# PLANT PATHOGEN LIFE-HISTORY TRAITS AND ADAPTATION TO ENVIRONMENTAL CONSTRAINTS

EDITED BY : Christophe Le May, Josselin Montarry, Cindy E. Morris, Omer Frenkel and Virginie Ravigné PUBLISHED IN : Frontiers in Plant Science and Frontiers in Microbiology

#### Frontiers eBook Copyright Statement

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

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

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

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

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

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

ISSN 1664-8714 ISBN 978-2-88963-530-6 DOI 10.3389/978-2-88963-530-6

### 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

# PLANT PATHOGEN LIFE-HISTORY TRAITS AND ADAPTATION TO ENVIRONMENTAL CONSTRAINTS

Topic Editors:

Christophe Le May, Agrocampus Ouest, France Josselin Montarry, INRAe Centre Bretagne-Normandie, France Cindy E. Morris, INRAe Centre Provence-Alpes-Côte d'Azur, France Omer Frenkel, Agricultural Research Organization (ARO), Israel Virginie Ravigné, Centre de Coopération Internationale en Recherche Agronomique pour le Développement (CIRAD), France

Citation: Le May, C., Montarry, J., Morris, C. E., Frenkel, O., Ravigné, V., eds. (2020). Plant Pathogen Life-History Traits and Adaptation to Environmental Constraints. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-530-6

# Table of Contents

*04 Editorial: Plant Pathogen Life-History Traits and Adaptation to Environmental Constraints*

Christophe Le May, Josselin Montarry, Cindy E. Morris, Omer Frenkel and Virginie Ravigné


Elise Vaumourin and Anna-Liisa Laine

*74 Striking Similarities Between* Botrytis cinerea *From Non-agricultural and From Agricultural Habitats*

Marc Bardin, Christel Leyronas, Claire Troulet and Cindy E. Morris

*85 Pathogenicity Traits Correlate With the Susceptible* Vitis vinifera *Leaf Physiology Transition in the Biotroph Fungus* Erysiphe necator*: An Adaptation to Plant Ontogenic Resistance*

Agnes Calonnec, Jerome Jolivet, Philippe Vivin and Sylvain Schnee

*102 A Trade-Off Between Sporangia Size and Number Exists in the Potato Late Blight Pathogen* Phytophthora infestans*, and is not Altered by Biotic and Abiotic Factors*

Nicolas Mariette, Alexander Kröner, Romain Mabon, Josselin Montarry, Bruno Marquer, Roselyne Corbière, Annabelle Androdias and Didier Andrivon

### *109 Responses to Hydric Stress in the Seed-Borne Necrotrophic Fungus*  Alternaria brassicicola

Guillaume Quang N'Guyen, Roxane Raulo, Muriel Marchi, Carlos Agustí-Brisach, Beatrice Iacomi, Sandra Pelletier, Jean-Pierre Renou, Nelly Bataillé-Simoneau, Claire Campion, Franck Bastide, Bruno Hamon, Chloé Mouchès, Benoit Porcheron, Remi Lemoine, Anthony Kwasiborski, Philippe Simoneau and Thomas Guillemette

## Editorial: Plant Pathogen Life-History Traits and Adaptation to Environmental Constraints

Christophe Le May 1\*, Josselin Montarry <sup>1</sup> , Cindy E. Morris <sup>2</sup> , Omer Frenkel <sup>3</sup> and Virginie Ravigné<sup>4</sup>

<sup>1</sup> IGEPP, INRA, Agrocampus-Ouest, Université de Rennes 1, Le Rheu, France, <sup>2</sup> INRA, UR Pathologie Végétale, Avignon, France, <sup>3</sup> Agricultural Research Organization, Volcani Center, Rishon LeZion, Israel, <sup>4</sup> BIOS Departement, PVBMT Laboratory, CIRAD, Montpellier, France

Keywords: coinfection, epidemic, host resistance, hydric stress, leaf maturity, non-agricultural habitats, survival, temperature

Editorial on the Research Topic

Plant Pathogen Life-History Traits and Adaptation to Environmental Constraints

#### Edited and reviewed by:

Brigitte Mauch-Mani, Université de Neuchâtel, Switzerland

\*Correspondence: Christophe Le May lemay@agrocampus-ouest.fr

#### Specialty section:

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

Received: 07 November 2019 Accepted: 09 December 2019 Published: 24 January 2020

### Citation:

Le May C, Montarry J, Morris CE, Frenkel O and Ravigné V (2020) Editorial: Plant Pathogen Life-History Traits and Adaptation to Environmental Constraints. Front. Plant Sci. 10:1730. doi: 10.3389/fpls.2019.01730 Life-history is a key concept in evolutionary biology and ecology. It corresponds to the narrative of the various events punctuating the existence of an organism from his birth to his death (Begon et al., 2006; Michalakis et al., 2016). Throughout life, living organisms acquire resources that they actively find or extract from their environment and then allocate to different functions: development, survival, dispersal and reproduction (Roff, 1992; Stearns, 1992; Michalakis, 2009). Traits involved in the timing and amplitude of these allocation dilemmas are defined as life history traits. Life-history traits are often determinant for individuals (Kingsolver and Pfennig, 2007). They influence spatial and temporal disease dynamics, and thus the genetic diversity and structure of pathogen populations (Barrett et al., 2008; Michalakis et al., 2016). These determinants are involved in the ability of pathogens to adapt to varying ecological factors including changes in the biotic and abiotic effect, but also direct or indirect interactions with other strains or species of pathogens co-occurring on the same host (Michalakis et al., 2016; Tollenaere et al., 2016). For plant pathogens, these traits can be grouped into two categories: those involved in the epidemic phase, onto or into the host, and those related to the survival phase, often outside the host. Trade-off can occurred between these two phases, leading to consequences on epidemic dynamics, evolution, and speciation (Pariaud et al., 2009; Hamelin et al., 2011). Understanding processes maintaining variation in plant pathogen lifehistory traits is a central question in evolutionary ecology and a major challenge for the design of disease control strategies (Galvani, 2003; Grenfell et al., 2004).

The aim of this Research Topic was to collect research papers with vast perspective on plant pathogen life-history traits and adaptation to environmental constraints. Life-history traits of both epidemic and survival phases were considered, and tested environmental constraints included abiotic (temperature, precipitation, hydric stress) and biotic (habitat of origin, cultivar, maturity of leaves, coinfection) factors. Among the 10 accepted papers, three types of plant pathogens were considered: fungi (seven papers), nematodes (two papers), and oomycetes (one paper).

Regarding the epidemic phase, four main topics were covered by the research papers accepted in this special issue: i) host adaptation, ii) impact of host's physiological status, iii) adaptation to climatic conditions, and iv) effects of coinfection.

How host adaptation affects life history traits evolution was particularly analyzed by studying the effect of resistant cultivars onto several life-history traits. In the ascomycete Colletotrichum gloeosporioides the impact of host quantitative resistance on pathogen evolution was studied by evaluating the diversity of populations at the field scale at both neutral markers and pathogenic traits (Frezal et al.). Population genetic structure revealed a significant influence of clonal reproduction in C. gloeosporioides evolution and a low migration rate between fields. Results of cross-inoculation tests showed that aggressiveness of the fungal clones seemed to have evolved through an accumulation of components specific to each water yam cultivar, suggesting an adaptation to their host cultivar. Despite the remaining marks of adaptation to the former widely cultivated host, adaptation to current cultivars was clearly depicted. Such pattern of local adaptation to the dominant cultivar was previously highlighted in the potato late blight pathogen Phytophthora infestans (Andrivon et al., 2007). In their study, Mariette et al. explored whether the now classical negative relationship between offspring size and number occurred in this oomycete, P. infestans, and whether the tradeoff was impacted by potato cultivar or host of origin of the pathogen (tomato and potato). They confirmed the existence of a trade-off and showed that it was not affected by any of these biotic factors. The observed polyphenism for these traits in P. infestans populations will favor the coexistence of distinct reproductive strategies.

Over the last years, intra-host dynamics has been identified as a major component of epidemiological and evolutionary dynamics of pathogens (Alizon et al., 2009). In plants it is known that host tissue characteristics vary a lot with tissue age. In particular, according the age of tissues, nutrient availability as well as defense levels may vary significantly, with opposing effects on the success of infection (Kus et al., 2002; Al-Naimi et al., 2005). Ontogenic resistance against powdery mildew (Erysiphe necator) on Vitis vinifera was studied by Calonnec et al. They showed that the three pathogenic traits studied (infection efficiency, sporulation, and mycelium growth) were affected, and their variation was strongly correlated with leaf age. Sporulation was more closely correlated with variations in sugar contents and the infection efficiency with leaf water content, suggesting that ontogenic resistance on grapevine leaves seems to be an immutable physiological process that E. necator is able to circumvent by restricting its development to sink tissue. Similarly, Maupetit et al. tested how nutrient availability and defense level impacted the development of Melampsora laricipopulina (infection efficiency, latent period, uredinia size, mycelium quantity, sporulation rate, sporulation capacity, and spore volume). They showed that M. larici-populina was more aggressive on more mature leaves as indicated by wider uredinia and a higher sporulation rate. In contrast, phenolic contents (flavonols, hydroxycinnamic acid esters, and salicinoids) were negatively correlated with uredinia size and sporulation rate suggesting that pathogen's fitness appeared to be more constrained by the constitutive plant defense level than limited by nutrient availability, as evident in the decrease in sporulation.

Like other organisms, pathogens need a particular range of temperature and humidity in order to invade and develop on their host plant (Agrios, 2005). These two parameters particularly can modify spore germination, host colonization, spore production and dispersal of fungal pathogens (Fitt et al., 2006). Decades of research have generated considerable knowledge and greater understanding of the seasonal effects of temperature, rainfall, and humidity on diseases affecting major food crops (Garrett et al., 2011; Pautasso et al., 2011; Donatelli et al., 2017). In this special issue, two main parameters were studied, change in temperature and effect of water stress. Regarding the effect of temperature, Vaumourin and Laine developed a study on Podosphaera plantaginis and showed that temperature had a significant effect on all measured life-history traits, suggesting that the effect of temperature on life-history traits was both direct as well as mediated through a genotype-bytemperature interaction. Similarly, Mariette et al. showed on P. infestans that the negative relationship between offspring size and number was maintained for all the tested temperatures. Using cyst nematodes, other studies showed an effect of temperature on different life-history traits. Indeed, Fournet et al. showed a strong temperature effect on life-history traits of the beet cyst nematode Heterodera schachtti. While nematode multiplication was not differentially affected by temperatures, as favorable conditions for the host are also favorable for the parasite, the effect of temperature on hatching depended on the origin of populations, separating southern from northern European populations. Using a genome scan approach, Gendron St-Marseille et al. evaluated the adaptive potential of the soybean cyst nematode Heterodera glycines. The genotyping by sequencing of 64 H. glycines populations, allowed identifying 15 loci under selection for climatic or geographic co-variables. Lastly, regarding the effect of water stress, a transcriptomic and metabolomic approach was developed on the Arabidopsis thaliana/Alternaria brassicicola pathosystem, to study the level of susceptibility of the fungus to water stress and its impact on its seed transmission ability (N'Guyen et al.). These approaches led to the identification of specific proteins (hydrolipin) implicated in the tolerance toward water stress, and seed transmission, but not in the aggressiveness of the pathogen.

The ability of a pathogen to establish and grow on its host may be drastically altered by simultaneous infection by other pathogens (other strains or species of pathogens co-occurring on the same host). Life-history allocation may change under coinfection, affecting the evolutionary potential and epidemiological dynamics of pathogens (Tollenaere et al., 2016; Laine and Mäkinen, 2018). In their study, Vaumourin and Laine found that coinfection only modified the number of sexual resting structures produced, but did not changed the production of asexual forms. Coinfection studies, assessment of fungi sharing common characteristics, and host species, create a challenge for conventional disease diagnosis and subsequent management strategies. Indeed, lack of powerful tools to simultaneously distinguish and quantify these types of pathogens constitutes a major limitation. In their study, Abdullah et al. developed a duplex real-time PCR assay for quantifying of co-infecting wheat pathogens, Pyrenophora triticirepentis and Parastagonospora nodorum. The utility of the method was demonstrated using field samples of a cultivar sensitive to both pathogens. While visual and culture diagnosis suggested the presence of only one of the pathogen species, the assay revealed not only presence of both co-infecting pathogens (hence enabling asymptomatic detection) but also allowed quantification of relative abundances of the pathogens as a function of disease severity.

Survival stage outside the host has often been much overlooked due to the difficulty of its study. It is a key stage in pathogen's life cycle. Pathogens are all dependent on the life cycle, where despite the existence of dormant forms, evolutionary processes may occur. Thus, according to the status of their host (perennial or annual host plant), pathogens have to develop a wide array of strategies to overcome the alternation of presence/absence (and/or modification of physiological activity) of their host (Agrios, 2005). This survival phase is consequently important as it may affect the population size of the pathogen and lead to a bottleneck process. In this, special issue, research developed by Bardin et al. have investigated the ubiquity of the broad host range necrotrophic fungus Botrytis cinerea, outside of agricultural settings and have determined whether the populations in these natural habitats can be distinguished phenotypically and phylogenetically from populations isolated from diseased crops. Their results showed that B. cinerea strains sampled on different non-agricultural substrates were genetically and phenotypically similar to

### REFERENCES


strains sampled in agricultural substrates. These results suggest that highly diverse populations of this plant pathogen persist outside of agriculture in association with substrates other than cultivated plants and that this component of their life-history is compatible with its capability to maintain its potential as plant pathogen. The survival phase was also considered in Vaumourin and Laine, as coinfection affects the number of sexual resting structures produced, and in Fournet et al., as they considered also the hatching after storage at different temperatures, simulating survival conditions during the inter-cropping period.

Considering plant pathogen traits in the context of life history trait evolution is a promising research avenue, still in its infancy. We hope that this research topic will help foster this line of research and provide a valuable resource for researchers working in the field of evolutionary ecology of plant pathogen populations.

### AUTHOR CONTRIBUTIONS

CL, JM, CM, OF, and VR have made substantial direct contributions to the work, and approved it for publication.

### ACKNOWLEDGMENTS

We thank all the authors involved in this research topic for their excellent and varied contributions.

Climate Change (2nd edn): Observed Impacts on Planet Earth. Ed. Trevor M. Letcher (Amsterdam, The Netherlands: Elsevier BV), 325–338.


Tollenaere, C., Susi, H., and Laine, A.-L. (2016). Evolutionary and epidemiological implications of multiple infection in plants. Trends Plant Sci. 21, 80–90. doi: 10.1016/j.tplants.2015.10.014

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 Le May, Montarry, Morris, Frenkel and Ravigné. 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.

# Genome Scans Reveal Homogenization and Local Adaptations in Populations of the Soybean Cyst Nematode

Anne-Frédérique Gendron St-Marseille1,2, Etienne Lord<sup>1</sup> , Pierre-Yves Véronneau<sup>1</sup> , Jacques Brodeur <sup>2</sup> and Benjamin Mimee<sup>1</sup> \*

<sup>1</sup> Saint-Jean-sur-Richelieu Research and Development Centre, Agriculture and Agri-Food Canada, Saint-Jean-sur-Richelieu, QC, Canada, <sup>2</sup> Institut de Recherche en Biologie Végétale (IRBV), Université de Montréal, Montréal, QC, Canada

Determining the adaptive potential of alien invasive species in a new environment is a key concern for risk assessment. As climate change is affecting local climatic conditions, widespread modifications in species distribution are expected. Therefore, the genetic mechanisms underlying local adaptations must be understood in order to predict future species distribution. The soybean cyst nematode (SCN), Heterodera glycines Ichinohe, is a major pathogen of soybean that was accidentally introduced in most soybean-producing countries. In this study, we explored patterns of genetic exchange between North American populations of SCN and the effect of isolation by geographical distance. Genotyping-by-sequencing was used to sequence and compare 64 SCN populations from the United States and Canada. At large scale, only a weak correlation was found between genetic distance (Wright's fixation index, FST) and geographic distance, but local effects were strong in recently infested states. Our results also showed a high level of genetic differentiation within some populations, allowing them to adapt to new environments and become established in new soybean-producing areas. Bayesian genome scan methods identified 15 loci under selection for climatic or geographic co-variables. Among these loci, two non-synonymous mutations were detected in SMAD-4 (mothers against decapentaplegic homolog 4) and DOP-3 (dopamine receptor 3). High-impact variants linked to these loci by genetic hitchhiking were also highlighted as putatively involved in local adaptation of SCN populations to new environments. Overall, it appears that strong selective pressure by resistant cultivars is causing a large scale homogenization with virulent populations.

