing the concentration of chlorophyll a, chlorophyll b, and chlorophyll a + b in leaves of Arabidopsis thaliana inoculated 28 dpi with Gluconacetobacter diazotrophicus (Gd+Rso−), inoculated 12 dpi with Ralstonia pseudosolanacearum GMI1000 (Gd−RsoGMI1000+), with both bacteria (Gd+ RsoGMI1000+) or mock inoculated plants (Gd−Rso−). Concentration of pigments values are means of two leaves per plant of three plant replicates from each treatment. The experiments were performed independently by triplicate. The error bars represent the standard deviation. Significant differences between treatments are represented by different letters (ANOVA, p < 0.05).

Gluconacetobacter diazotrophicus Pal 5 and subsequently inoculated with R. solanacearum A21. (E, F) are panel (D) detail of vascular bundles and areas of the stem cortical parenchyma. Each image is a representative result of observation of at least 10 sections from five biological replicates.

bacterium can promote plant growth through different mechanisms that include the biological fixation of nitrogen, the secretion of phytohormones, the solubilization of mineral nutrients, and antagonism toward phytopathogens (Eskin et al., 2014).

During the last decades, several studies have tried to identify genes and regulatory proteins involved in the plant-*G diazotrophicus* association (dos Santos et al., 2010; Lery et al., 2011; Galisa et al., 2012; Bertini et al., 2014). However, many of the crops of agronomic interest are slow-growing species with a complex genome, and often their large size hinder their growth in greenhouses under controlled conditions, which limits the detailed molecular characterization of the plant-bacteria interaction. *A. thaliana* has a short life cycle (6 weeks) but has the typical characteristics of the other angiosperms in terms of morphology, anatomy, growth, development, and responses to the environment. Therefore, results can be obtained in a shorter time. *A. thaliana* is susceptible to only a limited number of pathogens, including viruses, bacteria, fungi, nematodes, and insects, and responds to pathogen attack similarly to species of higher plants (Andargie and Li, 2016). As a result, *A. thaliana* seedlings can be easily manipulated to study plant-endophyte interactions.

FIGURE 13 | Arabidopsis thaliana plants inoculated 12 dpi with Ralstonia solanacearum A21. Root cross sections dyed with safranin-fast green. (A) Plants inoculated with R. solanacearum A21; (B) is panel (A) detail showing the epidermis, cortical zone, endodermis, and part of the vascular cylinder of the root. The black arrows indicate areas of damaged tissue. (C) Plants inoculated with Gluconacetobacter diazotrophicus Pal 5 and subsequently inoculated with R. solanacearum A21. (D, E) are panel (C) detail of xylem vessels of the vascular cylinder. The red asterisks show the highest lignification of the xylem vessels compared to panel (B). c, cortical zone; en, endodermis; ep, epidermis; x, xylem vessel. Each image is a representative result of observation of at least 10 sections from five biological replicates.

*R. solanacearum* is the bacterium responsible for the bacterial wilt of tomato and brown rot of potato (Agrios, 2005; Hayward, 1991). This bacterium is of great importance worldwide due to the enormous economic losses it causes, given that it infects a wide variety of crops and wild plants (Jones et al., 1991). The present study aimed to obtain *A. thaliana* seedlings with *G. diazotrophicus* and evaluate the protective potential of the endophytic bacterium against two phytophatogenic strains; *R. pseudosolanacearum* GMI1000 and *R. solanacearum* A21.

Our results show that the endophytic population of *G. diazotrophicus* was (2.57 ± 0.21) x105 CFU/g in roots of *A. thaliana* seedlings, while for the stem the value was (2.60 ± 3.66) x105 CFU/g at 28 dpi. The endophytic bacterial population of *G. diazotrophicus* in roots and leaves of *A. thaliana* was previously analyzed by Rangel de Souza et al. (2015) at different times after inoculation. They observed 1.5x106 , 3.1x106 , and 2.1x105 CFU/g of root at 14, 28, and 50 dpi, respectively. They did not detect bacteria in the leaf at these times post-inoculation. The endophytic nature of *G. diazotrophicus* was confirmed in Brazil by counting this bacterium in roots, stems, and aerial parts of sugarcane (Reis et al., 1994). Values in all parts of the plant were between 103 and 106 CFU/g fresh weight. High values (106 –107 CFU/g of fresh weight) were also found in sugarcane plants in India (Muthukumarasamy et al., 1999). There are other studies

on sorghum, wheat, and tomato species where the bacterial population of *G. diazotrophicus* remains highest during the first days of infection (around seven or more) and then decreases and remains constant (Luna et al., 2012). Faced with these observations, Rangel de Souza et al. (2015) postulated that as the plant-endophyte interaction progresses, the bacterial population reduces, probably as a consequence of the plant's defense system.

Rangel de Souza et al. (2015) reported inhibition in the growth of *A. thaliana* col-0 seedlings inoculated with *G. diazotrophicus* at 28 dpi. Unlike this, our observations did not register significant statistical differences in the growth parameters analyzed between the plants inoculated with *G. diazotrophicus* and the mock inoculated plants at 28 dpi (**Supplementary Figure 1**). Instead, an increase in xylematic tissue and a greater amount of sclerosed tissue were observed in the stems and roots of the inoculated plants compared to the mock-inoculated plants. Greater lignification was also observed in the xylem vessels of the inoculated plants (**Figure 2**). These anatomical changes induced by *G. diazotrophicus* in *A. thaliana* are part of the defense response of the plant primed by this bacterium. In addition, in this study a greater chloroplasts size was observed in the plants inoculated with *G. diazotrophicus* regarding to the mock inoculated plants at 28 dpi. Likewise, a significant difference was observed in the content of chlorophyll *a*, chlorophyll *b*, and chlorophyll *a+b* between plants inoculated with *G. diazotrophicus* and mock inoculated plants. Inoculated seedlings had a higher content of the pigments (**Figure 4**).

If a satisfactory symbiotic relationship is established with the endophyte, the biological fixation of nitrogen by the microorganism can supply a considerable part of the requirement of this nutrient. Nitrogen is essential for the synthesis of the Rubisco enzyme and for the synthesis of the light-harvesting complex which is strongly associated with chlorophyll molecules. About 70% of the nitrogen in the leaves exists in the chloroplasts and is mostly used to synthesize the photosynthetic machinery (Zhou et al., 2006). The biological fixation of nitrogen could stimulate the rate of photosynthesis through the increase of Rubisco activity and the speed of the photosynthetic electron transport chains (Harley et al., 1992). It is known that the microorganisms associated with plants stimulate photosynthesis because they use photosynthates as a carbon source for their growth, diverting them from their real destiny in the plant, which is why they are forced to increase the rate of photosynthesis to supply their requirements (Kaschuk et al., 2009). As occurs in soybean plants (Abu-Shakra et al., 1978), it could happen that the synergistic effect of the prolonged acquisition of nitrogen, and the stimulation of photosynthesis by these microorganisms, postpone the degradation of the proteins present in the leaves and also of the chlorophyll. This could explain the observed increase of photosynthetic pigments in plants inoculated with *G. diazotrophicus* without an increase in the size of the plants at 28 dpi.

It is known that most phytopathogenic microorganisms have biological antagonists that can be used for biological control. In recent years, the use of antagonistic bacteria and fungi in agricultural diseases treatment has become more relevant (Robles Carrión, 2012). In this work the behavior of the phytopathogenic strains of *R. solanacearum* and the endophyte bacteria *G. diazotrophicus* in the plant was analyzed in order to gain knowledge about benefits of *G. diazotrophicus* against the pathogen causing disease in solanaceous crops. For this, plants that were previously inoculated with *G. diazotrophicus* and in which the endophytic bacteria had already been established, were subsequently infected with the strains GMI1000 or A21 of *R. solanacearum*. Symptoms of chlorosis and dehydration began to be observed at 12 dpi in plants treated with *R. solanacearum* (GMI1000 or A21) (**Figures 5I** and **11I**). Symptoms were more pronounced when the strain used was GMI1000. Chlorophyll content of plants treated with *R. pseudosolanacearum* GMI1000 was the lowest regarding the other treatments (**Figure 10**); whereas for plants treated with *R. solanacearum* A21 it did not show statistical differences with the chlorophyll content of mock inoculated plants (**Figure 14**). These results would indicate that the A21 strain of *R. solanacearum* presents a different phenotype from that of the GMI1000 strain and therefore the responses of the *A. thaliana* plants to the infection is also different. Exomorphological signs observed in *A. thaliana* seedlings treated with GMI1000 or A21 strains of *R. solanacearum* were compatible with the symptoms caused by this phytopathogen in the *A. thaliana* species (Deslandes et al., 1998). Plants previously inoculated with *G. diazotrophicus* and infected later with GMI1000 or A21 strains of *R. solanacearum* did not show symptoms of disease (**Figures 5L** and **11L**).

Although between 6 and 9 dpi the phytopathogenic bacterium does not generate changes in the exomorphological aspect of the plants in any of the treatments, presence of bacteria could be detected along the stem in samples stained with toluidine blue in plants of *A. thaliana* treated only with *R. pseudosolanacearum* GMI1000 (**Supplementary Figures 2A–H**). Presence of a strain GMI1000-GFP was confirmed in roots and stems of *A. thaliana* seedling treated only with this strain by confocal microscopy (**Figures 6A, B** and **7A, B**).

The interaction of *G. diazotrophicus* Pal5 and *R. solanacearum* A21 and its anatomical effects on *A. thaliana* plants were analyzed. Stem cross sections stained with safranin-fast green of *A. thaliana* seedlings treated with endophytic and pathogenic bacteria showed tissue integrity with greater lignification of xylematic vessels and sclerosed cortical parenchyma between the vascular bundles (**Figures 12D, E**). The opposite was observed for plants only inoculated with *R. solanacearum* A21 (**Figures 12A–C**). The structural differences allowed the stems to conserve their structure better than those plants infected only with pathogenic bacterium. In this case, *A. thaliana* seedling with *G. diazotrophicus* elicited a resistance mechanism to *R. solanacearum* A21 infection. Roots of *A. thaliana* plants without *G. diazotrophicus* were more easily colonized as observed in **Figure 13**. Plasmolized epidermal, cortical, and endodermal root cells were evidenced according to Digonnet et al. (2012), who described the route of *R. solanacearum* colonization in *A. thaliana* roots. Greater lignification of xylematic elements of vascular cylinder and mayor integrity of cortical and endodermal root cells were observed in roots of plants with *G. diazotrophicus* (**Figures 13D, E**).

Bacterial counting assays were also performed to determine microbial populations resulting from the interaction of endophytic and phytopathogenic bacteria. Both the microbial population of *R. pseudosolanacearum* GMI1000 and *R. solanacearum* A21 decrease in roots of *A. thaliana* seedlings in the presence of *G. diazotrophicus*. Stem extracts and dilutions of *A. thaliana* seedlings treated with *G. diazotrophicus* and *R. solanacearum* (GMI1000 or A21) did not show growth in the selective medium mSMSA for pathogenic bacterium (**Table 1**).

This and previous results indicated that the bacterial population of *R. solanacearum* A21 and *R. pseudosolanacearum* GMI1000 decreased in the presence of *G. diazotrophicus* in *A. thaliana* seedlings, therefore this endophyte microorganism would be exerting a mechanism of biocontrol on the phytopathogen.

Biocontrol mechanisms by beneficial microorganisms include: i) direct interference with pathogens, this may be through competition for nutrients and space, the secretion of antibiotics, or the degradation of virulence factors; ii) the induction of resistance by the host plant, which is often related to the induced systemic resistance (ISR) that involves ET and JA phytohormones (Haas and Défago, 2005; Van Wees et al., 2008). In the case of beneficial microorganisms, ISR is usually associated with priming the defense routes for an enhanced response, rather than directly activating of the defense system (Van Wees et al., 2008; Zamioudis and Pieterse, 2012). Filgueiras et al. (2019) observed an increase in PR-10 a pathogenesis related protein, a marker of the JA/ET defense route in rice plants when they were inoculated with *G. diazotrophicus*. This endophyte activated a similar response in sugarcane plants since these plants inoculated with *G. diazotrophicus* were more resistant to infection with pathogens such as *X. albilineans*, *C. falcatum*, and *Meloidogyne incognita*. De Nogueira et al. (2001) and Cavalcante et al. (2007) demonstrated the activation of genes involved in the ET signaling pathway in sugarcane plants colonized by *G. diazotrophicus* Dong et al. (1995; 1997) reported the accumulation of polysaccharides and tannins in the parenchymal cells surrounding the metaxylem vessels of sugarcane plants inoculated with *G. diazotrophicus* suggesting that the plant's defense system is activated during the interaction with the bacteria. The increased lignification in xylem elements and sclerosis of diverse tissue in both stems and roots of *A. thaliana* col-0 inoculated with *G. diazotrophicus* are concordant with Dong et al. (1997) observations. In the present work confocal microscopy technique was used and *R. pseudosolanacearum* GMI1000 with GFP to observe the presence of this bacterium in roots and stems of *A. thaliana* Col 0. Roots of *A. thaliana* plants treated with *G. diazotrophicus* and *R. pseudosolanacearum* GMI1000-GFP showed the phytopathogenic bacteria arrested in the cells surrounding the metaxylematic vessels, without colonizing them (**Figures 7C–D**). This could be due to antibacterial compounds present in these cells or to cell wall modifications of the metaxylematic vessels. Nakaho et al. (2000) reported thicker electron-dense pit membranes in resistant tomato cultivars resulting in a limited movement of *R. solanacearum*. Meanwhile, Caldwell et al. (2017) proposed that the differential colonization of *R. solanacearum* in resistant and susceptible tomato roots was due to the ability of the resistant cultivars, through different mechanisms, to restrict bacterial root colonization in time and space. Therefore, xylem vessel structure could determine the plant response to this phytopathogen. On the other hand, Rangel de Souza et al. (2015) reported the participation of the defense mechanism mediated by SA when inoculated *A. thaliana* Col-0 with *G. diazotrophicus*. *NahG* mutant plants, which present a bacterial salicylate hydroxylase that degrades SA, showed no growth problems during the first stages of infection with *G. diazotrophicus* as did Col-0 plants of *A. thaliana*. Conn et al. (2008) reported that an endophytic actinobacteria in plants of *A. thaliana* Col-0, is able to priming for both defense routes, the SAR route, and the JA/ET route, regulating "upstream" genes in both pathways depending on the pathogen that later infects the plant. So, the resistance to the bacterial pathogen *Erwinia carotovora* subsp. *carotovora* required the JA/ET route and, on the other hand, the resistance to the fungal pathogen *Fusarium oxysporum* involved the SAR response.

Plants presented many specialized defense mechanisms; the plant cell wall represents a fundamental line of defense. The cell wall is reinforced with a complex structure, so-called papillae, at sites of interaction with foreign microorganisms. Papillae are formed between the plasma membrane and the cell wall. The biochemical composition of papillae may vary between plant species, but some commonly compounds include reactive oxygen species, phenolics, cell wall proteins, and polymers such as (1,3)-β-glucan callose (Voigt, 2014; Ellinger and Voigt, 2014). Ellinger et al. (2013) reported that timing of the different papilla-forming transport processes is important factor to slow or even stop pathogen invasion. In the present work, *G. diazotrophicus* prime the cellular defense response that involves the deposition of callose. The formation of a discrete number of small papillae in root hairs of *A. thaliana* plants treated with *G. diazotrophicus* was observed, indicating that this is the entry site chosen by this endophyte (**Figures 8 D**–**F**). The formation of papillae in plants treated with *R. pseudosolanacearum* GMI1000 was observed both in cells of the root surface (**Figures 8G–I**) and in the root hairs (**Figures 8G, H**). In plants previously treated with *G. diazotrophicus* and then with *R. pseudosolanacearum* GMI1000, the papillae were larger and more abundant, indicating a rapid response to their formation (**Figures 8J–L**). Similarly, callose deposition in trichomes and leaf tissue of *A. thaliana* Col 0, was greater in those plants that had previously been inoculated with the bacterial endophyte and then were infected with *R. pseudosolanacearum* GMI1000 (**Figures 9G, H**). The participation of the phytohormone SA in the biocontrol by *G. diazotrophicus* of *R. solanacearum* was confirmed using a *sid2* mutant of *A. thaliana* in the SA biosynthesis. In these plants no callose deposition on the development of papilla was observed in radical hairs of the roots (see **Supplementary Figure 6**). In addition the absence of callose was also evident in trichomes and areas of the epidermis in the *sid2* mutants with all treatments (**Supplementary Figure 5**). Ahn et al. (2007) demonstrated that a non-pathogenic rhizobacteria, *Pseudomonas putida*

LSW17S allowed a strong and rapid transcription of defense genes and the accumulation of hydrogen peroxide and callose in plants of *A. thaliana* Col 0 infected with *Pseudomonas syringae* pv. *tomato* DC3000. LSW17S prime the resistance to the disease in *A. thaliana* plants *via* the activation of SAR and JA/ET routes.

Endophytic bacterium produces a protective effect in the plant against the pathogen through the cell wall reinforcement and the increase in lignin and callose, preventing the successful colonization of the pathogenic bacteria which is evidenced by a smaller amount of bacteria in the plant and a delayed phenotype of the appearance of damage to plant tissue and chlorosis caused by the phytopathogenic bacteria. All these modifications lead to an increase in the plant defense response and are extremely linked with the production of SA. Altogether these results indicate that *G. diazotrophicus* colonizes *A. thaliana* plants through the radical hairs, inducing the resistance to *R. solanacearum* infection by mechanism such as papillae formation that contains callose and structural changes in xylem vessels

Our study also provides results about a new typified strain of *R. solanacearum*, A21, isolated from tomato crops of Argentinean northeast region. *G. diazotrophicus* also prevents the advance of the infection of this strain of *R. solanacearum.* Our work opens new insight in the integrated management of production/ protection of intensive agronomic crops of interest attacked by *R. solanacearum*.

### DATA AVAILABILITY STATEMENT

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

MR and EO conceived, designed and direct the work. MR and NA carried out the experiments, performed the measurements and analyzed the data. JT and AC contributed in carried out and planned the biocontrol assays. MSS contributed in carried out the colonization assays. MIS and VF performed molecular typing of Argentinean isolated strains. MM and AAC contribute to histological evaluations. All authors contributed to analysis and interpretation of results. MR, MIS and EO wrote the manuscript. JT, AC and MSS revised and corrected the manuscript. All authors have made substantial, direct and intellectual contribution to the work. All the authors have read the final manuscript and approved the submission.

## FUNDING

This work was supported by the Agencia Nacional de Promoción Científica y Tecnológica (ANPCyT PICT 2017-2242 to EO) and by Science and technology Secretary from Rosario National University (UNR) [Grant BIO432] to EO.

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

#### ACKNOWLEDGMENTS

Authors greatly acknowledge Ing. Agr. Paola Delaporte Quintana and Dr. Marc Valls by bacterial strains ceded. EO and MR are staff members and JT, AC and MSS are Fellows of the Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET, Argentina)

#### REFERENCES


greatly acknowledge CONICET. MM and AC are Professors at the Área de Biología Vegetal of the Facultad de Ciencias Bioquímicas y Farmacéuticas de la Universidad Nacional de Rosario.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019.01618/ full#supplementary-material


**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 © 2019 Rodriguez, Tano, Ansaldi, Carrau, Srebot, Ferreira, Martínez, Cortadi, Siri and Orellano. 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.*

## Biocontrol of Bacterial Wilt Disease Through Complex Interaction Between Tomato Plant, Antagonists, the Indigenous Rhizosphere Microbiota, and Ralstonia solanacearum

Tarek R. Elsayed1,2† , Samuel Jacquiod3,4† , Eman H. Nour<sup>1</sup> , Søren J. Sørensen<sup>3</sup> and Kornelia Smalla<sup>1</sup> \*

1 Institute for Epidemiology and Pathogen Diagnostics, Julius Kühn-Institut, Federal Research Centre for Cultivated Plants, Braunschweig, Germany, <sup>2</sup> Department of Microbiology, Faculty of Agriculture, Cairo University, Giza, Egypt, <sup>3</sup> Marine Microbiological Section, Department of Biology, Faculty of Natural and Life Sciences, University of Copenhagen, Copenhagen, Denmark, <sup>4</sup> Agroécologie, AgroSup Dijon, INRAE, Université Bourgogne, Université Bourgogne Franche-Comté, Dijon, France

#### Edited by:

Kalliope K. Papadopoulou, University of Thessaly, Greece

#### Reviewed by:

Sanushka Naidoo, University of Pretoria, South Africa Valeria Ventorino, University of Naples Federico II, Italy

#### \*Correspondence:

Kornelia Smalla kornelia.smalla@julius-kuehn.de †These authors have contributed equally to this work

#### Specialty section:

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

Received: 12 July 2019 Accepted: 22 November 2019 Published: 10 January 2020

#### Citation:

Elsayed TR, Jacquiod S, Nour EH, Sørensen SJ and Smalla K (2020) Biocontrol of Bacterial Wilt Disease Through Complex Interaction Between Tomato Plant, Antagonists, the Indigenous Rhizosphere Microbiota, and Ralstonia solanacearum. Front. Microbiol. 10:2835. doi: 10.3389/fmicb.2019.02835 Ralstonia solanacearum (biovar2, race3) is the causal agent of bacterial wilt and this quarantine phytopathogen is responsible for massive losses in several commercially important crops. Biological control of this pathogen might become a suitable plant protection measure in areas where R. solanacearum is endemic. Two bacterial strains, Bacillus velezensis (B63) and Pseudomonas fluorescens (P142) with in vitro antagonistic activity toward R. solanacearum (B3B) were tested for rhizosphere competence, efficient biological control of wilt symptoms on greenhouse-grown tomato, and effects on the indigenous rhizosphere prokaryotic communities. The population densities of B3B and the antagonists were estimated in rhizosphere community DNA by selective plating, real-time quantitative PCR, and R. solanacearum-specific fliC PCR-Southern blot hybridization. Moreover, we investigated how the pathogen and/or the antagonists altered the composition of the tomato rhizosphere prokaryotic community by 16S rRNA gene amplicon sequencing. B. velezensis (B63) and P. fluorescens (P142) inoculated plants showed drastically reduced wilt disease symptoms, accompanied by significantly lower abundance of the B3B population compared to the non-inoculated pathogen control. Pronounced shifts in prokaryotic community compositions were observed in response to the inoculation of B63 or P142 in the presence or absence of the pathogen B3B and numerous dynamic taxa were identified. Confocal laser scanning microscopy (CLSM) visualization of the gfp-tagged antagonist P142 revealed heterogeneous colonization patterns and P142 was detected in lateral roots, root hairs, epidermal cells, and within xylem vessels. Although competitive niche exclusion cannot be excluded, it is more likely that the inoculation of P142 or B63 and the corresponding microbiome shifts primed the plant defense against the pathogen B3B. Both inoculants are promising biological agents for efficient control of R. solanacearum under field conditions.

Keywords: Ralstonia solanacearum, biocontrol, latent infection, fliC, amplicon sequencing

## INTRODUCTION

fmicb-10-02835 January 9, 2020 Time: 18:36 # 2

The utilization of microbes to improve plant growth and health is gaining momentum. While significant knowledge on the links between plant traits and their microbiota was obtained from next generation sequencing technologies (Panke-Buisse et al., 2015), downstream applications of that knowledge are still difficult (Herrmann and Lesueur, 2013). Indeed, crop treatment with beneficial strains might be compromised by the poor survival rates of inoculants under field conditions (Dutta and Podile, 2010) and thus a better understanding of the ecology of inoculants is needed. Furthermore, deciphering the complex interaction of inoculants, pathogens, and the indigenous rhizosphere prokaryote community stand as one of the major challenges in understanding the ecology of plant– microbe interaction (Philippot et al., 2013; Shi et al., 2016; Berg et al., 2017).

Ralstonia solanacearum is a quarantine phytopathogen responsible for huge agricultural losses worldwide (Mansfield et al., 2012). R. solanacearum strains (Yabuuchi et al., 1995) form a species complex in the Burkholderiaceae family, divided into four phylotypes associated to geographic locations following human societies and agriculture expansion (I: Asia, II: America, III: Africa, IV: Pacific; Lowe-Power et al., 2018a,b). This soilborne phytopathogen can infect more than 200 plant species, including crucial commercial crops. R. solanacearum survives for long periods in the environment (Graham et al., 1979; Grey and Steck, 2001) and when stressed (e.g., by cold temperatures; van Elsas et al., 2000, 2001; Kong et al., 2014), R. solanacearum initiates a resistance phase, the so-called "viable but non-culturable state" (VBNC), making it readily prone for dissemination via surface irrigation or infested soils. It may colonize rhizospheres of numerous non-host crops and weeds, or even hide under latent infection forms in endophytic compartments (Ciampi et al., 1980; van Overbeek et al., 2004). Although warm areas favor the development of the R. solanacearum wilting symptoms (Bocsanczy et al., 2014), cold-tolerant strains belonging to the brown rot phylotype IIB1 (Cellier and Prior, 2010) can infect host plants in temperate zones (Milling et al., 2009), making R. solanacearum a major threat to agriculture worldwide.

Infection is initiated through primary root tissue penetration via wounds or naturally occurring openings (e.g., secondary root emergence spots), followed by aggressive colonization of the host plant root system before becoming systemic, with appearance of typical shoot symptoms (Lowe-Power et al., 2018a,b). Several factors may influence R. solanacearum virulence, including anoxic condition. Indeed, while preferring oxygen, nitrate assimilation and respiration can enhance R. solanacearum attachment to the roots and promote its virulence (Dalsing and Allen, 2014; Dalsing et al., 2015). Furthermore, R. solanacearum persistence and success is ensured by efficient responses through up-regulation of genes involved in: (i) response to root exudates and low-oxygen conditions in rhizospheres (Colburn-Clifford and Allen, 2010), (ii) degradation pathways against plant defense compounds (e.g., hydroxycinnamic acid) (Lowe et al., 2015), or (iii) the adaptation to the nutrient-deprived xylem environment (Brown and Allen, 2004; Jacobs et al., 2012). For review on the topic, see Lowe-Power et al. (2018a,b).

Different strategies were developed to control R. solanacearum, such as agrochemicals, soil disinfection, antibiotics, antimicrobial plant extracts, resistant cultivars, genetic modification, crop rotations, organic amendments, lytic bacteriophages, and bacterial antagonists (reviewed by Yuliar et al., 2015). The use of environmentally-friendly biocontrol strategies relying on bacterial inoculant strains to enhance the soil wilt suppressiveness and plant priming capacity is a promising strategy, particularly in areas where the pathogen is endemic (Xue et al., 2013).

The major objective of this study was to assess the rhizosphere competence, the efficiency of reducing bacterial wilt symptoms on tomato, and the effects on the indigenous rhizosphere communities under greenhouse conditions for the two strains Bacillus velezensis (B63) and Pseudomonas fluorescens (P142). Seed inoculation and drenching was done, and tomato plants were grown in soil infested with R. solanacearum B3B or not. An integrative approach coupling several methods was employed to investigate pathogen abundance, rhizocompetence of the inoculant strains, root colonization patterns of the gfptagged P142, and the treatment effects on the rhizosphere prokaryotic communities. We hypothesized that priming of tomato plants against R. solanacearum is achieved through a complex interplay between plant, inoculants, rhizosphere microbiome shifts, and the pathogen.

### MATERIALS AND METHODS

#### Plant Materials and Bacterial Isolates

Tomato plant (Lycopersicon esculentum Mill. cv. Money maker) was selected as a host plant susceptible to R. solanacearum (strain B3B, race 3 biovar 2). The two antagonists, B. velezensis (B63) and P. fluorescens (P142), were selected after a pre-screening of in vitro antagonists on tomato plants for the greenhouse experiments reported here. Strains P142 and B63 were isolated from the tuber endosphere of potato plants grown in Germany or Egypt, respectively. The genomes of both strains were recently sequenced and the taxonomic assignment is based on multi-locus sequence analysis (Elsayed unpublished).

#### Generation of Rifampicin Resistance Mutations and/or gfp-Tagged Antagonists

Rifampicin-resistant mutants (Rif<sup>r</sup> ) were generated for both antagonists by inoculating 100 µl of 24-h bacterial culture of each antagonist onto R2A medium supplemented with rifampicin (50 µg/ml) and incubated (28◦C). Rifampicin-resistant colonies were picked after 72 h and preserved at −80◦C in LB broth medium supplemented with 20% glycerol. The Rif<sup>r</sup> strain P142 was tagged with gfp gene encoding the green fluorescent protein (GFP) in a triparental mating (Haagensen et al., 2002). In brief, Escherichia coli CC118λpir was used as a donor for IncQ plasmid pSM1890 carrying the mini-Tn5-PA1/04/03-gfpmut3

cassette coding for the GFP as well as streptomycin (Sm<sup>r</sup> ) and gentamicin (Gm<sup>r</sup> ) resistance, E. coli CM544 carrying IncP-1β plasmid as a helper (Haagensen et al., 2002) and P142 as recipient. The presence of the gfp gene in P142 was tested by real-time PCR (Hajimorad et al., 2011) and the identity of the gfp-tagged antagonists was confirmed via comparing the BOX-fingerprints with the corresponding original isolate (Rademaker and De Bruijn, 1997). Antagonistic activity was re-tested for the gfptagged and/or Rif<sup>r</sup> strain P142 according to Xue et al. (2013). Primers, PCR conditions, and probes used are compiled in **Supplementary Table S1**. The Rif<sup>r</sup> B63 strain was not gfp-tagged as the IncQ plasmid pSM1890 could not stably replicate in B63.

#### Rhizosphere Competence and Biocontrol Efficiency

The Rif<sup>r</sup> antagonists P142 and B63 were grown in 50 ml LBbroth medium supplemented with corresponding antibiotics in an Erlenmeyer flask and incubated in a rotary shaker at 28◦C. Bacterial cells were harvested by centrifugation (4500 × g for 10 min) after 24 h, pellets were washed three times (sterile NaCl 0.85%) and the density of the resuspended cells was adjusted to OD<sup>600</sup> = 1.0 (about 10<sup>8</sup> CFU/mL in NaCl 0.85%). Tomato seeds were soaked in the bacterial suspensions (20◦C, 15 min) and airdried (10 min). Inoculated and non-inoculated seeds were sown in diluvial sand soil (DS; information on the bacterial community composition and the physicochemical composition were reported by Schreiter et al., 2014) soil mixed with a standard potting soil (1:1 v/v) and kept in a greenhouse (2 weeks, 16 h light, 28◦C). Uniformly developed seedlings were transferred to 15 cm pots filled with 300 g DS soil (four replicates per isolate, one plant per pot) under the same conditions. An additional drenching step was done one day prior to transplantation [14 days post sowing (dps)] with 4 ml bacterial culture suspension OD<sup>600</sup> = 1.0 (about 10<sup>8</sup> CFU/ml, Colony Forming Unit). Four plants treated with 4 ml saline solution served as control. Inoculated seedlings were transplanted to soil artificially infested by R. solanacearum B3B (TCR-B63; TCR-P142) or to control soil which was not infested (TC-B63; TC-P142). In addition, seedlings not inoculated with antagonists grown in non-infested soil served as control (TC), and as pathogen control (TCR) when grown in infested soil. Two different doses of R. solanacearum B3B were used, at a final population of 4.4 10<sup>4</sup> (low dose) or 1.8 10<sup>6</sup> CFU g−<sup>1</sup> of soil (high dose). Only non-inoculated plants grown in soil infested with the high dose developed wilting symptoms (**Figure 1**). Symptoms were recorded daily for 2 weeks post transplanting. Hence, the analysis of rhizocompetence, biocontrol efficiency, and prokaryotic community analysis was done only for tomato plants grown in high dose B3B-infested soils. Fourteen days after transplanting, tomato plants were harvested and rhizosphere samples were obtained and analyzed as described below.

#### Sampling and Sample Processing

Tomato plants were sampled 14 days after transplanting. Rhizosphere samples: the entire root system with the tightly adhering soil was transferred into a Stomacher bag, resuspended in 15 ml of 0.85% NaCl and treated with a Stomacher

400 Circulator (Seward Ltd., Worthing, United Kingdom) at middle speed. The supernatant was collected and the Stomacher treatment was repeated twice. A total of 45 ml of supernatant was collected in 50 ml Falcon tubes and used for plating and harvesting the rhizosphere cell pellet after centrifugation. To sample the endorhiza communities, the same root used for the rhizosphere analysis was surface-sterilized by dipping the root in sodium hypochlorite (5% active chlorine) for 3 min, followed by 3 min in hydrogen peroxide 3% according to Sturz et al. (1999), then three washing steps for 10 min each using sterilized saline. The root sterility was checked by pressing the roots on R2A medium. Surface-sterilized roots were blended using sterilized mortar and pestle. Serial dilutions were prepared from the rhizosphere and endorhiza bacterial suspension and plated onto King's B Agar medium (King et al., 1954), supplemented with Rif50, Sm50, Gm10, ampicillin<sup>100</sup> , chloramphenicol30, and cycloheximide<sup>100</sup> (Cyc) for P142, and PCA medium supplemented with Rif and Cyc for B63. CFU counts were enumerated after 48 h of incubation at 28◦C and related to gram root fresh weight (rfw). The CFU counts of B3B were determined using semi-selective medium from South Africa (SMSA) supplemented with suitable antibiotics as described by Engelbrecht (1994). CFU counts were recorded after 48 h incubation at 28◦C. Significant differences of the

CFU counts were analyzed by Tukey's LSD test at (p ≤ 0.05) using SAS software.

Total community DNA was extracted from the rhizosphere pellets (500 g) with the FastDNA spin kit for soil (MP Biomedicals, Heidelberg, Germany). The GENECLEAN SPIN Kit (MP Biomedicals, Heidelberg, Germany) was applied to purify the extracted DNA according to the manufacturer's instructions. DNA samples were diluted 1:10 by 10 mM Tris HCl pH 8.0 and stored at −20◦C for further analysis.

#### Confirmation of the in planta Biological Control of Ralstonia solanacearum and Latent Infection

The two antagonists P142 and B63 were tested in a second greenhouse experiment with more plants to confirm the results of the previous greenhouse experiment and to test for latent infection. Tomato seeds were treated with each antagonistic isolate culture suspension (OD<sup>600</sup> = 1.0), respectively. Seeds were germinated and grown in potting soil for 1 month; before transplanting, a drenching with 4 ml of each antagonist (OD<sup>600</sup> = 1.0) was applied 28 dps, control plants were treated with the same volume of NaCl 0.85%. Seedlings were transferred to pots filled with 300 g B3B-infested DS soil (32 replicates each) or non-infested soil. Untreated plants served as control. The soil was artificially infested with 4 ml B3B per pot (OD<sup>600</sup> = 1.0) to a final density of 1.3 10<sup>6</sup> CFU g−<sup>1</sup> of soil. The development of wilting symptoms was daily observed for one month. After 14 days, four tomato plants were harvested, rhizosphere samples were processed, and CFU counts of B3B, P142, and B63 were determined as described above. In addition, surface-sterilized tomato shoots (in sodium hypochlorite 5% for 3 min, followed by 3% H2O<sup>2</sup> for additional 3 min, and finally three washing steps in sterile water) were immediately frozen in liquid nitrogen and ground in sterilized mortar and pestle, then ground plant materials were kept at −80◦C for total community DNA extraction.

#### Real-Time PCR-Based Quantification of Target Genes From the Rhizosphere Total Community DNA

Bacterial 16S rRNA gene copies (rrn) were estimated in rhizosphere community DNA according to Suzuki et al. (2000). The copy numbers of gfp gene were determined in rhizosphere total community DNA of TCR-P142 and TC-P142-treated plants and related to 16S rRNA gene copies (Yankson and Steck, 2009). Primers targeting the UDP-3-O-acyl-GlcNAc deacetylase, proposed by Chen et al. (2010), were modified in order to improve the specificity for B3B (**Supplementary Figure S1**). The copy numbers of R. solanacearum B3B were quantified in total community DNA using the modified primers under the following conditions: 95◦C for 2 min followed by 40 cycles at 95◦C for 20 s, 62◦C for 25 s, 72◦C for 35 s, and finally 80◦C for 3 s before plate read, a melt curve step was included to verify the primer specificity. The primer pairs (B3B-RSF and B3B-RSR) were used under PCR conditions of 94◦C for 5 min, and 30 cycles of 94◦C for 1 min, 54◦C for 1 min and 72◦C for 1 min, and then 10 min at 72◦C were applied before cooling down to 4◦C, to amplify a fragment of 441 bp from B3B which was subsequently cloned in E. coli using the pGEM-T Easy Vector system I (Promega Corporation, Madison, WI, United States) according to the manufacturer's protocol. The pGEM-T Vector was reextracted using the GeneJET Plasmid Miniprep kit (Thermo Fisher Scientific, Vilnius, Lithuania) and used for serial dilutions to establish the standard calibration for the real-time PCR. All primers are listed in **Supplementary Table S1**.

#### Illumina Sequencing and Analysis of 16S rRNA Gene Amplicons From Total Community DNA

Amplicon sequencing was performed according to defined and acknowledged best practices as previously described (Schöler et al., 2017). Prior to tag-encoded 16S rRNA gene sequencing, the 24 samples of extracted DNA were subjected to an initial PCR amplification step using a set of primers, 341F and 806R (**Supplementary Table S1**), which flank the approximately 460 bp variable V3–V4 region of the 16S rRNA gene of the target group Prokaryotes including domains of Bacteria and some Archaea. A second amplification step of the corresponding 16S rRNA gene region using the same primers with attachment of adaptors and barcode tags was done as previously described (Jacquiod et al., 2017). Purification and size selection (removal of products of less than 100 bp) of the approximately 620 bp PCR amplicon products was performed using Agencourt AMPure XP beads (Beckman Coulter, Brea, CA, United States) according to the manufacturer's instructions. The concentration of purified amplicon samples was subsequently measured using a Qubit Fluorometer (Life Technologies, Carlsbad, CA, United States), the samples were pooled and adjusted to equimolar concentrations, concentrated using the DNA Clean and ConcentratorTM-5 kit (Zymo Research, Irvine, CA, United States), and finally subjected to 2 × 250 bp paired-end high-throughput sequencing on an Illumina <sup>R</sup> MiSeq <sup>R</sup> platform (Illumina, San Diego, CA, United States).

Amplicon sequences were analyzed using qiime\_pipe<sup>1</sup> with default settings, which performs sample demultiplexing, qualitybased sequence trimming, primer removal, and paired-end reads assembly prior to annotation workflow (Caporaso et al., 2010). Paired-end mating was applied with a minimum overlap length of 50 bp, maximum mismatches of 15, and a minimum quality of 30. Criteria for sequence trimming were based on: (1) reads shorter than 200 bp, (2) average quality scores lower than 25, (3) maximum number of ambiguous bases, and (4) six as maximum lengths of homopolymers. Chimera check was done with UCHIME (Edgar et al., 2011) and operational taxonomic units (OTUs) were picked at 97% sequence identity level. OTU representative sequences were selected by the highest abundance within the cluster and assigned to taxonomy using the RDP classifier (Cole et al., 2003), with a confidence threshold of 80%. Information regarding the sequence counts for each sample is provided in the **Supplementary Table S2**, and rarefaction curves are presented in **Supplementary Figure S2**. Communitylevel analysis was performed with a cluster dendrogram using

<sup>1</sup>https://github.com/maasha/qiime\_pipe

the unweighted pair group method with arithmetic mean (UPGMA, Euclidean distance). Significant changes in the relative abundance of dominant taxa were identified with an ANOVA under a generalized linear model, followed by Tukey's honest significance detection test (p < 0.05). Sequences were submitted for deposition at the public repository Sequence Read Archive (SRA<sup>2</sup> ) with the accession number PRJNA574588<sup>3</sup> .

#### PCR-Southern Blot Hybridization-Based Detection of R. solanacearum Specific fliC Gene

PCR amplification with primers targeting Rs-fliC gene was performed according to Schönfeld et al. (2003) from total community DNA from rhizosphere and shoots of tomato plants grown in B3B-infested soils. PCR products were analyzed by 1% agarose gel electrophoresis for 1 h (50 V), gels were checked by UV light after staining with ethidium bromide and Southernblotted as described by Binh et al. (2008). Hybridization was performed with Digoxygenin-labeled fliC probe generated from purified PCR products obtained with B3B by means of the DIG DNA labeling kit (Roche Applied Science, Mannheim, Germany).

#### Confocal Laser Scanning Microscopy (CLSM) Analysis

Tomato roots were analyzed by confocal laser scanning microscopy (CLSM) 5 days after drenching with gfp-tagged isolate P142 to detect and localize its colonization of tomato ecto- and endophytic root compartments. The tightly attached soil particles were removed by shaking the root vigorously, then cut into small pieces of ca. 2 cm, and mounted with a few drops of 0.85% NaCl. Root pieces were analyzed using Leica TCS SP2 CLSM. Argon/Krypton laser (excitation at 488 nm) was used to detect the excitation of the GFP combined with the transmitted light pictures. Detected GFP signals were confirmed by applying lambda scan.

#### RESULTS

#### Rhizocompetence of the Inoculant Strains

The potential of the two antagonists to colonize the rhizosphere of tomato root, expressed by means of CFU counts, was determined 14 days after transplanting. The CFU counts of P142 indicated efficient rhizosphere colonization with 5.9 Log<sup>10</sup> CFU g −1 rfm, while rather low CFU counts were detected for B63 with 3.1 Log<sup>10</sup> CFU g−<sup>1</sup> rfm (see **Figure 2A**).

The high relative abundance of the gfp-copy number (−1.49) determined by qPCR in total community DNA confirmed the high rhizosphere competence of P142 (**Figure 2B**). Noteworthy, the relative abundance of P142 in the rhizosphere was significantly higher for P142-treated plants grown in B3Binfested soils (−1.49 Log<sup>10</sup> gfp copy number/16S rRNA gene

<sup>2</sup>https://www.ncbi.nlm.nih.gov/sra

g −1 g of rfm) compared to non-infested soils (−2.95 Log<sup>10</sup> gfp copy number/16S rRNA gene g−<sup>1</sup> of rfm) (**Figure 2B**).

#### Biological Control of Ralstonia solanacearum in planta

For assessing the efficiency of P142 and B63 to reduce tomato wilt symptoms, plants transplanted into soil infested with the high dose of B3B were assessed daily for the appearance of wilting symptoms. All tomato control plants grown in soil infested with high B3B dose (TCR; 1.8 10<sup>6</sup> B3B CFU g−<sup>1</sup> soil) had collapsed 14 days post infection (dpi), while no uniform wilting symptoms were observed when the soil was infested with low B3B population density (4.4 10<sup>4</sup> CFU g−<sup>1</sup> soil). Thus, effects of P142 and B63 on the indigenous prokaryotic communities of the tomato rhizosphere were only assessed for the plants grown in soil with the high B3B density. Plants inoculated with P142 or B63 showed no wilting symptoms 14 dpi (**Figure 1**). The tomato plants treated with antagonists showed significantly lower B3B CFU counts compared to the TCR (8.6 Log<sup>10</sup> CFU g −1 root). Approximately three orders of magnitude lower B3B CFU counts were recorded for both TCR-P142 and TCR-B63 (5.2 and 5.1 Log<sup>10</sup> CFU g−<sup>1</sup> rfm, respectively; **Figure 2A**). The relative abundance of B3B was very high in the pathogen controls (TCR: -0.85 Log<sup>10</sup> copy number/16S rRNA gene g−<sup>1</sup> of rfm; **Figure 2B**) and significantly lower in DNA from the rhizosphere of inoculated tomato plants with relative abundance of B3B being reduced about two to three orders of magnitude in the treatments with P142 (−3.22 Log<sup>10</sup> copies/rrn) and B63 (−3.95 Log<sup>10</sup> copies/rrn) compared to TCR (**Figure 2B**).

For further confirmation, the R. solanacearum-specific fliC gene was detected by PCR with subsequent Southern blot hybridization. Very strong hybridization signals were obtained for TCR rhizosphere samples of tomato plants grown in soil infested with the high R. solanacearum population compared to the rhizosphere samples grown in soil infested with low densities (positive signals for three replicates out of four) (**Supplementary Figure S3**), while no hybridization signals were detected in the uninfected control samples (TC). The fliC hybridization patterns obtained from the rhizosphere samples concurred with results of R. solanacearum real-time qPCR data, as stronger hybridization signals corresponded to high relative abundance of R. solanacearum detected via qPCR. Weak or no hybridization signals were detected for TCR-B63 followed by TCR-P142.

#### Biological Control of R. solanacearum: Checking for Latent Infections

An additional greenhouse experiment with a larger number of plants was conducted in order to confirm the efficiency of both B63 and P142 antagonists against R. solanacearum and to check for latent infections. Wilting symptoms of inoculated and non-inoculated plants grown in soil infested with high R. solanacearum densities (3.9 10<sup>6</sup> CFU g−<sup>1</sup> of soil) was recorded 14 days post transplanting. Out of the total 32 replicates, 19 TCR plants (59%) had collapsed. The number of collapsed plants was similar in both TCR-P142 and TCR-B63, only six plants representing 18.8%. The CFU

<sup>3</sup>https://www.ncbi.nlm.nih.gov/bioproject/PRJNA574588

FIGURE 2 | (A) Significant reduction of Log<sup>10</sup> (colony forming units/g root fresh mass) for R. solanacearum B3B (dark gray) for inoculated treatments with the antagonists TCR-P142 and TCR-B63 14 days post transplantation compared to the pathogen control TCR as indicated by lower case letters (Tukey's HSD, p < 0.05). In light gray bars the Log<sup>10</sup> (colony forming units/g root fresh mass) of P-142 and B63 are given. The significant differences in the CFU counts of both antagonists are indicated by "∗∗∗" (Tukey's HSD, p < 0.05). (B) A significant reduction of Log<sup>10</sup> (target gene/rrn) for R. solanacearum B3B (dark gray) was also shown by real time quantitative PCR for R. solanacearum (dark gray) related to 16S rRNA gene (rrn) copy numbers 14 days post transplanting (Tukey's HSD, p < 0.05). The light gray bar represents the Log<sup>10</sup> (gfp/rrn) for P142. No quantitative PCR data four B63 as gfp labeling was not successful for B63.

antagonists TCR-P142 and TCR-B63 of the second greenhouse experiment 14 days post transplantation compared to the pathogen control TCR as indicated by lower case letters (Tukey's HSD, p < 0.05). In light gray bars the Log<sup>10</sup> (colony forming units/g root fresh mass) of P-142 and B63 are given.

counts of the antagonists in the rhizosphere of symptomless plants were 4.7 and 4.5 Log<sup>10</sup> (CFU/g rfm) for TCR-B63 and TCR-P142, respectively (**Figure 3**). While R. solanacearum populations reached 9.57 Log<sup>10</sup> (CFU/g rfm) in TCR samples, significantly lower pathogen levels were recorded for P142 and B63-inoculated plants (Log<sup>10</sup> (CFU/g rfm) root: 6.5 and 7.5, respectively).

Quantification of the R. solanacearum-specific gene confirmed significantly lower B3B copy numbers in total community DNA of TCR-P142 and TCR-B63 (6.8 and 4.0 Log<sup>10</sup> copies per g−<sup>1</sup> of rfm, respectively) compared to TCR samples (Average number of TCR samples = 9.4 Log<sup>10</sup> copies/g rfm) (**Figure 4**). Similarly, in tomato shoot samples, significantly lower B3B CFU counts were recorded in TCR-P142 antagonists (ranging from 5.0 to 8.6 copies per g−<sup>1</sup> of sfm) compared to TCR samples (9.9–10.1 copies per g−<sup>1</sup> of sfm) while it was below detection limit in shoot of tomato plants inoculated with B63. While the Log<sup>10</sup> (gfp/rfm) gene copy number of gfp g−<sup>1</sup> rfw was 7.9 ± 0.38 in the rhizosphere of P142-inoculated tomato plants (**Figure 4**), two out

of four replicates showed, additionally, colonization of tomato shoot endophytic compartments [5.6 ± 0.7 Log<sup>10</sup> (gfp copies g −1 sfm)].

Southern blot hybridization targeting fliC gene in the total community DNA of both rhizosphere and shoot samples showed patterns concurred with the R. solanacearum realtime PCR results (**Figure 5**). Stronger hybridization signals were obtained for TCR rhizosphere samples compared to TCR-P142 and TCR-B63. Regarding the shoot samples, compared to the strong signals obtained for the TCR, only one strong and two very weak signals were detected in TCR-P142, while no hybridization signals were detected in the shoots of TCR-B63 plants.

#### 16S rRNA Gene Amplicon Illumina Sequencing

Illumina amplicon sequencing of V3–V4 regions from the 16S rRNA gene was obtained from the rhizosphere of inoculated or non-inoculated tomato plants grown in B3B-infested or noninfested soil using 2 × 250 bp paired-end Illumina MiSeq. A total of 630,016 bacterial sequences were generated from six treatments (four replicates per each treatment; **Supplementary Table S2** and **Supplementary Figure S2**). The highest number of bacterial sequences was detected for TCR-B63 and TC-B63 (27,865 and 32,112 sequences, respectively), while the lowest number of bacterial sequences was detected for TCR-P142 and TC-P142 (21,171 and 21,325 sequences, respectively). The TC and TCR samples had 32,988 and 31,044 sequences, respectively. The bacterial sequences were affiliated with 10 phyla, 29 classes, 56 orders, 135 families, and 263 genera.


TABLE 1 | Relative abundance of dominant phyla and classes in the rhizosphere of tomato affected by the pathogen R. solanacearum and/or inoculation (average ± standard error of the mean, n = 4 per treatment).

Number shows the average percentage followed by ± standard deviation (n = 4). Treatments sharing the same letters are non-significantly different (p < 0.05, ANOVA under generalized linear model followed by Tukey's Honest Significant Detection test). Significant increases in abundance compared to the TC are highlighted in green, while significant decreases are highlighted in red.

#### Tomato Rhizosphere Bacterial Community Composition

In the rhizosphere of healthy tomato plants (TC), Proteobacteria were the most dominant phylum with relative abundance of 62.3%, followed by Actinobacteria, Bacteroidetes, and Firmicutes (14.6, 12.7, and 8.4%, respectively). Other phyla were detected in the rhizosphere with relative abundance of less than 1% such as Gemmatimonadetes, Planctomycetes, Nitrospirae, Verrucomicrobia, and Chloroflexi. Among the Proteobacteria, most OTUs were affiliated to Alphaproteobacteria (relative abundance of 31.2%), followed by Gammaproteobacteria and Betaproteobacteria (17.6 and 13.1%, respectively), while the relative abundance of Deltaproteobacteria was low (0.4%) (**Table 1**). At the genus level, Rhodanobacter was the most dominant genus with relative abundance of 9% followed by Shinella, Rhizobium, Arthrobacter, Massilia, Sphingobium, Sphingomonas, and Devosia (**Table 2**).

#### Inoculation and Infection-Dependent Prokaryotic Community Structure in Tomato Rhizosphere

Cluster dendrogram analysis (UPGMA) based on the relative abundance of all bacterial OTUs obtained from tomato rhizospheres revealed two major distinct clusters. The first included only the TCR samples, while the second combined the samples of TCR-B63, TCR-P142, TC-B63, and TC-P142 in addition to TC samples (**Figure 6**). Notably, the clustering of the four TCR replicates was correlated with the development of wilting symptoms, as TCR1 was the first plant showing wilting symptoms (4 days before harvest) followed by TCR2, while both TCR3 and TCR4 showed symptoms only 1 day before harvest. However, the second cluster was divided into two sub-groups based on inoculation, then each was further divided, attributed to the presence of B3B, forming a total of four separate clusters (TC-B63 and TCR-B63; TC-P142, TCR-P142, and TC).

#### Influence of R. solanacearum on the Prokaryotic Community Composition of Tomato Rhizosphere

Ralstonia solanacearum B3B strongly shaped the bacterial community composition in the rhizosphere of tomato plants (TCR). Significant changes in the relative abundance of dominant taxa were identified by Tukey's honest significance test under a generalized linear model. The analysis of rhizosphere microbiota of tomato plants grown in B3B-infested soil (TCR) was compared with the non-infected control plants (TC). The results showed that the pathogen massively dominated the rhizosphere microbiota. Thus, the relative abundance of Betaproteobacteria increased (46.7 ± 12.6%, mostly Ralstonia) compared to healthy non-infected tomato plants (13.1 ± 1.2%). In contrast, Gammaproteobacteria, Actinobacteria, Bacteroidetes, and Firmicutes (particularly Bacilli) decreased in DNA from TCR (**Table 1**). At OTU


TABLE 2 | Relative abundance of dominant responding OTUs (relative abundance ≥ 0.5%) detected two weeks after inoculation and/or infection in the rhizosphere of tomato (average ± standard error of the mean, n = 4 per treatment).

Number shows the average percentage followed by ± standard deviation (n = 4). Treatments sharing the same letters are non-significantly different (p < 0.05, ANOVA under generalized linear model followed by Tukey's Honest Significant Detection test). Significant increases in abundance compared to the TC are highlighted in green, while significant decreases are highlighted in red.

level, a decreased relative abundance was detected for 53 taxa, affiliated with 39 different genera. Rhodanobacter, Dyella, Arthrobacter, Rubrobacter, Sphingomonas, Bradyrhizobium, Curtobacterium, Salinibacterium, and Bacillus showed the highest decrease. Besides Ralstonia, OTUs affiliated with Sphingobium and Niastella increased in TCR compared to TC (**Table 2** and **Supplementary Tables S2, S4**).

#### Inoculation of Bacillus velezensis B63 or Pseudomonas fluorescens P142 Antagonists Changed the Tomato Rhizosphere Bacterial Community Composition

In the rhizosphere of tomato plants grown in non-infested soil and inoculated with either B63 or P142 (TC-B63; TC-P142),

FIGURE 6 | Cluster analysis of tomato rhizosphere bacterial communities responding to inoculation by antagonists and/or infection with R. solanacearum (clustering method = UPGM, distance = Euclidean, bootstraps = 1000).

a decrease in Proteobacteria, especially Betaproteobacteria and Gammaproteobacteria classes, was observed (**Table 1**). The phyla Actinobacteria, Gemmatimonadetes, Chloroflexi, and Verrucomicrobia increased in TC-B63 and TC-P142 samples. At OTU level, the highest number of responders was detected in TC-B63 followed by TC-P142. In TC-B63 rhizosphere DNA, a total of 116 OTUs changed, 85 OTUs increased while 31 OTUs decreased. For TC-P142, a total of 68 OTUs changed, 52 OTUs increased while 16 OTUs decreased (**Table 2** and **Supplementary Table S4**). A total of 56 OTUs commonly responded with similar patterns with either TC-B63 or TC-P142 (44 increased and 12 decreased responders). Regarding the strong responders (OTUs that increased or decreased more than two folds compared to TC), members from Alphaproteobacteria substantially increased (Ochrobactrum, Devosia OTU-93 and OTU-244, Pseudolabrys, Rhodospirillaceae, and Sphingomonas), except in the genus Rhizobium (for both antagonists), as well as Asticcacaulis and Devosia in TC-B63 which all decreased. OTUs affiliated to Gammaproteobacteria (Rhodanobacter and Dyella) decreased in both TC-B63 and TC-P142, while Rudaea decreased only in TC-B63. However, within the same class and/or genus, different OTUs showed variable responses to the inoculation with antagonists, as abundances of some OTUs increased while others decreased, compared to the TC samples (**Table 2**).

#### The Complex Interaction Between R. solanacearum, Antagonists, and Indigenous Rhizosphere Microbiota

Sequences affiliated to Ralstonia were about three orders of magnitude lower in TCR-P142 as well as in TCR-B63 compared to TCR. Thus, the relative abundance of B3B was only 0.1% in both TCR-B63 and TCR-P142 while it reached 35.8% in the TCR samples. Alpha-diversity analysis on the rhizosphere microbiota revealed that all tested indices were decreased when tomato plants grew in B3B-infested soil (TCR) due to the dominance of Ralstonia (**Supplementary Table S3**). Richness and evenness were higher when tomato plants grown in non-infested soil were inoculated with B63 (TC-B63; **Supplementary Table S3**).

Gammaproteobacteria and Bacilli were lower in both TCR-P142 and TCR-B63 compared to TC-P142 and TC-B63, respectively. Bacteroidetes, particularly Sphingobacteria, were lower in TCR-B63 compared to TC-B63 samples. The phyla Actinobacteria and Verrucomicrobia showed a higher relative abundance in both TCR-B63 and TCR-P142 samples compared to TC-B63 and TC-P142 (**Table 1**). At OTU level, the abundance of 90 responders changed in TCR-B63 samples (57 increased and 43 decreased), while 54 OTUs changed in TCR-P142 (35 increased and 19 decreased) compared to TC-B63 and TC-P142. A total of 35 OTU responders were shared between both TCR-B63 and TCR-P142 (21 increased and 14 decreased) (**Table 2**). OTUs affiliated to Arthrobacter, Gaiella, Niastella, and Ochrobactrum had a higher relative abundance in both TCR-B63 and TCR-P142, while Devosia, Shinella, Sphingomonas, Acidovorax, and OTUs affiliated with unclassified Rhodospirillaceae were higher only in TCR-B63. Bacillus, Dyella, and Rhodanobacter were lower in both TCR-B63 and TCR-P142 while those of Rhizobium and Chitinophaga decreased only in TCR-P142 (**Table 2** and **Supplementary Table S4**).

### Localization of gfp-Tagged P142 in Rhizosphere and Root Endophytic Compartments

Confocal laser scanning microscopy was used to obtain insights into the P142 root colonization patterns. Tomato root surfaces were efficiently colonized by gfp-tagged biocontrol bacteria. Strong signals were detected five days after drenching in inoculated plants while no signals were detected in control plants (besides auto-fluorescence that could be removed by narrowing

Root pieces were analyzed using Leica TCS SP2 CLSM. Argon/Krypton laser (excitation at 488 nm) was used to detect the excitation of the GFP combined with the transmitted light pictures.

the detection wavelength based on lambda scan). The gfptagged strain P142 was detected in lateral roots as well as in root hairs. Micro-colonies were observed along the root surface while the endophytic life style of P142 was confirmed by the colonization and invasion of epiphytic root cells as well as xylem vessels (**Figure 7**).

#### DISCUSSION

Here we investigated the efficacy of two in vitro antagonists of R. solanacearum to reduce wilting symptoms in tomato plants under greenhouse conditions and followed the abundance of the inoculant strains and the pathogen using cultivationdependent and independent methods. Field testing was no option as R. solanacearum is a quarantine organism. Strain P142 showed good survival in the rhizosphere of tomato plants. By means of CLSM, we could show that P142 was able to colonize tomato roots internally without adverse symptoms, indicating good potential for being true endophytes. Surprisingly, although strain B63 seems to have a rather weak rhizosphere competence its efficiency to reduce wilting symptoms and the abundance of B3B in the rhizosphere and in the tomato shoots was remarkable. Also the numbers of taxa with increased or decreased relative abundance in response to the inoculation of B63 was higher compared to P142. Inoculation of B63 strongly shaped the rhizosphere community in TC-B63 and TCR-B63. One possible explanation is that B63 colonized the soil fraction not that close to the root and thus was missed with the rhizosphere sampling protocol used. In addition, in the present study, vegetative B63 cells were inoculated and not spores, as typically done for Bacillus inoculants. Based on pre-experiments, we have used seed inoculation and drenching before transplanting, as recommended also by Götz et al. (2006).

A high B3B population density (1.78 10<sup>6</sup> g −1 soil) was required to observe uniform infection, and maximum symptom severity was recorded when B3B populations reached 8.44 Log<sup>10</sup> (CFU). The CFU counts of B3B in the rhizosphere were higher in the second greenhouse experiment, and this might explain the appearance of wilting symptoms despite a two to three orders of magnitude reduction of the B3B CFU counts in the plants treated with B63 or P142. Results confirmed that pathogenicity regulation is density-dependent (Schell, 2000), requiring high cell concentration for triggering virulence gene expression and accumulation of deleterious metabolites to cause acute infection (e.g., putrescine; Lowe-Power et al., 2018a). Meanwhile, although a significant decrease in wilted plants (18.8 vs 59%) and lower R. solanacearum densities were observed with both inoculants, only shoot endophytic compartments of TCR-B63 exhibited no latent infection. Therefore, B3B detection in asymptomatic plants showed that absence of visible symptoms of the disease is not a reliable proxy for pathogen eradication. This has important implications for trade with countries where R. solanacearum is endemic. However, it remains unclear at which density level expression of pathogenicity determinants and the subsequent development of wilting symptoms occur, as published studies are often done under control conditions where R. solanacearum densities are high (∼109−10; Lowe-Power et al., 2018a). In the second greenhouse experiment, higher CFU counts for B3B (9.57 Log<sup>10</sup> (CFU/g rfm); **Figure 3**) were detected in the rhizosphere compared to the first greenhouse experiment (8.63 Log<sup>10</sup> (CFU/g rfm); **Figure 2A**), and thus the detection of some plants with wilting symptoms for TCR-P142 and TCR-B63 compared to no wilting in the first greenhouse experiment was not too surprising. Several mechanisms are likely at play to explain the drastic reduction of B3B abundance and the absence of wilting symptoms in the first greenhouse experiment. Plant systemic resistance induction was investigated by Park et al. (2007) for R. solanacearum biological control via Bacillus vallismortis strain Iq EXTN-1, but this aspect is going beyond the scope of our study. The main focus of the present study was to decipher the relative abundance of inoculant, pathogen, and the indigenous prokaryotic community composition in the tomato rhizosphere and how they link to wilting symptoms. Competitive exclusion occurring between R. solanacearum and antagonists was previously suggested, resulting in unsuccessful establishment of the pathogen (Upreti and Thomas, 2015). Here, by CLSM localization of P142 cells on tomato lateral root hair, root surface, as well as xylem vessels, we demonstrated a highly heterogeneous colonization pattern of the gfp-tagged P142 and thus the likelihood that direct interactions play a role is rather low. More likely is a priming of the tomato plants through the presence of the inoculants and/or the prokaryotic community shifts. Both inoculant strains drastically reduced the abundance of R. solanacearum B3B as revealed by CFU counts, qPCR, fliC PCR, and subsequent Southern blot hybridization and amplicon sequencing by about three orders of magnitude.

Genome sequencing revealed for both strains the presence of numerous genes involved in plant beneficial interaction, e.g.,

P142 carries the phl and the phz gene (Elsayed, unpublished). The phl gene coding for 2,4-diacetylphloroglucinol (2,4-DAPG), was previously reported for in vitro and in vivo R. solanacearum suppression (Ramadasappa et al., 2012; Zhou et al., 2012) and in addition to protists predation escaping (Jousset et al., 2006). The phz gene, encoding phenazine production, might also play a role in R. solanacearum control (Hariprasad et al., 2014). Recently, a selection strategy of potential antagonists based on the number of biological control and/or plant growth promoting related function per inoculant candidate was proposed by Mota et al. (2017). They have shown a positive correlation between the number of in vitro functions per antagonist and their effects on the pathogen. Both inoculant strains used in the present study affected the prokaryotic community composition in the rhizosphere. Indeed, we suspect that priority effects are at play, where the chronology of whoever comes first is determining the subsequent community assembly rule (Vannette and Fukami, 2014). Inoculants may further change the recruitment (or not) of other rhizosphere microbial members by the plant, e.g., through changes of the root exudate composition, as previously reported by Windisch et al. (2017).

Illumina amplicon sequencing analysis of 16S rRNA gene fragments of TC, TCR, TC-B63, TCR-B63, TC-P142, and TCR-P142 community DNA revealed numerous "dynamic taxa." The most severe modulation of the rhizosphere prokaryotic community composition was observed for TCR compared to TC. Interestingly, two other genera, Sphingobium and Niastella, profited from the nutrient situation of the rhizosphere of the diseased plants. Most importantly, B3B was nearly suppressed under antagonist presence, as rhizospheres of TCR-B63 or TCR-P142 treatments had only 0.1% of Ralstonia-affiliated sequences, clearly demonstrating the strong biocontrol efficiency of both inoculants. Interestingly, the rhizosphere abundance of P142 was higher in the presence of R. solanacearum B3B (**Figure 2B**).

Pronounced and divergent responses in rhizosphere prokaryotic communities were found, with numerous and phylogenetically diverse OTUs showing either significantly increased or decreased relative abundance compared to controls (**Tables 1**, **2** and **Supplementary Table S4**). The most remarkable observation is the enrichment of Actinobacteria in inoculated treatments. Actinobacteria are recognized for their production of diverse bioactive compounds, their potential biological control activities, and plant growth promotion (Schrey and Tarkka, 2008; Joseph et al., 2012). Similar trends were recorded for the genus Gaiella that was previously described as member of the core microbiome of a disease-suppressive soil (Xue et al., 2015). The other striking observation was the pronounced enrichment of Arthrobacter only in TCR-B63 and TCR-P142. Egamberdieva et al. (2017) reported that Arthrobacter crystallopoietes had a plant growth promotion and protection effect on tomato plant, as it exhibited significant reduction of Fusarium infection while enhancing plant growth. Therefore, the similar responses of the prokaryotic community to the inoculation of B63 and P142 strongly suggest the indirect involvement of the plant itself, steering its root microbiota in a similar manner via the recruitment/stimulation of beneficial soil microbes.

Other dynamic OTUs from Gemmatimonas, Devosia, and Sphingomonas were enriched in response to B63 or P142, indicating that direct/indirect social interaction processes such as microbial facilitation may be involved in biocontrol. Sphingomonas is a strictly aerobic bacterium often characterized as an environmental oligotroph (Lauro et al., 2009; Jacquiod et al., 2017). Some Sphingomonas strains were shown to produce indole acetic acid (IAA), while others displayed phenazine degradation capabilities (Ma et al., 2012; Sukweenadhi et al., 2015). Sphingomonas was detected in lettuce rhizospheres (Schreiter et al., 2014) as well as in endophytic compartments of tomato plants (Khan et al., 2014). It is assumed that the majority of plant-associated Sphingomonas spp. can have a plantprotective effect (Innerebner et al., 2011; Sato et al., 2012). Moreover, genera such as Luteolibacter (Verrucomicrobiae) and Ochrobactrum were increased in all inoculated treatments. Nunes da Rocha et al. (2013) reported a rhizocompetence potential for members of Luteolibacter (Verrucomicrobiae) that might explain its increase. Potential antagonism of Ochrobactrum against phytopathogens was also reported through affecting the quorum sensing regulating the pathogen virulence factors (Czajkowski et al., 2011). However, it seems that the inoculation of antagonists tends to engineer the prokaryotic community toward enriching other beneficial bacteria. In contrast, OTUs from 20 genera were significantly decreased in TCR, mostly due to the dominance of B3B compared to TC.

The inoculation of B63 and P142 resulted in a complex response of the tomato rhizosphere bacterial communities as revealed by amplicon sequencing analysis, although it was far more pronounced for B63. This was also reflected on alpha-diversity, with an increased richness and evenness for TC-B63, indicating that it might be an important keystone species impacting the whole community via facilitation processes (**Supplementary Table S3**). Whether this effect was direct through social interactions with other species, or indirect via stimulation of the plant (e.g., rhizodeposition, plant defense molecules) has yet to be clarified.

## CONCLUSION

Control of bacterial wilt disease caused by R. solanacearum is an important challenge. Many strategies were proposed for controlling bacterial wilt disease. Among them, manipulating soil suppressiveness through organic amendments and managing soil suppressiveness via inoculant strains are considered the most promising and environmentally-friendly alternatives. Our results showed that the strains, B. velezensis B63 and P. brassicacearum P142, are promising candidates for future biocontrol of R. solanacearum under field conditions, through significantly lowered R. solanacearum densities in tomato shoots and in the rhizosphere. Amplicon sequencing revealed many dynamic taxa, likely indicating complex interactions between the inoculant strains, B3B, the prokaryotic community in the tomato rhizosphere and the plant itself. The inoculation with B63 or P142 significantly promoted specific taxa, with potential plant protection and/or growth promotion-related traits, respectively,

which might, in turn, affect soil suppressiveness and increase plant defense. For the first time, 16S rRNA gene amplicon sequencing was used to demonstrate R. solanacearum reduction through inoculation of in vitro antagonists which were correlated to the reduction of wilting symptoms. Combination between cultivation-dependent and independent methods correlated well and in particular Illumina sequencing of 16S rRNA gene fragments amplified from total community DNA allowed deeper insights into the complex interaction that might lead to pathogen suppression. Present research with focus on the plant strongly points to an induction of plant systemic resistance. In summary, this study revealed that both antagonists were efficient in controlling bacterial wilt disease, but likely shifts in the rhizosphere microbiota and the antagonists contributed to the efficient control of bacterial wilt.

#### DATA AVAILABILITY STATEMENT

The datasets generated for this study can be found in the Sequence Read Archive (SRA), accession: PRJNA574588.

### AUTHOR CONTRIBUTIONS

TE contributed to experimental work, data analysis, and manuscript writing. SJ contributed to data analysis and manuscript writing. EN contributed to experimental work, data

### REFERENCES


analysis, and manuscript editing. SS contributed to study design and manuscript editing. KS contributed to study design and manuscript writing.

### FUNDING

This work was funded by the BIOFECTOR project 312117. TE was granted a German Egyptian Research Long-Term Scholarship (GERLS) offered by the Ministry of Higher Education of the Arab Republic of Egypt (MoHE) and by Deutscher Akademischer Austauschdienst (DAAD). SJ was funded by the University of Bourgogne Franche-Comté via the ISITE-BFC International Junior Fellowship award (AAP3: RA19028.AEC.IS).

#### ACKNOWLEDGMENTS

The authors would like to thank Ilse-Marie Jungkurth for proofreading the manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.02835/full#supplementary-material



assays. Appl. Environ. Microbiol. 66, 4605–4614. doi: 10.1128/aem.66.11.4605- 4614.2000


Ralstonia solanacearum strains - an improved strategy for selecting biocontrol agents. Appl. Microbiol. Biotechnol. 97, 1361–1371. doi: 10.1007/s00253-012- 4021-4


**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 Elsayed, Jacquiod, Nour, Sørensen and Smalla. 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.

# Endophytic Fungi of Native Salvia abrotanoides Plants Reveal High Taxonomic Diversity and Unique Profiles of Secondary Metabolites

Yeganeh Teimoori-Boghsani<sup>1</sup> , Ali Ganjeali<sup>1</sup> , Tomislav Cernava<sup>2</sup> \*, Henry Müller<sup>2</sup> , Javad Asili<sup>3</sup> and Gabriele Berg<sup>2</sup>

<sup>1</sup> Department of Biology, Faculty of Science, Ferdowsi University of Mashhad, Mashhad, Iran, <sup>2</sup> Institute of Environmental Biotechnology, Graz University of Technology, Graz, Austria, <sup>3</sup> Department of Pharmacognosy, Faculty of Pharmacy, Mashhad University of Medical Sciences, Mashhad, Iran

Endophytic fungi are often embedded in their host's metabolic networks, which can result in alterations of metabolite production and higher amounts of active compounds in medicinal plants. This study reports the occurrence, diversity, and secondary metabolite profiles of endophytic fungi isolated from Salvia abrotanoides plants obtained from three geographically distinct sites in Iran. A total of 56 endophytic fungi were isolated from roots and leaves of S. abrotanoides; site-specificity and root-dominated colonization was found to be a general characteristic of the endophytes. Based on molecular identification, the endophytic fungi were classified into 15 genera. Mycelial extracts of these isolates were subjected to high-resolution mass spectrometry analyses and revealed a broad spectrum of secondary metabolites. Our results demonstrated that Penicillium canescens, P. murcianum, Paraphoma radicina, and Coniolariella hispanica are producers of cryptotanshinone, which is a main bioactive compound of S. abrotanoides. Moreover, it was shown that it can be produced independent of the host plant. The effect of exogenous gibberellin on S. abrotanoides and endophytic fungi was shown to have a positive effect on increasing the cryptotanshinone production in the plant as well as in endophytic fungi cultivated under axenic conditions. Our findings provide further evidence that endophytic fungi play an important role in the production plant bioactive metabolites. Moreover, they provide an exploitable basis to increase cryptotanshinone production in S. abrotanoides.

Keywords: Salvia abrotanoides, endophytic fungi, secondary metabolites, cryptotanshinone, gibberellin

### INTRODUCTION

Plants can be considered as holobionts that are embedded in multiple mutualistic networks connecting them with the environment and microbial communities of varying structure and diversity (Vandenkoornhuyse et al., 2015). Plant-microbe interactions can be very profound and versatile, especially with highly adapted endophytes (Hardoim et al., 2015). Plant endophytes can improve their host's resistance against biotic and abiotic stress by provision of various

#### Edited by:

Kalliope K. Papadopoulou, University of Thessaly, Greece

#### Reviewed by:

Carolina Chiellini, University of Pisa, Italy Michalis D. Omirou, Agricultural Research Institute (Cyprus), Cyprus

\*Correspondence: Tomislav Cernava tomislav.cernava@tugraz.at

#### Specialty section:

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

Received: 07 August 2019 Accepted: 16 December 2019 Published: 17 January 2020

#### Citation:

Teimoori-Boghsani Y, Ganjeali A, Cernava T, Müller H, Asili J and Berg G (2020) Endophytic Fungi of Native Salvia abrotanoides Plants Reveal High Taxonomic Diversity and Unique Profiles of Secondary Metabolites. Front. Microbiol. 10:3013. doi: 10.3389/fmicb.2019.03013

**Abbreviations:** GA3, gibberellin 3; HPLC-MS, high-performance liquid chromatography-mass spectrometry; ITS, internal transcribed spacer region; LC-MS, liquid chromatography-mass spectrometry; LSD, Least Significant Difference; MA, malt extract agar; PGB, potato glucose broth.

bioactive compounds (Gunatilaka, 2006). Previous studies have shown that endophytic microbial communities within medicinal plants have a great potential as producers of novel bioactive compounds and thus have a high potential for agricultural, pharmaceutical, and other applications (Köberl et al., 2013; Rai et al., 2014). Moreover, it is known that endophytes can produce distinct host plant metabolites or their precursors; e.g., taxol (Stierle et al., 1993), comptothecin (Kusari et al., 2009), azadirachtin (Kusari et al., 2012), tanshinone I/IIA (Ming et al., 2012), and maytansine (Wings et al., 2013). In addition, microbial inoculants can enhance the concentration of bioactive metabolites in medicinal plants as shown for Bacillus subtilis Co1-6 and Paenibacillus polymyxa Mc5Re-14, which enhanced apigenin-7-O-glucoside in chamomile (Schmidt et al., 2014), and Chaetomium globosum D38, which promotes bioactive constituent accumulation and root production in Salvia miltiorrhiza (Zhai et al., 2018). Despite the evident potential to improve the availability of active compounds for diverse health issues, plant-endophyte interactions and their metabolic interplay in medicinal plants are not yet fully understood.