Keywords: Bayesian outlier detection, genetic diversity, genotyping-by-sequencing, Heterodera glycines, isolation by distance

### INTRODUCTION

The introduction of an organism into a new environment can have unpredictable detrimental consequences, including public health problems, losses in biodiversity and ecosystem services or crop yield losses due to exotic weeds, insects and pathogens, altogether resulting in significant economic impact (Pimentel et al., 2005). Unfortunately, the steady increase in international trade

#### Edited by:

Christophe Le May, Agrocampus Ouest, France

#### Reviewed by:

Yessica Rico, Instituto de Ecología (INECOL), Mexico Sebastian Eves-van Den Akker, University of Cambridge, United Kingdom

> \*Correspondence: Benjamin Mimee benjamin.mimee@agr.gc.ca

#### Specialty section:

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

Received: 21 March 2018 Accepted: 18 June 2018 Published: 17 July 2018

#### Citation:

Gendron St-Marseille A-F, Lord E, Véronneau P-Y, Brodeur J and Mimee B (2018) Genome Scans Reveal Homogenization and Local Adaptations in Populations of the Soybean Cyst Nematode. Front. Plant Sci. 9:987. doi: 10.3389/fpls.2018.00987

**8**

facilitates the movement and introduction of new invasive species (Hulme, 2009). In addition, climate change is altering environmental conditions and could change the species' distribution range or favor their establishment in previously unsuitable habitats (Early and Sax, 2014). It is therefore imperative to carry out specific risk assessment in order to target species to be controlled. Alien invasive species, by definition, did not evolve in the biogeographic habitat in which they are introduced. Consequently, they are often poorly adapted to their new environment. Generally, the most successful invaders will have a high potential for rapid adaptation through phenotypic plasticity or microevolution (Novak, 2007). Understanding the genetic mechanisms of local adaptation is therefore critical to predict future species distribution.

Plant-parasitic nematodes are microscopic worms that reduce global annual food production by 12.3% and cause more than US\$157 billion in economic losses worldwide (Hassan et al., 2013). In North America, one of the most damaging species is the soybean cyst nematode (SCN), Heterodera glycines Ichinohe. Since its first detection in 1954 in Hanover County, North Carolina, SCN has been reported in almost every soybeanproducing county in the United States (Davis and Tylka, 2000), as well as in southwestern Ontario, Canada (Anderson et al., 1988). During the 2000s, SCN colonized several new northern localities in the United States (North Dakota in 2003; Mathew et al., 2015) and Eastern Canada (northeastern Ontario in 2007 and Québec in 2013; Mimee et al., 2014). A few studies have investigated SCN dispersal, which seems to follow a northward and eastward pathway in North America (Tylka and Marett, 2017). This dispersion pattern correlates with the expansion of soybean cultivation following the introduction of new soybean varieties with shorter maturity periods and improved tolerance to drought and cold (Shurtleff and Aoyagi, 2010; Yu, 2011).

It is expected that SCN would survive and multiply throughout the current North American soybean-growing area and complete at least two generations at its northern limit (Gendron St-Marseille, 2013). Climate warming should also favor the establishment of SCN at higher latitudes and increase the number of generations per year in most regions. However, these predictions are based strictly on temperature requirements and do not account for the intrinsic capacity of SCN to adapt to new environmental conditions. Genetic variations within a population reflect its evolutionary potential and result from four evolutionary forces that affect individual fitness: mutation, gene flow, selection, and genetic drift (Eizaguirre and Baltazar-Soares, 2014). For most organisms, including SCN, the relative weights of these forces can differ significantly. Mutations are rare events that should not contribute significantly to SCN adaptations in a short time frame. Gene flow depends on the dispersal ability of an organism, which for SCN is achieved mainly by means of human activities at the short spatial scale (Kristjansson, 2010). In addition, wind and flooding can carry SCN cysts over very long distances, and contribute to its dispersal at the regional and continental scales. Many different selection pressures can shape the genetic structure of SCN populations, but host plant is probably the strongest selection factor. The ability of SCN to reproduce on a given soybean genotype differs greatly depending on its resistance genes and the nematode's virulence profile (HG type) (Colgrove and Niblack, 2008; Niblack et al., 2008). Thus, management decisions by growers (for example, the systematic use of resistant cultivars) can result in a strong selection pressure. Finally, the influence of genetic drift will also depend on pest control strategies, because they contribute to dictate population size, although it was shown for cyst nematodes that genetic diversity can be very high within a single cyst (Green et al., 1970). Each nematode female can mate with several males and lay hundreds of eggs that can survive for at least a decade in the soil (Slack et al., 1972). Thus, even if the diversity appears reduced due to genetic drift under strong selection by resistant cultivars, most alleles probably persist for several years in the population in a "dormant" state.

In other cyst nematode species, genetic diversity at the population level has been studied by means of several techniques, including microsatellite markers, ITS-RFLP (internal transcribed spacer–restriction fragment length polymorphism), RAPD (random amplified polymorphic DNA), and 2-DGE (twodimensional gel electrophoresis) (e.g., Blok et al., 1997; Grenier et al., 2001; Plantard et al., 2008; Boucher et al., 2013). However, these methods focus only on specific sections of the genome, yield very few markers, or do not allow precise comparisons among populations. For example, Eves-van den Akker et al. (2015) recently published a metagenetic approach to genotype populations of the pale cyst nematode, Globodera pallida. While this method was shown powerful to rapidly evaluate genetic diversity and distribution of specific mitotypes, it is based on very few neutral markers, which prevent any assessment for selection. Genome scan is an interesting approach for the identification of genetic loci involved in adaptation to specific selection pressure. It was notably used in G. pallida to identify genomic regions associated with virulence on resistant potato cultivars (Eoche-Bosy et al., 2017).

Despite rapid advances in next-generation sequencing (NGS) technologies, sequencing a large number of individual genomes at high coverage in order to perform population genetic studies remains very expensive and may require important quantities of DNA from individuals. Elshire et al. (2011) developed a genotyping-by-sequencing (GBS) protocol to rapidly identify single-nucleotide polymorphisms (SNPs). The GBS technique and other restriction-site-associated DNA (RAD) sequencing methods produce large quantities of reads that do not cover the entire genome but have higher sequencing depth, thus reducing sequencing errors (Gautier et al., 2013; Anand et al., 2016). Loci generated by GBS can be present in both coding and non-coding regions and will be shared between all populations owing to the conservation of restriction sites (Cariou et al., 2013). Finally, GBS does not require any prior genomic information for the species being studied, which is an important consideration for SCN since there is no reference genome yet. The optimal gene coverage to reduce the amount of missing data depends on the choice of restriction enzyme (Fu et al., 2016). Fortunately, optimal gene coverage was already tested for the closely related species Globodera rostochiensis (Mimee et al., 2015).

The main objectives of this study were to (1) investigate the genetic relationships among SCN populations from United States and Canada, (2) detect isolation by geographical distance (IBD) between SCN populations from United States and Canada, (3) detect genetic loci under selection associated to environmental and climatic parameters, and (4) identify the putative gene functions contributing to the adaptation of SCN populations to specific environmental conditions.

### MATERIALS AND METHODS

### Soybean Cyst Nematode Populations Sampling and Genotyping-by-Sequencing

A total of 64 field populations of SCN, representative of the area currently infested in North America, were sampled or provided by collaborators from 11 US states (Delaware, Iowa, Illinois, Indiana, Kansas, Michigan, Minnesota, Missouri, North Dakota, Ohio, South Dakota) and one Canadian province (Ontario) (**Figure 1**). For DNA extraction, 40 cysts were randomly chosen from each population and pooled together. Eggs were extracted from each cyst and then washed twice in sterile filtered water. Total genomic DNA of each pool was extracted using the DNeasy Blood and Tissue Kit (Qiagen, Mississauga, ON, Canada) following the manufacturer's instructions. DNA extracts were quantified using Qubit fluorometric quantification (ThermoFisher Scientific, Burlington, ON, Canada) and normalized at 2 ng/µL prior to library preparation and sequencing. These steps were performed following standard protocols (Elshire et al., 2011; Poland et al., 2012) at the Institute for Integrative and Systems Biology (IBIS; Université Laval, Québec City, QC, Canada). Genotyping-bysequencing was performed using the method described by Mimee et al. (2015) with a combination of two restriction enzymes (PstI/MspI) (New England Biolabs, Whitby, ON, Canada). After the restriction enzyme treatment, all samples (one composite per field population) were barcoded and multiplexed to obtain a single library for the 64 samples, which was sequenced on three Ion Proton chips (ThermoFisher Scientific) at IBIS.

### Single-Nucleotide Polymorphism Calling

The UNEAK pipeline (Lu et al., 2013), which is part of the TASSEL 3.0 bioinformatics analysis package (Bradbury et al., 2007), was used to process the raw reads, since no complete reference genome is yet available for H. glycines. This pipeline is designed to call SNPs de novo, without a reference genome with high stringency. Only sequences containing a single putative SNP (1-bp mismatch) per sequencing read were kept. Before analysis, the SNPs were further filtered with a minimum coverage (minCov) of 20 reads, a maximum coverage (maxCov) of 10,000 reads, and a minimum allele frequency (MAF) of 0.01.

### Population Genetics

Clustering of SCN populations using PCA was carried out using the prcomp function from the stats package in R software (R Core Team, 2017). The poppr package (v2.4.1) in R (Kamvar et al., 2014) was used to investigate the genetic relationship between populations and to build a phylogenetic tree (Provesti's distance, 10,000 bootstraps, neighbor-joining algorithm counting missing data as equivalent in the distance computation). Visualization of sample coordinates and phylogenetic relationship analyses were carried out using the phytools 0.6–20 package in R (Revell, 2017).

Fixation index (FST) values (Wright, 1943) were calculated using the classical approach (Hartl and Clark, 2007) with the PoPoolation2 software (Kofler et al., 2011) to evaluate the genetic differentiation between each pair of populations. Furthermore, the effect of isolation by distance on population structure was tested using the correlation between the genetic distance ratio [FST/(1–FST)], as defined by Rousset (1997), and the geographic distance of population pairs in kilometers (km). The geographic distance between each sample location was calculated with the haversine formula using the geosphere package in R. To examine the significance of the relationship between the genetic distance ratio and the pairwise geographic distance distances, we performed a Mantel test (Spearman rank correlation) using the vegan package in R with 999 permutations. To evaluate the effect of time on population genetic differentiation, linear regression analyses were run in R software using pairwise FST distances based on a point of origin selected on the basis of the first reports of SCN in North America. In our dataset, the closest sample to the oldest population was located in Clarkton, Missouri (MO1) (Hegge, 1957; Tylka and Marett, 2017). To access the genetic isolation from the MO1 location, a Spearman rank correlation (rs) test was performed at a 0.95 confidence level.

Populations from three states (North Dakota, Minnesota, Illinois) were selected as case studies to evaluate the local (shortscale) genetic differentiation. These states were chosen because there were sufficient samples for comparison and because (i) North Dakota shows a recent introduction of SCN and a continuous northward dispersal of the nematode (Nelson and Bradley, 2003; Mathew et al., 2015); (ii) Minnesota has a longer history of SCN, with the nematode being first detected in 1978 and the infested area still expanding each year (Zheng et al., 2006); (iii) in Illinois SCN has been well established in every county for many years (Riggs, 2004; Tylka and Marett, 2017).

### Outlier Detection and Their Association With Environmental Variables

Two geographic and four climatic covariables were investigated as possible factors explaining loci under selection: latitude (LAT), longitude (LONG), annual mean air temperature (BIO1), maximum air temperature of the warmest month (BIO5), annual precipitation (BIO12), and total precipitation of the warmest quarter (BIO18). All climatic variables were retrieved from the WorldClim global climate database, version 1.4 (Fick and Hijmans, 2017), corresponding to historic conditions (1960– 1990). The spatial resolution used for the bioclimatic analysis was set at 30 s or 0.86 km<sup>2</sup> .

To detect correlations between variations in population allele frequencies of SNPs and environmental factors, we used three different Bayesian methods (software programs): BayPass, version 2.1 (Gautier, 2015), BayeScan, version 2.1 (Foll and Gaggiotti, 2008), and BayeScEnv, version 1.1 (de Villemereuil and Gaggiotti, 2015). For each program, triplicate runs with different random seeds were performed with a pilot run of

10,000 iterations to estimate starting parameters, a burn-in length of 50,000 iterations, and a minimum of 50,000 iterations, accounting for the small number of SNPs being investigated. Post-run diagnostics were carried out using the coda package in R software in order to ensure sufficient iterations and normality of the Markov chain (Plummer et al., 2006).

The first software, BayPass, identifies genetic markers subject to selection by covariates such as phenotypic or environmental variables associated with the population of interest (Gautier, 2015). This application is based on the BAYENV model proposed by Coop et al. (2010) and Günther and Coop (2013), but with several modifications detailed in Gautier (2015), including the reprogramming of the Markov chain Monte Carlo (MCMC) algorithm. For BayPass, we used a pool-size file with the – d0yij option set at 800 and 20 pilot runs with a length of 10,000 iterations, a burn-in length of 50,000 iterations, and a chain length of 50,000 iterations. For each SNP, BayPass generates a Bayes factor (BF), quantifying evidence against the null hypothesis, and an empirical Bayesian p-value (eBPis), a metric measuring the difference between observed data and a simulated set of data (posterior distribution) (Kass and Raftery, 1995; Andraszewicz et al., 2015; Gautier, 2015). To be considered under selection, a SNP had to meet two criteria: a BF greater than 10 (BF > 10) and an eBPis lower than 0.05 (eBPis < 0.05).

The second software, BayeScan, uses a Bayesian likelihood method that assumes a Dirichlet distribution of allele frequencies between populations (Foll, 2012). This program estimates the probability that each locus is subject to selection by using a logistic regression on the two locus-population FST coefficients. This Bayesian method uses a reversible-jump MCMC algorithm to calculate a posterior probability that each locus is under selection. The decision criterion to determine whether a locus is likely to be under a strong selection is the q-value (Foll, 2012), analog to a false discovery rate (FDR) p-value, that must be under 0.05. A second decision criterion was applied to further endorse the selected outliers: the ratio of posterior probabilities (PO). The PO threshold to affirm that a locus was under selection, in comparison with a neutral model, was set to 0.91, which corresponds to a strong relationship on the Jeffreys scale (Foll, 2012). We used BayeScan with the default parameters, but we set the minimum number of iterations to 50,000, the length of 20 pilot runs to 10,000 iterations, and the burn-in length to 50,000 iterations.

The third software, BayeScEnv, is similar to BayeScan and uses the FST index to detect loci with a high level of differentiation in comparison with the entire genome. This program allows a normalization vector to be applied to the environmental data instead of only a binary combination, thus generating a lower number of false positives, according to the creators of the software (de Villemereuil and Gaggiotti, 2015). We used BayeScEnv with the default parameters, with the number of iterations set to 50,000 and 20 pilot runs with a length of 10,000 iterations. We used the reported q-value, which is related to the FDR, as our decision criterion, considering only SNPs with a q-value less than 0.05.

### Localization of Outlier Loci, Gene Function, and Genic Environment

The SNPs identified in short reads were first retrieved from a draft SCN genome available from SCNBase (https://www. scnbase.org/) by means of BLASTN with the default parameters, except for a smaller word size of 4, with the Blast2GO application (Conesa et al., 2005). Many of the identified SNP-containing fragments matched multiple genes or genome locations (see Supplementary Tables 3–5). To assign a putative gene function to each SNP, we compared the aligned sequences to the National Center for Biotechnology Information (NCBI) protein database by means of BLASTX and BLASTP (Altschul et al., 1990) on a subset of sequences (nematodes, taxid: 6231) or to all of the NCBI non-redundant (nr) sequence database with an E-value significance cutoff of 1e−<sup>5</sup> .