Salvia abrotanoides (Kar.) Sytsma, which is part of the Lamiaceae family, was previously also known as Perovskia abrotanoides Kar. (Drew et al., 2017). It is a traditional medicinal plant, growing in various regions of Iran. This plant grows as a bush or semi-shrub with a height of about one meter and is propagated by seeds (Ghahreman, 1993). The roots of this rather unknown medicinal plant are mainly used for the treatment of leishmaniasis in Iranian folk medicine (Jaafari et al., 2007). There are some reports that imply leishmanicidal, antiplasmodial, anti-inflammatory, antibacterial, and cytotoxic pharmacological effects of S. abrotanoides (Hosseinzadeh and Amel, 2001). These effects are attributed to the presence of tanshinones as the most important and most abundant bioactive compounds in the roots of this plant (Sairafianpour et al., 2001). Although not all relevant biosynthetic pathways have been explored in detail, some common key enzymes were previously described (Hedden et al., 2001). Tanshinones are abietane-type norditerpenoid quinones that were first identified in 1930s from the roots of Salvia miltiorrhiza (Nakao and Fukushima, 1934). For this compound group, diverse pharmacological activities such as anticancer (Hu et al., 2015), antidiabetes (Kim et al., 2007), cardioprotective effects (Fu et al., 2007) and neuro-protective activity (Yu et al., 2007) have been reported. Moreover, cryptotanshinone as a prominent member of tanshinones is known for its antibacterial activity (Cha et al., 2013) and strong anticancer properties (Hu et al., 2015; Li et al., 2015; Wu et al., 2016). The effectiveness and potential usefulness of tanshinones led to a number of studies with the aim of increasing their concentration in planta by different approaches. Some of these studies explored the effects of biotic and abiotic elicitors on improvement of the accumulation of tanshinons in plants (Hao et al., 2015; Zaker et al., 2015). Recently, the implementation of endophytic microorganisms in order to discover novel, biologically active compounds was expanded (Bedi et al., 2018; Sharma et al., 2018). However, nothing is known about the occurrence, diversity and secondary metabolite profiles of endophytic fungi in Salvia abrotanoides and microbial producers of tanshinones.

In the present study, 56 endophytic fungal strains were isolated from native S. abrotanoides plants grown in different arid areas in Northern Iran to specifically screen for potential producers of tanshinones. In addition, an untargeted metabolite profiling approach was implemented in order to characterize isolates that produce a high diversity of secondary metabolites. Such isolates can serve as a valuable bioresource in the future to increases the concentration of distinct compounds in planta or in biotechnological applications. In a complementary approach, we explored possibilities to improve the accumulation of tanshinones in planta and the development of cultivation methods for fungi with the same aim. We expected that addition of end products from interconnected biosynthetic pathways might have favorable effects on the production of cryptotanshinone due to regulation mechanisms. A positive effect of gibberellin supplementation on the cryptotanshinone biosynthesis in S. abrotanoides and in different endophytic fungi was discovered and therefore studied in more detail.

### MATERIALS AND METHODS

#### Sample Collection and Isolation of Endophytic Fungi

Endophytic fungi were isolated from the roots of Salvia abrotanoides (Kar.) in flowering stage. Plant sampling was conducted at three different locations in the northeast of Iran (Zoshk, N 36◦ 160 5800, E 59◦ 070 0600, Kalat, N 36◦ 350 0700, E 59◦ 520 1200, Darrud, N 36◦ 100 5700, E 59◦ 100 1200), and a voucher specimen was deposited with the herbarium of Ferdowsi University of Mashhad, Iran, under voucher code 36763 (FUMH). The samples were kept at 4◦C and the isolation of fungal endophytes was conducted within 24 h after sample collection. For fungal isolation, all plant tissues were surfacesterilized using the procedure described by Fisher et al. (1993). The roots of plants were washed under running water and cut to 15 mm segments. Then the root segments were surface-sterilized by sequential immersion in 70% ethanol for 1 min, sterilized water for 1 min, 2.5% sodium hypochlorite for 3 min, sterilized distilled water for 1 min, and 70% ethanol for 30 s. The roots were then rinsed three times in sterilized water for 1 min to remove the remaining chemicals from their surface. Sterilization of leaves was conducted with analogous protocol as for the roots, where the sodium hypochlorite solution was diluted three times and the immersion time was reduced to 1 min. The effectiveness of the sterilization process, which resulted in the elimination of all epiphytic microorganisms was confirmed using aliquots of sterile, distilled water from the last rinse. Aliquots from all samples were inoculated on culture media in Petri dishes and checked for microbial growth. Two root segments were then evenly placed in Malt extract agar plates (30 g malt extract, 3 g soya peptone, 15 g agar, 1000 ml deionized water) augmented with 100 mg l−<sup>1</sup> Streptomycin to avoid bacterial contamination. The Petri dishes were sealed with Parafilm (PM-996) and incubated at 25 ± 2 ◦C in an incubator until the fungal colonies emerged from root sections. Hyphal tips of fungal colonies were transferred to new

Petri dishes with malt extract agar (MA) to obtain pure cultures of the fungal isolates.

#### DNA Extraction, Amplification of the ITS Region, and Sequencing

A molecular approach was implemented to identify the fungal isolates. The endophytic fungi were grown on fresh MA medium for 5–7 days. Mycelium of each isolate was then transferred into a tube with glass beads (250 mg of beads with a diameter of 0.25–50 mm and two beads of 2.85–3.45 mm) and 450 µl of DNA extraction buffer (200 mM Tris–HCl, 250 mM NaCl and 0.5% SDS). The mechanical disruption of mycelia was performed by shaking 2 × 30 sec in a FastPrep intstrument (MP Biomedicals, Solon, OH, United States). Subsequently, fungal DNA was extracted with the phenol/chloroform method. In the final step the ethanol was decanted and the DNA-containing pellet was dried and resuspended in 50 µl of nuclease-free H2O. DNA quality and quantity were checked by spectrophotometry using a UV-Vis spectrophotometer (NanoDrop 2000c; Thermo Fisher Scientific, Waltham, MA, United States) and stored at −20◦C until further processing. PCR amplification with genomic DNA from each isolate was performed in a 30 µl of PCR reaction mix with 1.8 µl MgCl<sup>2</sup> (25 mM), 6 µl Taq-&Go, 0.6 µl of ITS1f primer (CTT GGT CAT TTA GAG GAA GTA A), 0.6 µl of ITS4r primer (TCC TCG GCT TAT TGA TAT GC), 20 µl PCR grade water, and 1 µl of the DNA template (White et al., 1990). The amplification was conducted with an initial denaturation at 95◦C for 5 min, followed by 36 cycles of 95◦C for 30 s, 54◦C for 35 s, and 72◦C for 40 s with a final extension at 72◦C for 10 min using the TPersonal Combi, Biometra Thermocycler (Biometra GmbH, Germany). PCR products were purified using the Wizard SV Gel and PCR Clean-Up System (Promega, Madison, WI, United States) and quantified on a Nanodrop 2000c spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States). Subsequently, 14 µl of 20 ng µl −1 PCR product including one specific primer (ITS1f) was sent to LGC Genomics (Berlin, Germany) for sequencing. Sequences were identified using the BLAST algorithm against the NCBI Targeted Loci Nucleotide BLAST – Internal transcribed spacer region (ITS) database and were deposited in GenBank with the accession numbers MK367721- MK367776.

#### Extraction of Secondary Metabolites From Fungal Isolates

Fungal mycelial plugs were used to inoculate 100 ml potato extract glucose broth (Carl Roth GmbH, Germany) in 250-ml Erlenmeyer flasks. The flasks were incubated in the dark at 30◦C for 24 days with rotary shaking at 105 rpm. The fungal cultures were vacuum filtered through filter paper (Rotilabo <sup>R</sup> round filter, type 112 A, 47 mm, Carl Roth GmbH, Germany) to remove the biomass. The mycelia were dried in an oven at 40◦C to obtain the dry weight. The dried mycelia were then homogenized by mortar and pestle, suspended in 0.5 ml ethyl acetate and mechanically disrupted with glass beads (three beads with a diameter of 2.85–3.45 mm and 250 mg of 0.25–0.5 mm) for 2 × 40 s at 6 m/s in a FastPrep instrument (MP Biomedicals, Solon, OH, United States). Precooled ethyl acetate and glass beads at −70◦C were used for reproducible extraction and to avoid further degradation of metabolites. The homogenate was centrifuged for 10 min, 13000 rpm at 4◦C. Subsequently, 250 µl of each supernatant were collected and stored at 4◦C before the HPLC-MS analysis was conducted.

#### Detection of Bioactive Compounds by HPLC-MS Analysis

Extracts of fungal isolates were analyzed with a combined HPLChybrid quadrupole-orbitrap mass spectrometer (Q Exactive; Thermo Scientific, Bremen, Germany). Chromatographic separation of the extracts was performed on an Atlantis dc 18 µm, 2.1 × 100 mm column (Waters Corporation, United States) using H2O with 0.1% formic acid as solvent A and acetonitrile with 0.1% formic acid as solvent B with the following gradient elution program: at 0 min, the program started with 5% B and was increased to 40% B at 2 min, and subsequently to 100% B at 15 min, which was kept until the end of each run. The run time was 40 min at a solvent flow rate of 0.3 mL/min and a sample injection volume of 20 µl. Mass spectrometric detection was carried out in positive and negative mode using an electrospray ionization (ESI) source. The ESI conditions were set to 3.1 kV spray voltage and 330◦C capillary temperature. Scans were recorded in the range 100.0–1500 m/z with the AGC target set to 1 × 10<sup>6</sup> and maximal accumulation time of 200 ms. The resolution was adjusted to 70,000. Altering full MS-SIM and targeted MS<sup>2</sup> cycles were employed and a specific inclusion mass of 297.14807 amu was selected. Standard calibration was obtained with 0.9, 1.8, 9.0, 13.5, 18, and 27 µg cryptotanshinone standard (Sigma-Aldrich, United States). The obtained mass spectra were analyzed with Compound Discoverer 2.0 (Thermo Scientific) and the integrated mzCloud database<sup>1</sup> to detect secondary metabolites in an untargeted approach. For the initial analysis, the implemented HighChem HighRes algorithm was selected and the identity of the automatically detected compounds was confirmed by comparing their spectra with reference data available in the software package. Fungal isolates that produced either cryptotanshinone and/or ≥10 secondary metabolites that were detectable with the conducted experiment setup, were considered for further data visualizations and interpretations. For isolates assigned to the same fungal species, but originating from different isolation sites, the producer of the most diverse secondary metabolite profile was selected as the representative strain.

#### Plant Material, Growth Conditions, and Gibberellin Treatments

Seeds of S. abrotanoides used in this experiment were collected in the Zoshk area, Iran. The seeds were sown in pots filled with a mixture of sandy soil, vermiculite and compost in a proportion of 3:1:1, respectively. The plants grown in a growth chamber (approximately 22/20◦C day/night temperature, 55% relative humidity and a 15 h photoperiod) located at Graz University

<sup>1</sup>https://www.mzcloud.org

of Technology (Graz, Austria). A GA<sup>3</sup> (Duchefa Biochemie BV, Haarlem, Netherlands) solution was applied three times with a weekly interval as foliar spray to the plants after 87 days, at concentrations of 0, 50, 100, and 150 mg l−<sup>1</sup> in sterile water. All treatments were performed with low-pressure hand-wand sprayers and 5 ml for each pot. Sterile water without GA<sup>3</sup> was implemented as a control. Following the growth period, the roots of plants grown in the same pots were separately dried in an oven at 35◦C. Each of the treatments was conducted with three biological replicates. The roots were homogenized and subjected to extractions with the abovementioned method that was also used to extract metabolites from fungal samples.

#### Cultivation of Selected Endophytic Fungi in Combination With Gibberellin Treatments

The isolates which were shown to produce cryptotanshinone were selected for gibberellin treatments in order to assess stimulatory effects. The isolates were cultivated in PGB medium and treated with 5 ml filtered (0.45 µm syringe filter) GA<sup>3</sup> at a concentration of 50 mg l−<sup>1</sup> in Erlenmeyer flasks (95 ml PGB and 5 ml GA3). Flasks including the fungus and 100 ml PGB medium were included as a non-treated control for each isolate. The flasks were kept in an incubation room at 21◦C for 24 days shaking on a rotary shaker at 105 rpm. Each experiment was conducted with three biological replicates. The extraction of cryptotanshinone was conducted with the aforementioned method.

#### Statistical Analysis

Statistical analyses were performed with SPSS v.20.0.0 (SPSS Inc., Chicago, IL, United States). Analysis of variance for plant treatments with gibberellin was performed with one-way ANOVA and the significance of the results was assessed with the Duncan post hoc test. The Univariate General Linear model and LSD test were implemented to assess the significance of the effects of gibberellin treatments on cryptotanshinone production in each fungal isolate. P values <0.05 were considered to be significant. All experiments were performed in triplicates (n = 3) and the results were reported as means ± standard error.

#### RESULTS

#### Isolation and Identification of Endophytic Fungi

A total of 56 endophytic fungi were isolated from leave and root segments of Salvia abrotanoides at the three sites. Of these, only two isolates were recovered from the plant's leaves (Thielavia microspore and Aspergillus sp.), while the remaining isolates were obtained from root samples. All isolates were grouped based on the site of sampling, and included the fungal genera Penicillium, Paraphoma, Phaeoacremonium, Talaromyces, Aspergillus, Psathyrella, Trichoderma, Alternaria, Thielavia, and Acremonium originating from Zoshk, Fusarium, Talaromyces, Penicillium, and Coniolariella from Kalat and Fusarium, Paecilomyces, Simplicillium, and Monocillium from TABLE 1 | Overall occurrence of Salvia abrotanoides-colonizing fungi in plant samples collected in the Zoshk, Kalat, and Darrud areas in Iran.


The presence of a distinct isolate is indicated with "+," while "−" indicates the absence of a isolate in the specified areas.

Darrud (**Table 1**). Several of the isolates were represented by different species within the same genus. Penicillium was represented by four species including Penicillium canescens, P. chrysogenum, P. charlesii and P. murcianum. Moreover, one of the Penicillium isolates was not identifiable at the species level by the utilized molecular approaches. Talaromyces was represented by two different species including Talaromyces verruculosus and another species that was also not identified at the species level by molecular approach. In addition, the genus Fusarium was represented by two species including Fusarium dlaminii and Fusarium solani. The remaining isolates were represented by one species for each genus including Paraphoma radicina, Coniolariella hispanica, Phaeoacremonium rubrigenum, Aspergillussp., Psathyrella candolleana, Trichoderma asperellum, Alternaria chlamydosporigena, Thielavia microspore, Acremonium sclerotigenum, Paecilomyces lilacinus, Monocillium ligusticum, and Simplicillium cylindrosporum.

#### Secondary Metabolite Profiles

Phytochemical screening of fungus-derived ethyl acetatic extracts showed high chemical diversity of various secondary metabolites. Among the isolates, the genera Penicillium, Talaromyces, Fusarium, Paraphoma and Coniolariella produced the highest diversity of compounds, including terpens, fatty acid amids, vitamins, dicarboxylic acids, isoflavons, ketons, alcohols, phenols, lipids, alkaloids, catecholamins, and polyketides. The results of the qualitative assessment of phytochemical profiles of two prevalent fungal genera are shown in **Table 2**. TABLE 2 | Secondary metaboliteprofiles of endophytic fungal isolates assigned to Penicilliumspp. andTalaromycesspp. that were identified in the cultivation medium.


(Continued) fmicb-10-03013 January 8, 2020 Time: 18:33 # 5

Teimoori-Boghsani et al.

Secondary Metabolites of

S. abrotanoides

Endophytes

#### TABLE 2 | Continued


Squares indicate the excretion of specific compounds by distinct isolates. Biological functions were derived from available literature. They indicated that different species of Penicillium produce different profiles of secondary metabolites. suberic acid and pantothenic acid were the only two compounds that were produced by all Penicillium isolates (**Table 2**). Compounds that were produced by four Penicillium isolates included pyridoxine (P. canescens, P. chrysogenum, P. charlesii and P. murcianum), hexadecanamide (P. canescens, P. chrysogenum, P. charlesii and Penicillium sp.) and nicotinic acid (P. canescens, P. charlesii, P. murcianum and Penicillium sp.). In addition, azelaic acid and arabitol were were produced by three distinct Penicillium strains (P. canescens, P. charlesii, and Penicillium sp.). Compounds that were produced by two Penicillium isolates included cryptotanshinone (P. canescens and P. murcianum), manitol (P. canescens and P. charlesii), glutaric acid (P. canescens and P. murcianum), monoolein (P. canescens and P. chrysogenum) and paracetamol (P. chrysogenum and Penicillium sp.). The results revealed that indol-3-acetic acid, daidzein and nipecotic acid were produced by P. canescens, as well as itaconic acid and N-acetylanthranilic acid that were identified for P. murcianum and Penicillium sp., respectively. Talaromyces spp. produced compounds from the same chemical groups as Penicillium spp. (**Table 2**). Phenenthylamin, solanidine, and trigonelline which are all alkaloids, caffeic acid from the phenols group, grisoeofulvin (polyketides group), glycitein (isoflavones group), N-acetyldopamine (catecholamines group), acetophenone (ketones group), mevalonolactone (terpens group), and xylitol from the alcohols group were secondary metabolites produced by the Talaromyces isolates that differed from the profiles of Penicillium species. Two species within the Fusarium genus obtained from two different sampling sites showed substantially different profiles (**Supplementary Figure S1A**). Only Fusarium dlaminii from Darrud area produced stachydrine (alkaloid) that was specific for this species among other isolates. Except of mandelic acid, the remaining compounds produced by Paraphoma radicina were also common within the other isolates (**Supplementary Figure S1B**). Metabolic profiles of Coniolariella hispanica indicated that it produced secondary metabolites with less diversity; nevertheless cryptotanshinone was detected in the diterpenes group (**Supplementary Figure S1C**). Various isolates assigned to the genera Penicillium and Talaromyces were isolated from the sampling sites Kalat and Zoshk und thus subjected to a complementary comparison of secondary metabolite profiles. Isolates belonging to the same species shared 25–57% of the identified secondary metabolites in case of Penicillium and 23 – 40% in case of Talaromyces (**Supplementary Table S1**).

Moreover, LC-MS analyses showed that Penicillium canescens, P. murcianum, Paraphoma radicina, and Coniolariella hispanica can produce cryptotanshinone independent of the host plant (**Figure 1** and **Supplementary Figures S2, S3**). Analysis of the secondary metabolite profile of the host plant showed that azelaic acid and suberic acid (data not shown), which are dicarboxylic acids as well as cryptotanshinone were present in host plant extracts. The overall results of the phytochemical analysis indicated that endophytic fungi from S. abrotanoides can produce several phytochemicals that are either identical or structurally similar to those of the host

plant, as well as new bioactive compounds that are not present in the host plant.

#### Effects of Exogenous GA<sup>3</sup> on Cryptotanshinone Production in S. abrotanoides and the Isolated Fungi

In order to explore the effect of GA<sup>3</sup> supplementation on cryptotanshinone biosynthesis in S. abrotanoides, exogenous GA<sup>3</sup> was supplemented in three different concentrations. Spray application of sterile water without GA<sup>3</sup> was implemented as a control for comparative assessments. Analytical quantifications showed that the GA3-treated plants responded with a significant increase in cryptotanshinone biosynthesis in comparison to the control (**Figure 2**). When 50 mg l−<sup>1</sup> GA<sup>3</sup> were sprayed on S. abrotanoides plants, the cryptotanshinone concentration in the plant increased to 1.60 ± 0.06 mg g−<sup>1</sup> . In the 100 mg l−<sup>1</sup> GA<sup>3</sup> treatment, the cryptotanshinone concentration increased to 1.71 ± 0.13 mg g−<sup>1</sup> , while at the highest applied concentration of 150 mg l−<sup>1</sup> , the concentration increased to 1.87 ± 0.19 mg g−<sup>1</sup> .

In addition to the exogenous hormonal applications to S. abrotanoides plants, the effect of 50 mg l−<sup>1</sup> GA<sup>3</sup> on each fungus producing cryptotanshinone was investigated. Analytical quantifications showed an increase in the amount of cryptotanshinone production by the fungi after treatment with GA<sup>3</sup> (**Figure 3**). When GA<sup>3</sup> at a concentration of 50 mg l−<sup>1</sup> was supplemented to Paraphoma radicina cultures, the cryptotanshinone concentration showed a significant increase from 0.37 ± 0.02 mg g−<sup>1</sup> in the control to 1.09 ± 0.29 mg g−<sup>1</sup> for the GA<sup>3</sup> treatment. Moreover, when GA<sup>3</sup> was supplemented at a concentration of 50 mg l−<sup>1</sup> to different fungal isolates, again an increase in cryptotanshinone production was observed, however, the significance was not confirmed with statistical analyses. For Penicillium murcianum, Coniolariella hispanica and Penicillium canescens, the cryptotanshinone concentration increased to 0.86 ± 0.2 mg g−<sup>1</sup> (0.51 ± 0.008 mg g−<sup>1</sup> in the

control), 0.23 ± 0.04 mg g−<sup>1</sup> (0.21 ± 0.01 mg g−<sup>1</sup> in the control), and 0.31 ± 0.12 mg g−<sup>1</sup> (0.19 ± 0.01 mg g−<sup>1</sup> in the control), respectively.

#### DISCUSSION

In the present study, different arid areas in Northern Iran were selected to investigate secondary metabolite profiles of endophytic fungi isolated from indigenous S. abrotanoides plants. All Salvia plants harbored taxonomically distinct fungal endophytes, especially in their roots, that produced an unexpectedly broad spectrum of secondary metabolites. The results revealed isolate-specific secondary metabolite profiles, e.g., illustrated by distinct spectra obtained with Penicillium isolates. We identified a broad range of well-studied fungal secondary metabolites as well as such that were not yet described in fungi, e.g., cryptotanshinone a major bioactive diterpenoid previously isolated from Salvia species.

The untargeted profiling approach of secondary metabolites in the isolated plant endophytes resulted in the detection of a broad spectrum of compounds from various chemical groups. Compounds including itaconic acid and azelaic acid that were detected in Penicillium and Talaromyces isolates, were previously also detected in isolates assigned to the genus Aspergillus and connected to plant-microbe interactions (Okabe et al., 2009; Pineda et al., 2010). In the present study, only a low diversity of secondary metabolites were detected in the Aspergillus isolate (data not shown) and none of the two organic acids was present. Itaconic acid was shown to have a high potential as a biochemical building block for the production of synthetic resins, synthetic fibers, plastics, rubbers, surfactants, and oil additives (Okabe et al., 2009). Azelaic acid has profound anti-inflammatory, antioxidative effects, and is bactericidal against a range of Gramnegative and Gram-positive microorganisms as well, including antibiotic-resistant bacterial strains (Pineda et al., 2010). Most of the isolates produced a variety of sugar alcohols. Sugar alcohols including arabitol and mannitol that are considered as tracers for

the quantification of airborne fungal spores (Bauer et al., 2008), improve survival of fungi under drought conditions, and shelf life of encapsuled Metarhizium brunneum coupled with enhanced fungal virulence (Krell et al., 2018). Penicillium canescens was shown to produce nipecotic acid, a piperidinemonocarboxylic acid that belongs to beta-amino acids. Nipecotic acid is one of the most potent inhibitors of neuronal and glial γ-aminobutyric acid (GABA) and thus relevant for medicinal applications (Bonina et al., 1999). Penicillium canescens, two species of Fusarium, and Paraphoma radicina produced daidzein, a natural isoflavone known from the Leguminosae plant family. This metabolite has been shown to elicit myoblast differentiation and myotube growth (Lee et al., 2017). Daidzein was previously found in Trichoderma sp., an endophytic fungus of Azadirachta indica (Xuan et al., 2014). Although Trichoderma spp. are common endophytes in plants, only one isolate was recovered from Salvia abrotanoides in the present study and diadezein was not found in its secondary metabolite profile (data not shown). Furthermore, Penicillium canescens, Penicillium chrysogenum, Talaromyces sp. and Paraphoma radicina were shown to produce monoolein which is considered one of the most important lipids in the fields of drug delivery, emulsion stabilization and protein crystallization (Kulkarni et al., 2011; Esposito et al., 2018). In addition, a broad range of vitamins was identified in the metabolite profiles of the endophytic fungi. Vitamins including nicotinic acid, pyridoxine and pantothenic acid were previously reported in Fusarium proliferatum and Cercospora nicotianae (Harvais and Pekkala, 1975; Wetzel et al., 2004; Jin et al., 2013). This is in accordance with the findings in the present study, where Fusarium isolates were found to produce three different vitamins or precursors thereof. Paracetamol (acetaminophen) was only detected in Penicillium chrysogenum and Penicillium sp. metabolite profiles. This compound is one of the most popular and most commonly used analgesic and antipyretic drugs around the world, which also had antibacterial and antifungal properties (Refat et al., 2017). Different reports show that distinct microorganisms can utilize paracetamol as a carbon and energy source (Wu et al., 2012), however, fungal biosynthesis of this compound was not described so far. The degradation capacity indicates that it might be commonly found in natural environments, where fungi could be involved in its biosynthesis. Deepening analyses are required to clarify if distinct members of the genus Penicillium genus produce the bioactive molecule or a structural analog thereof. Indole-3-acetic acid as a plant hormone was identified in metabolite profiles of a broad range of fungal genera that were subjected to metabolic profiling in the present study. This plant hormone was previously found in a broad range of fungi and bacteria (Ek et al., 1983; Patten and Glick, 1996). Talaromyces spp. profiles also showed other chemically diverse compounds including trigonelline, solanidine and phenethylamine, they are used in treatments of diabetes, cancer and Alzheimer diseases, respectively (Sterling et al., 2002; Zhou et al., 2012; Zupkó et al., 2014). Talaromyces sp. and Paraphoma radicina produced mandelic acid, which is a monocarboxylic acid. Mandelic acid has been used as an antibacterial agent, particularly in the treatment of urinary tract infections (Putten, 1979). Recently, mandelic acid production was achieved by microbial fermentations using engineered Escherichia coli and Saccharomyces cerevisiae expressing heterologous hydroxymandelate synthases (Reifenrath et al., 2018). Moreover, caffeic acid was found in metabolite profiles of isolates assigned to the genera Talaromyces and Paraphoma. This well-known phenol has been previously found in endophytic fungi Cladosporium velox (Singh et al., 2016), Penicillium canescens and Fusarium chlamydosporum (Das et al., 2018). Antibacterial, antifungal and modulatory effects of caffeic acid have been shown in previous studies (Lima et al., 2016). Talaromyces sp. was the only isolate that produced griseofulvin that is a polyketide. Griseofulvin is an antifungal antibiotic widely used for the treatment of human and animal dermatophytic infections (Gull and Trinci, 1973). This compound has been previously detected in endophytic fungi including Penicillium griseofulvum (Zhang et al., 2017), Nigrospora sp. (Zhao et al., 2012), and Xylaria sp. (Park et al., 2005). Our results indicate that fungal isolates from S. abrotanoides, but different geographic isolation sources can have differing secondary metabolite profiles even if they belong to the same species. In this context, it must be taken into account that these isolates would likely show genetic differences when assessed at the strain level. Previous studies that addressed other plant-endophyte systems have shown that geographic as well as seasonal differences influence endophytic communities (Collado et al., 1999; Christian et al., 2016).

In general, the plant-endophytic isolates were shown to produce a broad variety of exploitable metabolites, however, one particularly important compound, that was found to be also synthetized during axenic fermentation approaches in the present study, was cryptotanshinone. Previous studies showed that cryptotanshinone has antibacterial activity (Lee et al., 1999), inhibits angiogenesis (Hur et al., 2005) and inhibits STAT3 in prostate cancer (Shin et al., 2009). Moreover, it was previously reported that cryptotanshinone can be used for the treatment of coronary heart disease (Yu et al., 2009) and diabetes (Kim et al., 2007). These promising medicinal properties, as well as the general demand for bioactive compounds that are produced under controlled settings, led us to explore possibilities to increase cryptotanshinone production in S. abrotanoides and its endophytic fungi. Previous studies have shown that there is an overlap between the tanshinone and the gibberellin biosynthesis pathways (Su et al., 2016). Detailed assessments showed two distinctive GA biosynthesis pathways are present in plants and fungi that have the copalyl diphosphate synthase and kaurene synthase enzymes in common (Hedden et al., 2001). Copalyl diphosphate synthase and kaurene synthase are also the main enzymes in the tanshinone biosynthetic pathway (Gao et al., 2014). We expected that supplementation of one end product might regulate conversion pathways in favor of the other compound. Therefore, we explored the potential of exogenous gibberellin supplementation to increase cryptotanshinone production by the plant and isolated endophytic fungi. The present study demonstrated that GA<sup>3</sup> significantly increases the amount of cryptotanshinone in S. abrotanoides. This finding is in consensus with reports that have shown gibberellins have been effective in increasing the levels of tanshinones in Salvia miltiorrhiza (Yuan et al., 2008).

Moreover, our finding showed an increase in cryptotanshinone production in different endophytic fungi when compared when they were cultivated under laboratory conditions. The overall findings provide an exploitable basis for the cultivation of S. abrotanoides plants with higher concentrations of its bioactive compounds and for biotechnological applications based on its endophytes.

#### DATA AVAILABILITY STATEMENT

The datasets generated for this study can be found in the GenBank under accession numbers MK367721–MK367776.

#### AUTHOR CONTRIBUTIONS

JA, AG, and YT-B designed the study. YT-B carried out the plant collection and fungi isolation under the supervision of AG. YT-B carried out the molecular identification of fungi, LC-MS, secondary metabolite profiles, GA<sup>3</sup> experiments, and analyzed the LC-MS data. HM and TC supervised the molecular identification and GA<sup>3</sup> experiments. AG subjected the GA<sup>3</sup> experiments data to statistical analyses. YT-B and TC wrote the

#### REFERENCES


final version of the manuscript. GB reviewed the final version of the manuscript. All authors read and approved the final version of the manuscript.

#### FUNDING

Financial support by the Department of Environmental Biotechnology, Graz University of Technology, Graz, is gratefully acknowledged. This work was supported in part by research grant (No. 3/44714) funded by Ferdowsi University of Mashhad.

#### ACKNOWLEDGMENTS

We appreciate the help of Angelika Schäfer (Graz) for her support during preparation and realization of LC-MS experiments.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.03013/full#supplementary-material


leishmanicidal assay using promastigotes of Leishmania major. Pharmacol. Online 1, 299–303.


and Sr (II) complexes with paracetamol drug. Arab. J. Chem. 10, S2376–S2387. doi: 10.1016/j.arabjc.2013.08.017


neuro-protective activity in experimentally stroked rats. Neurosci. Lett. 417, 261–265. doi: 10.1016/j.neulet.2007.02.079


**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 Teimoori-Boghsani, Ganjeali, Cernava, Müller, Asili and Berg. 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.

# Culture-Dependent and Culture-Independent Characterization of the Olive Xylem Microbiota: Effect of Sap Extraction Methods

Manuel Anguita-Maeso<sup>1</sup> , Concepción Olivares-García<sup>1</sup> , Carmen Haro<sup>1</sup> , Juan Imperial <sup>2</sup> , Juan A. Navas-Cortés <sup>1</sup> and Blanca B. Landa1\*

<sup>1</sup> Institute for Sustainable Agriculture, Spanish National Research Council (CSIC), Córdoba, Spain, <sup>2</sup> Institute of Agricultural Sciences, Spanish National Research Council (CSIC), Madrid, Spain

#### Edited by:

Donald L. Hopkins, University of Florida, United States

#### Reviewed by:

Philippe Eric Rolshausen, University of California, Riverside, United States David Gramaje, Institute of Vine and Wine Sciences (ICVV), Spain

> \*Correspondence: Blanca B. Landa blanca.landa@csic.es

#### Specialty section:

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

Received: 26 September 2019 Accepted: 04 December 2019 Published: 21 January 2020

#### Citation:

Anguita-Maeso M, Olivares-García C, Haro C, Imperial J, Navas-Cortés JA and Landa BB (2020) Culture-Dependent and Culture-Independent Characterization of the Olive Xylem Microbiota: Effect of Sap Extraction Methods. Front. Plant Sci. 10:1708. doi: 10.3389/fpls.2019.01708 Microbial endophytes are well known to protect host plants against pathogens, thus representing a promising strategy for the control of xylem-colonizing pathogens. To date, the vast majority of microbial communities inhabiting the olive xylem are unknown; therefore, this work pursues the characterization of the xylem-limited microbiome and determines whether the culture isolation medium, olive genotype, and the plant material used to analyze it can have an effect on the bacterial populations retrieved. Macerated xylem tissue and xylem sap extracted with the Scholander chamber from olive branches obtained from two cultivated and a wild olive genotypes were analyzed using culturedependent and -independent approaches. In the culture-dependent approach using four solid culture media, a total of 261 bacterial isolates were identified after performing Sanger sequencing of 16S rRNA. Culturable bacteria clustered into 34 genera, with some effect of culture media for bacterial isolation. The cultivated bacteria belonged to four phyla and the most abundant genera included Frigoribacterium (18.8%), Methylobacterium (16.4%), and Sphingomonas (14.6%). On the other hand, in the culture-independent approach conducted using Illumina MiSeq 16S rRNA amplicon sequencing [next-generation sequencing (NGS)] of the xylem extracts, we identified a total of 48 operational taxonomic units (OTUs) belonging to five phyla, being Sphingomonas (30.1%), Hymenobacter (24.1%) and Methylobacterium (22.4%) the most representative genera (>76% of reads). In addition, the results indicated significant differences in the bacterial communities detected in the xylem sap depending on the genotype of the olive tree studied and, to a minor extent, on the type of sap extraction method used. Among the total genera identified using NGS, 14 (41.2%) were recovered in the culture collection, whereas 20 (58.8%) in the culture collection were not captured by the NGS approach. Some of the xylem-inhabiting bacteria isolated are known biocontrol agents of plant pathogens, whereas for others little information is known and are first reported for olive. Consequently, the potential role of these bacteria in conferring olive tree protection against xylem pathogens should be explored in future research.

Keywords: microbiome, xylem, culture, next-generation sequencing, vascular pathogens

## INTRODUCTION

Olive tree (Olea europaea L.) is one of the oldest cultivated trees and has been part of traditional Mediterranean agriculture ever since Roman times. Totaling over 2,000 cultivars (Lavee, 1990), the culture spread from Asia along Syria, Iran, and Palestine to the rest of the Mediterranean basin ca. 6,000 years ago. Since then, olive oil has been a food staple of Mediterranean countries (Müller et al., 2015). Olive oil is considered a high-quality food with multiple beneficial effects for human health, mainly due to its high content in polyphenols, which has induced an increase in olive oil demand and trade worldwide (Landa et al., 2019). Nowadays, world olive cultivation is estimated to cover 10.8 million ha, which results in an olive oil production of about 3.05 million tons, of which ca. 9.5 million ha of olives are grown in the Mediterranean region, accounting for 95% of the cultivated olive area worldwide (i.e., about 98% of the olive oil and 80% of table olive production are form Mediterranean countries) (IOC, www.internationaloliveoil. org/; FAOSTATS, http://www.fao.org/faostat/). Olive crop is of great value within the Mediterranean Basin, being critical for the sustainment and maintenance of Mediterranean ecosystems as an integral part of their landscape. Moreover, olive crop is particularly well adapted to cultivation in less accessible areas, including mountain slopes and hillsides, where it helps control soil erosion by reducing the surface runoff of soil and increases soil fertility and nutrient retention (Gómez et al., 2014). Finally, as a major landscape player, it contributes to the establishment of ecological niches for different organisms, thus helping in the maintenance of biodiversity (Rey, 2011). Consequently, since olive is a "high natural value" agricultural system with relevant economic and environmental roles, representing an important part of the heritage and sociocultural life across the Mediterranean, its cultivation should be maintained and preserved.

Nowadays, the health of the olive groves is being seriously threatened, as consequence of a notable increase, both in extent and in severity, of diseases caused by various pathogens, which are capable of adversely affect their growth and production. Among olive diseases, those caused by the vascular plant pathogenic bacterium Xylella fastidiosa (specifically from subspecies multiplex and pauca) and the soilborne vascular fungus Verticillium dahliae are, without a doubt, global threats for olive production worldwide (Jiménez-Díaz et al., 2011; Saponari et al., 2018; Almeida et al., 2019; Landa et al., 2019; Landa et al., 2020). Xylem vessels are considered ideal niches for microbial endophytes (bacteria or fungi; both beneficial or pathogenic) by providing an effective internal pathway for dispersion throughout the plant and a continuous source of nutrients (McCully, 2001). However, despite significant progress in research on plant microbiota over past years, only a few number of publications have revealed the nature and role of the xylem microbiome, and its relationship to plant health and crop productivity (e.g., Deyett et al., 2017; Fausto et al., 2018; Deyett and Rolshausen, 2019). This fact may be due to microbiome complexity, technical difficulties in the isolation of the xylem-inhabiting microorganisms, as well as to the myriad of biotic and abiotic factors that may affect and determine the microbial community composition and their interaction within the host plant (Vandenkoornhuyse et al., 2015).

With olive, most microbiome studies have focused on determining the microbial composition of the rhizosphere (Mercado-Blanco et al., 2004; Berg and Hallmann, 2006; Mendes et al., 2007; Aranda et al., 2011; Prieto et al., 2011; Montes-Borrego et al., 2013; Caliz et al., 2015; Gómez-Lama Cabanás et al., 2018), whereas only a few studies have focused on the endosphere or xylem microbiomes (Müller et al., 2015; Fausto et al., 2018; Sofo et al., 2019). Unraveling the microbiome of the olive xylem sap should provide a better understanding of the microbes that systemically move throughout the plant or are horizontally transmitted from plant to plant during vegetative propagation. Combined use of culture-dependent and culture-independent approaches such as next-generation sequencing (NGS) technologies may provide a better characterization of the xylem microbiome composition (Turner et al., 2013; Berg et al., 2014) than that obtained when using each approach independently. An essential, parallel step would be to determine the proportion of the microbiome that can be easily isolated and cultured, in order to be subsequently tested and potentially exploited as biocontrol agents of the xylem-inhabiting plant pathogens, such as X. fastidiosa or V. dahliae. These resident microorganisms could act through direct inhibition or through niche displacement of those pathogens, as commensal and symbiotic organisms that can support the olive immune system, and/or as plant growth promoting agents.

Our research was designed to address this gap in the knowledge by characterizing the bacterial taxa that shape the olive xylem sap microbial communities by using culturedependent and culture-independent approaches. We also determined in which extent the culture isolation media and the type of plant extract or olive genotype used to isolate the xylem sap bacterial microbiome have an effect on the characterization of its composition, diversity, and structure. Knowledge of the effect of those factors may be essential to identify, isolate, and establish a bacterial collection that might be used subsequently as biocontrol agents against xylem-inhabiting plant pathogens.

### MATERIAL AND METHODS

### Sampling of Olive Trees

Branches (ca. 35 cm long, 2 years old) from 8-year-old cultivated olive (O. europaea var. europaea) trees of "Picual" and "Arbequina" cultivars were sampled from an experimental rainfed field located at the Institute for Sustainable Agriculture from the Spanish National Research Council (IAS-CSIC) in Córdoba (Southern Spain). Trees of both varieties have been submitted to the same agricultural practices throughout the years. Cv. Picual has been found to be highly susceptible to Defoliating (D) and susceptible to Nondefoliating (ND) V. dahliae isolates, respectively, while cv. Arbequina has been shown to be susceptible to D and moderately resistant to ND V. dahliae (Calderon et al., 2014). In addition, clones of a wild olive tree of O. europaea var. sylvestris "Acebuche" known to be moderately resistant to D V. dahliae infection from a collection of wild-olive genotypes at IAS-CSIC was also sampled. Three 35 cm-long terminal branches (one per tree) from independent "Picual" and "Arbequina" cultivated olive trees and from clone trees of "Acebuche" were sampled in December 2018 to perform xylem sap extraction with the Scholander chamber. Xylem sap was extracted at the end of Autumn, a season of the year when the olive stem water potential is more stable (Iniesta et al., 2009) and lower than 40 bar of pressure, the maximum pressure allowed by the Scholander chamber device that was used to extract the xylem sap (see below). Likewise, for wood chips collection, a similar number of branches (one branch per tree, with three trees in total per olive genotype) were sampled, and 6 cm-long stem portions were selected from each branch (three pieces per branch). All pruned branches were placed in sterile plastic bags, sprayed with distilled water and kept in a cold room at 4°C to avoid desiccation until later processing in the same day.

#### Microbiome Extraction From Xylem

The three 6-cm-long pieces of branches from each sample were debarked with a sterile scalpel. Bark tissue or xylem chips were obtained by scraping the debarked woody pieces with a sterile scalpel. A total of 0.5 g of xylem chips was weighted by mixing the chips obtained from all pieces sampled from same branch and tree and placed in a Bioreba bag containing 5 ml of sterile phosphatebuffered saline (PBS). Bioreba bags were closed with a thermal sealer and the content was homogenized with a hand homogenizer (BIOREBA, Reinach, Switzerland). Extracts were stored at 4°C until plating onto culture media, and then at -80°C, until DNA extraction. A total of three replicates per olive tree within each genotype was processed. All the processes described above took place under sterile conditions within a flow hood chamber.

Xylem sap extraction from olive branches was performed with a Scholander pressure chamber connected to a nitrogen cylinder following the Bollard process described by Alexou and Peuke (2013). An external port allowed switching from the internal Scholander chamber to an external 60-cm-long super chamber admitting a maximum of 40 bar of pressure. After inserting the branch in the super chamber, 5 cm of the main stem protruded to the exterior of the lid. At this point, it was important to preserve the branch with no cuts or loose leaves that could compromise the pressure within the chamber. To avoid microbial contamination of the xylem sap from bark and phloem, 2 cm of the main stem was debarked and disinfested. The edge between bark and xylem tissue was covered with parafilm to avoid leaks, and the pressure was increased gradually until xylem sap drops were observed to a maximum of 35 bars. The first drops of xylem sap were discarded to avoid external contamination. Xylem sap was collected within a 15 ml sterile falcon tube placed on ice. An average of 10 ml of xylem sap per genotype and branch was obtained and was preserved at 4°C until plating onto culture media, and then kept at -80°C until DNA extraction. All the processes described above took place under sterile conditions within a flow hood chamber.

#### Culture-Dependent Characterization of Xylem Bacteria

For the culture-dependent approach, four solid culture media were evaluated: BCYE (Wells et al., 1981), PD2 (Davis et al., 1981), R2A (Reasoner and Geldreich, 1985), and Nutrient agar (NA) (CONDALAB, Madrid, Spain). R2A and NA are general culture media of low- or high-nutrient contents, respectively. BCYE and PD2 are media specifically designed for isolation of the fastidious bacterium X. fastidiosa from xylem tissues.

Aliquots or 1/10 dilutions (100 µl each) of xylem extracts obtained by each of the two techniques indicated above were plated directly onto three plates of each medium and incubated at 28°C in the dark for 2 weeks. After incubation, all bacterial colonies were counted and a representative number of colonies from each medium and genotype was selected based on abundance and colony feature morphological criteria. Selected colonies were purified by triple serial colony isolation in the same medium. Purified isolates were grown at 28°C in the dark for 1 to 2 weeks depending on their rate of growth prior to DNA extraction.

DNA was extracted from a total of 261 bacterial isolates using the DNeasy kit (QIAGEN, Madrid, Spain). The near-complete 16S rDNA gene was amplified using primers 8f (5′- AGAGTTTGATCCTGGCTCAG-3′) and 1492r (5′- ACGGCTACCTTGTTACGACTT-3′) (Weisburg et al., 1991) as described in Aranda et al. (2011). Amplicons were purified with ExoSAP-IT (Thermo Fisher Scientific, Madrid, Spain), and directly sequenced in both directions using primers 8f and 1492r (STABVIDA, Caparica, Portugal). Sequences were assembled and manually corrected using DNASTAR software version 15.3.0.66 (Madison, WI, USA). Isolates were identified to genus/species level by the nearest neighbor in the GenBank "nt" database after alignment with reference 16S rRNA gene sequences using the BLAST algorithm according to Altschul et al. (1997).

#### Culture-Independent Characterization of Xylem Bacteria

Aliquots of xylem sap samples (0.5 ml) obtained from macerated xylem chips were placed in PowerBead tubes (DNeasy PowerSoil Kit, QIAGEN) and homogenized 7 min at 50 pulses s-1 with the Tissuelyser LT (QIAGEN). Sap extracts were incubated in the lysis buffer for 1 h at 60°C to increase cell lysis, and then processed following the DNeasy PowerSoil Kit manufacturer's instructions.

Aliquots (8 ml) of xylem sap samples extracted with the Scholander chamber were filtered through a 0.22-µm pore MF-Millipore™ filter (Merck Millipore, Madrid, Spain). Then, filters were placed into 1.5-ml Eppendorf tubes and the filtrate resuspended by vortexing for 5 min in the DNeasy PowerSoil Kit lysis buffer, incubated 1 h at 60°C and processed as described before. DNA obtained was quantified and outsourced to the Integrated Microbiome Resource (IMR) at Dalhousie University (Canada) to perform V5-V6 amplicon library sequencing with primers 799F (5'-AACMGGATTAGATACCCKG-3') and 1115R (5'-AGGGTTGCGCTCGTTG-3') and paired-end sequenced by using Illumina MiSeq sequencing platform (V3; PE 2x 300 bp). The ZymoBIOMICS microbial standard (Zymo Research Corp., Irvine, CA, USA) and water (no template DNA) were used as internal positive and negative controls, respectively, for library construction and sequencing. Raw sequence data were deposited in the Sequence Read Archive (SRA) database at the NCBI under BioProject accession number PRJNA574439.

#### Statistical and Bioinformatics Analysis

To determine the effects of the olive genotype, type of xylem sap extraction procedure, and culture media, data of culturable bacterial populations obtained in the culture-depended approach were subjected to analysis of variance (ANOVA) using the GLM (General Linear Models) procedure in Statistical Analysis System v. 9.4 (SAS Institute Inc.). Data of culturable bacterial population were log-transformed to fulfill ANOVA assumption. The experiment had a completely randomized design, with olive genotype, type of xylem sap, and culture media as factors with three replications (plates) per experimental unit. Data fulfilled the assumptions for ANOVA according to proper statistics. Orthogonal single-degree-offreedom contrasts were computed to test the effect of selected experimental treatment combinations.

The 16S rRNA sequences obtained were analyzed using the Quantitative Insights into Microbial Ecology bioinformatics pipeline, QIIME2 (version 2018.11; https://view.qiime2.org/) (Caporaso et al., 2010; Bolyen et al., 2018) with default parameters unless otherwise noted. DADA2 pipeline was used for denoising raw fastq paired-end sequences and filtering chimeras. Operational taxonomic units (OTUs) were obtained at 1% of dissimilarity and were taxonomically classified using RDP Bayesian classifier (Wang et al., 2007) against Silva SSU v.132 reference database. Singletons were discarded for taxonomy and statistical analyses.

Differences among bacterial communities derived from the culture-independent approach were calculated in QIIME2 using rarefaction curves of alpha-diversity indexes (including Shannon, Simpson, Faith\_PD, and Richness) at the genus level. Alpha and beta diversity as well as alpha rarefaction curves were conducted rarefying all samples to the minimum number of reads found. The Kruskal-Wallis test (P <0.05) with FDR correction (Benjamini and Hochberg, 1995) was used to find differences in alpha diversity indexes among the studied factors. Venn diagrams were generated using the "Venn diagram" online tool (http://bioinformatics.psb.ugent.be/webtools/Venn/) and were used to identify shared (core microbiome) or unique taxa according to the type of xylem sap extract and olive genotypes studied, and to compare the culture-dependent and cultureindependent approaches. We filtered bacterial taxa and retained those occurring in at least 50% of the samples in a given category. A heat tree summarizing main results was created using Metacoder package in R software (Foster et al., 2017). We unified the taxonomic affiliation derived from BLAST analysis (culture-dependent approach) and Silva SSU v.132 reference database (culture-independent approach) using the NCBI Taxonomy Browser (https://www.ncbi.nlm.nih.gov/Taxonomy/ Browser/wwwtax.cgi?mode=Root). Taxonomic abundances within each identified Phylum to genus level were visualized using Krona hierarchical data browser (Ondov et al., 2011).

A non-supervised principal component analysis (PCA) and multivariate hierarchical clustering analysis (using Pearson's correlation to measure distance and the Ward clustering algorithm) were performed using the OTU frequency matrixes at the genus level derived from the culture-dependent and cultureindependent approach with the online tool MetaboAnalyst 4.0 (http://www.metaboanalyst.ca; Chong et al., 2018).

### RESULTS

#### Bacterial Abundance and Alpha Diversity Measures

In the culture-dependent approach, bacterial population densities in xylem sap extracted from woody chips ranged from 40 to 1,920 colony forming units (cfu)/ml, whereas those obtained from xylem sap samples obtained with the Scholander pressure chamber ranged from 10 to 610 cfu/ml. Only two factors in the study were found to be significant [olive genotype (F=12.6, P < 0.0001) and type of xylem sap extraction method (F = 22.4, P < 0.0001), representing 16.6% and 29.7% of the mean square error (MSE) in the model, respectively]. Only the interaction type of xylem sap extraction method by olive genotype was significant (F = 35.67, P < 0.0001) representing 47.2% of the MSE in the model.

A significantly higher (F > 32.09, P < 0.0001) bacterial population density was estimated on xylem samples extracted from woody chips compared to xylem sap extracted with the Scholander chamber for "Acebuche" and "Picual," while the opposite occurred for "Arbequina" (F = 17.13, P = 0.0001) for both xylem extraction procedures (Figure 1). "Arbequina" showed a significantly higher (F > 10.45, P < 0.0021) bacterial population than "Picual" and "Acebuche" when xylem sap was obtained with the Scholander chamber, whereas "Acebuche" showed a significantly higher (F > 9.24, P < 0.0037) bacterial population than "Arbequina" and "Picual" when the xylem sap was extracted from woody chips macerates (Figure 1).

Maximal recoveries of bacterial population were observed with R2A medium (478.9 ± 131.9 cfu/ml), and decreased with BCYE (352.2 ± 101.3 cfu/ml), PD2 (335.0 ± 97.6 cfu/ml), and NA (270.0 ± 82.0 cfu/ml), in that order, although no significant differences were found among culture media (F = 1.98, P = 0.1298) (data not shown).

A total of 261 bacterial isolated were selected for further study from xylem sap and xylem chips extracts after cultivation in NA, PD2, R2A, or BCYE culture media (Figure 1; Table S1). Out of those, 137 bacterial isolates were from xylem sap extracted with the Scholander chamber while 124 were selected from xylem chips extracts. From the 261 isolates, 51 and 117 bacterial isolates were recovered from cultivated olives "Picual" and "Arbequina," respectively, and 93 bacterial isolates from the wild olive genotype "Acebuche" (Figure 1).

In the culture-independent approach, Illumina MiSeq sequencing analysis resulted in a total of 21,411 good quality reads with an average of 334 bp after removal of chimeras, unassigned or mitochondrial reads. No chloroplasts reads were detected in our samples. A total of 58 OTUs were identified for all treatments, out of which 48 OTUs were retained after

rarefying all data to 1,234 sequences (the minimum number of reads obtained in one of the samples) and singleton removal.

for each genotype the existence of significant differences at P≤ 0.05 between

both xylem sap extraction methods.

Rarefaction curves of observed OTUs (Richness) displayed significantly higher (H = 12.79; P < 0.0003) values in xylem sap extracted with the Scholander chamber as compared to the woody chips maceration method. Interestingly, although "Picual" was the olive cultivar showing the lowest number of culturable bacteria (Figure 1), it presented a significantly higher number of OTUs for both extraction methods (H = 20.19, P < 0.0001), followed by "Acebuche" and "Arbequina" (Figure S1). Maximum values of Good's coverage of 1.0 were obtained for all samples (data not shown).

Alpha diversity indices (Richness, Shannon, Faith\_PD and Simpson) did not show global significant differences between xylem sap extraction methods (H < 3.185, P > 0.074), among olive genotypes (H < 1.977, P > 0.372) or their interaction (H < 6.591, P > 0.253) (data not shown).

#### Composition of Xylem Sap Bacterial Communities

In the culture-dependent approach, a total of 4 phyla, 7 classes, 14 orders, 22 families, and 34 genera were identified (Figure 2). Significant differences (P < 0.05) were found in the number of genera isolated according to the culture media (Figure S2). Out of the total of 34 bacterial genera identified, 11 genera were detected in AN, 17 in BCYE and PD2, and 24 in R2A. Seven of the genera (Bacillus, Brachybacterium, Curtobacterium, Dermacoccus, Frigoribacterium, Methylobacterium, and Sphingomonas) were isolated in all culture media, whereas seven, four and four, and one genera were unique to R2A, PD2 and AN, and BCYE, respectively. Those unique genera corresponded normally to single isolates (Figure S2; Table S1).

Regarding olive cultivars, 13 genera were identified in "Picual," 22 in "Arbequina," and 21 in "Acebuche." Only four bacterial genera (11.8% of the total identified) were shared among the three olive genotypes and were present in more than 50% of all samples (Core bacterial genera: Bacillus, Frigoribacterium, Methylobacterium, Sphingomonas), six were shared between the two cultivated olive genotypes (Brevibacillus, Dermacoccus, Kineococcus, Marmoricola, Micrococcus, Staphylococcus), seven were shared between the wild olive genotype "Acebuche" and "Arbequina" (Amnibacterium, Curtobacterium, Frondihabitans, Microbacterium, Modestobacter, Nocardioides, Paenibacillus), and only one (Variovorax) between "Acebuche" and "Picual" (Figure 3). "Acebuche" showed the highest number of unique genera (nine in total). Finally, regarding xylem sap extraction method, 13 (38.2%) of the culturable bacterial genera (Bacillus, Brevibacillus, Curtobacterium, Dermacoccus, Frigoribacterium, Methylobacterium, Micrococcus, Modestobacter, Nocardioides, Paenibacillus, Sphingomonas, Staphylococcus and Variovorax) were isolated by both extraction methods, whereas 11 (32.4%) and 10 (29.4%) of bacterial genera were unique to the Scholander chamber extraction method (Acinetobacter, Amnibacterium, Clavibacter, Corynebacterium, Frondihabitans, Kocuria, Marmoricola, Novosphingobium, Patulibacter, Quadrisphaera, Roseomonas) or the woody chips extraction method (Brachybacterium, Cellulomonas, Enterococcus, Friedmanniella, Kineococcus, Microbacterium, Rhodococcus, Risungbinella, Spirosoma, Terribacillus), respectively (Figure 4).

When using the culture-independent approach, a total of 5 phyla, 8 classes, 17 orders, 23 families, and 31 genera were identified (Figure 2). Twenty-one of the genera were identified in "Picual," 14 in "Arbequina," and 16 in "Acebuche" (Figure 3). Only seven bacterial genera (22.6% of the total identified) were shared among the three olive genotypes (Core bacterial genera: Curtobacterium, Friedmanniella, Geodermatophilus, Hymenobacter, Kineococcus, Methylobacterium, Sphingomonas), two were shared between the two cultivated olive genotypes (Cutibacterium and Pseudomonas), two were shared between the wild olive genotype "Acebuche" and "Arbequina" (Amnibacterium and Mucilaginibacter), and two (Exiguobacterium and Novosphingobium) between "Acebuche" and "Picual." "Picual" showed the highest number of unique genera (10 in total) (Figure 3). Regarding xylem sap extraction method 14 (45.2%) of the bacterial genera identified were isolated by both extraction methods (Acidibacter, Curtobacterium, Cutibacterium, Exiguobacterium, Friedmanniella, Geodermatophilus, Hymenobacter, Kineococcus, Marmoricola, Methylobacterium, Pseudomonas, Roseomonas, Sphingomonas, Spirosoma), whereas 14 (45.2%) and 3 (0.10%) of bacterial genera were unique to the Scholander chamber extraction method (1174- 901-12, Acidiphilium, Amnibacterium, Arthrobacter, Brevibacterium, Deinococcus, Frigoribacterium, Mucilaginibacter, Nocardioides, Novosphingobium, Quadrisphaera, Rathayibacter, Streptococcus, Variovorax) or the woody chips extraction method (Bradyrhizobium, Pectobacterium, Rubellimicrobium), respectively (Figure 4).

Four of the phyla, Actinobacteria, Bacteroidetes, Firmicutes, and Proteobacteria, were obtained both by the culture-dependent and culture-independent approaches. Only the phylum Deinococcus-Thermus emerged exclusively when using the culture-independent

independent approaches and extracted with the Scholander chamber (SCh) or from woody chips (WC) macerates. The size and color of nodes and edges are correlated with the abundance of taxa. The central nodes are the total of all the other nodes in the tree for each phylum.

approach (Figure 2). At the Class level Actinobacteria, Bacilli, Alphaproteobacteria Cytophagia, Betaproteobacteria, and Gammaproteobacteria were identified following both approaches. Thermophilia class was detected only in the culture-dependent approach while Sphingobacteriia and Deinococci classes were identified only when using the NGS procedure (Figure 2). At the genus level, a total of 51 genera were identified combining both culture-dependent and culture-independent approaches; 20 and 17 genera were exclusive of the culture-dependent and cultureindependent approaches, respectively, and 14 bacterial genera were shared by both methodologies (Figure 2).

#### Bacterial Abundance Distribution

Actinobacteria phylum presented the highest relative abundance in the culture-dependent approach (46.37%) followed by Proteobacteria (34.09%), Firmicutes (19.15%), and Bacteroidetes (0.38%) (Figure S3). Differently, Proteobacteria showed more than half of the total bacterial genera in the culture-independent approach (56.25%), followed by Bacteroidetes (24.76%), Actinobacteria (13.44%), Firmicutes (0.77%), and Deinococcus-Thermus (0.21%) (Figure S4). At the family level, the most abundant bacterial families were Microbacteriaceae (34.48%), Methylobacteriaceae (16.86%), Sphingomonadaceae (31.44%), and Bacillaceae (13.03%) when using the culture-dependent approach, whereas Sphingomonadaceae (31.44%), Hymenobacteraceae (24.11%), and Methylobacteriaceae (22.36%) were the most abundant when using the cultureindependent approach. Focusing on the culture-dependent methodology, the most abundant genera were Frigoribacterium (18.77%), Methylobacterium (16.36%), Sphingomonas (14.56%),

FIGURE 4 | Prevalence Venn diagram showing the unique and shared bacterial genera obtained using culture-dependent approaches (upper panel) or cultureindependent approach (lower panel) in olive xylem sap samples when compared by xylem sap extraction method [Scholander chamber (SCh) or wood chips (WC) maceration].

Bacillus (12.64%), and Curtobacterium (11.88%), while when we used the NGS approach, the most abundant genera identified were Sphingomonas (30.10%), Hymenobacter (24.11%), and Methylobacterium (22.36%) (Figures S3 and S4).

#### Bacterial Community Structure

Hierarchical clustering analysis and PCA using OTU frequencies at the genus level differentiated xylem bacterial communities according to the sap extraction method or the olive genotype irrespective of the culture approach used (Figure 5; Figure S5). However, these differences were more noticeable when using the culture-independent approach, for which there was a clear trend to group the bacterial communities first by olive genotype and then by extraction method with only one exception, and with "Acebuche" showing the most distant bacterial communities as compared to the cultivated olive genotypes.

#### DISCUSSION

This study resulted in an initial characterization of the bacterial microbiome of the xylem sap from two cultivated and one wild olive genotypes following culture-dependent and -independent approaches, thus providing new insights into the olive microbiome profile, especially in a vascular system where some of the most destructive olive pathogens thrive. Our results suggest that the characterization of the xylem bacterial endophyte community composition is strongly dependent on the use of culture-dependent or -independent approaches and on the xylem sap extraction method used (Scholander chamber vs. woody chips). This initial characterization of olive xylem bacterial microbiome represents a first step making inroads into a potentially promising strategy for the identification of prospective biological control agents well adapted to the ecological niche in which these organisms would be potentially applied.

We found that the population of culturable xylem-associated bacteria ranged from 101 to 2×10<sup>3</sup> cfu/ml, depending on the olive genotype and the extraction method. These population ranges are in agreement with those obtained for other woody crops such as citrus and grapes (Bell et al., 1995; Gardner et al., 1982). Although the estimates obtained per gram of homogenized vessel xylem tissue are not strictly comparable to data obtained for the vacuum Scholander chamber, mainly due to differences in the amount of plant tissue sampled, the population levels obtained when using the former method appeared to be much higher in general, even when the amount of tissue sampled was smaller. Microscopic analysis of xylem samples from grapes and citrus (Bell et al., 1995; Gardner et al., 1982) revealed that many bacteria remain attached to xylem vessel walls with fibrillar material. These observations might also explain our differences in olive, not only quantitatively but also qualitatively, i.e., the differences noticed in bacterial genera retrieved by both extraction methods. It may be possible that certain groups of xylem-inhabiting bacteria show, or certain environmental conditions induce, a preference for planktonic growth as compared to a static-biofilm associated growth on xylem tissue. Since biofilm formation in plants, and specifically on xylem tissues, is associated with both symbiotic (beneficial) and pathogenic responses, these potential differences in growth styles or changes due to environmental conditions within the xylem vessels should be taken into consideration in future studies (Cruz et al., 2012; Bogino et al., 2013).

FIGURE 5 | Principal component analysis of the relative abundance of bacterial genera obtained using culture-dependent approaches (left panel) or cultureindependent approach (right panel) in olive xylem sap samples when compared by olive genotype (Ace: "Acebuche," Arb: "Arbequina," and Pi: "Picual") or by xylem sap extraction method [Scholander chamber (SCh) or wood chips (WC) maceration].

Olive is one of the most ancient domesticated plants, widely distributed in Roman times and its culture documented since ca. 6,000 years ago (Lavee, 1990). A key feature of cultivated plants is represented by the process of domestication and breeding, with a net reduction of genetic diversity and the promotion of growth with external inputs, which may interfere with the establishment of microbiome assemblages (Doebley et al., 2006). We found differences in the microbiome community composition associated to xylem-sap depending of the olive genotype, indicating a more different bacterial composition of wild olive as compared to both cultivated olive genotypes. Several studies with domesticated crops have indicated differences in microbiome community composition between wild accessions and modern cultivated varieties, mainly studied in herbaceous species (e.g., barley, bean maize, and rice; Peiffer et al., 2013; Bulgarelli et al., 2015; Edwards et al., 2015; Pérez-Jaramillo et al., 2017), less so in woody crops (e.g., Bullington and Larkin, 2015; Deyett et al., 2017; Yang et al., 2017), with only a few studies in olive (Aranda et al., 2011; Montes-Borrego et al., 2013; Montes-Borrego et al., 2014; Müller et al., 2015). Although this study shows some evidence on the existence of significant differences in population densities and community composition of the xylem-limited microbiome according to the olive genotype, this observation should be further explored including a higher number of olive cultivars growing under different environmental conditions.

Our NGS results with olive xylem indicated the presence of endophytic bacteria from the phyla Actinobacteria, Firmicutes, Bacteroidetes, Proteobacteria, and Deinococcus-Thermus. However, while Actinobacteria and Proteobacteria bacteria were present in all three olive genotypes, only "Picual" contained Firmicutes and Bacteroidetes, and only "Acebuche" contained bacteria from the Deinococcus-Thermus phylum. These results are in accordance with those obtained previously following similar approaches from olive leaves and xylem sap (Müller et al., 2015; Fausto et al., 2018; Sofo et al., 2019) as well as from other woody plants, like grapes, oak, poplar, and Pinus (Uroz et al., 2010; Bonito et al., 2014; Cregger et al., 2018; Deyett and Rolshausen, 2019). Interestingly, our results were more in agreement with those obtained by Deyett and Rolshausen (2019) that found similar proportions for Proteobacteria and Actinobacteria on grapevine xylem sap than those on olives obtained by Fausto et al. (2018), when using a similar xylem sap extraction method with a Scholander pressure chamber. Some differences were also found among those studies at the genus level. Thus, Deyett and Rolshausen (2019) found bacteria in the Enterobacteriaceae family, and belonging to Streptococcus, Bacteroides, Bacillus, Acinetobacter, and Pseudomonas genera as the most abundant on grapewine sap, whereas Fausto et al. (2018) identified some bacterial genera in the olive sap common to our study (e.g., Curtobacterium, Hymenobacter, Methylobacterium, and Sphingomonas), although they did not provide their relative abundance. Those differences may be due to the fact that different primers targeting different variable regions of the 16 rRNA were used for amplification in each study, or to the differences on environmental growing conditions and the plant genotypes evaluated.

In our study the bacterial genera detected in the highest proportions were, in this order, Sphingomonas, Hymenobacter, and Methylobacterium. Interestingly, despite the abundance of Hymenobacter determined by NGS, we did not isolate it in culture, although this bacterium has been readily isolated from stem and leaves of other plant species (Izhaki et al., 2013; Ding and Melcher, 2016; Durand et al., 2018; Ginnan et al., 2018). Other abundant genera identified through NGS were Kineococcus and Friedmanniella, both in xylem sap and woody chips. However, they were only isolated in culture from woody chips, which might indicate the presence of different species of cultivable or noncultivable bacteria of Kineococcus and Friedmanniella in the olive xylem microbiota or a preference of these genera for a stationary growth within the xylem. Cultivated members of these genera had been previously isolated from plant stems (Qin et al., 2009; Tuo et al., 2016) and, as members of the Actinobacteria phylum, they have been proposed to increase agricultural productivity through plant-growth promotion and to have the potential to be used as an alternative to chemical fertilizers (Palaniyandi et al., 2013; Hamedi and Mohammadipanah, 2015).

Traditional approaches for studying the diversity of plant microbial communities have relied on cultured-dependent approximations. Although the advances in -omic and NGS technologies over the last decade have enabled the cultureindependent study of plant microbiomes, we believe that both approaches should be used in parallel to provide deeper knowledge and exploitation of the plant-associated microbiome, especially for perennial crops (Gagliardi et al., 2001; Aranda et al., 2011; Mendes et al., 2011; Jackson et al., 2013; Dissanayake et al., 2018). With olive, our results using the culture-dependent approach indicated that endophytic bacteria from phyla Actinobacteria, Firmicutes, Bacteroidetes, and Proteobacteria can be readily isolated from the different plant genotypes. Considering the unique microbial communities found for each olive genotype, however, no Bacteroidetes could be isolated from "Arbequina," while for "Picual," unique bacteria belonging to Actinobacteria phylum were isolated. These results are in line with the gross taxonomic bacterial distribution of isolates already identified in the olive rhizosphere and phyllosphere (Ercolani, 1991; Aranda et al., 2011), as well as with the composition of bacterial endophytes identified in other woody plants, such as citrus, grapevine, or poplar (Bell et al., 1995; Compant et al., 2011; Azevedo et al., 2016; Durand et al., 2018). At the genus level, Sphingomonas, Methylobacterium, Curtobacterium, Frigoribacterium, and Bacillus were among the most abundant culturable genera. Frigoribacterium are endophytic bacteria isolated from poplar trees (Ulrich et al., 2008) and from potato, where they have an antagonistic capacity against phytopathogenic fungi (Berg and Hallmann, 2006). Bacillus bacteria are well-known rhizosphere and endosphere colonizers with diverse antagonistic activities against plant pathogens (Kumar et al., 2012). The widespread occurrence of the Sphingomonas and Methylobacterium genera has been described as crucial in plants due to a diverse range of functions, including degradation of certain contaminants and facilitation of soil nutrient cycling and plant growth (White et al., 1996; Chen et al., 2016). This may explain their ample distribution in the

plant kingdom, from legume species (Khan et al., 2014) to woody plants (Van Aken et al., 2004) and throughout different plant organs, such as seed, leaf, and flower tissues (Kim et al., 1998). Interestingly,Methylobacteriumisolates have been shown to exhibit antagonistic properties that induce resistance against the attack of diverse fungi (Rajendran et al., 2009; Ardanov et al., 2012) or that, more importantly, can modify the response of the xylem-inhabiting bacterial pathogen X. fastidiosa in citrus, resulting in a reduction of Citrus Variegated Chlorosis (CVC) symptoms (Azevedo et al., 2016). Similarly, Curtobacterium has been also described in citrus plants as an endophyte interacting with X. fastidiosa and reducing the severity of the CVC symptoms (Lacava et al., 2007). All these characteristics make several of the bacterial isolates from xylem sap of olive obtained in this study good candidates for testing their potential to confer protection against the xylem-inhabiting pathogens X. fastidiosa and V. dahliae in future research.

Although several studies on woody plant species have revealed the power of using culture-independent approaches for revealing previously unappreciated microbial diversity on xylem samples (Müller et al., 2015; Deyett et al., 2017; Fausto et al., 2018; Deyett and Rolshausen, 2019; Sofo et al., 2019), our study is unique in that culture-dependent and -independent methods were compared directly and simultaneously (i.e., applied to the same samples) in different genotypes of the host plant using two different extraction methods, to perform a quantitative and qualitative assessment of the xylem bacterial microbiome. Although knowing the specific limitations of each of the culture-dependent and culture-independent approaches, and the short number of olive genotypes and samples evaluated, our study is relevant in that it paves the way for future studies on the identification and characterization of olive xylem sap endophytic bacteria that could be involved in plant defense by acting as biocontrol agents against diverse xylem-inhabiting pathogens and by promoting plant health and growth promotion of olive plants. This information, in turn, will be generally useful for innovative microbiome research based both on culture-dependent and culture-independent approaches wherever vascular plant pathogens and nutrient requirements are involved.

#### DATA AVAILABILITY STATEMENT

The raw sequence data have been deposited in the Sequence Read Archive (SRA) database at the NCBI under BioProject accession number PRJNA574439.

### AUTHOR CONTRIBUTIONS

MA-M and BL: conceived research, performed statistical and bioinformatics analyses, interpreted results, and wrote the manuscript. CO-G, MA-M, and CH: prepared materials and equipment and performed the experiments. JN-C and JI: contributed to reviewing the manuscript and interpreting results. All authors viewed and approved the manuscript.

### FUNDING

This study was funded by project AGL2016-75606-R (Programa Estatal de I+D Orientado a los Retos de la Sociedad from the Spanish Government and the Spanish State Research Agency and FEDER-EU) and Project XF-ACTORS (grant 727987) from the European Union's Horizon 2020 Framework Research Programme. MA-M is a recipient of a research fellowship BES-2017-082361 from the Spanish Ministry of Economy and Competitiveness.

#### ACKNOWLEDGMENTS

"Martin Rivero" and "Zoco" secondary schools and M. P. Velasco, participants of Science IES 2018–PIIISA (Project to introduce research and innovation into secondary schools in Andalusia, Spain) are acknowledged for their interest in this research study.

We acknowledge support of the publication fee by the CSIC Open Access Publication Support Initiative through its Unit of Information Resources for Research (URICI).

#### SUPPLEMENTARY MATERIALS

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2019. 01708/full#supplementary-material

FIGURE S1 | Comparison of rarefaction curves of observed OTUs present in the xylem sap of 'Picual' (Pic), 'Arbequina' (Arb), and 'Acebuche' (Ace) olive genotypes extracted with the Scholander chamber (SCh) or from woody chips macerates (WC). Error bars represent standard derivation of three independent tree replicates. Data were rarified to 1,234 sequences.

FIGURE S2 | Prevalence Venn diagram and table of cases showing the unique and shared bacterial genera obtained using culture-dependent approaches when compared by culture isolation media.

FIGURE S3 | Bacterial taxonomic abundances (%) from phylum to genus level derived from culture-dependent approach.

FIGURE S4 | Bacterial taxonomic abundances (%) from phylum to genus level derived from culture-independent approach.

FIGURE S5 | Hierarchical cluster analysis and heat-map of the relative abundance of bacterial genera obtained using culture-dependent approaches (left panel) or culture-independent approach (right panel) in olive xylem sap samples when compared by olive genotype (Ace: 'Acebuche', Arb: 'Arbequina' and Pi: 'Picual') or by xylem sap extraction method (Scholander chamber (SCh) or wood chips (WC) maceration). The relative abundance of the bacteria are scaled by a color gradient bar.

TABLE S1 | Bacterial isolates obtained from xylem sap of olive in the present study, indicating the sap extraction method, culture media of isolation and their identification at genus level based on partial sequencing of 16S rRNA and BLAST analysis.

### REFERENCES


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 Anguita-Maeso, Olivares-García, Haro, Imperial, Navas-Cortés and Landa. 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.