As our GBS sequencing covers approximately 0.8% of the genome (see Results section), we explored the genomic regions around outlier loci for genetic hitchhiking in whole-genome sequences from four populations of different origin and distinct virulence profile (ON1, ON34, IL4, and KS2). Sequencing libraries were prepared using the Nextera XT DNA Library Preparation Kit (Illumina, San Diego, CA) and sequenced on a MiSeq sequencer (Illumina) using the MiSeq Reagent Kit v3 (600-cycle). Reads were demultiplexed using the Sabre software program (https://github.com/najoshi/sabre) and processed with Trimmomatic (v0.32) (Bolger et al., 2014) to remove adapters and barcodes. Alignment on the draft reference genome (see above) was done using SAMtools (v0.1.19) (Li et al., 2009) and BWA (v0.7.12) (Li and Durbin, 2009). Only a window of ± 50 kb around the 15 loci under selection was retained for the genic environment analysis. Variants were called with freebayes (v1.0.2) (Garrison and Marth, 2012) and snpeff (v4.2) (Cingolani et al., 2012). Distribution of allele frequencies in these four populations for each gene variant was compared with the allele frequency of the corresponding loci under selection in the same population in the genotyping-by-sequencing dataset. Only SNPs exhibiting a similar allele frequency and having a high impact on the predicted protein product were retained in our analysis.

### RESULTS

### Genotyping-by-Sequencing

A total of 192,576,709 short reads were obtained from the sequencing of the DNA from the 64 SCN populations, following digestion by the PstI/MspI restriction enzymes. The H. glycines effective genome length was estimated to be 96,752,286 bp, with an average of one PstI restriction site every 12,688 bases (E. Lord, private communication, 2017). On the basis of the sequence size used by the UNEAK (Universal Network-Enabled Analysis Kit) pipeline (first 64 bp of each read), the expected horizontal coverage was approximately 0.8% of the total SCN genome, and the vertical coverage at each locus, considering the number of reads obtained, was 400×. Before filtering, the UNEAK pipeline identified 3,172 variants. After filtering for low coverage and selecting only SNP variants, two datasets were generated. The first dataset contained 245 SNPs without missing data (loci sequenced in all populations), and the second contained 804 SNPs with missing data (Sequences and coverage in Supplementary Table 7).

## Population Structure

### Principal Component Analysis

A principal component analysis (PCA) was performed using the dataset without missing data, which contained 245 SNPs. The Gendron St-Marseille et al. Population Genetics of Heterodera glycines

PCA plot of the 64 North American SCN populations revealed a geographically ordered pattern (**Figure 1A**). Overall, the Ontario populations showed the greatest dispersion, indicating they are more genetically differentiated, while the central regions of the US showed more clustered populations. Also, the group containing populations from Minnesota and North Dakota (in green in **Figure 1A**) was clearly different from the Ontario populations by the first axis, which explained 21% of the total variation. All populations from the central states, although they originated from many states and covered a much wider area, were less diverse and more clustered together.

### Phylogenetic Tree

To better understand the population structure of each population, we used the 804-SNP dataset to conduct a phylogenetic analysis based on Provesti's distance. The inferred neighbor-joining phylogenetic tree (**Figure 1B**) showed that most of the Ontario samples (in red in **Figure 1B**) were different from the rest of the North American populations. Some samples from Ontario were found to be different from each others and more similar to those of central states. Such a pattern can also be observed in the PCA analysis (**Figure 1A**).

### Genetic Differentiation

Overall, we observed an average value for Wright's fixation index (FST) of 0.15 ± 0.07. However, pairwise genetic distance was highly variable, with a minimum FST of 0.005 and a maximum FST of 0.53 (Supplementary Table 1). The highest FST value was found between two of the most remote populations, ND6 from North Dakota and ON41 from Ontario, which were separately by approximately 1,720 km. On the other hand, the lowest FST value was observed between ON3 and ON5, which were separated by only 17.9 km. In general, higher FST values were found in distant populations, especially between populations from Minnesota (MN4 and MN5), North Dakota (ND3 and ND6), South Dakota (SD2), and Ontario (ON6, ON13, ON25, ON41, and ON46).

### Isolation by Distance

The analysis of the relationship between genetic distance [FST/(1– FST)] and geographic distance (km) for all pairs of populations revealed a significant (p = 0.012) but weak effect of isolation by distance, with a Mantel correlation (r) of only 0.135. However, when analyzing the effect of isolation by distance at the local scale, a different pattern was found. With respect to US states with recent SCN history, a strong and significant positive correlation was observed between the genetic and geographic distances in North Dakota (r = 0.430, p = 0.013) and Minnesota (r = 0.730, p = 0.006). In contrast, no correlation was found among the populations from Illinois (r = −0.170, p = 0.767).

To further explore the contribution of isolation by distance to the genetic differentiation of United States and Canada populations of SCN, we compared the genetic distance of each population from a theoretical ancestor population (MO1) using the 245-SNP dataset. This analysis included 38 independent populations that were separated by 273 to 1,367 km from MO1 (**Figure 2A**). The genetic-distance ratio to the "oldest population" varied from 0.06 for IL1 (386 km apart) to 0.52 for

ND6 (1,327 km apart) (Supplementary Table 2) and an overall moderate (r<sup>s</sup> = 0.535) but significant (p = 0.001) relationship was observed between genetic distance and geographic distance (**Figure 2B**).

### Signature of Genetic Adaptations Genome Scans

Using three Bayesian inference genome scan approaches, we identified 71 SNPs (out of 804) that were under selection. Among the different computational methods, 48 loci were identified by BayeScan, 23 loci were identified by BayPass, and 16 loci were identified by BayeScEnv (Supplementary Tables 3–5). Out of those 71 SNPs, only one was highlighted by the three methods, while 15 outliers were inferred by at least two of the methods (**Figure 3**). The geographic variables (LAT and LONG) were associated with seven and eight outlier loci, respectively, and the climatic variables (BIO1, BIO5, BIO12, and BIO18) were associated with nine, six, eight, and seven SNPs, respectively (**Table 1**). Three SNPs were strictly associated with

only one environmental variable: TP188858 was associated with maximum temperature, while TP252086 and TP364091 appeared to be linked to annual temperature. The allele frequencies analysis at these 15 loci revealed SNPs that were specific for a given region, such as TP364091, TP181514, and TP333913 for Minnesota and North Dakota and TP376687, TP146577, and TP226227 for Ontario (**Figure 4**).

### Localization, Effect, and Genic Environment of Outlier Loci

Of the 15 SNPs under selection identified by at least two Bayesian approaches, 10 SNPs were located in predicted genes (seven in exons and three in introns), and five SNPs were located in intergenic regions (**Table 2**). Two SNPs (TP380188 and TP364091) correspond to non-synonymous mutations that induce a change in the amino acid sequence of the resulting protein. The annotations of the genes putatively impacted by these modifications are SMAD-4 (mothers against decapentaplegic homolog 4) for TP380188 and DOP-3 (dopamine receptor 3) for TP364091. The sequence corresponding to TP364091 was retrieved in four different genes (g14639.t1, g14642.t1, g14644.t1, and g14654.t1) in the draft reference genome and introduced two different mutations (Supplementary Tables 3–5). All the other SNPs (13/15) were silent mutations (synonymous in exons or located in introns or intergenic regions) and thus probably not the cause of their selection. Among the 42 remaining SNPs that were under selection but identified by only one pipeline, 12 correspond to a non-synonymous modification impacting the predicted protein sequence. These SNPs include the gene UFM1 (Ubiquitin-fold modifier 1), the gene PLA2 (85/88 kDa calcium-independent phospholipase A2), and a pyruvate kinase (Supplementary Tables 3–5).

The whole-genome resequencing of four SCN populations for the exploration of the genic environment of outlier loci highlighted 257 genes containing high-impact variants in a ± 50-kb window around loci under selection. Annotation was available for 148 of these genes, corresponding to 132 different gene functions (Supplementary Table 6). Of these genes, 25 exhibited allele frequencies similar to their neighboring outlier loci (**Table 3**).

### DISCUSSION

Since its introduction into North America in 1954, the soybean cyst nematode has spread at a regular pace to most of the soybean-producing areas in the US and reached Canada in 1987. Its northern limit of establishment remains to be determined, as new populations are found each year. Soybean acreage is rapidly increasing in northern latitudes; for example, farmers in the province of Manitoba, Canada, sowed 2.3 million acres with soybean in 2017, a 40% increase in 1 year in a province where soybean was not grown a decade ago (Statistics Canada, 2017). This spectacular change in land use results from a coordinated breeding program for short-season and droughttolerant soybean cultivars (Tardivel et al., 2014). Although soybean is nowadays very profitable in Canada (Manitoba, Northern Ontario, Québec), managing a pest such as SCN could be challenging, as no resistant cultivars are currently adapted to these regions. Simulations based on thermal development have shown that SCN could theoretically survive in these new soybean growing areas (Gendron St-Marseille, 2013), but actual establishment is a different matter, as initial observations in the province of Québec indicated that even though the nematode is detected in many regions, populations weakly reproduced (Mimee et al., 2016). This phenomenon probably reflects SCN's very recent history in this part of the world and the poor fitness of the introduced populations but does not exclude possible adaptations of the nematode in the future. To better understand the evolution of SCN population genetics in North America, we compared 64 populations originating from 11 US states and the province of Ontario in Canada.

The dynamics of cyst nematode populations are intrinsically complex. Several thousand cysts from different females can form on a single plant, and each cyst contains hundreds of halfsibling individuals as a result of SCN's polyandrous mode of reproduction. Thus, the number of unique genotypes occurring in a field is massive. Isolating individual nematodes, even hundreds of them, to explore the genetic relationship between populations would be an arduous task and the approach would still be biased owing to limited sub-sampling. For these reasons, we opted for a Pool-Seq (sequencing of pooled DNA samples) approach. Sequencing DNA from pooled samples for each population also has the advantage of keeping the number of


TABLE 1 | Association of outlier loci [single-nucleotide polymorphisms (SNPs)] identified by at least two Bayesian genome scan methods in 64 populations of the soybean cyst nematode from North America with geographic or climatic co-variables.

redundant sequences low (Schlötterer et al., 2014). However, this approach does not allow the assignment of sequences to individuals and prevents some population genetic analyses such as linkage-disequilibrium estimation (Lynch et al., 2014; Anand et al., 2016). Nevertheless, a number of recent studies have overcome these limitations and have proposed new methods to

climatic co-variables. Black squares represent a significant association.

analyze genetic variations among populations based on Pool-Seq data (Van Tassell et al., 2008; Gautier et al., 2013; Navon et al., 2013; Mimee et al., 2015; Anand et al., 2016).

Both the PCA and phylogenetic analyses showed strong clustering of populations based on their geographic origin, which supports the "nearest-neighbor theory" movement as in the stepping-stone model (Kimura and Weiss, 1964; Hutchison and Templeton, 1999). We observed a geographic separation dividing SCN populations into two clusters, a northeastern one (Ontario) and a northwestern one (Minnesota, North Dakota) that both share genetic similarities with central populations. In northeast and northwest regions, SCN populations established only recently, and new detections continue to occur in the northern areas. In the central states, SCN populations have been present for a longer time and resistant soybean cultivars are routinely used. We believed that these two conditions contribute to explain the homogeneity of the central populations. On one hand, the continuous exchange of genetic material would have led to the homogenization of alleles in sympatric populations. On the other hand, the massive use of nematode resistant cultivars that are all derived from a single source of resistance (PI 88788) would have selected for the virulent SCN genotypes, thereby limiting the propagation of some alleles. It is thus surprising to observe a greater genetic differentiation in SCN populations from Ontario since the main assumption is that they originate from central states. These populations should contain less diversity, owing to the founder effect, which would normally result in reduced potential for differentiation. However, the selective pressures mentioned above are not yet present in Ontario, allowing the development of most genotypes. This finding concurs with the results of Faghihi et al. (2010), who observed a much greater diversity in SCN population phenotypes (HG types) in Ontario than in Tennessee, Indiana, or Illinois. A similar pattern was obtained in Minnesota by Zheng et al. (2006) and in China by Liu et al. (1997) when comparing populations from the north to populations from central regions. However, all these authors suggested that climatic conditions or local environmental factors may also play a role.

Frontiers in Plant Science | www.frontiersin.org

TABLE 2 | Identification and effect of outlier loci [single-nucleotide polymorphisms (SNPs)] identified by at least two Bayesian genome scan methods in 64 populations of the soybean cyst nematode from North America.


<sup>1</sup>Synonymous (S) or non-synonymous (N-S) mutations in the final protein product.

<sup>2</sup>a.a., amino acid substitution.

According to Hartl and Clark (2007), the overall average FST value of 0.15 that we obtained by comparing all pairs of SCN populations from North America corresponds to a high level of differentiation. However, when we examined specific pairs of populations, we found that the populations were very similar at the local scale. Also, the differentiation of recent populations (Ontario, Minnesota, North Dakota) from those of the central states was only moderate, whereas comparing these recent populations together generated very high FST values (up to 0.53). This corroborates results from the PCA analysis and clearly suggests that populations diverged following two distinct routes, one in the northwest toward North Dakota and one in the northeast toward Ontario. This hypothesis is further supported by the isolation-by-distance analysis. Overall, we observed a significant effect of distance on genetic differentiation at the continental scale. The effect of distance was even more noticeable in newly colonized areas like North Dakota and Minnesota or when each population was compared to a common ancestor. On the other hand, the absence of any effect of distance in Illinois and low FST in the central states clearly suggests that dispersal is frequent enough to weaken signs of isolation by distance after a short period of time. This continuous gene flow combined with the selection of alleles due to the general use of the same resistant soybean lines are thought to cause a rapid homogenization of SCN populations (Mitchum et al., 2007; Niblack et al., 2008; Zheng and Chen, 2011). If SCN is already adapted to environmental conditions prevailing in northern areas, this homogenization process would accelerate the development of virulent genotypes in these regions (in comparison with the central states) and would be independent from the use of resistant cultivars. This concurs with the finding of Faghihi et al. (2010) where 15% of SCN populations in Ontario were able to multiply on PI 548402 and 6% were able to multiply

on PI 90763, even though these sources of resistance were not present in commercial cultivars in Ontario.

Although 71 SNPs were found to be under selection among SCN populations, only one SNP was identified by the three Bayesian approaches tested. This SNP (TP333577) was strongly associated with precipitation; but the geographic distribution of allele frequencies (**Figure 4**) at that specific locus did not match with the general pattern of genetic differentiation, instead showing a very local effect (in populations from Minnesota but also in some populations from Iowa, Illinois, North Dakota, and Ohio). Thus, this selection for wetness was either lost in favor of a more stringent pressure or too costly to be maintained in the other regions. Overall, considering the broad range of environmental conditions in which these populations develop, finding only a few genes under selection suggests that SCN possesses an intrinsic capacity to evolve within a large range of temperature and wetness gradients.

Most of the 15 outliers identified by at least two pipelines correlated with temperature, reinforcing that temperature is a key factor that acts on the genetic diversity and natural selection of nematode populations. Nematodes are poikilothermic organisms and depend greatly on temperature to complete their life cycle. Thus, any adaptation that facilitates the development and reproduction of nematodes at lower temperatures would be advantageous. On the basis of the distribution of allele frequencies, however, this adaptation does not seem crucial for survival, as the northern populations did not all exhibit similar patterns. This trait is probably not yet fixed in these populations. Nevertheless, six SNPs under selection and associated with temperature were specific to populations from Ontario (TP376687, TP146577, and TP226227) or only retrieved from Minnesota and North Dakota (TP364091, TP181514, and TP333913), a finding that clearly suggests local adaptations.