# Effectiveness of Plant Beneficial Microbes: Overview of the Methodological Approaches for the Assessment of Root Colonization and Persistence

Ida Romano1† , Valeria Ventorino1,2\*† and Olimpia Pepe1,2

<sup>1</sup> Department of Agricultural Sciences, University of Naples Federico II, Naples, Italy, <sup>2</sup> Task Force on Microbiome Studies, University of Naples Federico II, Naples, Italy

#### Edited by:

Ofir Bahar, Agricultural Research Organization (ARO), Israel

#### Reviewed by:

Joelle Sasse Schlaepfer, University of Zurich, Switzerland Chanyarat Paungfoo-Lonhienne, University of Queensland, Australia

#### \*Correspondence:

Valeria Ventorino valeria.ventorino@unina.it

† These authors have contributed equally to this work and share first authorship

#### Specialty section:

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

> Received: 01 August 2019 Accepted: 06 January 2020 Published: 31 January 2020

#### Citation:

Romano I, Ventorino V and Pepe O (2020) Effectiveness of Plant Beneficial Microbes: Overview of the Methodological Approaches for the Assessment of Root Colonization and Persistence. Front. Plant Sci. 11:6. doi: 10.3389/fpls.2020.00006 Issues concerning the use of harmful chemical fertilizers and pesticides that have large negative impacts on environmental and human health have generated increasing interest in the use of beneficial microorganisms for the development of sustainable agri-food systems. A successful microbial inoculant has to colonize the root system, establish a positive interaction and persist in the environment in competition with native microorganisms living in the soil through rhizocompetence traits. Currently, several approaches based on culture-dependent, microscopic and molecular methods have been developed to follow bioinoculants in the soil and plant surface over time. Although culture-dependent methods are commonly used to estimate the persistence of bioinoculants, it is difficult to differentiate inoculated organisms from native populations based on morphological characteristics. Therefore, these methods should be used complementary to culture-independent approaches. Microscopy-based techniques (bright-field, electron and fluorescence microscopy) allow to obtain a picture of microbial colonization outside and inside plant tissues also at high resolution, but it is not possible to always distinguish living cells from dead cells by direct observation as well as distinguish bioinoculants from indigenous microbial populations living in soils. In addition, the development of metagenomic techniques, including the use of DNA probes, PCR-based methods, next-generation sequencing, whole-genome sequencing and pangenome methods, provides a complementary approach useful to understand plant–soil–microbe interactions. However, to ensure good results in microbiological analysis, the first fundamental prerequisite is correct soil sampling and sample preparation for the different methodological approaches that will be assayed. Here, we provide an overview of the advantages and limitations of the currently used methods and new methodological approaches that could be developed to assess the presence, plant colonization and soil persistence of bioinoculants in the rhizosphere. We further discuss

**232**

the possibility of integrating multidisciplinary approaches to examine the variations in microbial communities after inoculation and to track the inoculated microbial strains.

Keywords: bioinoculant, plant growth-promoting microbes, colonization, persistence, culture-dependent methods, microscopy-based techniques, metagenomic approach

#### INTRODUCTION

The increasing demand to reduce the use of chemical fertilizers and pesticides for the development of an agri-food system sustainable for environmental and human health, as well as the current shifting in the agricultural legislation of several countries, have led to an expanded use of bioinoculants. Chemical inputs usually alter the natural physico-chemical and biological equilibrium of soil, and microbial consortia used in agricultural management practices could return soil to its natural status (Lucy et al., 2004; Woo and Pepe, 2018). Although the manipulation of soil microbiomes to optimize crop productivity is an ancient practice, it is still little explored, especially regarding mechanistic studies of plant–microbe interactions and microbial persistence in heterogeneous communities in diverse locations, soils, and hosts (Finkel et al., 2017). Among the numerous bacterial or fungal strains used as bioinoculants, plant growth-promoting microbes (PGPM) are the most commonly applied. PGPM may affect plant performance through multiple mechanisms of action, operating directly by the production of specific substances that are able to promote plant growth and increase the availability and uptake of nutrients in soil (i.e., phosphate solubilization, siderophore and indole-3-acetic acid production, nitrogen fixation) or indirectly through the suppression of plant pathogens (Ribeiro and Cardoso, 2012). Several plant growth-promoting rhizobacteria (PGPR) have also been demonstrated to exert a beneficial effect on plant growth under nutritional and abiotic stress (Sharma et al., 2014; Singh and Sharma, 2016; Van Oosten et al., 2018) or during the restoration of polluted soils (Ventorino et al., 2014). Moreover, plants could also establish symbiosis with arbuscular mycorrhizal fungi (AMF), which increase the root surface area for nutrient acquisition (Wu et al., 2005).

A successful microbial inoculant has to colonize the external and/or internal part of plant tissues and establish a compatible interaction with the host as well as to persist in the soil against autochthonous microorganisms living in environment through its rhizocompetence traits (Finkel et al., 2017). In general, rhizosphere colonization occurs through several different mechanisms, such as bacterial movement, survival in the rhizosphere by competition against other microbes, adherence to and colonization of root surfaces, for instance by biofilm formation, and the creation of synergistic interactions with the host plant (Bhattacharya et al., 2017). Moreover, even if PGP inoculants colonize the plant initially, their persistence over time is not guaranteed. Measuring the persistence of microbial inoculants in soil poses technical difficulties, as the inoculant needs to be identified from within a complex community. The tracking and monitoring of the persistence of PGPM released in the environment have been widely studied (Brandt and Kluepfel, 1991; Kloepper and Beauchamp, 1992; Stahl and Kane, 1992; Gamalero et al., 2003; Podile and Kishore, 2006; Ahmad et al., 2011; Glick, 2015; Rilling et al., 2019) to understand their behavior in soil and which factors influence their survival under various conditions. Several sets of techniques are currently used to detect root colonization and persistence in the soils: microbial enumerations by culture-based methods, microscopy-based techniques, and DNA-based methods. The results may depend on the choice of technique since each has advantages and limitations, and each technique may have bias in favor of specific microbial taxa.

This review examines and presents an overview of the current methodological approaches that could be used to assess and detect plant colonization and soil persistence of microbial bioinoculants in the rhizosphere environment and considers multidisciplinary approaches to track and monitor inoculated microorganisms.

#### GOOD PRACTICES FOR RHIZOSPHERE SAMPLING AND SOIL PREPARATION

In natural ecosystems such as soils, several variables or factors can influence the results due to the highly heterogeneous distribution of microbial cells in the environment. Therefore, a well-organized experimental plan to investigate microbial populations from plant roots and soil is necessary. Usually, in field experiments, the simplest approach used to overcome spatial variables is a completely randomized design with replicates since the treatments are assigned completely at random, creating homogeneous treatment groups (Fiorentino et al., 2018; Lusiba et al., 2018).

To ensure good results in microbiological analysis, the first fundamental prerequisite is the correct soil sampling, both in laboratory and in greenhouse trials and in field experiments, to obtain representative samples for each treatment to be analyzed (Pennock et al., 2008). Temporal and spatial aspects could be considered during rhizosphere (soil area influenced by plant roots and their exudates; Barillot et al., 2013) or bulk soil (soil not adhering to roots and not influenced by exudates; Barillot et al., 2013) sampling since changes in microbial diversity over time are usually related to environmental changes. Therefore, soil or rhizosphere microbial diversity studies are usually carried out over years or seasons (Lombard et al., 2011). Moreover, it is known that other factors, such as plant age and developmental stage, could also influence plant microbial community structure (Compant et al., 2019); therefore, these variables could also be considered for soil sampling.

Soil and rhizosphere samples can be collected by different sampling approaches, as extensively detailed by Wollum (1994): i) simple random, which ensures that each sample has the same opportunity to be selected, usually by using a grid; ii) stratified random, similar to simple random, except the area to be sampled is broken into smaller subareas; or iii) systematic, which ensures that the entire area is sampled and represented by individual samples that are obtained by establishing predetermined points. The number of soil samples to take depends on the microbial population distribution and can be calculated using the formula suggested by Wollum (1994), which considers a prestudy sampling, the sample variance and the sample mean. However, it is recommended to brush away stone, rubbish, trash or grass from the soil surface before taking samples. Then, using a sanitized shovel, it is possible to take the samples from topsoil to an adequate depth (for instance, 0-20 cm) or to collect plant roots by excavating or uprooting plants to study microbial diversity in bulk soil and/or rhizosphere. For rhizosphere studies, after plant sampling, roots should be shaken vigorously by hand to remove bulk soil and to collect soil adhering to roots (Ventorino et al., 2012; Barillot et al., 2013). Moreover, during the sampling, it is necessary to avoid root damage. Manual excavation using spades and hand tools and working progressively in layers or sectors could minimize the corruption of soil architecture and ensure the safety of the roots. It is also fundamental to take a sufficient number of replications for data analysis (Neumann et al., 2009). Following this, the samples must be recovered in sterile polyethylene bags or vessels and stored at 4°C to avoid desiccation during transport to the laboratory.

To evaluate external and internal root colonization, which generally occur in the rhizoplane and endosphere, respectively, several steps for sample preparation are necessary (Figure 1). In particular, plant roots should be washed by agitation in sterile water or buffer [e.g., phosphate buffered saline (PBS) or physiological buffers] without tearing or cutting plant tissues to facilitate the separation between soil/root particles and microorganisms (Kloepper and Beauchamp, 1992). For instance, a good practice to detach the bacteria from the soil particles is shaking for 30 min at 120-130 rpm in an adequate volume of isotonic solution containing tetrasodium pyrophosphate (16% w/v) (Ventorino et al., 2014). Barillot et al. (2013) reported that after vigorously hand-shaking roots to separate bulk soil from rhizospheric soil, shaking the roots a second time in a sterile 0.9% NaCl solution allowed rhizosphere collection, and shaking the roots a third time in the same sterile solution containing Tween 80 (0.01% v/v) allowed the rhizoplane fraction (thin layer of soil strongly adhering to the roots; Barillot et al., 2013) to be collected (Figure 1). Indeed, to study microbial endophytes, it is necessary to surface sterilize the roots prior to grinding, chopping or blending them (McInroy and Kloepper, 1991). Several works describe a prior wash with 1% chloramine and cycles of washing/agitation treatments using ethanol and PBS (Ladha et al., 1997; Dennis et al., 2008; Richter-Heitmann et al., 2016). Cleaned roots to be analyzed by cultureindependent methods can be stored in a solution of PBS buffer and 70% ethanol (2:3 v/v) for a long time at -20°C (Dennis et al., 2008; Richter-Heitmann et al., 2016). However, fresh root samples used to evaluate the density of the cultivable microorganisms by plating on growth media should be analyzed within a short time (24–48 h).

#### MICROBIAL ENUMERATIONS BY CULTURE-DEPENDENT METHODS

Mainly because of their ease of use, culture-dependent methods are commonly used to estimate the persistence of inoculated microorganisms in soil and/or rhizosphere. However, these methods are limited since it is difficult to represent the high diversity of bacteria on culture media because only 0.1 to 1.0% of soil bacteria are cultivable (Daniel, 2005), and at the same time, it is difficult to differentiate inoculated organisms from native populations based on morphological characteristics (Lima et al., 2003).

To increase the likelihood of cultivating a high number of microbial strains, enrichment, selective and differential media are usually used as well as synthetic media mimicking the soil environment, typically containing soil extracts, are also developed. This approach has been successful, and it allowed the detection of a higher diversity of cultivable populations compared with other methods (Andreote et al., 2009). Although culture-dependent methods have been used to detect bioinoculants in different experimental conditions (growth chamber, greenhouse, open field), they are especially useful when the experiment is carried out in sterile conditions and interference by soil autochthonous microbial populations can be avoided. Therefore, advantages and limitations of culturedependent approaches will be discussed on the basis of experimental conditions (i.e. growth chamber, greenhouse, field).

#### Growth Chamber

Experiments conducted in growth chambers are usually performed using sterile synthetic substrates or hydroponic conditions for plant growth, allowing the control of all environmental parameters, such as temperature, relative humidity, light/dark cycle, and light intensity. Therefore, this approach is particularly suitable for the detection of inoculated strains in plant tissues by enumeration on culture media.

Castanheira et al. (2017) used viable counts to assess the colonizing abilities of a bacterial consortium composed of Pseudomonas sp. G1Dc10, Paenibacillus sp. G3Ac9, and Sphingomonas (S.) azotifigens DSMZ 18530 on the rhizoplane and surface-disinfected roots, stems and leaves of annual ryegrass plants grown under gnotobiotic conditions (Table 1). Sterile experimental conditions allow the use of a unique generic growth substrate to perform total bacterial counts and can allow three different bacterial strains to be distinguished on the basis of colony morphology.

Indirect viable counts on solid medium also allowed the assessment of the survival of endophytic trans-conjugant Pseudomonas sp. strains tagged with green fluorescent protein (GFP) in different tissues of poplar trees for 10 weeks (Germaine et al., 2004; Table 1). Since the plants were grown in a sterilized

FIGURE 1 | Schematic description of sampling collection, separation of different soil fractions, and methods (culture-dependent methods, microscopy-based techniques and molecular approaches) for the detection of microbial inoculants. After plant sampling, roots should be shaken vigorously by hand to collect bulk soil (soil not adhering to roots and not influenced by exudates). Shaking the roots a second time in a sterile 0.9% NaCl solution allowed rhizosphere (soil area influenced by plant roots and their exudates) collection, and shaking the roots a third time in the same sterile solution containing Tween 80 (0.01% v/v) allowed the rhizoplane (thin layer of soil strongly adhering to the roots) fraction to be collected. To study microbial endophytes, it is necessary to add a step of sterilization of the root surfaces prior to grinding, chopping or blending them. Root samples should be analyzed in a short time (24–48 h) to evaluate the density of the cultivable microorganisms by plating on growth media or they can be stored in a solution PBS buffer and 70% ethanol at -20°C for later analysis by culture-independent methods (microscopic and molecular methods).

substrate but were not maintained under sterile conditions throughout the experiment, a number of indigenous endophytic strains were also isolated on growth medium. Therefore, to exclusively count the inoculated strains, only the colonies expressing gfp were enumerated by examining the plates under an epifluorescence microscope (Germaine et al., 2004).

Similarly, Kandel et al. (2015) used trans-conjugant GFPtagged strains of Burkholderia sp., Rhizobium tropici PTD1, and Rahnella sp. WP5 to evaluate their colonization abilities in rice plants (Table 1). At 20 days after inoculation, the use of a selective growth medium allowed them to enumerate the total number of inoculated endophytes in the plant tissues. However,


the use of axenic experimental conditions ensures ease of study and that only inoculated strains will be recovered.

#### Greenhouse

Greenhouse experimental conditions could be considered a variation of farming in a controlled environment, which provides favorable growing conditions and protects crops from unfavorable weather and various pests. Therefore, this approach could be suitable for evaluating the viability of inoculated microorganisms by culture-dependent methods.

In pot greenhouse conditions, Wu et al. (2005) counted viable bacteria to demonstrate the successful colonization and the synergistic effect of beneficial rhizobacteria such as Azotobacter (A.) chroococcum and Bacillus (B.) (B. megaterium and B. mucilaginous) combined with mycorrhizal fungi belonging to the genus Glomus (G.) (G. mosseae or G. intraradices) in the rhizosphere of Zea mays plants (Table 1). The use of differential culture media allowed the detection and enumeration of groups of bacteria similar to the inoculants on the basis of their specific plant growth promoting activities, such as nitrogen fixation, phosphate and potassium solubilization.

Similarly, culture-dependent methods, based on the use of differentiation media for plant growth-promoting properties, were also useful to assess the persistence of bacterial (A. chroococcum, B. megaterium and B. mucilaginous) and fungal (G. mosseae or G. fasciculatum) consortia (Khalid et al., 2017; Table 1). The use of this approach demonstrated that the microbial concentration and root colonization of Spinacia oleracea L. was improved by the application of a consortium of microorganisms, suggesting the synergistic behavior of the strains.

The plate count method was also used to analyze the survival of five Azotobacter strains (ST3, ST6, ST9, ST17, and ST24) at different stages of wheat (Triticum aestivum L.) plant growth. These strains were inoculated in earthen pots containing saline soil under greenhouse conditions. The results of rhizosphere soil monitoring showed that the concentration of the inoculated strains increased up to 60 days of sampling (Chaudhary et al., 2013; Table 1). However, this approach did not allow the identification of microorganisms present in the culture at genus and species level in nonsterile condition. In fact, it is difficult to distinguish bioinoculants from indigenous microbial populations living in soils based on morphological characteristics.

Van Oosten et al. (2018) used viable microbial counts to assess the persistence of the inoculated A. chroococcum 76A in the rhizosphere of tomato plants cultivated under abiotic stress conditions (Table 1). A differentiating culture nitrogen-free medium for N fixers allowed them to demonstrate that the strain A. chroococcum 76A, inoculated at a concentration of approximately 106 CFU/g, was able to grow in all experimental conditions, increasing by approximately one order of magnitude at the end of the experiment.

Interestingly, Solanki and Garg (2014) described a novel technique to enumerate viable cells of A. chroococcum in the unsterilized rhizoplane of Brassica campestris using a transconjugant strain of A. chroococcum Mac 27 containing a lacZ fusion (A. chroococcum Mac 27 L; Table 1). Using this approach, it was possible to monitor the growth and survival of the LacZtagged bacteria that formed blue-colored colonies on Burks medium containing X-gal.

#### Field

Although the field represents the natural and real condition for assessing the effectiveness of a microbial consortium or biofertilizer in soil, it is difficult to differentially enumerate inoculated microorganisms in this experimental state by culturedependent methods. However, some works have reported general results on the variation of microbial concentration in the rhizosphere of plants grown in agricultural fields.

Sharma et al. (2011) used a culture-dependent approach to assess microbial changes due to the application of a consortium formed by A. chroococcum AZ1 and AZ2 in association with G. fasciculatum and G. mosseae on apple plants grown in rainfed fields. As a general result, an increase in the concentration of bacteria and/or fungal strains in the inoculated tests was observed, although the results were more or less significant depending on the inoculant (used alone or in combination) and experimental conditions (Table 1).

A field experiment was also conducted to evaluate the inoculation effect of Azotobacter, Azospirillum (Az.), and AMF, either alone or in combination, on seedlings of apple cultivars. The viable counts of A. chroococcum and Az. brasilense in the rhizosphere were significantly higher in all the treatments than in the controls. In fact, the microbial concentration in the treatment with multi-inoculation of all the strains was significantly higher than those in all the other biological treatments but lower than that of the chemical fertilizer treatment (Singh et al., 2013; Table 1).

Culture-dependent methods have several advantages such as they are practical and useful techniques to quantify bioinoculants especially in sterile experimental conditions, and they allow to detect only viable cells and therefore bacterial inoculants that are competitive and able to persist overtime. Moreover, as reported in several works (Pitkäranta et al., 2007; Al-Awadhi et al., 2013; Ngom and Liu, 2014), it is difficult to detect the inoculated strain in unsterilized conditions. Culture-dependent methods cannot provide a comprehensive analysis of the endophytic ability of selected strains in unsterilized conditions since a portion of epiphytes that are resistant to sterilizing agents could determine an overestimation of their counts (Kandel et al., 2017). To explain the behavior of the bioinoculants in the natural soil ecosystem, culture-based methods should always be complemented with culture-independent approaches to examine the variations in the microbial community after inoculation treatment and to track the inoculated microbial strains.

### MICROSCOPY-BASED TECHNIQUES

Today, a wide range of microscopy-based techniques are available and have been used to detect microorganisms inoculated on plant tissues and to evaluate the colonization patterns of bacterial endophytes through molecular interactions and dynamics within living cells in specific vegetative tissues (Kandel et al., 2017).

Root colonization by bacteria and AMF has been studied by several types of microscopy, which can be divided into three major groups: light microscopy, electron microscopy and fluorescence microscopy.

#### Optical Microscopy

Light microscopy is the most common microscopic technique for assessing microorganisms in root systems due to its low costs of purchasing, maintaining, and servicing (Hulse, 2018).

Bright-field light microscopy was employed by White et al. (2014), who developed a combination of stains to evaluate the bacterial colonization of seedling root tissues. This approach was based on the use of 3,3'-diaminobenzidine tetrachloride (DAB) to stain hydrogen peroxide associated with bacterial invasion of eukaryotic cells followed by counterstaining with aniline blue/ lactophenol to stain protein in bacterial cells. This elementary technique allowed the visualization of bacteria and their eventual lysis in seedling roots, providing information on the defensive response of host cells and the bacterial degradation process (White et al., 2014).

Microscopy techniques that use different dyes are also usually used to assess mycorrhizal relationships with host plants. A wide number of staining procedures, which each have advantages and disadvantages, have been developed for studying AMF colonization, as extensively reported by Hulse (2018). Among these is a very simple, nontoxic, reliable and inexpensive staining technique for AMF colonization in root tissues; this technique is based on the use of an ink-vinegar solution after adequate clearing with KOH (Vierheilig et al., 1998). This solution stains all fungal structures, rendering them clearly visible by brightfield light microscopy.

The level of root colonization by mycorrhizal strains is usually evaluated using the microscopic procedure described by Phillips and Hayman (1970) and by Giovannetti and Mosse (Newman's intersection method, 1980). This method requires a stereomicroscope for observation; randomly dispersed roots are stained, placed on a grid in a 9-cm Petri plate and quantified by counting the number of intersections between grid lines and colonized roots. Although this method is strongly influenced by operator skill, it could provide sufficient information to evaluate the mycorrhizal colonization level. In fact, the gridline intersect method has been extensively used in many works to assess and quantify root colonization of mycorrhizal fungi (Sharma et al., 2009; Sharma et al., 2011; Sharma et al., 2012; Singh et al., 2013).

#### Electron Microscopy

Electron microscopy was further developed into scanning electron microscopy (SEM), which can be used to examine plant surfaces and microorganisms at high resolution, highlighting the adhesion of microbial cells to plant tissues. SEM was used to observe chickpea root colonization by A. chroococcum and Trichoderma viride (Velmourougane et al., 2017; Table 2). The plants were cultivated in sterile media composed of sand and vermiculite (1:1), and samples were taken at 40 days post inoculation. SEM microphotographs revealed the proliferation of Azotobacter cells, both individually and attached to the fungal mycelia. SEM observations have also highlighted the production of exopolysaccharides by A. chroococcum. These polymers improve the survival of EPSproducing microbial cells in natural ecosystems, exhibit beneficial effects in plant growth promotion and abiotic stress (Gauri et al., 2012; Van Oosten et al., 2017) and could be interesting for biopolymer production (Ventorino et al., 2019). Although SEM produces 3D images, it provides information only on surface morphology and colonization and is not as powerful as transmission electron microscopy (TEM). Although TEM is not considered a user-friendly technique since sample preparation is complex and time consuming, it is the most powerful microscopy technique, with a maximum potential magnification of 1 nanometer. TEM allows 2D ultrahigh resolution images to be obtained, providing information about the internal structure of a root sample; therefore, it is useful to establish endophytic interaction as reported by Singh and Sharma (2016). Hairy roots of Arnebia hispidissima were inoculated in vitro with five different A. chroococcum strains (Table 2). After 10 days of incubation, TEM showed that A. chroococcum strains were only inside hairy roots of inoculated plants, revealing the endophytic ability of A. chroococcum strains. However, since TEM allows only a small area of a sample to be explored, which provides information about the inner part of a sample, and SEM can explore a larger external area, these two techniques could be used in combination to obtain better detailed results about the rhizosphere environment and inoculant colonization (Thokchom et al., 2017).

Environmental scanning electron microscopy (ESEM) is another powerful method to evaluate the survival of a bacterial inoculant and its ability to colonize plant tissues. It provides new possibilities compared to conventional SEM and enables the investigation of nonconductive and hydrated samples without complex histological preparation steps (i.e., air drying, chemical fixation, dehydration, and coating), which are critical in conventional SEM (Stabentheiner et al., 2010). This approach was recently used by Dal Cortivo et al. (2017) to evaluate the colonization level of a commercial biofertilizer containing a bacterial consortium on wheat in sterile conditions (Table 2). ESEM imaging revealed good survival rates as well as external and internal colonization of leaf and root tissues by a bacterial consortium.

Although electron microscopy allows clear visualization of cells outside and inside plant tissues at a very high resolution, this technique can be used only in limited sterile conditions since it is unable to distinguish bioinoculants from indigenous microbial populations living in soils.

#### Fluorescence Microscopy

Fluorescence microscopy has become an essential technique in biology for the study of living tissues or cells. Although this method requires more complex and expensive instrumentation than conventional transmitted-light microscopy, it is widely used for the detection of bacteria inside plant tissues. This is possible because fluorescence microscopy reveals the position of fluorescent substances that were previously introduced into living cells. Several fluorescent dyes and protein tags and other methods to fluorescently label cells can be employed, providing a range of tools to track a microbial inoculant.

Narula and coworkers (2007) proposed the use of serological methods such as double-antibody sandwich enzyme-linked immunosorbent assay and immunofluorescence as potential techniques for investigating the colonization behavior of bioinoculants. They revealed the presence of A. chroococcum Mac 27 L in root fragments of hydroponically grown wheat plants using immunofluorescence (Table 2). However, one of the most commonly used methods for tracking endophytic inoculated bacteria within plant tissues is the use of GFP, which emits fluorescent green light when irradiated with blue light or near-ultraviolet (UV) light (Wang et al., 2015). The detection and quantification of GFP-tagged strains is possible using epifluorescence microscopy (Leff and Leff, 1996), confocal laser-scanning microscopy (CLSM) (Götz et al., 2006; Fan et al.,



FRET, fluorescence resonance energy transfer; SEM, scanning electron microscopy; TEM, transmission electron microscopy; ESEM, environmental scanning electron microscopy; GFP, green fluorescent protein; FISH, fluorescence in situ hybridization.

2011; Krzyzanowska et al., 2012), flow cytometry (Elvang et al., 2001), and UV exposure for solid agar plates (Errampalli et al., 1999). The use of GFP allowed the evaluation the colonization abilities of tagged Burkholderia sp., Rhizobium tropici PTD1, and Rahnella sp. WP5 in rice plants grown in N-free MS agar for twenty days in a growth chamber (Kandel et al., 2015; Table 2).

The presence of three inoculated GFP-tagged endophytic Pseudomonas sp. strains in different poplar tree tissues (leaf, stem and root) was verified by Germaine et al. (2004) using an epifluorescence microscope (Table 2). An innovative transparent soil made of a polymer with a low refractive index was used by Downie et al. (2014) to evaluate the abundance of GFP-tagged P. fluorescens SBW25 on Lactuca sativa roots (Table 2). The transparency of the substrate allowed them to capture images using confocal microscopy, which showed a high bacterial abundance on the root tips and at root branching zones. Although the use of GFP-tagged microbial strains has various advantages, such as no influence of autochthonous bacteria and the possibility of in situ detection, it can be used only in laboratory/greenhouse experiments since this method requires that the microbe be transformed before any application (Compant and Mathieu, 2013). In addition, the visualization of GFP expression is sometimes difficult due to the autofluorescence of the plant cell walls (Germaine et al., 2004), and it is difficult to detect inoculated microbes in situ because of interference by soil particles (Quadt-Hallmann and Kloepper, 1996). Finally, the procedure for the transformation of the GFPplasmid involves exposure to CaCl2, which promotes cyst formation in some endophytic strains, such as A. chroococcum; therefore, the procedure is unsuccessful in certain organisms. This is the main reason for developing an alternative procedure based on fluorescence resonance energy transfer to visualize endophytes inside plant tissues when the use of GFP is restricted. This technique is based on the use of a novel specific rhodamine-pyrene conjugate as an Al3+ selective colorimetric and fluorescence sensor to visualize the endophytes with minimum interference of background autofluorescence, unlike GFP tagging. The fluorescence resonance energy transfer-based technique was used by Banik et al. (2016) to track the A. chroococcum Avi2 strain after inoculation on sterile rice seedlings (Table 2). The results showed intracellular root colonization by the A. chroococcum Avi2 strain since a clear and stable green fluorescence was emitted by bacterial cells and detected by fluorescence microscopy, whereas a blue fluorescence was emitted by root tissues, proving the feasibility of this approach. In fact, the authors demonstrated that the rhodamine–pyrene conjugate was an excellent fluorescence ligand that was green-shifted only by the Al3+-treated bacterial cells since it was able to detect only intercellular Al3+ (Banik et al., 2016).

The fluorescent Al3+-siderophore complex produced by A. chroococcum strains was used by Viscardi et al. (2016) in combination with CLSM to assess the rhizocompetence of inoculated bacteria on tomato plants under sterile conditions in vitro, demonstrating the ability of the two selected bacteria to colonize plant roots (Table 2).

To determine the colonization ability of microbes on and inside plants, other methods, such as fluorescence in situ hybridization (FISH), have been employed. FISH is a molecular method based on the use of fluorescently tagged oligonucleotide probes, which are able to bind ribosomal RNA sequences to target metabolically active and intact cells (Moter and Gobel, 2000), combined with microscopy techniques such as epifluorescence microscopy (Compant and Mathieu, 2013) or CLSM (Rothballer et al., 2003; Wu et al., 2008). The range of available and developed probes for the detection of microbial cells using universal probes or strain-specific probes limits this technique. In addition, the long and complex sample preparation protocol (Moter and Gobel, 2000) could represent a disadvantage of this approach. Recently, the colonization ability of a multistrain inoculant composed of Pseudomonas sp. G1Dc10, Paenibacillus sp. G3Ac9 and S. azotifigens DSMZ 18530 on annual ryegrass plants was analyzed using FISH combined with CLSM (Castanheira et al., 2017; Table 2). However, in plant tissues, FISH showed several limitations due to weak and/ or unsuccessful hybridization signals of the probe. In fact, it was reported that in the FISH method, a low signal intensity of some of the detected microbes can occur due to a low cellular concentration of the target molecules or due to the low in situ accessibility of rRNA regions for singly labeled probes, thus preventing their successful visualization in plants (Wagner et al., 2003; Compant and Mathieu, 2013). Therefore, to overcome this problem, a combination of FISH, GFP-labeling methods and CLSM was employed. In detail, the use of FISH to detect a GFPlabeled S. azotifigens strain increased the signal, improving the visualization of bacterial cells and enabling the visualization and localization of inoculated strains in different parts of plants (Castanheira et al., 2017).

Although bioinoculants inside plant tissues can be clearly visualized by microscopy-based techniques, these techniques can suffer from several limitations (Pantanella et al., 2013; Emerson et al., 2017). For example, it is not always possible to distinguish living cells from dead cells by direct observation, and the autofluorescence of the plant cells sometimes makes it difficult to visualize microbial cells inside different plant tissues. Moreover, tagged microbial cells should be used only in limited and controlled experimental conditions (growth chamber and greenhouse) since it is not always permitted the dispersion of modified microorganisms in the environment, preventing the evaluation of survival and colonization ability of the bioinoculant in natural real ecosystems.

#### MOLECULAR APPROACHES

Methods based on the analysis of nucleic acids extracted directly from soil/rhizosphere samples have been developed to overcome cultivation limitations. In fact, the development of molecular tools allows new species of unculturable microorganisms associated with the root system to be discovered or helps to understand the ecological function of several microbial species (Lebeis et al., 2012; Bulgarelli et al., 2013). The total genetic material recovered directly from soil samples represents the soil metagenome (Daniel, 2005), and metagenomics is the field of molecular genetics and ecology that studies this "collective" genome to determine the phylogenetic and functional gene complements of a sample (Pershina et al., 2013; Jansson, 2015). The development of metagenomic techniques, including the use of DNA probes (Bouvier and del Giorgio, 2003), polymerase chain reaction (PCR)-based techniques (Ruppel et al., 2006) and next-generation sequencing (NGS, Mardis, 2008), has greatly increased the ability to track microorganisms in natural environments (Ahmad et al., 2011). However, considering the high microbial diversity and the complex environmental matrix, DNA extraction is a fundamental step that could affect the detection and quantification of microbial taxa inferred from metagenomic sequences in all molecular methods; therefore, specific microbial groups can be underrepresented (Morgan et al., 2010; Montella et al., 2017). Currently, two main approaches are used for microbial DNA extraction from soil (Lombard et al., 2011): i) direct extraction, based on the direct lysis of microbial cells inside the soil matrix followed by DNA extraction and purification; and ii) indirect extraction, based on the initial recovery of microbial cells from the soil samples followed by lysis and DNA extraction and purification. Although both DNA extraction approaches are suitable for metagenomic analysis, they have different advantages and drawbacks in terms of DNA quantity and quality, even when starting from the same matrix (Ventorino et al., 2015; Montella et al., 2017), as extensively reported by Lombard et al. (2011), depending on the soil type. Therefore, when beginning a metagenomic analysis of soil, it is critical to define which DNA extraction method will be optimal by considering the subsequent genomic analysis (Lombard et al., 2011). For a more detailed discussion on this topic see Lombard et al. (2011).

#### PCR-Based Methods

In recent decades, several molecular approaches, such as quantitative real-time PCR (qPCR), denaturing gradient gel electrophoresis (DGGE), automatic ribosomal interspace spacer analysis, amplified ribosomal DNA restriction analysis, and NGS, have been used to investigate the presence of microbial inoculant in the soil system and to determine its impact on the rhizosphere community (Ciccillo et al., 2002; Steddom et al., 2002; Gamalero et al., 2003). These approaches allow the detection of specific microorganisms and/or the abundance of different microbial populations or species on the basis of the amplification of specific genes. Among these techniques, qPCR is a sensitive and suitable approach for determining the abundance of functional genes from soil-derived DNA and RNA (Fiorentino et al., 2016), and it has therefore been extensively used to track and quantify inoculated strains in soil systems (Providenti et al., 2009; Timmusk et al., 2009). For instance, Sorte et al. (2014) used this method to design specific PCR primers targeting a 16S rRNA variable region to specifically measure the abundance of Gluconacetobacter diazotrophicus following coinoculation with other diazotrophic strains in sugarcane plants grown under field conditions (Table 3). The validation of employed species-specific primers allow the use of this method to evaluate the occurrence of endophytic diazotrophic G. diazotrophicus species in any soil type and plant tissue. A qPCR protocol was also developed by Couillerot et al. (2010) for the strain-specific quantification of Az. brasilense UAP-154 and CFN-535 in the maize rhizosphere using BOX-based sequence characterized amplified region markers, although the detection limit ranged from 10<sup>4</sup> to 10<sup>8</sup> CFU g-1 (Table 3). The success of this approach has led other authors to use it. In fact, strain-specific primers recovered from draft genome sequence analysis were employed for qPCR to quantify Az. brasilense FP2 in wheat roots as well as to assess its competitiveness following coinoculation with other PGPR (Stets et al., 2015; Table 3). All of these works demonstrate the high effectiveness and specificity of this culture-independent approach based on the use of strain-specific primers, allowing rapid and inexpensive detection of bioinoculants in the plant rhizosphere for monitoring and quantification purposes, which is also useful in nonsterile and uncontrolled conditions.

The addition of bioinoculants in a soil could determine variations in the native microbial community structure, as recently reported by Fiorentino et al. (2018). PCR-DGGE followed by sequence analysis of bands is a metagenomic approach able to describe changes in soil microbial communities after inoculation of bacterial or fungal strains as well as to test the persistence of microbial inoculant in the soil. By DGGE and gene sequence analyses, Chen et al. (2013) detected heavy metal-resistant Burkholderia sp. J62 and P. thivervalensis Y-1-3-9 in both root interiors and rhizosphere soil of Brassica napus L., demonstrating their influence on the rape-associated bacterial community structures in artificially Cdcontaminated soil (Table 3). The presence of Az. brasilense Cd (DSM 1843) in the rhizosphere of sorghum plants was monitored by Lopez et al. (2013) by gene sequencing of DGGE bands for three crop cycles (Table 3), highlighting its rhizocompetence against indigenous populations. However, since DGGE allows us to distinguish microbial populations at the species level, when the experiments are carried out in nonsterile soil, it is difficult to ensure that a sequence of bands originated from inoculated microbial strains or from other autochthonous strains belonging to the same species. Therefore, DGGE analysis is usually performed in combination with other techniques, such as FISH (Lopez et al., 2013), GFP (Piromyou et al., 2013), SEM, and TEM (Thokchom et al., 2017). In some cases, the combination of DGGE and qPCR is a suitable approach to investigate the abundance of specific microbial groups and the survival of bioinoculants in the soil, as recently reported by Kumar et al. (2018) in a pot trial-based study (Table 3). In this case, DGGE was a useful approach to check bioinoculants because no band corresponding to inoculated Dyadobacter sp. was recovered in the control soil.

### Next-Generation Sequencing

In recent decades, the development of massive DNA sequencing technology, known as NGS, and bioinformatic tools has provided a powerful alternative to other molecular studies of microbial ecology in natural environments, enabling the study of taxonomic diversity at a high resolution (Ventorino et al., 2018). Indeed, analyzing the rhizosphere microbiome with the high-throughput sequencing approach has different prospective results that could allow understanding the community structure of root-associated bacteria and, as a consequence, novel bacteria with plant growth promoting traits to be discovered. This approach could also help to understand changes in the microbial community dynamics and structure after inoculation treatments. NGS could be performed following two different approaches: i) amplicon sequencing based on the amplification of phylogenetic marker genes, usually hypervariable regions from small-subunit ribosomal RNA genes (i.e., 16S rRNA), followed by bioinformatic analysis; ii) shotgun sequencing based on random sequencing across entire genomes followed by genome



SCAR, sequence characterized amplified region; qPCR, quantitative real-time PCR; DGGE, denaturing gradient gel electrophoresis; NGS, next-generation sequencing; WGS, wholegenome sequencing.

assembly and bioinformatics analysis. The construction of environment-based libraries was a major advance in soil metagenomics, and these libraries could be screened by functional and sequence-based approaches to clarify several functions of organisms in soil communities and to simplify genomic analyses of uncultured soil microorganisms (Garza and Dutilh, 2015). Recently, NGS of 16S rRNA genes was used to evaluate the behavior of the strain Streptomyces sp. AH-B after it was inoculated in quinclorac-contaminated soil, as well as its influence on soil microbial communities (Lang et al., 2018). After alignment, sequences were clustered into operational taxonomic units (OTUs) at 97% identity, which revealed that Streptomyces sp. AH-B became the dominant species following inoculation and that the bacterial and fungal diversity in treated soil was higher than that in the control, probably due to the degradation activity of inoculant that could reduce quinclorac toxicity to microorganisms. However, due to the high and complex biodiversity of soil microbial communities and the presence of various PCR and library preparation inhibitors, such as humic substances, full coverage of the soil metagenome is a difficult task. Moreover, the identification of OTUs at 97% identity thresholds allow to discriminate microbial populations at the species level but not at the strain level, so different strains with different plant growth promoting activities could be pooled together. In addition, identical OTUs do not necessarily mean the same species, since there are several databases for microbial identification, and it could be difficult to compare different studies, since the determination of sequences depends on sequences entered into DNA collections. Finally, high-quality DNA extraction for NGS is challenging for soil studies and is dependent on the extraction method and soil characteristics (Daniel, 2005).

#### Whole-Genome Sequencing and Pangenome

The determination of the entire genomic DNA sequence at a single time sequence [whole-genome sequencing (WGS)] of a microbial strain could be a powerful approach to investigate the potential PGP activities of a strain as well as its plant colonization and survival efficiency in the rhizosphere, leading to the identification of specific genes related and involved in plant– microbe interactions. In recent years, this approach was used to characterize new PGPR strains. Functional annotation of WGS of the strain B. aryabhattai AB211 revealed the presence of common genes involved in PGP activities and in abiotic/biotic stress tolerance as well as genes conferring resistance to oxidative stresses in plants demonstrating its high potential as a PGPM (Bhattacharya et al., 2017). However, the presence of PGPrelated genes is essential but not sufficient for a bacterium to exert beneficial effects on plant growth in a real environment. In fact, although the presence of key attributes essential for possible colonization and interaction with the host plant were recovered in two Rhodopseudomonas palustris strains (PS3 and YSC3), these strains exhibited different expression patterns of genes related to PGP activities, probably due to the different physiological responses of these strains to specific compounds in the root exudates that act as signal molecules (Lo et al., 2018). Therefore, the effectiveness of PGP activities of a specific strain could also be affected by the different exudates released into the soil by different plants.

WGS could also be used in combination with metagenomic studies to identify microbial strains in the soil metagenome. Using this approach, the presence of the plant-associated strain B. amyloliquefaciens FZB42 on lettuce was assessed by Krö ber et al. (2014; Table 3). Fragment recruitments of metagenome sequence reads on the referenced genome sequence of B. amyloliquefaciens FZB42 following shotgun sequencing of whole rhizosphere microbial communities of inoculated plants evidenced that the strain was present for over 5 weeks. Therefore, the combination of WGS and shotgun sequencing could be a suitable approach to identify the persistence of a microbial inoculant in the rhizosphere of plants grown in a natural environment.

Another method for the detection and identification of key genes responsible for the adaptation and evolution of a microbe as an endophyte is the pangenome. The pangenome can be defined as the entire genetic repertoire of a species; it comprises a core genome, which is composed of the genes present in all strains of the species, and an accessory genome, comprising the genes that are unique to specific strains (Mira et al., 2010; De Maayer et al., 2014). By analyzing the pangenome of eight sequenced Pantoea ananatis strains isolated from different sources, De Maayer and coworkers (2014) identified proteins with a potential role in plant– microbe interactions. Despite the large amount of information that could be retrieved from the pangenome, this method is still rarely used for studying the genetic traits of endophytes since it is based on the cultivation of microbial strains; therefore, nonculturable endophytes remain unexplored (Kaul et al., 2016).

Recently, Albanese and Donati (2017) proposed a novel method (StrainEst) based on the use of single-nucleotide variant profiles of the referenced available genomes of selected species to identify and quantify the strains of interest present in metagenomic samples. This novel approach could be useful to highlight differences at the strain level that could allow us to track a microbial inoculant in the rhizosphere.

The increasing database of sequenced microbial genomes also allows genome-wide computational searches for clustered, regularly interspaced short palindromic repeats (CRISPRs) in microbial species (Sorek et al., 2008). These repetitive sequences have been detected in a wide number of bacterial and archaeal genomes (Horvath and Barrangou, 2010), including PGPM. CRISPRs are usually used as molecular markers for the detection of pathogenic microbes or for the evaluation of phageresistance mechanisms in bacteria (Sorek et al., 2008). Although the CRISPR approach has been applied to plant-soil environments only to detect plant pathogenic strains such as Erwinia amylovora (McGhee and Sundin, 2012), it could be exploited in the future for developing molecular markers to monitor PGPR for plant– microbe interactions (Rilling et al., 2019).

The development of molecular techniques based on the analysis of nucleic acids provides an approach useful to understand plant–soil–microbe interactions. These methods have greatly increased the ability to track microorganisms in natural environments and some of them allow a rapid and inexpensive detection of bioinoculants in the plant rhizosphere for monitoring and quantification purposes overcoming cultivation limitations. The use of one or a combination of these methods allow the investigation of the abundance of specific microbial groups and the survival of bioinoculants in the soil as well as variations in the native microbial community dynamics and structure (Kumar et al., 2018). Although DNAbased approaches have improved our knowledge of microbial ecology, they are not able to differentiate between live and dead cells. Therefore, it is recommended to use them in combination with conventional methods, such as culture enumerations, for investigating bacterial ecology in natural habitats. Finally, molecular methods are highly influenced by DNA quality and quantity that is dependent on the extraction method and soil characteristics (Daniel, 2005; Lombard et al., 2011).

## CONCLUSION

Assessing the root colonization of inoculants with beneficial effects on plant growth as well as their persistence over time in a soil is a critical issue in sustainable agriculture. Currently, several approaches that use culture-dependent, microscopic and molecular methods have been developed to follow bioinoculants in the soil and on the plant surface. However, to ensure good results in microbiological analysis, the first fundamental prerequisite is the correct soil sampling and sample preparation for the different methodological approaches that will be assayed.

Although plant colonization of bacterial endophytes can be assessed by microscopy-based techniques through molecular interactions and dynamics within living cells in a specific vegetable tissue, the measurement of the persistence of inoculants in soil poses technical difficulties, as the inoculant needs to be identified from a complex community. Methods to detect persistence include cultural enumeration or molecular approaches using PCR-based methods and next-generation sequencing. Culture-dependent methods are commonly used to estimate the persistence of inoculated bacteria in soil and/or rhizosphere, mainly for their ease of use, but this analysis is limited since it is difficult to represent the high diversity of bacteria on culture media and, at the same time, it is difficult to differentiate inoculated organisms from native populations based on morphological characteristics. Therefore, culture-dependent methods are especially useful when the experiment is carried out in sterile conditions to avoid interference by native microbial populations living in the soil. Molecular analysis allows the detection of bioinoculants or their activity in soil and contemporaneous evaluation of the effect of rhizosphere engineering on native microbial communities. However, most of the molecular techniques are based on the preliminary genomic characterization of the microbial strain used as inoculant and the specific molecular markers of the strain for its detection in the soil metagenome. Molecular approaches help to improve our knowledge of microbial ecology, but they cannot be considered as a substitute for more conventional methods, such as culture enumerations. In fact, if DNA is analyzed, there is the disadvantage of the inability to differentiate between live and dead cells; therefore, these methods should be considered complementary for investigating bacterial ecology in natural habitats. Future perspectives in the assessment of colonization and soil persistence should have a polyphasic approach combining several molecular and microbiological techniques to allow the tracking of inoculated strains or microbial consortia.

Moreover, a microscopy-based approach allows us to obtain a picture of bacterial colonization outside and inside plant tissues, but it is not possible to always distinguish living cells from dead cells by direct observation. The autofluorescence of the plant cells and interference by soil particles make it difficult to visualize microbial cells inside different plant tissues. Tagged microbial cells should be used only in limited and controlled experimental conditions (growth chamber and greenhouse), and the

### REFERENCES


evaluation of the survival and colonization ability of an inoculant in a natural real ecosystem cannot be performed because the strains could be released into the environment.

All the described methods have advantages and disadvantages and provide only partial results, and most of them are timeconsuming, expensive and unable to detect specific inoculated microbial strains. Therefore, to better explain the behavior of bioinoculants in the natural soil ecosystems, culture-dependent and culture-independent (molecular and microscopic approaches) methods should be used in combination to examine the variations in microbial communities after inoculation treatment and to track the inoculated microbial strains in different systems.

The main challenge for the application of PGPM as bioinoculants in unsterilized greenhouse or field conditions is the establishment of effective methods for the assessment of plant colonization and soil persistence. Moreover, modern soil microbiology lacks efficient methods for the detection and estimation of the effective PGP activities that inoculated strains have on the soil. This is another main bottleneck in the use of microbial inocula for rhizosphere engineering. Therefore, the development of specific and easy methodologies for the evaluation of PGP activities could help to understand what actually occurs in a natural soil system during plant–soil– microbe interactions.

#### AUTHOR CONTRIBUTIONS

IR and VV wrote the manuscript in collaboration. VV and OP conceived the concepts of the manuscript.

## FUNDING

This work was supported by FSE-FESR PON R&I 2014-2020, PhD program on "Sustainable agricultural and forestry systems and food security" - XXXIII cycle.


Bacillus amyloliquefaciens FZB42, engineered to express green fluorescent protein. J. Biotechnol. 151, 303–311. doi: 10.1016/j.jbiotec.2010.12.022


Azotobacter chroococcum and Pantoea agglomerans using serological techniques. Plant Soil Environ. 53, 167–176.


solarization and chemical disinfestation with mulch practices for sustainable nursery management. Sci. Hortic. 130, 164–174. doi: 10.1016/j.scienta.2011.06.032


Multidisciplinary Approach to Innovative Methods. Ed. A. Piccolo (Berlin, Heidelberg: Springer), 145–177. doi: 10.1007/978-3-642-23385-2\_6


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 Romano, Ventorino and Pepe. 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.

# Biocontrol by Fusarium oxysporum Using Endophyte-Mediated Resistance

Francisco J. de Lamo and Frank L. W. Takken\*

Molecular Plant Pathology, Faculty of Science, Swammerdam Institute for Life Sciences, University of Amsterdam, Amsterdam, Netherlands

#### Edited by:

Kalliope K. Papadopoulou, University of Thessaly, Greece

#### Reviewed by:

Víctor Flors, University of Jaume I, Spain Marta Berrocal-Lobo, Polytechnic University of Madrid, Spain

> \*Correspondence: Frank L. W. Takken F.L.W.Takken@uva.nl

#### Specialty section:

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

Received: 25 September 2019 Accepted: 13 January 2020 Published: 06 February 2020

#### Citation:

de Lamo FJ and Takken FLW (2020) Biocontrol by Fusarium oxysporum Using Endophyte-Mediated Resistance. Front. Plant Sci. 11:37. doi: 10.3389/fpls.2020.00037 Interactions between plants and the root-colonizing fungus Fusarium oxysporum (Fo) can be neutral, beneficial, or detrimental for the host. Fo is infamous for its ability to cause wilt, root-, and foot-rot in many plant species, including many agronomically important crops. However, Fo also has another face; as a root endophyte, it can reduce disease caused by vascular pathogens such as Verticillium dahliae and pathogenic Fo strains. Fo also confers protection to root pathogens like Pythium ultimum, but typically not to pathogens attacking above-ground tissues such as Botrytis cinerea or Phytophthora capsici. Endophytes confer biocontrol either directly by interacting with pathogens via mycoparasitism, antibiosis, or by competition for nutrients or root niches, or indirectly by inducing resistance mechanisms in the host. Fo endophytes such as Fo47 and CS-20 differ from Fo pathogens in their effector gene content, host colonization mechanism, location in the plant, and induced host-responses. Whereas endophytic strains trigger localized cell death in the root cortex, and transiently induce immune signaling and papilla formation, these responses are largely suppressed by pathogenic Fo strains. The ability of pathogenic strains to compromise immune signaling and cell death is likely attributable to their host-specific effector repertoire. The lower number of effector genes in endophytes as compared to pathogens provides a means to distinguish them from each other. Coinoculation of a biocontrol-conferring Fo and a pathogenic Fo strain on tomato reduces disease, and although the pathogen still colonizes the xylem vessels this has surprisingly little effect on the xylem sap proteome composition. In this tripartite interaction the accumulation of just two PR proteins, NP24 (a PR-5) and a b-glucanase, was affected. The Fo-induced resistance response in tomato appears to be distinct from induced systemic resistance (ISR) or systemic acquired resistance (SAR), as the phytohormones jasmonate, ethylene, and salicylic acid are not required. In this review, we summarize our molecular understanding of Fo-induced resistance in a model and identify caveats in our knowledge.

Keywords: Fusarium, endophyte, induced resistance, biocontrol, PR protein

## INTRODUCTION

The Fusarium oxysporum species complex embraces a variety of strains ubiquitously present in soils. Most of these strains are saprotrophs and despite their ability to colonize plant roots the majority represents commensal endophytes not affecting plant fitness (Bao et al., 2004). Some F. oxysporum (Fo) strains, such as Fo47 and CS-20, are actually beneficial to the host and can provide protection against root pathogens (Table 1). Biocontrolconferring Fo strains, such as Fo47, have been identified in vascular wilt-disease suppressive soils (Alabouvette, 1986). Identification of the causal microbes in wilt suppressive soils is typically done by sterilizing the soil following subsequent reinoculation with the original microbes and screening for isolates that restore the suppressive effect against Fusarium wilt (Tamietti et al., 1993).

Fusarium wilt is one of the major diseases caused by pathogenic Fo strains. Wilts are a major threat for agriculture (Fisher et al., 2012) and Fo ranks among the 10 most devastating fungal plant pathogens worldwide (Dean et al., 2012). Besides wilt disease some strains can also cause foot- or root-rot resulting in serious yield losses in affected crops (Michielse and Rep, 2009). Fo produces micro- and macroconidia and chlamydospores that can remain viable in infected soils for decades, thereby frustrating crop rotation schemes (Nelson, 1981). Pathogenicity of Fo is host-specific, as typically strains infecting one plant species do not cause disease in others. Based on this hostspecificity, pathogenic strains have been classified into socalled formae speciales (ff.spp.), of which over 100 have currently been described (Armstrong and Armstrong, 1981). An explanation for the emergence of host-specific pathogenic strains may be the extensive use of monocultures with limited crop rotation serving as breeding grounds for pathogens (Xiong et al., 2016). The evolved Fo pathogens can give rise to devastating crop losses, Fusarium wilt disease of banana, caused by Fo f.sp. cubense, being a prime example (García-Bastidas et al., 2014; Ordoñez et al., 2016).

To control wilt diseases different strategies are currently being employed in agriculture. One of these is chemical control, which includes broad-spectrum biocides like methyl bromide, benomyl, or carbendazim applied before planting. These chemicals can prevent infection, but do not cure a plant once infected. A caveat of these compounds is that they also affect beneficial soil microbiota and some accumulate in the food chain and for this reason many of these products are, or will be, banned (Lopez-Aranda et al., 2016). Heat sterilization of soils overcomes some of these drawbacks, but has the disadvantage that it is non-selective and it can have undesired effects on soil quality (Mahmood et al., 2014). Use of resistant plant varieties, e.g. plants carrying resistance genes is currently the most effective in terms of economy, ecology, and disease control. However, genetically encoded resistance is seldom durable and sooner or later new races emerge that overcome resistance in a never-ending arms race between Fo and its host (Takken and Rep, 2010; de Sain and Rep, 2015). Furthermore, Fo resistance genes are not available in the germplasm of all crops or they cannot be introgressed by breeding (Ploetz, 2015).

The limitations of the current approaches of wilt disease control urges the need to develop alternatives. An interesting alternative strategy is the use of beneficial Fo strains that confer biocontrol and thereby reduce disease incidence. A major advantage of biocontrol is the relatively broad-spectrum- and non-race specific protection conferred by endophytic strains (Table 1). A limitation is that the protection provided by these biological agents is highly variable and not consistent between seasons, crops, or fields. As illustration, even in greenhouse trials using tomato plants artificially co-inoculated with a pathogenic and a biocontrol Fo strain significantly different degrees of protection where observed in subsequent years (Fuchs et al., 1999). Furthermore, biocontrol observed under controlled lab conditions is not necessarily scalable to field conditions. For example, controlled soil co-inoculation of asparagus with Fo47 and Fo f.sp. asparagi (Foa) resulted in partial disease protection under lab conditions, but application of Fo47 in Foa-infested greenhouses did not reduce wilt disease (Blok et al., 1997).

A better understanding of the molecular mechanisms underlying biocontrol conferred by endophytic Fo strains may help to unleash the full potential that these organisms harbor to control disease conferred by their brothers in crime. In this review we mostly focus on two endophytic Fo strains, Fo47 and CS-20, as these are the best studied strains. We assess the differences between pathogenic and endophytic strains at their root colonization behavior, at the genome level and the responses they trigger in plants. Endophyte-mediated biocontrol consists of two components. The first is based on a direct activity on the pathogenic strain via parasitism and antibiosis (Benhamou et al., 2002; Le Floch et al., 2009) or by competing for nutrients or root niches. Several excellent reviews are available describing these non-plant mediated processes (Fravel et al., 2003; Alabouvette et al., 2009; Vos et al., 2014; Latz et al., 2018). In this review we focus on the other component of biocontrol, the indirect plantmediated resistance response triggered by Fo endophytes, called endophyte-mediated resistance (EMR).

### F. OXYSPORUM CONFERS BIOCONTROL IN VARIOUS PLANT SPECIES AGAINST ROOT PATHOGENS

The ability of a large variety of endophytic Fo strains to confer biocontrol has been reported in many independent studies implying that it is a generic feature for Fo (Table 1). This idea is supported by a study in which over 200 different nonpathogenic Fo strains isolated from a tomato field were able to confer biocontrol in tomato, albeit to various degrees (Bao et al., 2004). Another observation is that Fo-based biocontrol is effective in a wide variety of plant species including both monocot and dicot species. This suggests that biocontrol is an ancient property as these families diverged over 200 million years ago (Wolfe et al., 1989). A number of oomycete-caused diseases can also be suppressed by Fo. For instance, Fo47 is reported to reduce disease incidence caused by Pythium oligandrum in tomato (Le Floch et al., 2009), Pythium ultimum in cucumber

#### TABLE 1 | Fusarium oxysporum (Fo)-mediated biocontrol in various plant species.


(Continued)

#### TABLE 1 | Continued


Fo endophytes are depicted in the first column; root or soil inoculation is abbreviated as root- or soil inoc; when endophytes and pathogen have been inoculated separately they are abbreviated as "Fo" and "P" respectively; otherwise as co-inoc, ± split root system indicate that both setups have been used in the study. Inoculations are typically performed by incubating roots in a spore suspension or by adding spores to the soil as indicated.

\*Screening including several Fo strains.

\*\*Fo47 inoculum also tested in combination with Pseudomonas sp.

\*\*\*Fo47 inoculum mixed with Trichoderma harzianum.

\*\*\*\*Fo47 inoculum also tested in combination with different actinomycete bacteria.

(Benhamou et al., 2002) and Phytophthora capsici in pepper (Veloso and Díaz, 2012). A common property among these pathogens is that they infect roots, but unlike most pathogenic Fo strains not all colonize the vasculature, implying that EMR is not vasculature-specific.

Two studies report on Fo-induced biocontrol that is not exclusively targeted against a root pathogen (Table 1). Preinoculation of watermelon roots with Fo f.sp. cucumerinum (Foc) (a pathogen on cucumber) reduced lesion sizes of Colletotrichum lagenarium infected leaves (Biles and Martyn, 1989). The other example details enhanced tolerance to Botrytis cinerea in pepper plants pre-inoculated with Fol (a tomato pathogen) (Díaz et al., 2005). Based on our literature survey Fo-induced EMR appears to be mostly root-confined.

#### THE ROOT COLONIZATION PATTERN OF F. OXYSPORUM PATHOGENS DIFFERS FROM THAT OF ENDOPHYTES

Root colonization by Fo endophytes and pathogens has been extensively studied. In this chapter, we compare the root colonization process of Fo endophytes with that of pathogens. In our comparison we include colonization of 'incompatible interactions' in which a pathogenic Fo strain colonizes a resistant host, which does not result in disease emergence.

#### Spore Germination

The first stage of the colonization process starts with Fo spores or hyphae that grow in the vicinity of a root. Addition of sugar to the soil induces chlamydospore germination of pathogenic Fol and Fo f.sp. basilici (Larkin and Fravel, 1999). In natural settings root exudates presumably provide these carbohydrates as exudates from different crops enhance germination of microconidia of Fol and Fo f.sp. radicis-lycopersici (Forl) pathogens (Steinkellner et al., 2005). Hyphal exudations may also play a role in conidia germination as Fo uses autocrine pheromone signaling to control germination in a conidialdensity dependent manner (Vitale et al., 2019). Some root pathogens effectively grow towards roots using chemotropism (Yao and Allen, 2006). While earlier studies found no evidence of chemotaxis toward tomato roots by Fo47 or the pathogens Fol or Forl (Steinberg et al., 1999; Olivain et al., 2006) a recent study showed that peroxidases secreted by tomato roots elicit Fol chemotropism towards roots (Turrà et al., 2015; Nordzieke et al., 2019). Altogether, it seems that root exudations trigger spore germination and induces directional mycelial growth. At this pre-colonization stage, no differences are noticeable between endophytic and pathogenic Fo strains.

### Host Colonization

Upon germination, both Fo endophytes and pathogens colonize the root surfaces of host and non-host plants. Contact with the root triggers hyphal branching, after which Fo produces hyphal swellings to invade the root. The fungal hyphae enter plant roots via wounds, cracks in the epidermis, lateral root emergence points, or by direct penetration of the root tip depending on the Fo strain and plant species involved. Hyphae reach the vascular stele via the apoplast of the root cortex. In some cases, intracellular growth is noticeable along with local host cell-death, a phenomenon observed more often among nonpathogenic strains (He et al., 2002; Olivain et al., 2003; Humbert et al., 2015; Gordon, 2017). Both pathogenic and non-pathogenic strains colonize the root cortex (Gordon, 2017), but although the initial colonization pattern is similar, the extent and pattern of colonization differs during later stages. The amount of biomass of a pathogenic strain in the root is typically higher than that of an endophyte. This difference is already apparent at early stages. At 48 h post-inoculation (hpi) higher amounts of fungal biomass for the Fo40 pathogen were detected in roots of soybean plants than for the endophytic Fo36 (Lanubile et al., 2015). A similar difference was reported for other systems, like the interaction between tomato and Fo47 or Forl. Two weeks post-inoculation Fo47 biomass was 10-fold less than that of the pathogen (Validov et al., 2011). These observations imply that in early stages of the interaction Fo endophytes are less efficient root colonizers than pathogens.

Besides the amount of fungal biomass also the root colonization pattern differs between Fo pathogens and endophytes. Typically, only pathogenic strains are able to reach the xylem vessels from where they colonize above-ground tissues. The induced occlusions of the xylem vessels, aimed to restrict pathogen progress, results in the classical wilting symptoms of infected plants (Gordon, 2017). A well-studied example is the interaction between pea roots, Fo47, and the pathogen Fo f.sp. pisi. Whereas Fo47 colonization is restricted to the root surface and outermost cell layers of the cortex, the pathogen massively invades the deeper root tissues including the vasculature (Benhamou and Garand, 2001). A similar pattern is seen upon Fo colonization of tomato. When grew in hydroponics, Fo47 and Fol8 both efficiently colonized the surface of tomato tap roots following attachment of the microconidia to the root hair zone. Subsequently, both strains grew towards the elongation zone until they reached the root apex. Whereas Fol8 intensively colonized the deeper root tissues and eventually reached the vasculature, the endophyte was confined to the epidermis and cortex (Nahalkova et al., 2008). In line with these observations, Fo47 proteins were not identified in xylem sap of Fo47-inoculated soil-grown tomato plants, whereas Fol proteins were detected in sap of Fol-infected plants (de Lamo et al., 2018). This difference indicates that even at later stages of root colonization Fo47 does not reach the vasculature. Contrarily, Fo47 was reported to colonize xylem vessels of eucalyptus (Salerno et al., 2000) and the Fo endophyte CS-20 was found to colonize xylem vessels of cucumber (Pu et al., 2014). An explanation for the vascular presence of the latter two endophytic strains could be the clipping of the roots prior to inoculation, providing direct vascular access to the endophyte.

In incompatible interactions, the first stages of the root colonization pattern are similar to that of a compatible interaction. The pathogen colonizes the root cortex, but in contrast to a purely endophytic strain, the pathogen frequently reaches the vasculature of a resistant host. Examples of vascular colonization of a resistant host are chickpea and tomato roots inoculated by either avirulent Fo f.sp. ciceris or Fol (Mes et al., 2000; Jimenez-Fernandez et al., 2013; van der Does et al., 2018). Recently vascular colonization of tomato plants carrying three different classes of resistance (R) gene types by Fol was compared. Although the plasma membrane-localized immune receptors (I and I-3) restricted colonization to a larger extent than the intracellular receptor (I-2), vascular colonization was observed in all cases (van der Does et al., 2018). The amount of fungal biomass in a resistant plant, however, is low and fungal proteins cannot be detected in the xylem sap of infected plants (de Lamo et al., 2018). These findings are in line with the low number of hyphae observed in vessels of a resistant tomato variety (Mes et al., 2000). Inoculation of resistant cabbage roots with Fo f.sp. conglutinans also resulted in marginal colonization of the vasculature and fungal proteins were not identified in the xylem sap either (Pu et al., 2016).

The general pattern is that Fo endophytes, similar to endophytes such as Serendita indica (Jacobs et al., 2011) and arbuscular mycorrhizal fungi (Gadkar et al., 2001), are mostly de Lamo and Takken Fusarium Induces Root-Specific Resistance

root surface- and cortex-colonizers. Extensive colonization of the root cortex and vasculature is typically restricted to pathogens, a property that correlates with enhanced secretion of cell walldegrading enzymes by these strains (Jonkers et al., 2009). Pathogenic strains are also able to enter and, to a limited extent, colonize the vasculature of a resistant host. Upon (co-) inoculation of an endophyte and a pathogen both strains coincide at the same root tissues during the early stages of the interaction, but become spatially separated when the pathogen invades the vascular bundle. Therefore, disease protection induced by Fo endophytes at these later stages is likely plant-mediated.

### F. OXYSPORUM ENDOPHYTISM AND PATHOGENICITY ARE GENETICALLY DETERMINED BY THE FUNGUS–HOST COMBINATION

Bioassays can reveal whether a strain is pathogenic on a specific host, or on a specific variety of that host, but these assays cannot establish whether a strain is non-pathogenic. Given the narrow host-range of pathogenic Fo strains (mostly restricted to single plant species) it would be necessary to inoculate a particular strain on all possible plant species and varieties to conclude that it is likely a non-pathogen in case of a negative outcome. Giving the impracticality of such an approach, and the limitation of classical taxonomic features, there has been ample focus on identifying molecular features that could be used to distinguish ff.spp. and to diagnose pathogenic isolates and discriminate them from non-pathogenic strains.

Phylogenetic analyses of the Fo species complex using conserved gene sequences such as those encoding elongation factor 1a typically result in phylogenetic trees in which Fo pathogens and endophytes are distributed together over different clades (Wong and Jeffries, 2006; Ellis et al., 2014; Pinaria et al., 2015). Likewise, trees based on genomic markers such as restriction fragment length polymorphism of the ribosomal intergenic spacer regions, or on mating type results in trees in which the ff. spp. are polyphyletic and cluster together with Fo endophytes in various clades (Alves-Santos et al., 1999; Abo et al., 2005; Ellis et al., 2014; Nirmaladevi et al., 2016). Even a multiple-sequence alignment of 441 conserved core genes from various Fo genomes did not result in a tree that enabled differentiating ff. spp. or allowed unambiguous identification of non-pathogenic strains (van Dam et al., 2018).

A decade ago, however, it was reported that the presence of a lineage-specific chromosome determines pathogenicity of Fol toward tomato (Ma et al., 2010). Horizontal transfer of a pathogenicity chromosome from Fol to Fo47 turned the endophyte into a tomato pathogen (Ma et al., 2010). Subsequent studies revealed that chromosome transfer from a cucurbit-infecting Fo strain could transform Fo47 into a cucurbit pathogen (van Dam et al., 2017). Vice versa, loss of a dispensable pathogenicity chromosome from a Fol strain resulted in loss of pathogenicity (Vlaardingerbroek et al., 2016). Hence, pathogenicity appears to correlate with the presence of a pathogenicity chromosome. These pathogenicity chromosomes differ from core chromosomes by a high content of transposable elements and a low gene density (Ma et al., 2010). Genetic analysis of Fol revealed that its pathogenicity chromosome carries the genes encoding the putative host-specific virulence proteins (effectors) that the fungus secretes in the tomato xylem sap following infection (Schmidt et al., 2013). Some of these Secreted In Xylem, or SIX, proteins such as SIX1 (Avr3) (Rep et al., 2004), SIX3 (Avr2) (Di et al., 2017b), SIX4 (Avr1) (Houterman et al., 2008) and SIX6 (Gawehns et al., 2014) are genuine effectors and contribute to fungal virulence on tomato. Many of these effectors were found to be specific for the tomatoinfecting strain, providing the means to identify this pathogen based on its effector profile. Based on the features of these Fol SIX effector genes an effector prediction pipeline could be constructed in which putative effector genes in Fo can be identified based on: 1) a relatively small size (>25 aa and <300 bp), 2) presence of a signal peptide for secretion, and 3) proximity to a "miniature impala" transposable element (van Dam et al., 2016). Analyses of predicted Fo effectoromes revealed that these are shared between strains from a f.sp. infecting the same host, while they are divergent for those infecting other plant species, thereby allowing distinction of ff.spp. based on their effector profiles (Lievens et al., 2009; van Dam et al., 2016; van Dam et al., 2018). Hence, in contrast to phylogenetic analyses, Fo effectorome exploration proves a powerful tool to predict pathogenicity and the potential plant host for a given strain. Whether a potential pathogen is indeed able to cause disease ultimately depends on the corresponding genotype of the host. If the host carries a resistance gene recognizing a specific effector of the pathogen this may result in activation of gene-forgene-based resistance response restricting host colonization (Flor, 1956; Jones and Dangl, 2006). For instance, the Fol effector proteins Avr1, Avr2, or Avr3 are recognized by the tomato resistance proteins I, I-2, or I-3 (Simons et al., 1998; Catanzariti et al., 2015; Catanzariti et al., 2017) resulting in the activation of a resistance response in plants carrying these genes (Houterman et al., 2008). Similarly, the effector protein AvrFom2 from the melon pathogen Fo f.sp. melonis can be recognized by the melon resistance protein Fom2 thereby conferring avirulence to the fungus (Risser et al., 1976; Schmidt et al., 2016).

Non-pathogenic strains share a set of conserved putative effector genes with pathogenic strains, but typically carry much fewer effector candidates and no or few host-specific effectors (van Dam et al., 2016). This notable difference provides a means to distinguish potential pathogens from non-pathogens by the number of candidate effectors they carry. It is tempting to speculate that the candidate effectorome of non-pathogens determines their capacity to colonize roots and confer EMR. Unfortunately, little is known of the role of effectors for Fo endophytes. One putative effector, CS20EP, of the EMRconferring CS-20 strain was reported to trigger a defense response in tomato against Fol when applied prior to inoculation with the pathogen (Shcherbakova et al., 2015). The protein was identified in the culture filtrate of in vitro-grown fungus. However, whether the CS20EP gene is actually expressed during host root colonization awaits future study, as does its role in EMR, for which a knockout strain should be assessed. Altogether, the predicted effector profile from a Fo strain allows its classification as a likely endophyte or as a putative (a)virulent pathogen on a given host. The increasing number of Fo genomes becoming available allows f.sp.-specific effector candidates to be identified and to more precisely predict hostspecific pathogenicity of a given strain. Functional analysis of these effectors, and identification of their host targets, could provide new leads to combat pathogens (Gawehns et al., 2013).

#### THE TIMING AND AMPLITUDE OF ROOT RESPONSES UPON COLONIZATION BY ENDOPHYTIC OR PATHOGENIC F. OXYSPORUM DIFFER

Plant roots are typically exposed to a highly diverse soil microbiota (Hacquard et al., 2017). Plants recognize microorganisms via microbe-associated molecular patterns (MAMPs) that are present in both pathogens and nonpathogens (Henry et al., 2012). Well-known fungal MAMPs are chitin (Kaku et al., 2006) and ß-glucan (Cheong and Hahn, 1991). MAMP recognition is mediated by pattern recognition receptors (PRRs) located at the cell surface (Macho and Zipfel, 2014) such as the CERK1 chitin receptor of Arabidopsis (Miya et al., 2007). Forward genetics in Arabidopsis identified the receptor-like kinase MIK2 as a potential PRR and as a crucial component to recognize and respond to MAMPs from Fo (Coleman et al., 2019). PRRs are mainly expressed in root zones vulnerable to pathogen entry resulting in a heterogenic and tissue-specific responsiveness to different MAMPs (Chuberre et al., 2018). Responsiveness to chitin, for instance, is mostly confined to the mature zone and other parts of the root system are relatively insensitive to this MAMP and do not mount immune responses upon exposure to chitin (Millet et al., 2010; Coleman et al., 2019). This heterogeneity could explain why Fol typically does not penetrate mature root zones (Mes et al., 2000). In the responsive zones MAMP recognition results in activation of pattern-triggered immunity (PTI), which confers resistance to a wide variety of potential pathogens (Jones and Dangl, 2006; Bigeard et al., 2015).

Activation of PTI induces a variety of early signaling responses, such as a cellular Ca2+ and H+ influx resulting in extracellular alkalinization, production of reactive oxygen species (ROS), and phosphorylation of mitogen-associated protein kinases (MAPKs) (Bigeard et al., 2015; Chuberre et al., 2018). A number of PTIassociated signaling responses that have been monitored in cell cultures upon Fo treatment are depicted in Figure 1. Cell cultures have been instrumental to study early plant responses to Fo (Olivain et al., 2003; Humbert et al., 2015). Flax cell cultures exposed to germinated microconidia of Fo47 show a stronger extracellular alkalinization response than cells treated with pathogenic Fo f.sp. lini (Foln) (Figure 1) (Olivain et al., 2003). Also the Ca2+ influx was higher upon Fo47 exposure than to Foln

application (Olivain et al., 2003). Ca2+ influx activates calmodulin (CaM), and in cucumber roots the CaM signal transduction pathway was more strongly induced upon CS-20 colonization than when treated with pathogenic Foc (Pu et al., 2014), implying a weaker PTI induction by pathogenic strains.

ROS, besides being signaling molecules, have direct toxic effects on microbes (O'Brien et al., 2012) and can induce cell death thereby limiting progression of biotrophic pathogens (Mur et al., 2008). Within minutes Fo47 and Foln induce a similar early ROS burst in flax cells (Olivain et al., 2003). Fo47, however, also triggers a second more vigorous burst 3 h post-exposure, which is absent upon Foln treatment. Fo47 also induced more cell death than Foln especially at 14 hpi. A similar observation was made using tomato cell cultures incubated with germinated microconidia of Fo47 or Fol (Humbert et al., 2015). Analogously, inoculation of non-pathogenic Fo that triggers EMR against pathogenic Foa in asparagus induced a cell death response (≈10% cell death) in roots while no cell death was detected when Foa alone was inoculated (He et al., 2002). Transcriptome analysis of soybean roots infected by pathogenic Fo pathogen revealed upregulation of several MAPKs at a relative late stage (72 hpi) of infection, while none was induced by an endophytic strain (Lanubile et al., 2015). Whether MAPKs are differentially phosphorylated in an interaction between roots and Fo endophytes or pathogens remains a question for future study. Taken together, whereas both endophytes and pathogenic Fo strains trigger early PTI signaling, these responses are typically less pronounced in the presence of the latter, suggestive of stronger immune suppression by pathogenic strains.