TABLE 3 | Putative hitchhiking gene variants with high-impact neighboring loci under selection and exhibiting a similar allele frequency in Heterodera glycines.

<sup>1</sup>Expected value for annotations of each putative hitchhiking gene.

<sup>2</sup>Variant in bold indicates a stop codon gain. More details on types of variants and position in the gene are provided in Supplementary Table 6. SNP, single-nucleotide polymorphism; MNP, multiple-nucleotide polymorphism.

The SNP corresponding to TP364091 was retrieved in four out of six populations from Minnesota and in one population from North Dakota. That SNP, which is a nonsynonymous mutation in the gene DOP-3 (dopamine receptor 3) that replaces an aspartic acid with an asparagine, was strongly associated with the annual mean temperature and annual precipitation. This modification was retrieved in two distinct gene sequences (g14642.t1 and g14654.t1), and a synonymous mutation (Pro/Pro) was retrieved in two other sequences (g14639.t1 and g14644.t1) but in a different amino acid, indicating a reading frame shift and a putative upstream indel. These four genes were all annotated as DOP-3 and were probably the result of gene duplication. Gene duplication has frequently been hypothesized to play an important role in adaptation to the environment (reviewed in Kondrashov, 2012). In Caenorhabditis elegans, the DOP-3 protein is well known to modulate chemosensory functions, such as mating and foraging, and to be involved in locomotion (Chase et al., 2004; Wood and Ferkey, 2016). The other SNP that introduces a nonsynonymous mutation and was highlighted by two genome scan methods, namely, TP380188, was located in the gene SMAD-4 (mothers against decapentaplegic homolog 4) and was retrieved in only a few populations from Minnesota, North Dakota, and Ontario. The SMAD-4 protein is a regulator of different cellular processes, including cell differentiation, apoptosis, migration, and proliferation (Nikolic et al., 2011). Mutations of SMAD-2, SMAD-3, SMAD-4, or SMAD-6 were previously shown to result in a 30% reduction in body size in C. elegans (Savage-Dunn et al., 2003; Watanabe et al., 2007). Also of interest is the locus TP122360, identified only by BayPass, that was located in the UFM1 (ubiquitin-fold modifier 1) genes (g7250.t1 and g7228.t1). A deletion in this gene in C. elegans, although reducing reproduction rate and life span, increased the survival of this nematode under oxidative or heat stress (Hertel et al., 2013).

For the majority of the loci, their selection probably results from genetic hitchhiking rather than a direct contribution to adaptation. To explore that possibility, we analyzed the genic environment in a ± 50 kb window around each SNP in four SCN populations. Genes containing high-impact genetic variations and exhibiting the same allele frequencies as the associated SNP under selection were of particular interest. Two of these genes (g18327.t1 and g15502.t1) coded for G-protein-coupled receptors (GPCR). These proteins are members of a large and very diverse multigene family with hundreds of occurrences in the C. elegans genome (Bockaert and Pin, 1999; Bargmann, 2006). The GCPR are crucial in sensing the local environment and were shown to evolve following alterations in habitat or foraging behavior (Nei et al., 2008). Four genes coding for proteins implicated in metabolite transport into the cell were also identified (g12205.t1, g14656.t1, g18812.t1, and g18825.t1), as was one gene involved in the regulation of those proteins (g18365.t1). All these proteins are required to maintain homeostasis in the cell and to respond to local environmental changes. In a similar study using the fungus Fagus sylvatica, Pluess et al. (2016) found that a version of a potassium transporter was associated with lower precipitation and could contribute to the regulation of growth under dry conditions. Structural changes in the body of an organism as a result of microevolution can also confer a significant advantage in terms of resisting more adverse environmental conditions, such as drought or high temperatures (Hazel and Williams, 1990). In our study, mutations were observed in two genes involved in cell-membrane and cell-wall stability (g18328.t1 and g8311.t1). Lastly, modifications in genes involved in the regulation of transcription (g2288.t1 and g3633.t1) and in the maturation of proteins (g14633.t1 and g18358.t2) have the potential to radically change the proteome of the adapted organism.

Although some SCN genes were found to be under selection and local adaptation was found to be underway in this study, our results also indicate that there is no critical adaptive event required for SCN establishment in northern latitudes.

### REFERENCES


Consequently, all populations should theoretically survive and multiply at high latitudes. The risk is thus real for new soybean areas where cultivars resistant to SCN are not available at this time. Of course, the method we used does not explore the entire genome, and a pan-genomic study of these populations could reveal other loci under selection in an evolutionary process.

### 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

BM, A-FG, and JB conceived and designed the experiments. A-FG, EL, and P-YV performed the experiments and analyzed the data. A-FG, EL, and BM wrote the paper. All the authors revised the manuscript.

### ACKNOWLEDGMENTS

The authors thank Drs. Thomas Baum and Andrew Severin for graciously making SCN genome sequences available to us before public release on scnbase.org. The authors also thank Tom Welacky (AAFC) and Terry Niblack (OSU) for providing SCN populations. The authors wish to acknowledge the contribution of Génome Québec, Genome Canada, the Government of Canada, the Ministère de l'Économie, Science et Innovation du Québec, Semences Prograin Inc., Syngenta Canada Inc., Sevita Genetics, La Coop Fédérée, Grain Farmers of Ontario, Saskatchewan Pulse Growers, Manitoba Pulse and Soybean Growers, the Canadian Field Crop Research Alliance and Les Producteurs de grains du Québec.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018. 00987/full#supplementary-material


of complex traits in diverse samples. Bioinformatics 23, 2633–2635. doi: 10.1093/bioinformatics/btm308


R. D. Riggs (Marceline, MO: Schmitt and Associates of Marceline), 9–40.


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

Copyright © 2018 Her Majesty the Queen in Right of Canada, as represented by Agriculture and Agri-Food Canada (AAFC). 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.

# Real-Time PCR for Diagnosing and Quantifying Co-infection by Two Globally Distributed Fungal Pathogens of Wheat

Araz S. Abdullah<sup>1</sup> \*, Chala Turo<sup>1</sup> , Caroline S. Moffat<sup>1</sup> , Francisco J. Lopez-Ruiz<sup>1</sup> , Mark R. Gibberd<sup>1</sup> , John Hamblin<sup>2</sup> and Ayalsew Zerihun<sup>1</sup>

<sup>1</sup> Centre for Crop and Disease Management, School of Molecular and Life Sciences, Curtin University, Bentley, WA, Australia, <sup>2</sup> Institute of Agriculture, University of Western Australia, Crawley, WA, Australia

Co-infections – invasions of a host-plant by multiple pathogen species or strains – are common, and are thought to have consequences for pathogen ecology and evolution. Despite their apparent significance, co-infections have received limited attention; in part due to lack of suitable quantitative tools for monitoring of co-infecting pathogens. Here, we report on a duplex real-time PCR assay that simultaneously distinguishes and quantifies co-infections by two globally important fungal pathogens of wheat: Pyrenophora tritici-repentis and Parastagonospora nodorum. These fungi share common characteristics and host species, creating a challenge for conventional disease diagnosis and subsequent management strategies. The assay uses uniquely assigned fluorogenic probes to quantify fungal biomass as nucleic acid equivalents. The probes provide highly specific target quantification with accurate discrimination against non-target closely related fungal species and host genes. Quantification of the fungal targets is linear over a wide range (5000–0.5 pg DNA µl −1 ) with high reproducibility (RSD ≤ 10%). In the presence of host DNA in the assay matrix, fungal biomass can be quantified up to a fungal to wheat DNA ratio of 1 to 200. The utility of the method was demonstrated using field samples of a cultivar sensitive to both pathogens. While visual and culture diagnosis suggested the presence of only one of the pathogen species, the assay revealed not only presence of both co-infecting pathogens (hence enabling asymptomatic detection) but also allowed quantification of relative abundances of the pathogens as a function of disease severity. Thus, the assay provides for accurate diagnosis; it is suitable for high-throughput screening of co-infections in epidemiological studies, and for exploring pathogen–pathogen interactions and dynamics, none of which would be possible with conventional approaches.

Keywords: co-infections, qPCR, hydrolysis probes, molecular disease diagnosis, pathogen interactions

### INTRODUCTION

Co-infections, whereby a host-plant is invaded by multiple pathogens or multiple strains of the same pathogen, are common in the field and can have major consequences for disease ecology and pathogen evolution (Alizon et al., 2013). Despite the recognition of the significance of co-infection (Savary and Zadoks, 1992; Zadoks, 1999; Tack et al., 2012), empirical studies are still few, due mainly to the complexity of distinguishing/quantifying multiple pathogens, which often requires

#### Edited by:

Christophe Le May, Agrocampus Ouest, France

#### Reviewed by:

Martin Chilvers, Michigan State University, United States Syama Chatterton, Agriculture and Agri-Food Canada (AAFC), Canada

> \*Correspondence: Araz S. Abdullah araz.abdullah@curtin.edu.au

#### Specialty section:

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

Received: 23 April 2018 Accepted: 05 July 2018 Published: 09 August 2018

#### Citation:

Abdullah AS, Turo C, Moffat CS, Lopez-Ruiz FJ, Gibberd MR, Hamblin J and Zerihun A (2018) Real-Time PCR for Diagnosing and Quantifying Co-infection by Two Globally Distributed Fungal Pathogens of Wheat. Front. Plant Sci. 9:1086. doi: 10.3389/fpls.2018.01086

**22**

suitable molecular tools (Tollenaere et al., 2012; Tollenaere et al., 2016). As a result, the theoretical understanding of co-infection has largely outpaced experimental studies in natural and agricultural systems (Hood, 2003; Abdullah et al., 2017) and our capacity to understand and manage the processes and consequences of coinfection is limited until current and future theoretical models can be validated by reliable data.

Methods used for distinguishing multiple pathogens from coinfected tissues involve detection of nucleic acid targets using metagenome sequencing and/or analysis of melt-curves from PCR primers with various annealing properties (Brandfass and Karlovsky, 2006; Tollenaere et al., 2012; Gelaye et al., 2017). Such methods enable multiple pathogen detections but provide limited quantitative information on the relative abundance of each pathogen and their effects on overall disease (Bernreiter, 2017). The quantification of pathogen abundance is critical, as many pathogens are naturally present within plants, but their infection levels, pathogenicity and hence relative impacts can differ vastly (Bakker et al., 2014). At present, real-time quantitative PCR (qPCR) is the most reliable technique for measuring disease load in a sample while providing species specificity (Holland et al., 1991; Bates et al., 2001; Schena et al., 2006). Successful qPCR-based multiplexed assays have been developed for detection of Phytophthora diseases of soybean (Rojas et al., 2017), but the accuracy of quantification is often compromised in samples with unbalanced target ratios (Klerks et al., 2004; Atallah et al., 2007). Thus, to ensure reliable quantification, co-infection studies often analyze the abundance of multiple pathogens in separate reactions. Due to the cost of labor and resources required, this approach is limited to small-scale investigations limiting our capacity to investigate co-infections over large-scale host populations.

In order to address the limitations of current approaches, we investigate the use of dual-labeled species-specific probes to simultaneously detect and quantify two globally distributed fungal pathogens of wheat: Pyrenophora tritici-repentis and Parastagonospora nodorum. These two foliar fungal pathogens cause tan (yellow) spot and septoria nodorum blotch in wheat, respectively, damaging photosynthetically active leaf area and causing substantial yield losses (Bhathal et al., 2003). Symptoms caused by these two diseases are difficult to distinguish and may be misdiagnosed as other unrelated physiological abnormalities (Moffat et al., 2015), creating a challenge for conventional disease diagnosis and subsequent management strategies. Symptoms can also vary markedly depending on host genotype, further hampering accurate disease diagnosis (Lamari and Bernier, 1989). Interestingly, P. tritici-repentis carries a pathogenicity/virulence gene, ToxA, thought to have been acquired laterally from Pa. nodorum (Friesen et al., 2006). This suggests that co-infection by these two fungi is likely to occur in nature and may have consequences for disease management strategies. Despite the high likelihood of co-infection by these two fungi and difficulties of identification, no tools have been developed for the diagnosis of their abundance in co-infected host materials.

Here, we report a duplex assay that distinguishes specific DNA sequences unique to each of these fungi and quantifies their presence by direct comparison to standards amplified in parallel reactions. The basis of this method is the uniquely assigned fluorogenic reporters for each species sequence. The fluorogenic reporters, when bound to the target, quantify fungal b as nucleic acid equivalents. We report on a series of experiments designed to demonstrate the specificity and sensitivity of the method using both a simulated genomic DNA matrix of both fungi at varying ratios as well as naturally infected wheat leaves collected from the field. We also examine the capacity of the technique to show that a species that is present at very low level can be detected and quantified, with very limited interference, in the presence of an abundant species. Finally, we correlate the amount of fungal DNA quantified using the duplex qPCR assay developed here with the level of infection measured by conventional disease scores.

### MATERIALS AND METHODS

### PCR Primers and Conditions

P. tritici-repentis primers were designed to target a speciesspecific multicopy genomic region described previously (Moffat et al., 2015; See et al., 2016). Briefly, a 4.65-kb region of the P. tritici-repentis isolate Pt-1CBFP carrying the ToxA gene (supercontig\_1.4) was aligned to an orthologous genomic region of the Pa. nodorum isolate SN15 (scaffold\_55). All known isolates of the targeted fungi were included in this step to ensure that the primers amplify all known isolates of the pathogens of interest. A pair of primers that target a 99-bp fragment located 701-bp upstream of the ToxA coding region was designed to detect P. tritici-repentis (**Table 1**). These primers amplify a short fragment (99-bp) within the promoter region of a low molecular weight host-selective toxin. This multicopy region has been detected in a number races of P. tritici-repentis isolates collected from around the world (Moolhuijzen et al., 2018). The size of the primer pair was restricted to 99-bp to ensure comparable amplicon sizes between P. tritici-repentis and Pa. nodorum.

Pa. nodorum DNA was amplified using primers modified from those previously described by Oliver et al. (2008). Additional 3-bp was included at the beginning of each primer sequence to allow the probe to overcome the issue of primer– dimer association (Barbisin et al., 2009). Pa. nodorum primers amplify a 112-bp fragment of a highly conserved anonymous


<sup>∗</sup>Ptr and Pn refer to P. tritici-repentis and Pa. nodorum, respectively. Sequences in bold are pre-published primers modified from Oliver et al. (2008). Other sequences are designed in this study. Primers and probes were designed to amplify short fragments of the targeted pathogen DNA.

gene (SNOG\_01116·1). This gene has no significant similarity to any other sequences in the publically available genome databases (Oliver et al., 2008). All primers were designed using OligoArchitectTM primer analyzer (Sigma, Life Science) and scanned against the National Center for Biotechnology Information GenBank database using basic local alignment search tools to ensure their specificity.

The designed primers were subjected to conventional PCR to confirm their specificity. Each PCR reaction contained 1xMyTaq buffer, 250 nM forward primer, 250 nM reverse primer, 1 unit MyTaq DNA polymerase (Bioline), and 5 ng DNA template. Reactions were performed as follows: 3 min initial denaturation at 95◦C, 35 cycles of 30 s denaturation at 95◦C, 30 s of annealing at 58◦C and 1 min extension at 72◦C. Electrophoresis of PCR products was performed on 2% agarose gels stained with SybrSafe (Life Technologies) and visualized under UV light. Reproducibility of the results was confirmed by running the PCR with negative controls in duplicates. The PCR step also included DNA samples from five common fungal pathogens of cereals as negative controls. Colonies of Pyrenophora teres f. sp. maculata, Pyrenophora teres f. sp. teres, Blumeria graminis f. sp. tritici, Alternaria alternata, and Fusarium graminearum were kindly provided by Steven Chang, Curtin University/Centre for Crop and Disease Management. Species identity was confirmed by sequencing the internal transcribed spacer of the ribosomal DNA (Schoch et al., 2012). All sequence analyses and multiple sequence alignments were carried out using Geneious version R6.1.6.

### Real-Time PCR Probes and Conditions

Two dual-labeled probes were custom-designed and assigned to hybridize with a complementary region between the forward and the reverse primers. The P. tritici-repentis and Pa. nodorum probes were 23-bp and 20-bp in length, respectively. The length of each probe was chosen to ensure probe-primer hybrids are formed in a complementary manner with the length of the primers. Each probe was labeled with a unique fluorogenic reporter to ensure that target sequences of both pathogens were amplified simultaneously but detected independently. The P. tritici-repentis probe was labeled with 6-carboxyfluorescein (FAMTM; Sigma-Aldrich). The Pa. nodorum probe was labeled with CAL Fluor Gold <sup>R</sup> (CFG; Sigma-Aldrich). FAM has emission maxima between 494 nm and 518 nm and CFG emission peaks between 538 nm and 559 nm. The fluorogenic reporters were selected based on the capacity of the CFX96 detection system, the instrument used in this study, to resolve overlapping spectra. This was determined prior to carrying out the experiments using a spectra overlay tool available at http://www.Qpcrdesign. com/spectral-overlay. Both probes were paired with the nonfluorescent black hole quencher-1 (BHQ-1 <sup>R</sup> ; Sigma-Aldrich).

Probes and their matching primers were run on a 96-well spectrofluorometric thermal cycler (Bio-Rad CFX96) with the following conditions: 15 min at 95◦C, 15 s denaturation at 95◦C, 20 s at 72◦C followed by 40 cycles of 15 s at 95◦C and 30 s at 58◦C. Each 20 µl reaction volume contained 5 µl of the sample and 12 µl iQTM Multiplex Powermix (Bio-Rad). For a fixed amount of target template, P. tritici-repentis DNA was amplified faster than Pa. nodorum DNA. Hence, probe and primers of P. tritici-repentis were restricted to obtain comparable quantification cycles (Cq) to that of Pa. nodorum. In a preliminary experiment, the concentrations of primers and probes were gradually increased from 50 up to 450 nM at 50 nM intervals. P. tritici-repentis DNA was amplified using 200 nM forward primer, 200 nM reverse primer and 100 nM probe. Pa. nodorum DNA was amplified using 250 nM forward primer, 250 nM reverse primer and 150 nM probe. Unless specified, reactions were carried out in duplex where primers and probes for both species were applied together. A preliminary experiment showed a comparable amplification efficiency between duplex and singleplex reactions and no evidence of cross-amplification among primers and probes of the two species was observed (**Supplementary Table 1**). Presence of any non-specific amplicon was examined using post-PCR melt curve analysis.

### DNA Extraction and Quantification

Pure genomic DNA from fungal colonies was extracted using the Bio-sprint 15 plant DNA kit (Qiagen) as per the manufacturer's protocol. Fungal cultures were maintained on agar plates as described elsewhere (Moffat et al., 2015). Mycelia were harvested from these cultures and ground into a fine powder in liquid nitrogen. Subsamples (40 mg ground tissue) were placed into 1.5 ml microtubes, which were then used for DNA extraction. The concentration of DNA in each subsample was determined using a NanoDrop 2000 UV-Vis spectrophotometer (Thermo Scientific) and diluted to 50 ng µl −1 in ultrapure PCR grade water. DNA was stored at −20◦C until used. Aerosol protected pipette tips were used throughout the extraction and quantification steps to prevent DNA contamination. DNA from infected wheat leaves was extracted using the same DNA extraction kit and quantified as described above.

### DNA Spiking and Field Validation

In two independent experiments, a fixed concentration of fungal DNA of one pathogen (5 ng µl −1 ) was spiked into a progressively decreasing DNA concentration of the other pathogen (5, 0.5, 0.05, and 0.005 ng DNA µl −1 ). This generated a ratio of DNA concentration from one pathogen to the other in the sample ranging from 1:1 down to 1:10000. The spiking aimed to simulate the analysis of samples derived from infection conditions whereby the two species occur at different relative abundance, which is typical of many foliar fungal infections. Recovery of fungal DNA for the respective pathogens was expressed as the ratio of the total concentration of fungal DNA quantified using the qPCR method to the amount of DNA added × 100.

In a third experiment, recovery of fungal DNA was measured in fungal DNA samples spiked with abundant wheat DNA. The concentration of fungal DNA in the samples was progressively decreased from 5 down to 0.05 ng µl <sup>−</sup><sup>1</sup> while background wheat DNA was increased from 5 to 100 ng µl −1 . The experiment simulated analytical conditions where fungal DNA is present in small concentrations against a background of ample wheat DNA, such as would occur where infection severity and/or incidence are low.

To evaluate how well the assay works for field samples, diseased leaves from the wheat variety Scout were collected from

a site in the southwest of Western Australia (31◦ .74S, 116◦ .70E). Scout is rated as susceptible to very susceptible to both P. triticirepentis and Pa. nodorum (DAFWA, 2016). Sampled leaves were visually inspected for diseased leaf area and given scores on 0- 100 scale. Leaves were then split into two groups; one group (n = 9) was surface-sterilized in 2% chlorine and incubated on agar Petri-dishes in an attempt to characterize the causal agent of the disease. Leaves from the second group (n = 9) were ground in liquid nitrogen and used for DNA extraction. 50 ng µl −1 gDNA from the infected leaves, along with 50 ng µl −1 gDNA from uninfected leaves from glasshouse-grown plants (negative controls), were first analyzed by conventional PCR. A further 50 ng µl −1 gDNA from the same infected leaves and controls were then analyzed using qPCR.

### Detection and Quantification of Fungal Biomass

Fluorescence data from the qPCR machine were retrieved during the annealing step of every Cq. Threshold fluorescence was set automatically by the instrument manager system (CFX Manager Version 2.0.8) before carrying out the assay. A loglinear standard curve of a 10-fold dilution series corresponding to 5000 to 0.5 pg µl <sup>−</sup><sup>1</sup> was generated by plotting the logarithms of known concentrations of fungal DNA against the Cq values. The resulting regression equations were used to calculate fungal DNA in unknown samples. No-template controls, where sterile water was added instead of DNA, were included in each reaction. Limit of detection of the duplex assay, the lowest concentration at which reliable detection can be achieved, was determined following Armbruster and Pry (2008). All reactions were run with three replicates and samples that gave positive fluorescence before no-template controls were considered positive.

## RESULTS

### Specificity of the Assay

To test how well P. tritici-repentis and Pa. nodorum could be distinguished from each other, as well as from other common wheat pathogens and the host, we carried out two conventional PCR assays each using 5 ng gDNA. Five non-target controls of DNA from closely related cereal fungal pathogens and DNA from the wheat cultivar Scout were included in this step. Reactions were run in separate wells (i.e., singleplex). Conventional PCR provided amplicons of the expected sizes, and P. tritici-repentis and Pa. nodorum were distinguished based on the size of the amplicons (**Supplementary Figure 1**). P. tritici-repentis and Pa. nodorum primers only amplified the respective pathogen DNA. None of the five non-target controls or DNA from wheat gave specific amplicons that could be detected by conventional PCR (**Figures 1A,B**).

We then tested the specificity of the probes and their matching primers in two-singleplex real-time qPCR assays each containing 5 ng gDNA. Detection only occurred for probes that were complementary to the expected sequences, and none of the five non-target controls or the wheat DNA

Pn-Reverse, were tested against P. tritici-repentis DNA in two separate PCR reactions. Reactions included negative controls of DNA samples from five cereal fungal pathogens. Reactions including positive and negative controls were electrophoresed on 2% agarose with two technical replicates.

had specific amplification during 40 real-time qPCR cycles (**Figures 2A,B**). Samples that included no-template DNA or those that contained wheat DNA were negative during the course of the reaction (**Figures 2A,B**). P. tritici-repentis and Pa. nodorum were distinguished based on the emission spectra of the fluorogenic reporters that were used to label each pathogen probe.

### Dynamic Range, Efficiency, and Reproducibility of the Assay

We carried out two duplexed standard curve experiments on DNA of P. tritici-repentis and Pa. nodorum mixed at equal ratios. A 10-fold dilution series in the range of 5000 to 0.5 pg µl <sup>−</sup><sup>1</sup> was generated, analyzed and plotted against the number of Cq required to detect fluorescence signals. Each dilution was prepared with three replicates except the no-template control, which was analyzed using 10 replicates. Fitting log-linear standard curves between Cq and fungal DNA resulted in correlation coefficients (r) −0.997 and −0.996 for P. tritici-repentis and Pa. nodorum, respectively. The standard curves for both species were linear over 10,000-fold dilution, and the calculated amplification efficiency was 91.28% for P. tritici-repentis and 99.25% for Pa. nodorum (**Figures 3A,B**). The slopes and intercepts of the log-linear curves were comparable between P. triticirepentis and Pa. nodorum (p > 0.05). Mean Cq value (±standard deviation bars, where visible) of three replicated reactions was highly reproducible with relative standard deviation ≤ 10%

(**Figures 3A,B**). The variation around the mean was independent of template concentration, and the Cq value for 10 no-template controls averaged 36.618 with a relative standard deviation of 0.35%.

### Sensitivity and Limit of Detection of the Assay

We carried out two spiking experiments using a simulated matrix of P. tritici-repentis and Pa. nodorum DNA mixed at various ratios. Results of the spiking experiments demonstrated that reducing DNA ratio of one species while keeping DNA of the other species constant did not affect the detection limit of the assay. We were able to accurately quantify fungal DNA of either pathogen down to a relative ratio of 1:100 (**Figures 4A,B**). However, further reduction of the DNA ratio to 1:1000 underestimated the amount of P. tritici-repentis by 11.75% and Pa. nodorum by 18.55%. This underestimation became more pronounced with further dilution, and at 100 ppm (i.e., DNA ratios of 1:10000), neither pathogen was quantifiable in the presence of the other (**Figures 4A,B**). The minimal DNA concentration for detection was 0.059 pg µl −1 for P. tritici-repentis and 0.036 pg µl −1 for Pa. nodorum.

We also tested the ability of the assay to detect and quantify fungal DNA in a sample that included abundant wheat DNA. We were able to quantify both species with high accuracy up to a P. tritici-repentis: Pa. nodorum: wheat DNA ratio of 1:1:200 (**Figure 5**). Recovery of fungal DNA, total amount of DNA quantified using the qPCR method, was high at 95.69% for P. tritici-repentis and 94.20% for Pa. nodorum when the fungal to wheat DNA ratios dropped to 1:1:2000. At these ratios, however, relative standard deviations were 10.23% and 13.90% for P. triticirepentis and Pa. nodorum, respectively, affecting the reliability of fungal DNA quantification.

## Field Evaluation of Naturally Infected Plants

. Data represent means ± standard deviation (n = 6).

We collected eighteen naturally infected wheat leaves from a site in the southwest of Western Australia. These leaves showed typical symptoms of tan spot with distinct yellow halos and tan chlorotic lesions (**Figure 6A**). No apparent symptoms of Pa. nodorum were visible on these leaves. Leaves were split into two groups; one group (n = 9) was incubated on agar Petri-dishes in an attempt to characterize the causal agent of the disease. Incubation on agar only yielded colonies typical of P. triticirepentis (data not shown).

) of one pathogen DNA, with the other held constant at 5

DNA was extracted from the second leaf group (n = 9) and analyzed using conventional PCR. Only one out of nine duplicated PCR reactions tested positive to Pa. nodorum. All reactions were positive to P. tritici-repentis. DNA from the same leaves that were used in the conventional PCR was then analyzed using qPCR. The duplex qPCR assay was able to detect signals corresponding to sequences of P. tritici-repentis and Pa. nodorum despite the absence of any visible symptoms of

0.0005 ng µl

ng µl −1 −1

Pa. nodorum on these leaves (**Figure 6B**). There was a good agreement (r = −0.82) between total fungal DNA quantified by qPCR and conventional disease scores that were collected in the field (**Figure 6B**; black circles/line). Conventional disease scores were significantly correlated with the increase in P. tritici-repentis DNA, and P. tritici-repentis DNA was greater contributor to the disease scores than Pa. nodorum (**Figure 6B**; blue and orange circles/lines).

## DISCUSSION

Quantification of the relative abundance of co-infecting pathogens requires selection of primers and probes that are compatible with each other whilst distinguishing between the coinfecting species. Recent work has identified a 235-bp multicopy region present in the P. tritici-repentis genome, and primers designed in this region were able to detect P. tritici-repentis from wheat leaves, even prior to the visible appearance of tan spot lesions (See et al., 2016). However, these primers cross-hybridized with Pyrenophora teres f. sp. teres, which the authors suggested may be overcome by the addition of fluorescence-labeled probes in the middle of the amplicon.

In this study, primers were chosen to produce relatively smaller amplicons (99–112 bp) that are more efficient for amplifying specific products than longer primers (Bustin et al., 2009). Our primers successfully distinguished targeted pathogens from non-target closely related fungal species and host genes. However, for increased specificity, fluorogenic probes were designed to hybridize to sequences within the primers enabling the assay to simultaneously distinguish and quantify DNA associated with P. tritici-repentis from that associated with Pa. nodorum. The length of the probes was designed to ensure that stable primer-probe hybrids were formed at the same annealing temperature for both species, irrespective of the length of the primers. Furthermore, probes were labeled with a unique fluorogenic-reporter allowing DNA sequences from P. triticirepentis and Pa. nodorum to be amplified simultaneously but independent of each other. The choice of the fluorogenic reporter was decided based on instrument capability in resolving overlapping spectra. FAM-labeled probes are excited at lower emission spectra than CFG-labeled probes and hence are expected to produce stronger signals. Nevertheless, restricting the concentration of the FAM-labeled probe in the reaction optimized the fraction of the amplicon that is bound to the probe and produced comparable signals for P. tritici-repentis and Pa. nodorum. In addition to the fluorogenic reporters, probes were labeled with dark-hole quenchers to inhibit probe signals when probes are free in the solution (Holland et al., 1991). The dark-hole quencher was sufficient to inhibit signals from probes that are not in perfect contact with the target, resulting in accurate species-specific quantifications.

A series of experiments were conducted demonstrating the specificity, efficiency, and reproducibility of this assay for simultaneous detection and quantification of P. tritici-repentis and Pa. nodorum. The method reported here resolved two signals each unique to the assayed fungi. No cross-amplification with either host plant DNA or DNA from other closely related fungal

respectively.

fpls-09-01086 August 9, 2018 Time: 11:57 # 8

pathogens of cereals was observed. The assay amplified the targeted sequences with high specificity, and we were able to detect the presence of P. tritici-repentis at 0.059 pg µl −1 and Pa. nodorum at 0.036 pg µl −1 in mixed fungal DNA matrices. Furthermore, in mixed gDNA matrices of both targeted species, the assay was able to detect fungal signals at a relative abundance as low as 1 in 10,000 with ≤10% RSD. Similarly, the assay was able to quantify the presence of Pa. nodorum DNA down to 23-pg in leaf samples naturally co-infected by abundant P. tritici-repentis up to 2192-pg. Nonetheless, there was a slight interference to the quantification of the less abundant target when DNA of the other target and/or wheat DNA was highly abundant. This may occur because reaction components including DNA polymerase, dNTPs, and MgCl<sup>2</sup> become limiting in later qPCR cycles. The more abundant target may compete for reaction components with the less abundant target, delaying its amplification (Bustin et al., 2009). As a result, the quantification of the less abundant target, although remaining specific

shown on the plot. ns and ∗∗ refer to not significant and significant (P < 0.01),

and within the detection limit, may be compromised. One way to minimize this interference is to optimize concentrations of reaction components sequentially (Bustin et al., 2009). Regardless, the assay is highly stable and reproducible. The reproducibility was confirmed by the small standard deviations of triplicates samples in the standard curves (≤10%). The amplification curves of replicate runs showed strong overlap indicating stability and reproducibility.