An effective PTI response results in a transient, local and systemic transcriptional reprogramming of the host (Boller and Felix, 2009; Millet et al., 2010; Bigeard et al., 2015; Chuberre et al., 2018). For instance, in Fo47-inoculated pepper roots a transient expression of a PR-1 protein, a chitinase and a sesquiterpene cyclase (involved in capsidiol synthesis) was observed at 48 hpi, after which expression returned to basal levels at 120 hpi (Veloso and Díaz, 2012). In tomato roots, Fo47 and Fol did not differentially affect expression of a set of PR marker genes when monitored at 48, 72, or 96 hpi: two chitinases (CHI9 and CHI3), two glucanases (GLUB and GLUA), a lypoxygenase (LOXD) and PR-1a (Aimé et al., 2013). However, during later stages of infection at 6 to 22 days post-inoculation (dpi), Fo47, unlike Fol, did not trigger accumulation of PR transcripts (Aimé et al., 2008). In contrast, in cucumber roots CS-20 did induce major transcriptional changes, and at 72 hpi there was a strong induction of PR3, LOX1, PAL1, and NPR1 and of CaMs, CsCam7 and CsCam12 being the strongest induced. At the same time point pathogenic Foc induced NPR1 and to a lesser extent PR3 and PAL1 expression (Pu et al., 2014). RNA-seq analysis of soybean roots inoculated with endophytic or pathogenic Fo revealed that the latter induced more, and stronger, transcriptional changes at 72 and 96 hpi (Lanubile et al., 2015). The literature is ambiguous regarding transcriptional reprogramming in plant–Fo interactions, which might originate from dissimilarities in experimental setup, sampling time, and/or plant-endophyte combination (Table 1). Together the data shows that transcriptional reprogramming following Fo endophyte colonization varies depending on the strain, but typically is transient and returns to basal levels within days. The observation that CS-20 affects transcriptional responses more strongly than Fo47 correlates with CS-20 being a more potent EMR-inducer (Larkin and Fravel, 1999). Xylem sap proteome analysis of susceptible tomato plants showed a significant change in abundance of up to 92% of the identified proteins at 2 weeks post-inoculation of Fol (Gawehns et al., 2015; de Lamo et al., 2018). Among the proteins showing the highest induction are the PR proteins PR-1 and PR-10. Contrarily, at the same time point no significant changes were detected in the proteome of Fo47-inoculated plants as compared to mock treatment (de Lamo et al., 2018). In summary, the transcriptional reprogramming in response to Fo endophytes is confined to the first days of the interaction, while pathogenic strains induce changes mostly during later stages when disease symptoms emerge. At these later stages, major changes are also detected in the xylem sap proteome of diseased tomato plants.

PTI is hypothesized to result in establishment of physicochemical barriers such as callose depositions at the cell walls and exudation of phytoalexins aimed at restricting microbial invasion (Millet et al., 2010). In Fo47-inoculated pea roots, host cellwall penetration attempts by the endophyte appear constrained by callose-containing papillae depositions (Benhamou and Garand, 2001). Similar observations have been made in Fo–cucumber (Benhamou et al., 2002), Fo–flax (Olivain et al., 2003), and Fo– tomato interactions (Le Floch et al., 2009). Fo47 also induced accumulation of the phenolic compound caffeic acid in pepper roots at 48 hpi (Veloso et al., 2016), whereas in tomato roots CS-20 induced accumulation of ferulic acid at 72 hpi (Panina et al., 2007). Both compounds have in vitro antimicrobial activity to Verticillium dahliae (Veloso et al., 2016). Pea roots colonized by Fo47 respond by formation of an osmiophilic compound coating the secondary wall and the pit membranes of the vessel lumen (Benhamou and Garand, 2001). Inoculation with pathogenic Fo f.sp. pisi did not trigger these types of responses in pea (Benhamou and Garand, 2001).

Taken together, both Fo endophytes and pathogens trigger local PTI responses but these appear suppressed/evaded by the latter, likely by the secretion of host-specific effectors. Indeed, Fo effectors have been identified that suppress PTI, a prime example being Fol Avr2 that suppresses ROS production, callose deposition, MAPK phosphorylation, and growth-inhibition upon MAMP application (Di et al., 2017b). Recently, a chitin deacetylase (PDA1) has been found to be required for pathogenicity of Fo f.sp. vasinfectum to cotton (Gao et al., 2019). This provides evidence of a PTI avoidance strategy as de-acetylation of chitin converts it into chitosan, which is a poor inducer of PTI (Gao et al., 2019). Another strategy to evade PTI activation is masking fungal MAMPs. LysM-containing effectors in Cladosporium fulvum are involved in chitin-binding, thereby preventing their perception by the host (Bolton et al., 2008). LysM domain-containing effector genes are also present in Fo genomes (de Jonge et al., 2010; de Sain and Rep, 2015) and a LysM-containing protein secreted by Fol has been identified in tomato xylem sap (Gawehns et al., 2015; de Lamo et al., 2018). However, further research should clarify whether its role in pathogenicity is similar to that of C. fulvum. In summary, successful suppression of PTI by Fo pathogens seems required to cause disease and strains unable to do so, e.g. because they lack the proper host-specific effectors, do not cause disease and exert endophytic lifestyles.

### EMR INVOLVES LOCALIZED CELL DEATH AND ACCUMULATION OF SPECIFIC PR PROTEINS IN THE XYLEM SAP

The molecular and physiological changes in roots during EMR have been studied in some detail and were mostly focused on changes in transcriptome, metabolome, and xylem sap proteome. Whereas Fo endophytes typically trigger an early, minor, and transient change in gene expression, pathogenic strains induce a major and later (days) transcriptional reprogramming during the onset and development of disease (see The Timing and Amplitude of Root Responses Upon Colonization by Endophytic or Pathogenic F. oxysporum Differ). In tri-partite interactions surprisingly little changes in gene expression have been reported. One study of Fo47-inoculated tomato roots challenged with Fol revealed induction of transcripts encoding an acidic extracellular chitinase (CHI3), an acidic extracellular ß-1,3-glucanase (GLUA), and PR-1a 48 h after inoculation (Aimé et al., 2013).

Metabolomic studies revealed that pre-treatment of pepper plants with Fo47 2 days prior to V. dahliae inoculation enhanced the accumulation (at 8 and 24 hpi) of a phenolic acid, chlorogenic acid, in the roots in response to the latter (Veloso et al., 2016). Phenolic acids are involved in fortification of cell walls when crosslinked to cell wall polymers by a ROS-catalyzed process (McLusky et al., 1999; Bubna et al., 2011; O'Brien et al., 2012). Phenolics, together with callose, ROS, peroxidases, and structural proteins form the major constituents of the papillae depositions that are proposed to block cell entry of Fo (Underwood, 2012). In agreement, pre-treatment of cucumber roots with Fo47 resulted in more papillae depositions preventing P. ultimum to penetrate host cells (Benhamou et al., 2002). Another physiological aspect of EMR is the endophyte-induced host cell death during early stages of colonization. This phenomenon seems to be common among non-pathogenic strains as Fo endophytes typically induce host cell death in the root cortex to a larger extent than Fo pathogens during early stages of infection (He et al., 2002; Olivain et al., 2003; Humbert et al., 2015; Gordon, 2017). Noteworthy, in a mutagenesis screen of different Fo endophytes, those losing their ability to trigger biocontrol also showed a reduced induction of host cell death in cell cultures despite retaining its host colonization capabilities (Trouvelot et al., 2002; L'Haridon et al., 2007; Alabouvette et al., 2009).

Pathogenic Fo strains show reduced vasculature colonization upon EMR induction, which might be caused by a change in the xylem sap proteome. To address this hypothesis the xylem sap proteome of tomato plants inoculated with Fo47 and/or Fol was compared (de Lamo et al., 2018). Of the 388 quantifiable proteins, the abundance of only two proteins was strongly increased in the tri-partite interaction as compared to the mock controls. Accumulation of these two proteins, a b-glucanase and NP24, was induced 45- and 33-fold respectively as compared to the control. b-Glucanases exert its antimicrobial activity by hydrolyzing glucan molecules, one of the most abundant polysaccharides in fungal cell walls (Stintzi et al., 1993). Furthermore, the released ß-1,6-glucans act as fungus- and oomycete-specific MAMPs triggering host immune responses (Fesel and Zuccaro, 2016). NP24 is a member of the PR-5 family that includes osmotin and thaumatin-like proteins (Stintzi et al., 1993; Liu et al., 2010). PR-5 proteins exert in planta antimicrobial activity against the pathogens Phytophthora infestans (Woloshuk et al., 1991), P. capsici, and Fo (Mani et al., 2012) by disrupting their plasma membrane integrity via the formation of pores (Vigers et al., 1992). In addition, some PR-5 proteins exert ß-1,3-glucanase activity that could contribute to their antimicrobial activity (Grenier et al., 1999; Menu-Bouaouiche et al., 2003). Besides a direct effect on the pathogen, overexpression of a plum PR-5 in Arabidopsis activated the production of the phytoalexin camalexin (El-kereamy et al., 2011). The correlation between EMR and NP24 abundance is intriguing, as the only differentially accumulated protein in the xylem sap of resistant tomato plants inoculated with an avirulent Fol strain is also a PR-5 family member. Accumulation of this xylem sap-specific PR-5x protein was induced 158-fold upon inoculation of the avirulent pathogen. In a compatible interaction the abundance of the protein also increased, but to a much lower extent (de Lamo et al., 2018). The finding that PR-5 isoforms also specifically accumulate in xylem sap of susceptible and resistant Brassica oleracea infected with Fo f.sp. conglutinans (Pu et al., 2016) further indicates a role for these proteins in controlling the proliferation of pathogenic Fo strains in the vasculature. The observation that pathogenicity-compromised Fol strains in which specific effectors are deleted trigger an >200 fold induction of NP24 in the xylem sap provides additional support for this hypothesis (Gawehns et al., 2015). How these Fol effectors affect accumulation of PR-5 isoforms in tomato is unknown, but various plant pathogens, including V. dahliae (Zhang et al., 2019), Blumeria graminis (Pennington et al., 2016), and B. cinerea (Gonzalez et al., 2017), secrete effectors that directly target PR-5 proteins, stressing their importance in plant fungal interactions.

Altogether, Fo-based EMR seems to be a root-mediated response that triggers, among other responses, specific accumulation of xylem sap-localized PR-5 and ß-glucanase proteins and secretion of phenolic compounds that together with ROS are involved in cell wall lignification and callose depositions. Furthermore, host cell death induced by Fo endophytes correlates with the induction of an effective EMR response.

### EMR IS DISTINCT FROM INDUCED SYSTEMIC RESISTANCE AND SYSTEMIC ACQUIRED RESISTANCE RESPONSES

Many studies attribute Fo-induced resistance response in plants as the main contributor to biocontrol (Table 1). Split-root systems, in which the Fo endophyte is spatially separated from the pathogen, have shown that EMR can act systemically in root tissues (Kroon et al., 1991; Fuchs et al., 1997; Duijff et al., 1998; Larkin and Fravel, 1999; Kaur and Singh, 2007; Pantelides et al., 2009; Aïcha et al., 2014). The mechanism that transduce this signal to distant root tissues is unknown. Root colonization by endophytes such as S. indica or Trichoderma spp. triggers an induced systemic resistance response (ISR) that relies on the phytohormones jasmonic acid (JA) and ethylene (ET) (Shoresh et al., 2010; Franken, 2012). Other microbes, especially avirulent pathogens, can trigger a salicylic acid (SA)-dependent immune response, which results in systemic acquired resistance (SAR) (Pieterse et al., 2014). Both systemic responses prime the plant to respond faster and stronger to subsequent pathogen attack, thereby reducing it susceptibility to foliage-attacking pathogens (Durrant and Dong, 2004; Fu and Dong, 2013; Pieterse et al., 2014). The observation that Fo endophytes typically do not confer protection to pathogens attacking above-ground tissues (Table 1) raises the question whether EMR mechanistically differs from ISR or SAR. Only two reports describe Fo-induced resistance to a foliar pathogens (Biles and Martyn, 1989; Díaz et al., 2005). It was reported that a Fol strain that is pathogenic on tomato reduced susceptibility to B. cinerea in pepper (Díaz et al., 2005). Pre-treatment with 1-methylcyclopropene, an inhibitor of ET perception, compromised this Fol-induced plant protection to the fungus. The involvement of ET in this response implies that the nonhost pathogen Fol can trigger ISR in pepper. Remarkably, Fo47 inoculation did not confer protection against B. cinerea in the same experimental setup, although this strain triggered EMR (Veloso and Díaz, 2012), suggesting that Fol triggers both. The other example details the cucumber-pathogen Foc that induced systemic responses in aerial tissues in watermelon (Biles and Martyn, 1989). It will be interesting to investigate whether both pathogenic Fo strains carry effectors that are recognized by these non-host plants responsible for triggering a systemic ISR-type immune response. These examples imply that non-host pathogenic Fo strains can induce both ISR and EMR, while purely endophytic strains trigger only the latter response.

Whereas tomato mutants compromised in SA signaling are hypersensitive to Fusarium wilt disease, an increased tolerance was observed in mutants affected in ET biosynthesis or perception (Di et al., 2017a). In contrast, susceptibility of tomato mutants deficient in JA biosynthesis towards Fol was unaffected, showing that these three phytohormones have distinct roles in the interaction between tomato and pathogenic Fo (Di et al., 2017a). The interaction between these phytohormones and Fo is complex and differs for different pathosystems (Di et al., 2016). To elucidate the role of these defense phytohormones in EMR, Constantin and co-workers analyzed Fo47-induced immune responses in wild-type tomato plants and in mutants compromised in ET, JA or SA signaling (Constantin et al., 2019). Expression of ET marker genes (Pti4 and ETR4) was not induced in a tri-partite tomato–Fol–Fo47 interaction suggesting that ET is not involved in EMR (Constantin et al., 2019). Indeed, EMR was intact in tomato lines affected in either their ability to sense- (never-ripe mutant) or produce ET (transgenic lines constitutive expressing ACC deaminase) (Constantin et al., 2019). Also tomato plants with a defect in JA biosynthesis (def1) were still capable of mounting EMR upon co-inoculation with Fo47 and Fol. These findings make involvement of ISR in EMR unlikely, as this response requires intact ET/JA signaling pathways (Pieterse et al., 2014). Likewise, SAR, which requires SA, appears not to be involved as tomato lines compromised in SA accumulation (expressing NahG) exert a functional EMR response against Fol (Constantin et al., 2019). Together, these findings support a model in which EMR induced by Fo47 is distinct from ISR and SAR, as these responses require either JA/ET or SA and result in induced resistance in shoots, unlike EMR that is mostly root confined.

#### DISCUSSION

Based on the data presented we propose a mechanistic model on how Fo-induced EMR prevents disease. Figure 2 illustrates an early (≈ two dpi) interaction between a root and Fo. Both pathogenic and nonpathogenic Fo strains colonize the root epidermis and cortex. Whereas pathogenic Fo strains effectively compromise immune signaling by secreting effector proteins (Figure 2A) endophytes are unable to do so and trigger immune activation (Figure 2B). The transient induction of immune signaling confines the non-pathogenic fungus to the root cortex and restricts its growth by preventing entry into host cells by the formation of papillae and cell wall fortifications. Together these responses prevent the fungus from reaching the vasculature and causing disease. Localized cell death induced by Fo endophytes (He et al., 2002; Alabouvette et al., 2009) appears to be involved in the induction of EMR, because Fo mutants that lost their ability to induce cell death are also unable to trigger EMR even though they can still colonize the roots (Alabouvette et al., 2009). Endophytes such as Harpophora oryzae or the phylogenetically distant basidiomycete S. indica, are also known to trigger localized cell death upon root colonization (Deshmukh et al., 2006; Su et al., 2013). It is tempting to speculate that induction of host cell death by endophytes may be a generic property required for EMR induction. Possibly cell death primes, or potentiates, immune responses to an extent that they can no longer be mitigated by the effectors secreted by the pathogen. The potentiated immune responses restrict pathogen development in a tri-partite interactions and results in a reduced xylem sap colonization (Figure 2C). Although pathogenic Fo strains colonize the vasculature in tri-partite interactions their proliferation is reduced as are the disease symptoms. We speculate that the reduced ability to colonize the vasculature is in part due to the increased abundance of PR-5 protein family members and plant-produced ßglucanases. Assessing the biocontrol properties of Fo in plants in which these genes are knocked-out can put this hypothesis to the test.

#### CONCLUSION

Although being studied for over more than three decades the mechanism underlying EMR remains elusive. Understanding this inducible defense mechanism, which confers protection against root-invading vascular pathogens, holds potential for improved control of wilt diseases without affecting the conventional defense pathways. Future studies focusing on the nature of the systemic signal, the role of secondary metabolites, PR protein production, and papillae formation in tri-partite

or by both in a tri-partite interaction in which EMR is triggered (C). The drawings depict an interaction around 2 days after inoculation.

interactions will be instrumental to get a better understanding of the mechanism underlying EMR. Elucidating the relation between localized host cell death and EMR will reveal whether damage merely amplifies, or is essential, to trigger this immune response. Studying the potential role of host-specific and generic effector candidates in modulating EMR will increase our understanding of the endophytic side of the interaction, possibly allowing selection of endophytic strains conferring robust biocontrol in agricultural settings. A concern is that horizontal chromosome transfer from pathogenic Fo strains to the applied Fo endophytes could turn the latter into pathogens. Whether chromosome transfer occurs in natural setting should be investigated before agricultural application.

### AUTHOR CONTRIBUTIONS

FL and FT conceived and wrote the manuscript.

### REFERENCES


#### FUNDING

FL and FT are supported by the BestPass project. This project is funded by the European Union's Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement No. 676480 (International Training Network BestPass). FT also is supported by the NWO-Earth and Life Sciences funded VICI project No. 865.14.003.

#### ACKNOWLEDGMENTS

The authors are grateful to Martijn Rep and Ben Cornelissen for their feedback on the manuscript, to Maria Constantin and Babette Vlieger for providing the microscopy pictures used in Figure 1, and to both reviewers for their constructive, critical, and detailed comments.


inducing Fusarium oxysporum. New Phytol. 184 (3), 529–544. doi: 10.1111/ j.1469-8137.2009.03014.x


independent of ET, JA, and SA. Front. Plant Sci. 10. doi: 10.3389/ fpls.2019.00979


colonization of Arabidopsis roots by the Fungus Piriformospora indica. Plant Physiol. 157 (1), 531–531. doi: 10.1104/pp.111.900420


BEC1054 and barley proteins identify candidate host targets. J. Proteome Res. 15 (3), 826–839. doi: 10.1021/acs.jproteome.5b00732


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 de Lamo and Takken. 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.

# Beneficial Endophytic Bacterial Populations Associated With Medicinal Plant Thymus vulgaris Alleviate Salt Stress and Confer Resistance to Fusarium oxysporum

#### Edited by:

Massimiliano Morelli, Italian National Research Council, Italy

#### Reviewed by:

Zhansheng Wu, Shihezi University, China Petra Lovecka, University of Chemistry and Technology in Prague, Czechia

#### \*Correspondence:

Osama Abdalla Abdelshafy Mohamad Osama@Aru.Edu.Eg Wen-Jun Li liwenjun3@mail.sysu.edu.cn Li Li lili.bobo@outlook.com † These authors have contributed equally to this work

#### Specialty section:

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

Received: 13 July 2019 Accepted: 14 January 2020 Published: 14 February 2020

#### Citation:

Abdelshafy Mohamad OA, Ma J-B, Liu Y-H, Zhang D, Hua S, Bhute S, Hedlund BP, Li W-J and Li L (2020) Beneficial Endophytic Bacterial Populations Associated With Medicinal Plant Thymus vulgaris Alleviate Salt Stress and Confer Resistance to Fusarium oxysporum. Front. Plant Sci. 11:47. doi: 10.3389/fpls.2020.00047 Osama Abdalla Abdelshafy Mohamad1,2,3\*† , Jin-Biao Ma1† , Yong-Hong Liu<sup>1</sup> , Daoyuan Zhang<sup>1</sup> , Shao Hua<sup>1</sup> , Shrikant Bhute<sup>3</sup> , Brian P. Hedlund<sup>4</sup> , Wen-Jun Li 1,5\* and Li Li 1\*

<sup>1</sup> CAS Key Laboratory of Biogeography and Bioresource in Arid Land, Xinjiang Institute of Ecology and Geography, Urumqi, China, <sup>2</sup> Department of Biological, Marine Sciences, and Environmental Agriculture, Institute for Post Graduate Environmental Studies, Arish University, Al-Arish, Egypt, <sup>3</sup> Department of Environmental Protection, Faculty of Environmental Agricultural Sciences, Arish University, Al-Arish, Egypt, <sup>4</sup> School of Life Sciences, University of Nevada, Las Vegas, NV, United States, <sup>5</sup> State Key Laboratory of Biocontrol and Guangdong Provincial Key Laboratory of Plant Resources, School of Life Sciences, Sun Yat-Sen University, Guangzhou, China

As a result of climate change, salinity has become a major abiotic stress that reduces plant growth and crop productivity worldwide. A variety of endophytic bacteria alleviate salt stress; however, their ecology and biotechnological potential has not been fully realized. To address this gap, a collection of 117 endophytic bacteria were isolated from wild populations of the herb Thymus vulgaris in Sheikh Zuweid and Rafah of North Sinai Province, Egypt, and identified based on their 16S rRNA gene sequences. The endophytes were highly diverse, including 17 genera and 30 species. The number of bacterial species obtained from root tissues was higher (n = 18) compared to stem (n = 14) and leaf (n = 11) tissue. The endophytic bacteria exhibited several plant growth-promoting activities in vitro, including auxin synthesis, diazotrophy, phosphate solubilization, siderophore production, and production of lytic enzymes (i.e., chitinase, cellulase, protease, and lipase). Three endophytes representing Bacillus species associated with T. vulgaris such as EGY05, EGY21, and EGY25 were selected based on their ex-situ activities for growth chamber assays to test for their ability to promote the growth of tomato (Solanum lycopersicum L.) under various NaCl concentrations (50–200 mM). All three strains significantly (P < 0.05) promoted the growth of tomato plants under salt stress, compared to uninoculated controls. In addition, inoculated tomato plants by all tested strains decreased (P < 0.05) the activity of antioxidant enzymes (superoxide dismutase, catalase, and peroxidase). Six strains, representing Bacillus and Enterobacter species EGY01, EGY05, EGY16, EGY21, EGY25, and EGY31 were selected based on in vitro antagonistic activity to F. oxysporum for pot experiments under salt stress. All tested strains reduced the disease severity index (DSI) of tomato plants at all tested salt concentrations. Gas-chromatography/mass-spectrometry analysis of cell-free extracts of B. subtilis (EGY16) showed at least ten compounds were known to have antimicrobial activity, with the major peaks being benzene, 1,3-dimethyl-, p-xylene, dibutyl phthalate, bis (2-ethylhexyl) phthalate, and tetracosane. This study demonstrates that diverse endophytes grow in wild thyme populations and that some are able to alleviate salinity stress and inhibit F. oxysporum pathogenesis, making them promising candidates for biofertilizers and biocontrol agents.

Keywords: environmental microbiology, medicinal plants, endophytes, biofertilizer, biocontrol, Fusarium oxysporum, Bacillus subtilis, Thymus vulgaris

#### INTRODUCTION

Our life is dependent on plants as they produce oxygen and staple foods for domestic animals and humans. The beginning of 21st century is marked by global scarcity of water resources and a 10% annual increase in salinized areas; The Food and Agricultural Organization (FAO) expects an increasing human population to reach 8–9 billion by 2030 (FAO, 2010). As a result of climate change, it has been estimated that nearly 800 million hectares of global land and 32 million hectares of agricultural land has been affected by salinization (El-Beltagy and Madkour, 2012; Ahmad et al., 2015; FAO, 2015). Increased salinity in arid and semiarid regions is a devastating abiotic stress that inhibits normal growth and development of plants, and increases susceptibility of plants to soil-borne pathogens, resulting in considerable reductions in agricultural production and disruption of ecological balance in natural areas (Egamberdieva et al., 2011; Egamberdieva et al., 2013; Li et al., 2018; Mohamad et al., 2018).

In order to enhance agricultural productivity, it is crucial to understand the physiological and biochemical properties evolved by plants to alleviate salt stress (Ahmad et al., 2010a). Plants are equipped with a variety of enzymatic and non-enzymatic mechanisms against salt stress. The antioxidant defense system plays a major role in plant adaptation to salt stress and includes enzymes such as superoxide dismutase (SOD), ascorbate peroxidase (APX), catalase (CAT), and glutathione reductase (GR); non-enzymatic antioxidants include ascorbic acid (ASA) and glutathione (GSH) (Mittler, 2002; Ahmad and Sharma, 2008; Ahmad et al., 2010b; Rasool et al., 2013). Plant cells generate reactive oxygen species (ROS) under normal conditions, but the excessive production of ROS due to osmotic stress causes oxidative damage and affects the functional integrity of cells (Ahmad et al., 2010a; Ahmad et al., 2012c).

Recently, several strategies to alleviate the toxic effects caused by high salinity have been developed, including the use of beneficial microorganisms that act through various direct and indirect mechanisms (Wang et al., 2003; Nia et al., 2012; Ramadoss et al., 2013). Detailed knowledge of interactions between bacteria and plants under salt stress is limited, and the role of microbes in the management of biotic and abiotic stresses is gaining importance for alleviating salinity stress in saltsensitive crops (Yao et al., 2010; Egamberdieva et al., 2013).

The establishment of successful interactions between microbial symbionts and plants facilitates Induced Systemic Tolerance (IST) response, which has been suggested to be important for eco-friendly and sustainable agriculture (Grover et al., 2011; Beneduzi et al., 2012; Schlaeppi and Bulgarelli, 2015). Endophytes live inside plant tissues without damaging and causing any disease and in many cases, they directly or indirectly improve plant performance under stress by providing them with phytohormones, fixed nitrogen, and nutrients such as soluble iron, potassium, and phosphate (Hayat et al., 2010; Khan et al., 2017; Li et al., 2018).

In the last decade, a vast number of studies have supported the hypothesis that plant growth-promoting bacteria (PGPB) enable plants to maintain productivity under different abiotic stresses by various mechanisms. These bacteria belong to genera such as Rhizobium, Bacillus, Pseudomonas, Pantoea, Paenibacillus, Burkholderia, Achromobacter, Azospirillum, Microbacterium, Methylobacterium, Variovorax, and Enterobacter (Hamdia et al., 2004; Grover et al., 2011; Ahmad et al., 2012a; Bharti et al., 2013; Li et al., 2018). In addition, numerous studies have shown that PGPB enhance root and shoot growth and induce the antioxidant defense system in crops such as lentil (Lens esculenta) (Faisal, 2013), pea (Pisum sativum L.) (Meena et al., 2015), cucumber (Cucumis sativus), (Egamberdieva et al., 2011), rice (Oryza sativa) (Yadav et al., 2014) and soybean (Glycine max) (Egamberdieva et al., 2016). Therefore, endophytic plant symbionts provide excellent models for understanding the plant-microbe interactions that could potentially be engineered to maintain crop productivity to cope with climate change-induced stresses (Grover et al., 2011).

Medicinal plants have a long history in the development of human culture, and in many developing countries modern medicines are produced from traditional medicinal plants (Tian et al., 2014; Egamberdieva et al., 2017a). However, studies linking the growth of traditional medicinal plants to specific microbial interactions remain in incipient stages (Köberl et al., 2013b). Indeed, the composition of bioactive secondary metabolites synthesized by medicinal plants widely depending on the plant species; these metabolites can strongly affect their association with endophytic microbes (Qi et al., 2012; Chaparro et al., 2014; Cushnie et al., 2014). Hence, understanding the response of microbial communities associated with medicinal plants to alterations in the physicochemical environment of the rhizosphere may provide valuable insights into the ecology of plant-associated bacteria (Köberl et al., 2013a; Egamberdieva and da Silva, 2015; Mohamad et al., 2018). Thymus is a traditional medicinal plant in the mint family, Lamiaceae, and has a wide distribution in Africa, Europe, and Asia (Hosseinzadeh et al., 2015). Thymus vulgaris is used worldwide as an infusion to treat respiratory ailments such as colds, congestion, sore throat, and both upper and lower respiratory infections, diabetes, and intestinal infections and infestations. It also has been described to have antiseptic, antibacterial, and antifungal properties (Ekoh et al., 2014; Prasanth Reddy et al., 2014). However, to date Thymus has not been investigated with respect to microbial communities.

Tomato is another economically important crop worldwide. Tomato plants are sensitive to vascular wilt diseases by Fusarium oxysporum, particularly under salt stress. (Inami et al., 2014; McGovern, 2015). Symptoms of F. oxysporum include yellowing of older leaves and browning of vascular tissues (Heitefuss, 2012). Management of Fusarium wilt of tomato is challenging and globally important. Therefore, the objectives of the present study were to 1) isolate and identify beneficial endophytic bacteria associated with wild T. vulgaris; 2) evaluate their growth-promoting and salt stress alleviating ability on tomato plants; 3) evaluate their biological control ability against F. oxysporum in vitro/vivo; and 4) identify prevalent volatile organic compounds (VOCs) produced by endophytes only in the presence of F. oxysporum, which are likely to be effectors of the antifungal properties. To the best of our knowledge, this is the first report on the isolation, identification, and characterization of endophytic bacteria associated with the wild medicinal plant T. vulgaris.

### MATERIALS AND METHODS

#### Sample Collection, Plant Material, and Sterilization

Symptom-free T. vulgaris plants were collected in the summer of 2016 from their natural arid habitats in North Sinai Province of Egypt. The study sites were within the desert of North Sinai districts Rafah (31°18′16.4″N 34°13′13.3″E) and Sheikh Zuweid (31°05′26.2″N 33°59′48.9″E). T. vulgaris populations are abundant at these sites due to their adaptation to poor sandy soil and arid climatic conditions. At each location, three healthy plants located 2–7 m apart were harvested. Whole plants, including root systems (10–15 cm depth), were aseptically harvested, placed in Zip-loc bags, and stored at 4°C during transportation to the laboratory until further processing.

#### Isolation of Endophytic Bacteria

Each plant sample was thoroughly washed under tap water and subsequently sterile double distilled water to remove adhering epiphytes and sand debris. Plants were then separated into leaves, stems, and roots, and successively cut into small pieces by using sterile scissors. Approximately 5–7 g of plant tissue was excised from each tissue sample and was surface-sterilized under a laminar airflow cabinet by immersing it sequentially with shaking in 0.1% Tween 20 for a few seconds, 70% ethanol for 3 min, and 5% NaOCl for 5 min, with three rinses with sterile, distilled water between each step of surface sterilization in order to remove the residues and smell of the chemicals used in the surface sterilization protocol (Li et al., 2018). Subsequently, all sterilized samples were cut aseptically into 0.5 cm-long segments using sterile blades or scissors and placed on a piece of sterile filter paper in a laminar air flow chamber for 2–3 h. One gram of each tissue sample was weighed separately and macerated with a sterile mortar and pestle, along with 9 ml of sterile phosphate buffered saline (PBS) (20 mM sodium phosphate, 150 mM NaCl, pH 7.4). The homogenate was then transferred to a sterile polypropylene tube and vortexed for 2 min. The tissue homogenate was centrifuged at 2200 × g for 5 min and the supernatant was collected, serially diluted (10–<sup>2</sup> –10–<sup>4</sup> ), and then 150 μl aliquots from appropriate dilutions were spread onto ten different isolation media in triplicate (Table S1) (Li et al., 2018). The agar plates were sealed with parafilm, incubated at 28 °C, and monitored every five days for microbial growth. Several distinct colony morphologies on each isolation medium were re-streaked for purification on ISP2 media. To test the efficiency of plant surface sterilization, controls were also set up by plating 100 μL of sterile ddH2O used for the final step of surface sterilization onto the ten media. No microbial growth was detected on the isolation media after 7 days of incubation at 28 °C. This result indicated that the three-step surface sterilization protocol was successful in killing or at least inhibiting the growth of the epiphytic bacteria. Microbial isolates were therefore considered to be true endophytes. All bacterial isolates were stored in 25% glycerol at −80 °C.

### Genotypic Identification

Molecular identification of bacterial isolates was performed after extraction of DNA using Chelex® 100 sodium following the manufacturer's instructions (SIGMA-ALDRICH). The extracted DNA was dissolved in 20 mL TE buffer and used as the template for PCR. The 16S rRNA gene was amplified using the universal primers 27F (5′-CAGAGTTTGATCCTGGCT-3´) and 1492R (5′-AGGAGGTGATCCAGCCGCA-3´) (Li et al., 2018). Reactions were performed in a Biometra Thermal Cycler. The PCR mixture (25 mL) contained 12.5 mL 2×Taq PCR Master Mix procured from TIANGEN BIOTECH (Beijing, China), 2 mL DNA template (50 ng), and 1 mL of each primer (10 pmol). PCR was performed under the following conditions: initial denaturation step at 95°C for 6 min, followed by 35 cycles of denaturation at 94 °C for 45 (s), annealing at 57 °C for 1 min and extension at 72 °C for 1:30 min, with a final extension step at 72 °C for 10 min. The fragment was sequenced by Shanghai Sangon Biological Engineering Technology & Services Co., Ltd. All 16S rRNA gene sequences were compared with the GenBank database by using the EzBiocloud server (http://www.eztaxon. org) (Chun et al., 2007; Li et al., 2018). All near full-length 16S rRNA gene sequences have been deposited in GenBank under Accession Numbers MH764457–MH764573.

#### In Vitro Screening for Plant Beneficial Traits

#### Indole Acetic Acid (IAA) Production

Salkowski's colorimetric method was used to determine the ability of bacterial endophytes to produce indole-3-acetic acid (IAA). Isolates were grown in 25 ml of TYC broth (3 g L–<sup>1</sup> yeast extract; 5 g L–<sup>1</sup> tryptone; and 0.872 g L–<sup>1</sup> CaCl2∙2H2O) with 0.1% (w/v) L- tryptophan for 2–4 days at 28 ◦ C at 125 rpm (Egamberdieva and Kucharova, 2009; He et al., 2019). After incubation, the broth was centrifuged at 9,302 × g for 5 min, and 1 ml of supernatant was mixed with Salkowski reagent (2 ml of 0.5 M FeCl3, and 98 ml of 35% HClO4) (1:1 v/v) and incubated in the dark for 30 min. Development of pink color indicated indole production. Subsequently, results were confirmed by measuring the optical density (OD) at 530 nm and compared comparing with known amounts of IAA using Salkowski reagent and sterile TYC broth with tryptophan as blanks (Li et al., 2018).

#### Phosphate Solubilization

All endophytic bacterial isolates were screened for solubilization of inorganic phosphate was evaluated qualitatively on solid Pikovskya's medium supplemented with Ca3(PO4)2 (5 g/L) and Bromophenol Blue (0.025 g/L) as described (Paul and Sinha, 2017; Li et al., 2018) with some modifications. After seven days of incubation at 28◦ C, the formation of yellow halos and/or clearing zones was evaluated. The change of color from blue to yellow or the formation of a clear halo around the colonies was indicative of the utilization of tricalcium phosphate present in the agar medium.

### Nitrogen Fixation Activity

To test nitrogen fixation activity, bacterial isolates were tested on two nitrogen-free media: Ashby's mannitol agar, composed of (L−<sup>1</sup> ) (0.2 g KH2PO4; 0.2 g MgSO4; 0.2 g NaCl; 5.0 g CaCO3; 10.0 g mannitol; 0.1 g CaSO4; 15.0 g agar; pH 7.0) and NFC medium (0.2 g KH2PO4; 10.0 g mannitol; 0.2 g MgSO4∙7H2O; 0.2 g NaCl; 0.2 g CaSO4∙2H2O; 5.0 g CaCO3; 15.0 g agar; pH 7.2) (Liu et al., 2016; Li et al., 2018). Isolates were incubated at 28 ◦ C for seven days and nitrogen fixation activity was observed based on colony growth on the agar plates.

#### Production of Siderophores

Siderophore production was screened based on competition for iron (Fe) between ferric complexes of universal chrome azurol S (CAS) agar media as described (Alexander and Zuberer, 1991; Li et al., 2018). Isolates were incubated at 28 ◦ C for 5–7 days. Change of the blue color of the medium surrounding the colony and appearance of an orange/purple or purple/red halo zone was scored as positive for production of siderophores (Li et al., 2018).

#### Assays for Proteolytic, Lipolytic, Cellulolytic, and Chitinolytic Activity

The bacterial strains were checked for proteolytic activity using the spot inoculation technique on skim milk agar 5% (v/v) medium (Tiru et al., 2013). The skim milk agar plates were incubated for 48 h. Proteolytic activity was identified by the formation of a clear halo around the bacterial colonies due to hydrolysis of skim milk (Li et al., 2018).

Lipolytic activity was assayed using the spot inoculation technique using modified Sierra lipolysis agar supplemented with beef extract (3 g L–<sup>1</sup> ) and ferrous citrate (0.2 g L–<sup>1</sup> ). After autoclaving, 50 mL of Victoria Blue B solution (0.1 g per 150 mL) and 10 mL of Tween 80 was added to the medium. After 5–6 days of incubation, white calcium precipitates around bacterial colonies indicated a positive reaction (Li et al., 2018).

Cellulolytic activity was assayed with modified DSMZ medium 65 (http://www.dsmz.de/microorganisms/medium/ pdf/DSMZ\_Medium65.pdf) without CaCO3 and supplemented with carboxymethyl cellulose (5 g L–<sup>1</sup> ; Sigma) in place of glucose by using the spot inoculation technique. After incubation for 5-6 days, plates were stained with a Congo red solution and destained with a NaCl solution (Teather and Wood, 1982; Li et al., 2018; Mohamad et al., 2018). A clear or lightly colored halo around the colonies indicated a positive reaction.

Colloidal chitin was prepared from commercial chitin by the method of Agrawal and Kotasthane (Agrawal and Kotasthane, 2012; Mohamad et al., 2018). Chitinase detection medium consisted of (L−<sup>1</sup> ) 0.3 g of MgSO4.7H2O, 3.0 g of (NH4)2SO4, 2.0 g of KH2PO4, 1.0 g of citric acid monohydrate, 15 g of agar, 200 mL of Tween-80, 4.5 g of colloidal chitin and 0.15 g of bromocresol purple per liter and then autoclaved at 121◦ C for 15 min. Chitinolytic activity was assessed using the spot inoculation technique by observation of colored zones around colonies. All screening experiments for plant beneficial traits were performed twice with three replicates for each individual isolate.

#### Plant Growth-Promoting Activity in Saline Soil

Pot experiments were carried out in a growth chamber to investigate the symbiotic effects of bacterial isolates on plant growth in non-saline and saline soils. Tomato seeds (Solanum lycopersicum. cv. Fuji Pink) were surface-sterilized by immersion in sodium hypochlorite (2% v/v) for 2 min and subsequently rinsed five times with ion-free distilled water (Cao et al., 2004). Sterilized seeds were germinated on a wet filter placed in a Petri dish (9 cm diameter). The Petri dishes were covered with a polyethylene sheet to prevent evaporation and kept in the plant growth chamber at 25◦ C for 4–5 days.

Endophytic strains were grown overnight in ISP2 broth, and the cell suspension was centrifuged. The cell pellets were resuspended at a final concentration of 10<sup>8</sup> CFU/ml with phosphate-buffered saline PBS and adjusted by using Densicheck plus (Biomerieux, USA) (Errakhi et al., 2007; Egamberdieva et al., 2011). Tomato seedlings were inoculated by soaking roots for 15 min in a 1 mL solution each bacterial suspension and shaking gently with sterile forceps (Botta et al., 2013). The inoculated seedlings were aseptically transplanted into plastic pots (12 cm high × 10 cm in diameter) filled with autoclaved compost: sand: perlite: peat (1:1:1:1, v/v/v/v) and placed in a growth chamber at 25 ± 2°C and a 14-h photoperiod. After 3 days, 10 ml of the bacterial suspension (108 CFU/mL) was added near the root zone and salinity was increased gradually by applying a sodium chloride solution to each pot on alternative days to avoid osmotic shock to reach final salt concentrations of 50, 100, 150, and 200 mM and the desired salt concentrations of 50, 100, 150, and 200 mM were achieved after 2, 4, 6, and 8 days, respectively (Nejad and Johnson, 2000; Chatterjee et al., 2017). Parallel controls were maintained by cultivating tomato plants with autoclaved compost without inoculation of endophytes and irrigating with tap water. Each treatment contained three pots and each pot including four tomato seedlings. After 8 weeks, the tomato plants were harvested for growth and antioxidant enzyme assays. Plants were uprooted and washed to remove adhering peat. Shoot and root length and fresh weight were recorded.

#### Plant-Microbe Response Defense Under Saline Condition

#### Determination of Antioxidant Enzymatic Activity

Plant extracts were prepared from tomato leaves after 45 days and antioxidant enzymes were assayed as described by Ahmad et al. (2015). Briefly, fresh tomato leaves were ground using liquid nitrogen and the ground leaf samples were stored at –80 ◦ C. The ground leaf samples (approximately ∼1 g) were homogenized on ice using 10 mL of 50 mM phosphate buffer (pH 7.8) and then incubated for 10 min at 4 ◦ C. Subsequently, the homogenate was filtered using Advantech Qualitative Filter Papers (110 mm) and centrifuged at 4,000 × g for 15 min at 4 ◦ C. The supernatant was used for the determination of enzyme activities. The activities of superoxide dismutase (SOD, EC 1.15.1.1), catalase (CAT) (EC 1.11.1.6), and peroxidase (POD, EC 1.11.1.7), were measured using assay kits (kit Numbers. A001-1, A007-1, A084-3, respectively; Nanjing Jiancheng Bioengineering Institute, China), following the manufacturer's instructions (http://elder.njjcbio.com/index\_en.php) (Liang et al., 2017). This experiment was conducted in triplicate.

#### Photosynthetic Pigments

Chlorophyll content was quantified by using the Leaf Chlorophyll Meter (SPAD 502 Plus). Readings were taken on the uppermost fully expanded leaf with a visible collar during vegetative growth, and from the ear leaf (20 leaves per treatment) as suggested (Dwyer et al., 1995).

#### In Vitro Screening for Antifungal Activities on Solid Medium

Antifungal activity was screened in vitro against the following pathogenic microorganisms F. oxysporum f. sp. (F1), Fulvia fulva (Cooke) Cif (F2), and Alternaria solani Sorauer (F3) (Table S2). Briefly, fungal strains were grown in potato dextrose agar (PDA) plates for 6 days and the mycelial disc (5 mm) was transplanted into the center of a fresh PDA plate. Bacterial cultures pre-grown in ISP2 media for 3 days were symmetrically spotted onto the four corners of the plate, 2.5 cm from the plate periphery. All plates were wrapped with parafilm and incubated at 26 ± 2 ◦ C for 3 days and observed for the inhibition of the pathogen. Activity was quantified by measuring the zone of inhibition of the pathogen's growth (Velusamy et al., 2006; Kilani-Feki et al., 2011) and the percent inhibition was calculated using the following formula (Yasmin et al., 2016; Mohamad et al., 2018).

$$\text{Percent inhibition (\%)} = \frac{\text{F}\_{\text{cd}} - \text{T}\_{\text{fcd}}}{\text{F}\_{\text{cd}} - \text{F}\_0} \times 100\%$$

where Fcd is the fungal colony diameter on the control PDA base plate, Tfcd is the fungal colony diameter on the treatment PDA base plate, and F0 is the diameter of the test fungus agar discs (approximately 5 mm). This experiment was conducted twice in triplicate.

#### Biological Control of Tomato Root Rot by Endophytic Bacterial Strains In Vivo

Bacterial isolates with antagonistic activity against at least two of the tested fungal pathogens were tested for their ability to control tomato root rot caused by F. oxysporum in saline soils. Tomato seedlings were aseptically planted in plastic pots filled with nonsaline and saline soil as described above. F. oxysporum from 5 day-old potato dextrose broth was filtered with 6 layers of sterile gauze to remove mycelia and the spores were washed with sterile water and diluted to a concentration of 10<sup>7</sup> conidia/ml (Biomerieux, USA) (Fakhouri and Buchenauer, 2003). Bacterial suspensions were prepared as described above. The treatments were: (i) control, without pathogen and bacteria; (ii) control with pathogen only; and (iii) pathogen and bacteria. Plants at the four-true-leaf stage were wound-inoculated with spores by puncturing the stem at the first internode above the soil line with a 22-gauge needle by sterile syringe (Hanson, 2000; Fakhouri and Buchenauer, 2003; Chen et al., 2014; Mohamad et al., 2018). 10 mL of the bacterial suspension (10<sup>8</sup> CFU/mL) was applied near the root zone as described above 3 days after wounding.

Disease symptoms were recorded for 45 days following pathogen challenge, including leaf yellowing and chlorosis. Disease severity was classified into six grades (i.e., grades 0, 1, 2, 3, 4, and 5) according to the visible symptoms on the cotyledons and true leaves (Amini, 2009; Zheng et al., 2013; Mohamad et al., 2018). The scale estimated the percentage of affected leaves using five main categories or quarters (≤35, 35–55, 55–65, 65–85 and 85–100%) with five values for each category. The disease index (DI) was calculated according to the following formula: DI = [(∑ disease grades × number of infected)/(total checked plants ×5)] × 100 (Zheng et al., 2013; Mohamad et al., 2018). The disease index (DI) represents a comprehensive and objective measure of plant health, with high DI values corresponding to serious infections. In this experiment, each treatment contained three pots and each pot including four tomato seedlings. Average disease index and standard deviation were calculated for each isolate at each salt concentration; statistical analysis (ANOVA followed by Tukey's Honest Significant Difference post-hoc test) was conducted to identify significant differences.

#### Extraction of Metabolites

The antibiosis experiment was carried out by co-cultivation of strain EGY16 with F. oxysporum in 500 mL–<sup>1</sup> of broth medium at 28◦ C for 12 days with agitation at 150 rpm in triplicate. Cells were collected by centrifugation at 5,000 × g for 10 min. The cellfree supernatant was mixed with an equal volume (1:1) of ethyl acetate by vigorous shaking for 60 min and allowed to settle. The organic solvent phase was evaporated at 40◦ C under vacuum, using a rotary evaporator (IKA, HB10 basic). The ethyl acetate extract was dissolved in 5 mL of Tris–Cl buffer (0.02 M, pH 7.0) and used for gas-chromatography/mass spectrometry (GC-MS) (Mohamad et al., 2018).

### Identification of Metabolites

GC-MS analysis of the cell-free extracts was performed using a gas chromatograph (Model 7890A, Agilent, Palo Alto, CA, USA) equipped with a split-splitless injector, an Agilent model 7693 autosampler, and an Agilent HP-5MS fused silica column (5% Phenyl-methylpolysiloxane, 30 m length, 0.25 mm I.D., film thickness 0.25 mm). Injecting volume was 1 μL, and the GC conditions included programmed heating from 50 to 300 °C at 10 °C/min, followed by 10 min at 300 °C. The injector was maintained at 280 °C. Helium was the carrier gas, at 1.0 mL min−<sup>1</sup> , and the split mode was 5:1. The GC was fitted with a quadrupole mass spectrometer with an Agilent model 5975 detector. The MS conditions were as follows: ionization energy, 70 eV; electronic impact ion source temperature, 230 °C; quadrupole temperature, 150 °C; scan rate, 3.2 scans/s; mass range, 50–1000 u. The compounds were identified based on the match with their mass spectra and retention indices with the NIST/Wiley 275 library (Wiley, New York). The Relative abundance of each feature was calculated from Total Ion Chromatogram (TIC) computationally (Mohamad et al., 2018).

#### Intelligent Live Digital Imaging of Antibiosis Endophytes Strains and F. oxysporum

The morphological response of selected bacterial strains Bacillus sonorensis(EGY05), Bacillus subtilissubsp. subtilis (EGY01), Bacillus tequilensis (EGY21), Bacillus mojavensis (EGY25), Enterobacter xiangfangensis (EGY31), and Bacillus subtilis subsp.inaquosorum (EGY16) had demonstrated antagonism to F. oxysporum in vitro and in vivo were observed under a laser microscope (Olympus SZX2- ILLT, Japan) at different magnifications.

Six selected antagonistic bacterial strains were incubated on ISP2 medium. F. oxysporum was grown on ISP2 medium. A sixday-old mycelial disc (5 mm) was placed at the center of a 7 cm modified culture ISP2 plate. The bacteria were placed at four corners on the bacterial lawn at equidistant points 2.5 cm from the plate periphery. All plates were wrapped with parafilm, incubated at 28 ± 2◦ C for 6 days, and observed for the inhibition of the pathogen. Plates with pathogenic fungi alone served as control (Mohamad et al., 2018).

#### Statistical Analysis

The data represent mean of at least 10–12 replicates ± standard error (SE). One-way ANOVA was used to compare the means of root length, root fresh weight, shoot length, shoot fresh weight, SOD, POD, CAT, and Chlorophyll content for each salt concentration separately and Tukey's HSD post-hoc test used for multiple comparisons at alpha level = 0.05. Statistical analyses were conducted in R (R Core Team). (Pinheiro et al., 2018). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

#### RESULTS

#### Isolation and Identification of Endophytic Bacteria Associated With T. vulgaris

Bacterial endophytes were isolated from roots, stems, and leaves of six wild T. vulgaris plants collected from two different locations in North Sinai, Egypt. A total 117 strains were isolated based on unique colony morphologies on ten different isolation media. Based on nucleotide identity and phylogenetic analysis of near-complete 16S rRNA gene sequences, the isolates belonged to 17 different genera (Figure 1A) and 30 species (Figure S1). Most isolates belonged to Bacillus (47%), followed by Microbacterium (20%); Enterobacter (5.5%); Streptomyces, Rhodococcus, and Klebsiella (4.6% each); Arthrobacter, Escherichia, Micrococcus, and Shigella (1.85% each); and Kocuria , Dietzia , Lysinibacillus, Blastococcus, Cellulosimicrobium, Pseudomonas, and Micromonospora (0.93% each).

The diversity of the isolates varied based on the plant tissue. The highest number of distinct bacterial species was recovered from root tissue (n = 18), compared to stems (n = 14) and leaves (n = 11). Almost all isolates were belonged to genus Bacillus or to various genera within the phylum Actinobacteria. Pseudomonas, Cellulosimicrobium, Dietzia, Shigella, and Lysinibacillus species were only associated with root tissues. Arthrobacter and Kocuria species were only associated with stem tissues. Micrococcus species were only isolated from leaf tissue (Figure 1B).

The diversity of endophyte isolates was similar at the Sheikh Zuweid (n = 22 species) and Rafah (n = 17 species) sampling sites; however, the species composition at each site was distinct. In particular, Bacillus, Cellulosimicrobium, Enterobacter, Klebsiella, Micrococcus, Micromonospora, Pseudomonas, and Shigella species were only isolated from the Sheikh Zuweid site and Blastococcus, Dietzia, Kocuria, Lysinibacillus, Micrococcus, and Rhodococcus species were only isolated from the Rafah site (Figure 1C). Meanwhile, the highest number of endophytic species was isolated on M1 medium (n = 13), including members of the genera Bacillus, Microbacterium, Cellulosimicrobium, Enterobacter, Klebsiella, Streptomyces, and Micromonospora. The lowest number of species were obtained on M3 and M10 (n = 7 each) (Figure S2).

#### Beneficial Plant Traits of Endophytic Bacteria

All isolates were screened for multiple beneficial traits in vitro (Figure S3). The majority of isolates was able to fix nitrogen (84%) based on growth on both NFC and Ashby's media. Many

FIGURE 1 | The geographic distribution and identity of 117 endophytes from Thymus vulgaris based on 16S rRNA gene sequences. (A) A summary of genera presents at all sites. (B) Genus assignments for isolates from different tissues, showing high diversity in the root and low diversity in the leaf. (C) Genus assignments according to location, showing similar diversity of isolates at the two sites.

isolates produced siderophores (54%), solubilized phosphate (20%), and synthesized IAA (15%). Strains have more beneficial traits belonged to the genera Bacillus, Enterobacter, Pseudomonas, Klebsiella, and Microbacterium (Figure 2). In addition, all the isolates were screened for the presence of digestive enzymes that may be involved in lysis of fungal pathogens. Most of the endophytic strains produced one or more hydrolytic enzymes: cellulase (66%), lipase (47%), protease (46%), and chitinase (30%) (Figure 2). The endophytes producing at least three digestive enzymes belonged to the genera Bacillus, Enterobacter, and Micrococcus.

#### Antifungal Activity

All isolates were individually tested against three common tomato pathogens in vitro. Endophytes belonging to 7 genera and 12 species were antagonistic to all the tomato pathogens (Figure 3). Bacillus genus showed the highest antagonistic activity. In addition, the ability to inhibit the growth of the

tested fungi varied by the percentage of inhibition, ranging from 40 to 77%. Enterobacter xiangfangensis strain EGY31, Bacillus sonorensis strain EGY11, and Bacillus subtilis subsp. inaquosorum strain EGY15 showed the largest zones of inhibition against F. oxysporum f. sp. (F1; 70%), F. fulva (F2; 71%), and A. solani (F3; 77%), respectively.

#### Stimulation of Tomato Growth by Endophytes Under Salt Stress

To test whether the endophytes promote plant growth, three bacterial strains, Bacillus sonorensis (EGY05), Bacillus tequilensis (EGY21), and Bacillus mojavensis (EGY25), that were positive for at least four plant-beneficial traits were selected to test their plant growth stimulation properties in pot experiments with tomato plants under salt stress. Shoot and root length and weight decreased with increasing the salt concentrations from 50 mM to

200 mM (Figure 4). Tissues from plants inoculated with each of the endophytes were generally larger (P < 0.05) than uninoculated controls but the precise pattern of growth stimulation varied by plant tissue and salinity (Figure S4). Strain EGY05 showed the strongest stimulation of root growth, increasing both root length and root fresh weight significantly (P < 0.05) at most concentrations, compared to the uninoculated controls (Figure 4A). In particular, inoculation with strain EGY05 increased the root length by 15.9, 24.4, 25.4, and 23.8% under 50, 100, 150, and 200 mM NaCl treatments, respectively, compared to the uninoculated controls. Moreover, strain EGY05 increased the root fresh weight (Figure S5) by 27.6, 19.2, 21.9, and 40.2% under 50, 100, 150, and 200 mM NaCl treatments, respectively (Figure 4B).

For shoot fresh weight and shoot length, strain EGY25 showed the strongest stimulatory effects at all salinities tested. Strain EGY25 increased shoot length significantly (P < 0.05) by 18.5, 20.3, 23.3, and 20.0% under 50, 100, 150, and 200 mM of NaCl treatment, respectively, compared to the uninoculated control (Figure 4C). It also increased shoot fresh weight over the controls by 25.4, 24. 8, 21.0, and 32.3% under 50, 100, 150, and 200 mM of NaCl treatment; however, this effect was only significant at 50 mM NaCl (Figure 4D).

#### Effect of Endophytes on Antioxidant Enzyme Activity and Chlorophyll Content Under Salt Stress

The activity of the antioxidant enzymes superoxide dismutase (SOD), catalase (CAT), and peroxidase (POD) were measured in the tomato tissues (Figure 5). Salinity stress caused a significant increase in activities of all antioxidant enzymes in this study (P < 0.05). However, each of the three endophytes significantly (P < 0.05) decreased the activity of SOD, CAT, and POD under most conditions. For example, plants inoculated with strain EGY05 showed 6.6, and 17.7, 25.7, and 33.0% decreases in superoxide (SOD) activity at 50, 100, 150, and 200 mM NaCl, respectively, in comparison to the uninoculated control plants (Figure 5A). The same strain reduced POD activity by 20.3, 14.9, 32.5, and 28.7% at 50, 100, 150, and 200 mM NaCl (Figure 5B). Catalase activity (CAT) (Figure 5C) showed a similar pattern where inoculation with strain EGY05 significantly reduced the activity at 50, 100, 150, and 200 mM salt to 40.1, 71.5, 58.3, and 21.4%, respectively, as compared to the control. Salinity also affected chlorophyll content, but inoculation with the endophytes increased chlorophyll content under all conditions; however, this effect was only significant (P < 0.05) for strain EGY05 at 150 and 200 mM NaCl (25.4, and 31.5% increase, respectively) (Figure 5D).

#### In Vivo Biological Control of F. oxysporum Under Salt Stress

To evaluate the role of endophytic bacteria on resistance to pathogens in vivo, six strains, Bacillus sonorensis (EGY05), Bacillus subtilis subsp. subtilis (EGY01), Bacillus tequilensis (EGY21), Bacillus mojavensis (EGY25), Enterobacter xiangfangensis (EGY31), and Bacillus subtilis subsp. inaquosorum (EGY16), were selected for

their high antagonistic activity to F. oxysporum in pot experiments under salt stress.

Plant disease signs under salt stress developed within ten days of inoculation. The symptoms included yellowing of leaves, and leaf chlorosis, which began with older leaves and progressed to younger leaves and continually increased throughout the experiment over an 8-week period. The distribution of disease grades varied dramatically at different salt concentrations compared to uninoculated controls (Figure 6A). Severe signs reaching 76.4% (DI) were seen on the leaves of the tomato plants challenged with F. oxysporum in the absence of endophytes at different salt concentrations (Figure 6B). All selected antagonistic bacterial isolates were able to control tomato disease caused by F. oxysporum under NaCl treatment, with statistically significant (P < 0.05) effects, at different salt concentrations, in comparison to the pathogen-infected not treated controls. The results showed a dramatic decrease in disease symptoms and enabled plants to survive at different levels of salt stress (Figures 6A, B).

#### Detection of Bioactive Compounds by GC-MS Analysis

To evaluate the volatile compounds produced by the most bioactive strain, B. subtilis EGY16, and F. oxysporum were grown in coculture. GC-MS analysis of ethyl acetate crude extracts showed 20 compounds (Table S3) at pH 7 and all features were tentatively identified based on a comparison of spectra available through the National Institute of Standards and Technology (NIST) database. Six major features were obtained at RT 3.444, 4.582, 7.131, 7.312, 46.345, 60.315, and 62.898, suggesting the presence of acetic acid, butyl ester; benzene, 1,3-dimethyl-; propane, 1-ethoxy-; propane, 1 ethoxy-; dibutyl phthalate; tetracosane; and bis (2-ethylhexyl) phthalate. In addition, several minor peaks were present, including 2,2-Dimethylthiirane; 2-Methylpropanoic acid, TMS derivative; benzene, 1,3-dimethyl-; p-Xylene; hexane-1,3,4-triol, 3,5-dimethyl-; phenylethyl alcohol; heneicosane; nonadecane; tricosane; decanedioic acid, bis (2-ethylhex; Propenone, 1-(3 bromophenyl)-3; heneicosane, 11-cyclopentyl-; and cyclohexanecarboxylic acid, 2 (Figure 7).

FIGURE 6 | Effect of endophytic isolates on disease grades and disease index of Solanum lycopersicum L. plants to Fusarium oxysporum over eight weeks under salt stress. (A) Signs for individual plants rated from "0" to "5" (0: no signs, 1: ≤35, 2: 35–55, 3: 55–65, 4: 65–85, 5: 85– 100%); (B) Disease index of tomato plants with and without endophytes under salt stress. Bar height represents the mean disease index for each isolate at each salt concentration, whiskers represent standard deviation. Shared letters between two isolates at a given salt concentration indicate no significant difference in disease index as calculated using ANOVA followed by Tukey's HSD post-hoc test, at alpha level = 0.05.

#### Defense Response of Endophytic Strains to F. oxysporum via Laser Microscopy

The results showed that F. oxysporum could not grow normally after 5 days of incubation with antagonistic strains (Figures 8A– F). The morphological response of tested strains EGY05, EGY01, EGY21, EGY25, EGY31, and EGY16 to F. oxysporum under a laser microscope were observed at magnifications of 0.63X (Figures 8G–L) and 1X (Figures 8M–T). The microscopic characteristics of tested strains based on laser microscopy at 0.63X showed that the endophytic strains were able to control the growth of the fungal mycelium after 5 days of cultivation, and at 1X, a white powder appeared around bacterial cells. Thus, we hypothesize that these antagonistic strains may secrete antifungal compounds (Figures 8M–T).

### DISCUSSION

As a result of climate change, salinity has become one of the major abiotic stresses that reduce plant growth and crop productivity worldwide. A better understanding of the impacts of endophytic bacteria on plant health under conditions of salinity stress can lead to insights into improved cultivation techniques under stressful conditions.

Thyme has pharmacological properties such as antiinflammatory, anti-bacterial, anti-viral, antioxidant, and insecticidal activities (Prasanth Reddy et al., 2014; Mandal and DebMandal, 2016). However, to date, Thymus has not investigated with respect to microbial communities. Thus, in this study, a collection of 117 endophytic strains associated with roots, leaves, and stems of the medicinal plant T. vulgaris was obtained from two different wild populations in North Sinai, Egypt. Isolates were identified by 16S rRNA gene sequencing. The isolates belonged to 17 genera and 30 species mainly belonging to the phyla Firmicutes, Actinobacteria, and Proteobacteria. These results confirm the rich microbial diversity in plants grown in arid lands, in agreement with previous studies done by our group (Liu et al., 2016; Liu et al., 2017; Li et al., 2018). In addition, the bacterial taxa observed were similar to endophytes observed using cultivation-independent techniques (Qin et al., 2012; Kaplan et al., 2013; Jin et al., 2014; Egamberdieva et al., 2017b). The dominant bacterial genus was Bacillus, which is known for its beneficial effects on plant growth and health (Radhakrishnan et al., 2017; Rais et al., 2017).

FIGURE 8 | Intelligent live digital imaging of the response of endophytic strains EGY05, EGY01, EGY21, EGY25, EGY31, and EGY16 to Fusarium oxysporum under a laser microscope (A–F). In vitro evaluation of antagonistic activity of tested strains againist strain F. oxysporum (G–L). Response of tested strains to F. oxysporum under a laser microscope at 0.63X magnification (M, N, Q–T). Response of tested strains to F. oxysporum under a laser microscope at 1X magnification.

The microbial diversity differed based on plant tissue. The highest diversity was isolated from T. vulgaris roots. The diversity of endophytes from the two locations was similar but slightly higher at the Sheikh Zuweid site. It seems that endophytic bacteria isolated from arid land with very low soil nutrient content depend on the type of symbiosis interaction and availability of nutrients in the plant tissue, as has been reported previously (Li et al., 2012; Jin et al., 2014; Liu et al., 2016; Liu et al., 2017; Li et al., 2018).

A major objective of this study was to better understand the interactions between beneficial endophytic bacteria and medicinal plants in arid lands. Therefore, we screened the endophyte collection for beneficial plant traits in vitro with a goal of detecting the most promising microorganisms. Many of the endophytes produced several plants promoting traits, in agreement with our previous investigations dealing with medicinal plants in arid lands (Liu et al., 2016; Liu et al., 2017; Li et al., 2018). In addition, similar investigations reported that endophytic bacteria exhibited multiple plant beneficial traits (Egamberdieva and Kucharova, 2009; Egamberdieva et al., 2017b; Paul and Sinha, 2017).

To support the in vitro results, three bacterial strains, Bacillus sonorensis (EGY05), Bacillus tequilensis (EGY21), and Bacillus mojavensis (EGY25), were selected for plant growth stimulation properties in a pot experiment under salt stress. Each strain significantly promoted tomato plants at different salt concentrations, compared to un-inoculated controls (Figure 3), but each endophyte had a unique pattern of plant growth promotion. For example, B. sonorensis EGY05 had the strongest stimulatory effects on roots, but B. mojavensis EGY25 had the strongest effects on shoots. These activities are likely due to one more activities, such as nitrogen fixation (Zakry et al., 2012), IAA production (Leveau and Lindow, 2005), or phosphate solubilization (Hameeda et al., 2008; Lemanceau et al., 2009). Additionally, Bacillus is particularly resistant to environmental stresses that are common in the desert, such as UV, heat, high alkalinity, and high salinity (Horikoshi, 2008). Several other investigations have also reported that Bacillus strains belonging to Bacillus megaterium and Bacillus insolitus enhanced the length and biomass of shoot, roots, and leaves under salt stress (Ashraf et al., 2004; Radhakrishnan and Lee, 2016; He et al., 2018).

Under conditions of high salinity, plants face disorders in many metabolic pathways, such as those related to redox system and photosynthesis, which leads to reductions in plant growth (Munns and Tester, 2008). To understand the mechanism of plant-microbe responses defense under saline conditions, we investigated the antioxidant enzymes (SOD, CAT, and POD) activity of tomato plants under salt stress. The selected endophytes decreased SOD, POD, and CAT activities at different salt concentrations, compared to the uninoculated controls (Figure 5). Enhancement of plant growth during salt stress may be related to increases in the relative water content as well as the osmotic and turgor potential to improve plant growth under salinity conditions (Hashem et al., 2016; Yang et al., 2016). For example, Bacillus pumilus reduced the activity of caspase of salt-stressed rice plants (Jha and Subramanian, 2014). However, decreases in the activities of antioxidant enzymes could be due to a general decrease in biosynthesis and lower oxidative stress, as has been suggested by others (Egamberdieva et al., 2017b; Rais et al., 2017). In this study, the total chlorophyll in tomato leaves was slightly increased at different levels of salt concentrations when the plants were subjected to endophytes, but most of these effects were not statistically significant (Figure 5D). The reduction in the pigment content is attributed to the negative effect of salt stress on chloroplasts (Ahmad et al., 2012c; Ebrahim and Saleem, 2017).

Tomato (S. lycopersicum L.) is one of the most important and widely grown plants in the world. The control of F. oxysporum disease has been based almost solely on the application of chemical fungicides. However, for the management of plant diseases, recent efforts have focused on alternative methods of control to circumvent this situation and to develop environmentally friendly and sustainable procedures to control tomato Fusarium wilt disease. Thus, one objective of this study was to assess the biological control activities of the endophytes on Fusarium in vitro and vivo. Our results suggest that antibiotics produced by Bacillus endophytes are active against tested phytopathogens. Bacillus is well known for its diverse range of secondary metabolic products, including antibiotics (Ryan et al., 2008; Radhakrishnan et al., 2017; Mohamad et al., 2018). Interestingly, all six strains evaluated here could suppress Fusarium wilt disease at different levels of salt concentrations, compared to un-inoculated controls (Figure 6). Indeed, strain EGY16 was able to reduce disease severity up 26% at all tested salt concentrations. Other investigations reported that Bacillus spp. suppress pathogen growth and protect olive and pepper plants (Krid et al., 2012; Yi et al., 2013). In addition, strain EGY16 showed a proteolytic, cellulolytic, and chitinolytic activity that may involve attachment to the mycelial cell walls. These results are in general agreement with (Egamberdieva et al., 2017b) who reported that endophytic bacteria associated with the medicinal plant Ziziphora capital were able to produce chitinolytic enzymes. In accordance with these results, several previous studies have shown that endophytic Bacillus species control fungal pathogens, including B. mojavensis (Bacon et al., 2005), B. velezensis (Gao et al., 2017), B. megaterium (Gao et al., 2017), and B. atrophaeus (Mohamad et al., 2018). Bacillus spores are well-known for their ability to survive for a long time under unfavorable environmental conditions. Bacillus vegetative cells are able to secrete several metabolites to alleviate abiotic and biotic stresses with antifungal activities and improve plant growth (Chowdhury et al., 2015; Radhakrishnan et al., 2017). In recent years, the development of biological control agents derived from Bacillus isolates, such as "Custom" (Baugh and Escobar, 2007), "Avogreen" (Janisiewicz and Korsten, 2002) and "Shemer" (Droby, 2005), has been shown to be effective in plant fungal disease.

In this investigation, among all tested endophytes in vivo, B. subtilis EGY16 was selected for exometabolomic studies by GC-MS, based on its ability to decrease the disease severity index (DSI) and suppress the growth of F. oxysporum. 20 compounds were identified by GC-MS. Among these compounds, five of them were major peaks in cell-free extracts from the co-culture, suggesting that they very likely play a role in antagonism of fungal pathogens. These compounds are known as an antimicrobial compound such as benzene, 1,3-dimethyl-, p-xylene (Fernandes et al., 2003), dibutyl phthalate (Khatiwora et al., 2012), bis(2-Ethylhexyl) phthalate (Kavitha et al., 2009), and tetracosane (Naragani et al., 2016). In addition, several putative antimicrobial compounds were identified as minor peaks: phenylethyl alcohol (Strobel et al., 2001), heneicosane (Sharma et al., 2016), nonadecane (Hsouna et al., 2011), bis(2-Ethylhexyl) phthalate (Kavitha et al., 2009). These results suggesting that these compounds play an important role in antimicrobial activities.

### CONCLUSION

Globally, crop productivity is decreasing due to climatic change, and human populations are increasing daily. The Food and Agricultural Organization (FAO) predicts an increasing human population to reach 8–9 billion by 2030. Here, we showed that Bacillus spp. associated with T. vulgaris such as Bacillus sonorensis (EGY05), Bacillus tequilensis (EGY21), and Bacillus mojavensis (EGY25) produced plant growth-promoting substances including auxin, fixed nitrogen, solubilized phosphate and iron, and produced lytic enzymes (i.e., chitinase, cellulase, protease, and lipase). These bacteria may provide new strategies to mitigate salt stress and also develop new ways to enhance the tolerance and growth of plants such as tomato. In addition, all tested strains decreased the activities of antioxidant enzymes (SOD, CAT, and POD) of tomato plants at different salt concentrations, compared to the un-inoculated controls. The selected antagonistic isolates such as Bacillus sonorensis (EGY05), Bacillus subtilis subsp. subtilis (EGY01), Bacillus tequilensis (EGY21), Bacillus mojavensis (EGY25), Enterobacter xiangfangensis (EGY31), and Bacillus subtilis subsp. inaquosorum (EGY16) were able to control tomato root rot significantly (P < 0.05) caused by F. oxysporum under greenhouse conditions, suggesting they could be powerful biological control agents. To the best of our knowledge, this is the first study of the diversity of the microbial community associated with medicinal plant T. vulgaris and their plantgrowth promoting and biocontrol abilities. Further field investigations are ongoing for future application in tomato plant growth promotion and crop productivity as well as antifungal activities against Fulvia fulva (Cooke) Cif, and Alternaria solani Soraue. These results support the development of natural products that may minimize the need for the application of chemical fertilizer and fungicides, which would be an environmentally friendly approach and preserve biological resources in a sustainable agricultural system.

### DATA AVAILABILITY STATEMENT

Publicly available datasets were analyzed in this study. This data can be found here: GenBank under Accession Numbers MH764457–MH764573.

#### AUTHOR CONTRIBUTIONS

OAAM participated in the design of the study, performed all experiments, interpretation of results and wrote the manuscript. LL participated in the isolation and identification of endophytic bacteria. Y-HL helping for preparing enzymes activity test, and screening for plant beneficial traits. J-BM and DZ conducted the plant growth chamber experiments and data analysis. SH did the GC-MS and data analysis. SB did statistical analysis. BH revised and improved the manuscript. W-JL and OM revised the manuscript and supervised all experiments. All authors edited and critically revised the manuscript.

#### FUNDING

This research was supported by the National Key Research and Development Project (2016YFC0501502), National Natural Science Foundation of China (Grant No. U1703106 and 31650110479), Xinjiang Uygur Autonomous Region Regional Coordinated Innovation Project (Shanghai Cooperation Organization Science and Technology Partnership Program) (Grant No. 2017E01031). OAAM was supported by the Available Position Talented Young Scientists Program funded by Chinese Academy of Sciences President's International Fellowship Initiative (Grant No. 2016PB024).

#### ACKNOWLEDGMENTS

We would like to thank Dr. Chad L. Cross, School of Medicine, University of Nevada for helping on detailed statistical analysis.

#### REFERENCES


#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2020.00047/ full#supplementary-material

FIGURE S1 | A summary of species present at all sites of 117 culturable endophytes from Thymus vulgaris. based on 16S rRNA gene sequences.

FIGURE S2 | Distribution of endophytic isolates and species isolated from Thymus vulgaris on different media.

FIGURE S3 | Confirmation of plant growth promotion traits by a color change or halo zone on the selective medium (A: Protease; B: Cellulase; C: Lipase; D: Siderophore; E: Chitin; F: Nitrogen). (G) response of endophytic strains to F. oxysporum (H) response of endophytic strains to F. fulva, (I) response of endophytic strains to A. solani.

FIGURE S4 | Effect of endophytic isolates on the growth of tomato (Solanum lycopersicum L.) compared with an uninoculated control without salt (CK+ ) and inoculated control with salt (CK- ).

FIGURE S5 | Effect of inculation of endophytic strains on the root of tomato plants under salt stress.

TABLE S1 | Compositions of the different media used for the isolation of endophytic bacteria from Thymus vulgaris.

TABLE S2 | Fungal pathogens used in this study.

TABLE S3 | GC-MS identified components of the antibiosis crude extract of EGY16 and F. oxysporum mixture at pH7. (Volatile compounds are listed in ascending order of Retention Time).


Pseudomonaden Isolaten zur Bekämpfung der Tomatenwelke, verursacht durch Fusarium oxysporum f. sp. lycopersici. Zeitschrift für Pflanzenkrankheiten und Pflanzenschutz. J. Plant Dis. Prot., 143–156.


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 Abdelshafy Mohamad, Ma, Liu, Zhang, Hua, Bhute, Hedlund, Li and Li. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) 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.