A limitation of most nucleic acid-based quantification assays is an inability to distinguish between DNA from living and dead cells (Pinheiro et al., 2016). Although fungal cell viability was not assessed in this work, the high correlation coefficient, 0.826, especially between P. tritici-repentis DNA and the conventional disease score, suggests that viability may not have a significant impact on measuring disease as DNA equivalent. Quantification of fungal DNA has been shown to provide accurate measurement of disease severity for P. tritici-repentis (See et al., 2016) and Pa. nodorum (Oliver et al., 2008). Another limitation more specific to qPCR is that only a few targets can readily be accommodated in a single homogeneous assay. The number of targets that can be fitted in one assay depends on the design of the assay and the ability of the instrument to resolve overlapping spectra. The instrument used in this study has a blue monochromatic laser source for the excitation of fluorophores, and can simultaneously detect five targets on five different channels. This limits the fluorophores that could be efficiently excited by the blue laser to those that are distant from the far-red wavelengths. Instruments with high resolutiondetectors and variable-wavelength light sources may allow more targets to be included in an assay. Instruments with such specifications are available and can be employed for detection and quantification of large-scale screening of co-infection by multiple pathogens.

Detection of multiple pathogens has also been achieved by the use of post-PCR dissociation curve analysis. Such methods use a set of primers with different melting temperatures and G + C contents. Using this technique up to 10 microbial pathogens of humans have been distinguished in a single assay (Gelaye et al., 2017; Jiang et al., 2017). Post-PCR methods, although specific and accommodating several targets in a single assay, do not provide quantitative information on the level of disease. Plant tissues can be inhabited by a large number of pathogenic and non-pathogenic microbes. Nevertheless, infection levels of these microbes and the damage they cause to the plant can vary greatly (Bakker et al., 2014). Our assay offers several advantages over the dissociation curve analysis and many other methods for diagnosis of co-infection. First, no post-reaction processing is required. Reactions are driven to completion allowing higher levels of sensitivity even when target sequences are present at low concentrations. Second, the assay provides quantitative data on the level of disease load in a sample, which other methods including metagenome sequencing do not offer. Finally, the costeffectiveness, although not directly estimated, of testing multiple agents in a single test allows more testing using the same amount of reagents and staff time. When coupled with a 96-well capacity, this assay offers a sensitive and quantitative high throughput methodology for detection of co-infection in plants. The method

can also be used to study epidemiological consequences of coinfection in the field.

### AUTHOR CONTRIBUTIONS

AA contributed to the conception of the idea, defined/selected the primers, conducted the experiments, analyzed the data, and drafted the manuscript. CT assisted in setting up the conventional PCR and testing of the primers. CM, FL-R, AZ, MG, and JH supervised the work and provided critical suggestions on the article. All authors read and approved the final version of the article.

### FUNDING

AA is a recipient of Research Training Program (RTP), an Australian government funded scholarship for Ph.D. students. The Centre for Crop and Disease Management (CCDM) defrayed the publishing costs for this review. CCDM is a national research

### REFERENCES


centre co-funded by Curtin University and the Grains Research and Development Corporation (project grant CUR00023). The funders had no role in ideas, design, decision to publish, or preparation of the manuscript.

### ACKNOWLEDGMENTS

We thank Richard Oliver, Centre for Crop and Disease Management, for critically reading the manuscript and contributing to the genesis of the research. Belinda Cox and Weiwei Deng provided methodological and technical assistance, and Steven Chang shared some of the fungal isolates used in this study.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018.01086/ full#supplementary-material


species-specific detection of Phytophthora sojae and P. sansomeana root rot pathogens of soybean. Plant Dis. 101, 1171–1181. doi: 10.1094/PDIS-09-16- 1225-RE


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

Copyright © 2018 Abdullah, Turo, Moffat, Lopez-Ruiz, Gibberd, Hamblin and Zerihun. 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.

# Defense Compounds Rather Than Nutrient Availability Shape Aggressiveness Trait Variation Along a Leaf Maturity Gradient in a Biotrophic Plant Pathogen

Agathe Maupetit<sup>1</sup> , Romain Larbat<sup>2</sup> , Michaël Pernaci<sup>1</sup>† , Axelle Andrieux<sup>1</sup> , Cécile Guinet<sup>3</sup> , Anne-Laure Boutigny<sup>3</sup>† , Bénédicte Fabre<sup>1</sup> , Pascal Frey<sup>1</sup> and Fabien Halkett<sup>1</sup> \*

#### Edited by:

Christophe Le May, Agrocampus Ouest, France

#### Reviewed by:

Sabine Banniza, University of Saskatchewan, Canada Fournet Sylvain, INRA (Paris), France

> \*Correspondence: Fabien Halkett fabien.halkett@inra.fr

#### †Present address:

Michaël Pernaci, PalmElit SAS, Montferriez-sur-Lez, France Anne-Laure Boutigny, ANSES – Laboratoire de la Santé des Végétaux, Unité de Bactériologie, Virologie et OGM, Angers, France

#### Specialty section:

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

Received: 20 April 2018 Accepted: 03 September 2018 Published: 28 September 2018

#### Citation:

Maupetit A, Larbat R, Pernaci M, Andrieux A, Guinet C, Boutigny A-L, Fabre B, Frey P and Halkett F (2018) Defense Compounds Rather Than Nutrient Availability Shape Aggressiveness Trait Variation Along a Leaf Maturity Gradient in a Biotrophic Plant Pathogen. Front. Plant Sci. 9:1396. doi: 10.3389/fpls.2018.01396 <sup>1</sup> Université de Lorraine, INRA, IAM, Nancy, France, <sup>2</sup> Université de Lorraine, INRA, LAE, Nancy, France, <sup>3</sup> ANSES, Laboratoire de la Santé des Végétaux, Unité de Mycologie, Malzéville, France

Foliar pathogens face heterogeneous environments depending on the maturity of leaves they interact with. In particular, nutrient availability as well as defense levels may vary significantly, with opposing effects on the success of infection. The present study tested which of these factors have a dominant effect on the pathogen's development. Poplar leaf disks of eight maturity levels were inoculated with the poplar rust fungus Melampsora larici-populina using an innovative single-spore inoculation procedure. A set of quantitative fungal traits (infection efficiency, latent period, uredinia size, mycelium quantity, sporulation rate, sporulation capacity, and spore volume) was measured on each infected leaf disk. Uninfected parts of the leaves were analyzed for their nutrient (sugars, total C and N) and defense compounds (phenolics) content. We found that M. larici-populina is more aggressive on more mature leaves as indicated by wider uredinia and a higher sporulation rate. Other traits varied independently from each other without a consistent pattern. None of the pathogen traits correlated with leaf sugar, total C, or total N content. In contrast, phenolic contents (flavonols, hydroxycinnamic acid esters, and salicinoids) were negatively correlated with uredinia size and sporulation rate. The pathogen's fitness appeared to be more constrained by the constitutive plant defense level than limited by nutrient availability, as evident in the decrease in sporulation.

Keywords: constitutive defense, aggressiveness traits, plasticity, nutrient availability, rust fungus

### INTRODUCTION

Plant diseases and pests are responsible for crop damage that may account for up to 40% of yield losses worldwide (Boonekamp, 2012). Pathogen aggressiveness, the quantitative component of pathogenicity (Pariaud et al., 2009a), has been widely studied in order to forecast the evolutionary potential of plant pathogens and to better design disease control strategies (Lannou, 2012). Aggressiveness can be evaluated at different scales: by measuring the epidemic rate at the field scale or through elementary phenotypic traits at the individual scale (Pariaud et al., 2009a). The environment plays a key role in developmental processes of the pathogen and influences the individual's phenotype and the scale of fitness (Scheiner, 1993; Jarosz and Davelos, 1995).

**32**

Biotrophic fungal pathogens, such as rust and powdery mildew fungi, spend most of their life cycle within living plant tissues, so their fitness is directly affected by leaf physiology (Agrios, 2005; Gamm et al., 2011; Duplessis et al., 2013). On the one hand the leaf tissue provides the pathogen with the necessary nutrient resources for its development. On the other hand, it impedes pathogen development through a battery of plant defense compounds (Ullah et al., 2017).

Studies on interactions between plants and their foliar pathogens have shown that plant age, and especially leaf age, leads to different responses of the pathogen. Coleman (1986) postulated from a series of experiments that leaves pass through a peak of susceptibility to disease and pest attack. He proposed that for many stresses, including infection by biotrophic fungi, this peak coincides with changes in leaf properties during the transition from metabolite sink to source. This stage corresponds to a maximal nutrient availability. After this sinksource transition, free sugar content and thus nutrient resources decrease with increasing leaf maturity (physiological stage at a relative leaf age). Many studies support this view, demonstrating a maximum disease level before leaves are fully expanded and a decrease of disease severity on older leaves, which is also referred to as ontogenetic resistance (Ficke et al., 2002; Xavier et al., 2015; Farber and Mundt, 2017). In agreement, Merry et al. (2013) demonstrated that the peak in the severity of powdery mildew along a primary shoot of Vitis vinifera corresponds to the position of the first leaf that does not import carbohydrates. However, some studies have reported a contrasting pattern, e.g., Knott and Mundt (1991) found that infection efficiency of the wheat leaf rust fungus Puccinia recondita f. sp. tritici is 25% higher on older than on younger leaves in five wheat cultivars. The opposite trend may involve the dark side of plant-pathogen interaction: defense compounds. Their actions on a pathogen's development have been studied in several pathosystems, including rust fungi infecting poplar (Johnson and Kim, 2005), black currant (Vagiri et al., 2017), and willow (Hakulinen et al., 1999). Phenolic compounds present in leaves appear to highly influence the general plant resistance to fungi (Adandonon et al., 2017; Ullah et al., 2017). Most studies have focused on plant susceptibility to the disease (i.e., infection efficiency of the pathogen) or focused on overall disease severity. To dissect which fungal function is most influenced by leaf maturity and how, we propose to measure a set of elementary traits underlying components of aggressiveness.

Foliar pathogens of woody plants are good biological models to study the influence of leaf age on disease development, because they can infect leaves at various levels of maturity on the same host plant (Cellerino et al., 1978; Sharma et al., 1980; Turechek and Stevenson, 1998). Poplar is particularly well suited for studying the effect of leaf age on pathogens, since poplar plants grown from cuttings in a glasshouse produce new leaves at a constant rate, giving access to a wide range of leaves of increasing maturity on the same plant (Larson and Isebrands, 1971). These authors introduced the concept of leaf plastochron index (LPI) to refer to a position of a leaf along a poplar shoot. The youngest fully expanded leaf corresponds to LPI 1. Older leaves are numbered with increasing LPIs, according to their position along the shoot. In most experimental conditions the sink to source transition in poplar is established around LPI 6 (Coleman, 1986), which thus defines the first mature leaf. The effect of leaf maturity on the development of poplar pathogens (Marssonina brunnea, Melampsora medusae, and M. larici-populina) has been widely studied (Cellerino et al., 1978; Sharma et al., 1980; Coleman, 1986; Giorcelli et al., 1996). In most of these experiments, leaves were spray-inoculated with spore suspensions. In these conditions, hundreds of spores compete for infection, and as a result, inoculum density has been shown to influence aggressiveness traits (Giorcelli et al., 1996; Pei et al., 2003; Pariaud et al., 2009a). A way to eliminate any possible interaction among spores and to remove the effect of lesion density on trait variation is to use a single-spore inoculation protocol. Moreover, the single-spore inoculation enables the examination of phenotypic variation at the individual lesion level and not by means of average trait values measured on a cohort of lesions.

The aim of this study was to assess the effect of leaf age, and the associated content in nutrients and defense compounds, on the variation in aggressiveness traits in the poplar rust fungus M. larici-populina. We quantified trait variation along the leaf maturity gradient using the same fungal genotype to establish a so-called reaction norm (Gavrilets and Scheiner, 1993). We measured the effect of LPI using a single-spore inoculation method for a suite of aggressiveness and morphological traits: infection efficiency, latent period, uredinia size, sporulation rate, sporulation capacity, in planta mycelium quantity, and urediniospore volume. In a first step, we examined the differences in pathogen development along a leaf maturity gradient in order to study the variation in fungal traits and to assess the correlations between them. In a second step, we estimated the influence of the most important leaf compounds (sugars, total carbon and nitrogen, and phenolics) on the variation in aggressiveness traits.

### MATERIALS AND METHODS

### Plant and Fungal Material, and Experimental Design

Characterization of quantitative traits was performed on excised leaf disks of Populus deltoides × P. nigra 'Robusta'. Poplar plants were grown from dormant cuttings in 10-L pots containing a sand-peat (1:1, v/v) mixture, with an initial fertilization of 3.5 g L−<sup>1</sup> CaCO3 and 6 g L−<sup>1</sup> of slow release 13-13-13 N-P-K fertilizer (Nutricote <sup>R</sup> T100, Fertil). The plants were grown in a growth chamber regulated at 20/22◦C (night/day temperatures) and 16 h photoperiod in order to ensure optimal growth, and were watered daily with deionized water. After four months, plants were about 1.2 m high and exhibited 25 to 30 fully expanded leaves.

On the same day, we harvested eight leaves of different maturity (LPI 6, 8, 10, 12, 14, 16, 18, and 20) from each of three plants (total of 24 leaves). From each poplar leaf, 24 disks (12-mm-diameter) were excised, and placed in flotation, each in a different cell culture plate. Each plate consists in 24 wells and thus contained one sample (leaf disk) of all leaves.

A plate thus corresponded to a complete statistical block: Twentyfour replicate plates were prepared, each containing the 24 combinations of LPI and plants. Within plates, the positions of the samples were randomized (**Supplementary Figure S1**). After disk excision, the remaining part of each leaf (about half a leaf) was stored individually in an aluminum wrapper at −80◦C for sugar, nitrogen, carbon, and phenolic compound quantification. We hereafter refer to this plant material as "uninfected leaves."

Because the single-spore inoculation protocol is highly time-consuming, we were not able to inoculate more than six plates per day. Inoculations of the 24 plates were thus distributed over four consecutive days (each inoculation day formed a series of six plates). In other words, all leaf disks were excised the same day (day 1), but inoculated after 0 to 3 days depending on the series they belonged to.

Our experimental design (randomized complete block design) thus consisted in three experimental factors: LPI, plant and day of inoculation. Given an infection efficiency of approximately 25%, we expected six replicates per combination.

### Single-Spore Inoculation Protocol

Single-spore inoculation was performed with pure M. laricipopulina isolate 98AG31. This isolate was collected in 1998 in Moÿ-de-l'Aisne (France) on Populus trichocarpa x Populus deltoides 'Beaupré' leaves and serves since as a reference isolate (Duplessis et al., 2011). Urediniospores were cryopreserved at −80◦C at Institut National de la Recherche Agronomique Nancy (France) and multiplied on fresh poplar leaf disks before inoculation to maximize their germination and infection rates (Pei et al., 2002). To this aim, 1 mg of urediniospores was dispersed in 200 µL of water-agar (0.1 g L−<sup>1</sup> ). The resulting urediniospore suspension was applied as 2 µL droplets on the abaxial surface of two 30-mm-diameter leaf disks of 'Robusta'. Leaf disks were incubated by floating on deionized water in six-well polystyrene cell culture plates, with the abaxial surface upside, at 19 ± 1 ◦C, and under continuous fluorescent light (25 µmol m−<sup>2</sup> s −1 ). After a 10 to 13 days incubation period, the sporulating leaf disks were gently tapped over a microscope slide to release and collect urediniospores.

A germination test was performed just before inoculation to ensure the quality of urediniospores used. To this aim, a few hundred urediniospores were dispersed on the surface of a Petri dish containing agar (20 g L−<sup>1</sup> ). After an overnight incubation at 19 ± 1 ◦C, the ratio of germinated/total urediniospores was evaluated under a light microscope (100× magnification) to determine the germination rate. This test ensured that nearly all spores used for the experiment were physiologically able to germinate (mean germination rate of 93%).

For single-spore inoculation leaf disks were inoculated one by one. One single urediniospore was picked with a human eyelash under a stereomicroscope (63× magnification) and deposited into a 5 µL water-agar droplet placed at the center of the poplar leaf disk (abaxial surface). After inoculation the plates were incubated for 13 days in the same conditions as described above.

### Quantitative Trait Measurements

Several aggressiveness traits classically studied in plantpathogen interactions were measured: infection efficiency, latent period, uredinia size, sporulation rate, and sporulation capacity (Lannou, 2012). In addition, we measured in planta mycelium quantity, which is less commonly studied.


Beyond these six quantitative aggressiveness traits, we also measured a morphological trait:

• Volume of urediniospores was computed from length and width of urediniospores obtained from image analysis (see below).

We monitored the infections for 13 days, allowing measurements of latent period and infection efficiency. Emergence of uredinia was scored twice a day, at 9 a.m. and at 4 p.m. Infection efficiency was assessed for each leaf disk and was recorded as a binary variable: 1 if the deposited urediniospore produced one uredinia, 0 if not.

At day 13 (end of the monitoring period), leaf disks were harvested. We harvested all infected leaf disks (total: 158 samples out of 576 leaf disks inoculated, mean infection efficiency of 27.4%), giving from 4 to 12 biological replicates per LPI and per plant (mean: 6.58 replicates). The first step of the harvest consisted in separating the urediniospores produced

from the leaf disk. Infected leaf disks were placed in 2 mL Eppendorf tubes with 2 mL of Isoton <sup>R</sup> II isotonic buffered diluent (Beckman Coulter). The tubes were vigorously shaken for 20 s at 4 m s−<sup>1</sup> on a MP FastPrep <sup>R</sup> -24 homogenizer, in order to release all urediniospores from the uredinia. The second step consisted in measuring the size of the uredinia (or lesion size): photographs of the spore-free uredinia were taken under a stereomicroscope (25× magnification) coupled to a digital camera. Pictures were analyzed using ImageJ version 1.5i with a dedicated script (**Supplementary Note S1**). To measure the number of yellow pixels corresponding to the sporulating area, the Hue Saturation and Brightness was used: pixels with a hue under 37◦ were counted. Last, infected leaf disks were stored at −80◦C and lyophilized before DNA extraction.

Measurements of both sporulation rate and urediniospore morphological parameters were performed on the urediniospore suspensions using an Occhio <sup>R</sup> Flowcell FC200+ optical morphogranulometer, which allows simultaneous particle counting and image analysis of the particles. Before counting, urediniospore suspensions were vortexed to disperse clusters of urediniospores. For each sample, counting was performed using the following settings: 0.15 mL of priming, 0.85 mL of volume analysis, and 7% of volume sampling (i.e., 7% within 0.85 mL were really analyzed, evenly distributed along the analyzed volume). Two filters parameter values were applied:


These filters were applied on the raw data to compute sporulation rate expressed in spores per day, length, and width of urediniospores produced by each uredinia. The volume of urediniospores was computed following the formula of the volume of an ellipsoid (Philibert et al., 2011):

$$\text{Spore volume} = \frac{4}{3}\pi \times \frac{length}{2} \times \left(\frac{width}{2}\right)^2$$

### DNA Extraction and qPCR Quantitation of in planta Mycelium

DNA was extracted from infected poplar leaf disks using the DNeasy96 <sup>R</sup> DNA plant kit (Qiagen). The Frozen Plant Tissue protocol (DNeasy 96 Plant Handbook, June 2015) was followed except that 20 1-mm-diameter glass beads were added to one 3-mm-diameter tungsten carbide bead in the 2 mL Eppendorf tubes. The 158 samples were randomly processed in two DNA extraction plates, which were stored at 4◦C until qPCR analysis.

In order to correct for amplification differences among qPCR, three technical replicates were performed.

In order to quantify M. larici-populina mycelium in planta, quantitative Taqman <sup>R</sup> PCR was performed using new primers and probe designed for the rDNA ITS (Internal Transcribed Spacer). These primers and probe allowed a significant gain in sensitivity compared to a previously used primer pair, which was designed for a single-copy gene (Guinet et al., 2016). Reagent mix contained 3.1 µL of pure water, 6.5 µL of Mastermix <sup>R</sup> , 0.13 µL of primer ITS-Mlp-F (Primer sequence 5<sup>0</sup> -30 : TGACTCTTTGTATAAACCATTACCC), 0.13 µL of primer ITS-Mlp-R (TCAAAGTTGCCTTTGAGATACG), and 0.13 µL of probe ITS-Mlp-P (6-FAM-TGCATTGTGGCCCGTCAAAA-BHQ1) per sample. A volume of 3 µL of extracted DNA was added to 10 µL of reagent mix. Standard samples consisted of calibrated quantities of plasmid DNA containing the cloned ITS amplicon. Two series of eight 10-fold dilutions of plasmid DNA (from 1.47 × 10<sup>7</sup> to 1.47 fg µL −1 ) were included in each qPCR run to set a standard curve. Quantitative PCR runs were performed on a QuantStudio 6 Flex (Life Technologies) machine and consisted in 10 min at 95◦C followed by 45 cycles of 15 s at 95◦C and 45 s at 62◦C. DNA quantity was analyzed using the QuantStudioTM Real-time PCR software (version 1.1).

### Leaf Content Measurements: Total Nitrogen and Total Carbon Contents, Sugars, Phenolic Compounds

All leaf content measurements were conducted on the uninfected leaves, stored at −80◦C just after disk excision.

Total nitrogen and total carbon concentrations in uninfected leaves were analyzed using an elemental analyser (Thermo Quest, Type NCS 2500). Measurements, data analysis, data record, and weighing record (Mettler Tolédo MT5 microbalance) were performed using the EAGER 200 software.

Soluble sugars were extracted from 20 mg of dry leaf powder with 1 mL of 70% (v/v) methanol/water for 10 min and then centrifuged at 17,000 g for 5 min at 4◦C. This step was repeated twice and the resulting supernatants were pooled together. The supernatants, containing soluble sugars, were dried overnight with a vacuum evaporator (Maxi-Dry Plus; Hetomodel DW1, 0-110, Heto-HOLTEN A/S) to eliminate methanol. Dried extracts were rehydrated in 1.5 mL of distilled water, dissolved by sonication and then filtered at 0.2 µm (Acrodisc <sup>R</sup> Supor <sup>R</sup> filter; 0.2 µm, Van Waters Rogers). Undiluted aliquots of 20 µL were injected into a high-pressure liquid chromatography system (DionexTM ICS-5000+ HPLCTM) to determine soluble sugar composition. Soluble sugar separation was achieved on a DionexTM CarboPacTM SA10 IC separation column (6 µm 250 × 4 mm; Thermo ScientificTM). The flow rate was maintained at 1 mL min−<sup>1</sup> , and the column temperature at 40◦C. Individual carbohydrates were eluted from 0 to 30 min after injection. The sugar peaks were detected by light scattering (Sedex 45 ELSD system, Seder) then identified according to retention time and commercial standards. Sucrose, glucose, and fructose concentrations (% of dry weight) were determined by peak height and standard calibrations.

Phenolics were extracted from 30 mg dry leaf powder using the method described by Royer et al. (2013). Total phenolic content

was determined according to the Folin–Ciocalteu method. Briefly, each extract was diluted 1:16 in distilled water, and 20 µL of each diluted extract was distributed in a 96-well plate. Then 100 µL of Folin–Ciocalteu's reagent (Sigma), diluted 1:10 in distilled water, and 80 µL of 7.5% sodium carbonate (w/v) were added. The microplate was incubated 30 min at 25◦C and shaken 5 s every 10 min. Absorbance of the sample was measured with a microplate-reader (SynergyHT, BioTek) at 760 nm. Quantification of total phenolics was performed according to a calibration curve of gallic acid, and the results were expressed as milligram of gallic acid equivalent per gram of leaf dry weight.

In addition to total phenolic content, the concentration of major phenolics was analyzed on ultra-high performance liquid chromatography (U-HPLC) system (Shimadzu) equipped with a photo diode array detector (DAD) and a mass spectrometer. Each sample (3 µl) was separated on a C18 kinetex (100 mm × 2.1 mm) column (Phenomenex). The mobile phase consisted in 0.1% formic acid in ultra-pure water (solvent A) and 0.1% formic acid in methanol (solvent B). The molecules were eluted with a flow rate of 300 µL min−<sup>1</sup> , through a gradient elution from 1 to 50% B for 10 min, then to 99% B in 3 min, which was maintained for three additional minutes. The column was then re-equilibrated to 1% B for 4 min prior to the next run. Mass spectrometry analysis was carried out in ESI positive and negative modes. Quantification was performed by measuring the area under each peak at 280, 320, or 350 nm, depending on the lambda max of each molecule. Due to coeluting signal at 280 nm, catechin quantification was realized by measuring the area under peak at m/z 289, corresponding to the [M-H]– of the molecule. Experimental exact masses and MS fragments were compared to metabolomics databases (Respect<sup>1</sup> ; MassBank<sup>2</sup> , DNP<sup>3</sup> ) and data available in the literature in order to identify the nature of the metabolites. Sixteen phenolic compounds were identified. According to their structure, these 16 metabolites could be grouped into four phenolic subclasses, which are Flavan-3-ols (catechin, proanthocyanidin dimer), hydroxycinnamic acid esters (chlorogenic acid, caffeoyl shikimate, caffeoyl glucose isomers (×2), coumaroyl glucose isomers (×2)), flavonols (rutin, quercitrin), and salicinoids (salicine, salicortin, acetylsalicortin, homaloside D, tremulacin, and a putative populoside) (**Supplementary Figure S2**).

### Data Analysis

Principal component analyses (PCA) and linear models were used to explore correlations between phenotypic traits. Each independent phenotypic trait was analyzed separately using a generalized linear model evaluating effects of LPI, plant, day of inoculation, and the interaction between LPI and plant. To normalize the error, infection efficiency was analyzed with a binomial error family; latent period and sporulation capacity with a Gamma error family; uredinia size, spore volume, sporulation rate, and mycelium quantity with a Gaussian error family; and uredinia size and sporulation rate were square root transformed. An analysis of variance (ANOVA) with an alpha error <5% was performed on each model. Differences of means were checked with Tukey tests. PCA on leaf characteristics was performed to explore their variations. Spearman correlation coefficients between phenotypic traits and leaf content were computed on mean phenotypic traits per LPI and plant based on 4 to 12 replicates. Spearman correlation coefficients' associated p-values were adjusted with a False-Discovery-Rate correction for multiple tests. All data analyses were performed using R version 3.2.5 (see details in the RMARKDOWN document, **Supplementary Note S2**).

### RESULTS

### Variation in M. larici-populina Quantitative Traits

To describe the variation in quantitative traits of M. laricipopulina in this experiment, a PCA was performed on individual trait measurements assessed for each infected leaf disk (latent period, uredinia size, sporulation rate, sporulation capacity, mycelium quantity, and mean volume of produced spores). Infection efficiency cannot be included in the PCA analysis because it is set to one for infected leaf disk. Experimental factors (LPI, plant harvested, and day of inoculation) were considered as supplementary variables. The first principal component (PC1) explained 31% of the total inertia and was related to uredinia size, sporulation rate and to a lesser extent to sporulation capacity (**Figure 1**). PC2 and PC3 accounted for almost 20% and 17% of the total variance, respectively. PC2 was related to latent period which varied independently from uredinia size and sporulation rate, and PC3 was related to spore volume. However, there was large dispersal of individual measures that remained unexplained by experimental factors (**Supplementary Note S2**).

In accordance with PCA analysis, LPI was the main structuring variable in GLM analyses and had a significant effect on uredinia size (Padjusted = 1.22 × 10−21), sporulation rate (Padjusted = 7.67 × 10−15), and sporulation capacity (Padjusted = 1.40 × 10−05) (**Table 1**). Uredinia size and sporulation rate increased with the LPI, while sporulation capacity decreased (**Figure 2**). Uredinia size exhibited a threefold increase (from 0.272 to 0.737 mm<sup>2</sup> ) and sporulation rate a twofold increase (from 333 to 611 urediniospores day−<sup>1</sup> ) from LPI 6 to 20 (**Figure 2**). Sporulation capacity was reduced more than twofold from LPI 6 to 20 (from 12,420 to 5,800 urediniospores day−<sup>1</sup> mm−<sup>2</sup> ) (**Figure 2**). The sporulation rate was also significantly, although less strongly, explained by the plant used for the experiment (Padjusted = 4.63 × 10−05). Sporulation rate was significantly higher on plant No. 3 (mean rate = 526 urediniospores day−<sup>1</sup> ) and lower on plant No. 2 (mean rate = 396 urediniospores day−<sup>1</sup> ) (**Figure 2**), but no interaction was found between plant and LPI (**Table 1**). Other fungal traits varied but not in accordance with LPI, plant and day (**Supplementary Figure S3**).

We found no significant effect of the day of inoculation on any aggressiveness traits. As there was no difference in fungal traits across days of inoculation, we assumed that the incubation

<sup>1</sup>http://spectra.psc.riken.jp/

<sup>2</sup>http://massbank.jp/

<sup>3</sup>http://dnp.chemnetbase.com/

TABLE 1 | Summary of the analysis of variance of each experimental variable in generalized linear models (Chi-square test probability adjusted with a False-Discovery-Rate correction for multiple tests).


P-values < 0.05 are indicated in bold characters.

first and second axes.

conditions during the pre-inoculation period described above kept the leaf disks in their initial physiological state.

### Variation in Leaf Content Characteristics

To assess leaf content variations among the range of leaf maturity, a PCA was performed on leaf characteristics (**Figure 3**). Leaf characteristics were measured on one uninfected half-leaf of each plant and each LPI, so only plant and LPI supplementary variables were taken into account. PC1 explained almost 53% of the total inertia and was highly related to sugars (sucrose, fructose, glucose) and phenolic compounds (total phenolics, and all the phenolic subclasses) variations. Explaining 18% and 12% of the inertia, PC2 and PC3 were associated to sucrose and nitrogen, respectively. The total carbon content of leaves did not vary among the range of leaf maturity studied. Leaf maturity, characterized by the content in phenolic compounds, appeared to structure more leaf characteristics than plant did.

### Correlation of Fungal Phenotypic Trait Variation and Leaf Content Characteristics

To assess how fungal and leaf data were correlated, pairwise Spearman's correlation coefficients and their adjusted p-values were calculated.

Spearman's coefficient for uredinia size and both sporulation rate and sporulation capacity were highly significant (rhô = 0.87 and −0.73, Padjusted = 3.37 × 10−<sup>6</sup> and 4.66 × 10−<sup>4</sup> , respectively) (**Figure 4**), which is consistent with PCA results on individual traits. No other Spearman's coefficient appeared significant between fungal traits.

Spearman coefficients showed a negative correlation between total phenolic content and uredinia size and sporulation rate

(rhô = −0.77 and −0.61, Padjusted = 1.62 × 10−<sup>4</sup> and 1.07 × 10−<sup>2</sup> , respectively), as well as a negative correlation between hydroxycinnamic acid esters content and uredinia size and sporulation rate (rhô = −0.83 and -0.68, Padjusted = 2.25 × 10−<sup>5</sup> and 1.85 × 10−<sup>3</sup> , respectively), and a positive correlation between hydroxycinnamic acid esters content and sporulation capacity (rhô = 0.68, Padjusted = 1.85 × 10−<sup>3</sup> ) (**Figures 4**, **5**). Three of the four phenolic subclasses (salicinoids, hydroxycinnamic acid esters, and flavonols) showed the same trends as total phenolic content, with the exception that the correlation between sporulation rate and sporulation capacity, on one side, and salicinoids and flavonols, on the other side, were not significant (Padjusted > 0.05). In accordance with the PCA on leaf characteristics, all sugar contents were correlated with each other. However, no correlation was found between fungal phenotypic traits and sugars, carbon and nitrogen contents. Salicinoids, flavonols, and hydroxycinnamic acid esters were all positively correlated to each other. Flavan-3-ols were, however, negatively correlated to the three other phenolic subclasses, although this was only significant with flavonols.

### DISCUSSION

In this study, we analyzed the variation in seven aggressiveness and morphological traits of the poplar rust fungus when inoculated on poplar leaves of varying maturity levels. A single-spore inoculation technique was used, which enabled

us to study trait variance and covariance at the individual lesion level. Applying this method, we aimed at assessing the effect of leaf maturity on aggressiveness traits of M. larici-populina. In a second step, we explored correlations between the variation in fungal phenotypic traits and physiological features of the leaves in order to determine whether variation in phenotypic traits was influenced by nutritive or defensive compounds of leaves.

### Effect of Leaf Maturity on Aggressiveness and Morphological Traits

Regarding the variation in aggressiveness traits, our first result was a positive correlation between uredinia size and sporulation rate, whereby both increased with leaf maturity. That is, on more mature leaves, uredinia were wider with a higher production of urediniospores. Some authors found that the breaking down of sporulation rate into uredinia size and sporulation capacity enabled to explain differences in aggressiveness levels: an increase in sporulation rate can either be due to wider uredinia or higher sporulation capacity (flux of urediniospores) (Pariaud et al., 2009b). In our study, we found that the increase in sporulation rate is linked to wider uredinia, but not to higher sporulation capacity. The correlation between uredinia size and sporulation rate has already been identified by several authors, including Dowkiw et al. (2003) on poplar rust, Pei et al. (2002) on willow rust and Robert et al. (2002) on wheat leaf rust. Based on our data, we add to this that sporulation capacity is not necessarily even as assumed by Robert et al. (2002). Indeed, sporulation capacity significantly decreased with leaf maturity, in contrast to uredinia size and sporulation rate. Even if the flux of urediniospores was lower on mature leaves, the amount of urediniospores produced was higher, leading to an increased aggressiveness on mature leaves compared to younger ones.

In this study, we also quantified individual variation in mycelium quantity through qPCR analysis. This trait has received little attention as a component of aggressiveness, mostly for methodological reasons. However, it can serve to measure the relative investment in sporulation vs. mycelium growth, the two main functions to which host extracted resources are allocated (Gilchrist et al., 2006; Yegorov et al., 2017). Notably, the quantity of in planta mycelium did not vary along the leaf maturity gradient and was correlated to neither sporulation rate nor lesion size. Hence, lesion size does not equate to mycelium

quantity. This makes sense as for biotrophic fungi like rusts, lesion (uredinia) size is only the size of the aperture through which urediniospores are extruded, and may thus not be directly linked to the size of the area within a leaf that is colonized by the mycelium.

Neither leaf maturity, nor the two other experimental factors (plant and day of inoculation) had an effect on latent period, infection efficiency, mycelium quantity and spore volume, even though substantial variances were observed for those traits. The latent period and spore volume varied independently from each other and independently from the main structuring traits (uredinia size and sporulation rate). Variation in latent period and spore volume might have been affected by an unidentified environmental factor. Even though inoculated leaf disks were incubated under controlled conditions of temperature and light, micro-scale variation may have occurred for these crucial parameters influencing fungal development and infection dynamics (Cochrane, 1959; Rotem et al., 1978; Trapero-Casas and Kaiser, 1992; Turechek and Stevenson, 1998; Turechek et al., 2001; Pap et al., 2013). Moreover, the initial position of the disk in the leaf might have influenced infection outcome (Katsuya and Green, 1967; Johnston et al., 2016). Singlespore inoculation is more prone to inter-replicate variance, in contrast to spray-inoculation which buffers micro-environmental variations. Apart from this drawback, this method allowed us to accurately disentangle the variation in each trait and to properly assess putative correlations. In our study, the dissection of the infection process into elementary traits enabled us to assess that the leaf maturity gradient only influences the sporulation function.

### Disentangling Nutritive Limitation vs. Defense Reaction Effects

Two hypotheses were considered to explain variation in uredinia size and sporulation rate. On the one hand, sporulation could be enhanced by a higher nutrient availability in older leaves. On the other hand, higher levels of defense compounds could in contrast prevent fungal development in younger leaves.

Leaf sugars have long been known to be essential to the development of plant pathogenic fungi (Bilgrami and Verma, 1978; Madan and Thind, 1998). Using Populus deltoides leaves

inoculated with Melampsora medusae f. sp. deltoidae, Coleman (1986) showed that the maximum level of susceptibility, measured by the number of lesions (uredinia) as a proxy for infection efficiency, is reached when the sugar availability is maximal, i.e., at a LPI of 5. This finding is inconsistent with our study, in which the infection efficiency did not vary with leaf maturity and the two responsive traits (uredinia size and sporulation rate) steadily increased from LPI 6 to 20. The discrepancy between Coleman (1986) experiments and our results may lie in differences in the growth conditions of poplar plants, which may lead to different levels of nutrient availability. The light quality available in growth chambers has been significantly improved during the last decades. In Coleman (1986), the total sugar content decreased from LPI 6 to 9 (the highest LPI measured). In our experiment, even though sugar contents displayed some variance among leaves, there was no consistent pattern along LPI. Our growth conditions may have led to sugar availability sufficient for steady infection success even at higher LPI. We therefore conclude that leaf sugar content may not be linked to leaf susceptibility per se.

Leaf nitrogen content also plays a role in fungal growth and sporulation (Bilgrami and Verma, 1978; Madan and Thind, 1998). Bilgrami and Verma (1978) stated that nitrogen sources are used by fungi for both structural and metabolic functions. These functions require the assimilation of nitrogen into amino acids, proteins, and some peptides. For example, Jensen and Munk (1997) and Robert et al. (2004) found that higher nitrogen fertilization enhanced barley powdery mildew and wheat leaf rust sporulation. In our experiment, although there was a substantial variance in leaf nitrogen contents [same order of magnitude as in Robert et al. (2004)], this variation did not significantly affect M. larici-populina development. The leaf nitrogen content only slightly correlated with latent period and sporulation capacity (urediniospores produced par day and per area).

Overall, we did not observe consistent variation in nutrient content along the leaf maturity gradient and, moreover, the variation in nutrient content was not correlated with the two responsive traits, uredinia size, and sporulation rate. Consequently, nutrient availability was apparently not the environmental factor best explaining fungal development in our experiment.

By contrast, the defense hypothesis, tested by assessing the constitutive level of phenolics (total and individual) appeared to better explain fungal development related to leaf age. Indeed, the total phenolic content decreased with leaf age and correlated negatively with uredinia size and sporulation rate. Although higher phenolic concentration in younger leaves is common in poplar and other plants (Coleman, 1986; Johnson and Kim, 2005; Royer et al., 2013; Ullah et al., 2017), the correlation between phenolic concentrations and resistance to fungal infection has not been demonstrated. Coleman (1986) and Coleman et al. (1987) did not detect a correlation between leaf phenolic content and the amount of uredinia produced by M. medusae on poplar trees. By contrast, in the same pathosystem, a negative correlation was observed between total phenolic content and urediniospore germination efficiency and hyphae length (Johnson and Kim, 2005). Other studies on black currant (Ribes nigrum) and willow (Salix myrsinifolia) identified negative correlation between the concentrations of phenolic compounds and development traits of rusts (Hakulinen et al., 1999; Vagiri et al., 2017). By analyzing specific phenolics from poplar leaves, we were able to precisely identify groups of phenolic compounds that correlate with poplar resistance to rust. Among the four phenolic subclasses (from a total of 16 compounds), three subclasses (hydroxycinnamic acid esters, flavonols, and salicinoids) showed the same effect as total phenolic contents and correlated negatively with uredinia size and sporulation rate. The hydroxycinnamic acid ester chlorogenic acid has been shown to correlate to rust resistance in black currant (Vagiri et al., 2017). In addition, several studies highlighted the role of leaf surface flavonoids in the resistance to rust in poplar and birch (Shain and Miller, 1982; Valkama et al., 2005). In our study, we did not address the leaf location (inside or on the surface) of the two flavonols, rutin and quercetrin, however, the fact that they negatively correlate with rust development is in agreement with these previous studies. The only phenolic subclass that did not correlate with resistance to rust was flavan-3-ol which integrates catechin and proanthocyanidin dimers. A role of this phenolic subclass – especially catechin – in resistance to fungal pathogens, including rust fungi, has been proposed in several studies (Brignolas et al., 1998; Hakulinen et al., 1999; Vagiri et al., 2017). Moreover, a recent study on the M. larici-populina – Populus nigra pathosystem demonstrated that catechin is an effective chemical defense compound against rust (Ullah et al., 2017). This demonstration was based on the fact that (i) catechin is strongly induced upon rust infection, (ii) catechin exhibit a strong antifungal activity against M. larici-populina at in planta concentration, and (iii) poplar engineered to accumulate high or low levels of catechin were more resistant or more susceptible to rust, respectively. In our study, the concentrations of flavan-3-ols increased with leaf age. The same trend was observed in the study by Ullah et al. (2017) (see **Supplementary Figure S2**). The absence of a correlation between flavan-3-ol concentrations and rust resistance does not indicate that flavan-3-ols are not important factors in leaf rust resistance. Indeed, the compounds were assayed in uninfected leaves; hence reflect constitutive (i.e., non-induced) defense levels. In addition, the concentrations measured in our experiment (from 0.2 to 0.5 mg.g DW−<sup>1</sup> ) were below the concentrations proven to efficiently impair rust development on black poplar leaves (Ullah et al., 2017).

### CONCLUSION

Melampsora larici-populina development after single-spore inoculation onto poplar leaf disks was more constrained by the level of plant constitutive defenses than limited by nutrient availability. For the tested strain, uredinia size and sporulation rate, both components of the sporulation function, were the only traits responsive to reaction norms along a gradient of leaf maturity. This experiment – conducted on a single fungal isolate and on a single plant genotype – can nonetheless serve as a pilot study. We report the correlation of a fungal fitness component (sporulation) to an environmental characteristic imposed by the leaf i.e., the constitutive level of phenolic content. In the future, it would be worthwhile to assess the generality of this finding with more fungal isolates and plant genotypes and decipher further the interplay between constitutive and induced defense compound levels. From a broader perspective, it would be interesting to determine the plasticity and heritability of sporulation-related traits to better assess how the complexity of leaf environment shapes the evolutionary trajectory of a plant pathogen.

### AUTHOR CONTRIBUTIONS

fpls-09-01396 September 28, 2018 Time: 13:23 # 12

FH, PF, and BF designed the study. AM conducted the experiment and analyses. MP developed the single-spore inoculation and trait measurement methods. AA developed the qPCR method based on new primers and probe for Melampsora larici-populina in the rDNA ITS (Internal Transcribed Spacer) designed by CG and A-LB. RL conducted leaf phenolics analyses. AM and FH wrote the first draft of the manuscript. PF, MP, and RL contributed to the manuscript. All the authors approved the final version of the manuscript.

### FUNDING

This work was supported by grants from the French National Research Agency (Grant No. ANR-13-BSV7-0011, FunFit

### REFERENCES


project; ANR-11-LABX-0002-01, Cluster of Excellence ARBRE). AM and MP were supported by Ph.D. fellowships from the Region Lorraine and INRA, and from the French Ministry of Education and Research (MESR), respectively.

### ACKNOWLEDGMENTS

Jérémy Pétrowski is acknowledged for his excellent technical help in poplar plant production. We thank Jacqueline Marchand and Carole Antoine (Université de Lorraine, AgroParisTech, INRA, SILVA, Nancy, France) for the poplar leaf content analyses, Aude Fauvet for the analysis of total phenolics, and Didier Le Thiec for the lyophilization of the samples. We also thank Irène Hummel for advices in statistical analyses, Rantanplan for inspiration, Josselin Montarry, Elisabeth Fournier, Stéphane De Mita, Claire Fourrey, and Katherine Hayden for valuable comments on earlier drafts of this manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018.01396/ full#supplementary-material



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

Copyright © 2018 Maupetit, Larbat, Pernaci, Andrieux, Guinet, Boutigny, Fabre, Frey and Halkett. 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.

# Adaptation of a Fungal Pathogen to Host Quantitative Resistance

Lise Frézal1,2, Guy Jacqua<sup>1</sup> and Claire Neema<sup>3</sup> \*

1 INRA-URPV, Guadeloupe, France, <sup>2</sup> CNRS – ENS – INSERM, Institut de Biologie de l'Ecole Normale Supérieure, Paris, France, <sup>3</sup> UMR BGPI, Montpellier SupAgro, Campus International de Baillarguet, Montpellier, France

Impact of host quantitative resistance on pathogen evolution is still poorly documented. In our study, we characterized the adaptation of the pathogenic fungus Colletotrichum gloeosporioides, to the quantitative resistance of its host, the water yam (Dioscorea alata). Genetic and pathogenic diversities of C. gloeosporioides populations were specified at the field scale. We used nuclear markers to describe fungal population structuring within and between six fields of three cultivars differently susceptible to the fungus. Strain aggressiveness was then quantified in the laboratory through crossinoculation tests. The high level of genetic diversity and significant linkage disequilibrium revealed a significant influence of clonal reproduction in the C. gloeosporioides evolution. The recorded fungal migration between fields was weak (evidence for a dispersion mode via tubers rather than splashing dispersal), which provides the first molecular evidence for limited C. gloeosporioides migration via yam tuber exchanges. C. gloeosporioides's populations are adapted to their host resistance. The aggressiveness of the fungal clones seems to have evolved toward an accumulation of components specific to each host cultivar. Despite the remaining marks of adaptation to the former widely cultivated host, adaptation to current cultivars was clearly depicted.

#### Edited by:

Josselin Montarry, INRA Centre Bretagne-Normandie, France

#### Reviewed by:

Fabien Halkett, INRA Centre Nancy-Lorraine, France Sébastien Guyader, INRA Centre Antilles-Guyane, French Guiana

#### \*Correspondence:

Claire Neema claire.neema@supagro.fr

#### Specialty section:

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

Received: 17 June 2018 Accepted: 03 October 2018 Published: 31 October 2018

#### Citation:

Frézal L, Jacqua G and Neema C (2018) Adaptation of a Fungal Pathogen to Host Quantitative Resistance. Front. Plant Sci. 9:1554. doi: 10.3389/fpls.2018.01554 Keywords: local adaptation, aggressiveness, quantitative resistance, Colletotrichum gloeosporioides, Dioscorea alata

### INTRODUCTION

Discernment on the adaptive potential of plant pathogen species is essential to designing sustainable integrated pest management framework in which, durability of crop resistance is one crucial point. The adaptive potential of plant pathogen species highly depends on the forces shaping the pathogen evolution. Among these forces, the migration (spatial dispersal), the recombination (reproductive mode), genetic drift and above all, the selection by the host-plant resistances, are of particular importance (McDonald and Linde, 2002). Pathogen populations can either adapt to the most common host (general adaptation), or be divided into several entities, each adapted to one particular host genotype (local adaptation, Gandon et al., 1996; Gandon and Van Zandt, 1998). The theory of polygenic traits' evolution under divergent selection has not been much investigated, especially while investigating host pathogen interactions (Kawecki and Ebert, 2004). The few studies focusing on the host quantitative resistance impact on pathogen populations reported either the local adaptation of pathogen populations (Kaltz and Shykoff, 1998; Zhan et al., 2003), or a directional selection toward an increase in pathogen aggressiveness on all host cultivars (Andrivon, 1993; Andrivon et al., 2007; Montarry et al., 2007). In the present study, we described the population diversity of the plant pathogen Colletotrichum gloeosporioides in the Guadeloupean (French West Indies)

agrosystem. We also investigated the pathogen aggressiveness as an adaptive trait of the fungi to the quantitative resistance of its host, Dioscorea alata.

The filamentous ascomycete and plant pathogen C. gloeosporioides (Penz.) Penz. & Sacc. [telemorph: Glomerella cingulata (Stonem.) Spauld. & Schrenk] causes anthracnose disease on various crops (Bailey and Jeger, 1992; strawberry, Freeman et al., 1998; olive, Talhinhas et al., 2005; avocado, mango, stylo, Chakraborty et al., 1999; ornamental lupinines, Elmer et al., 2001; Salsola tragus, Berner et al., 2006; peach tree, Kim and Hong, 2008), among which the water yam (D. alata). The life cycle of this haploid and hemibiotroph fungus includes both sexual and asexual reproduction. If the two reproductive modes are observed on water yam, the asexual mode prevails during yam anthracnose epidemics. Conidia (asexual spores) disseminate from plant to plant by wind-directed-rain splashing (Green and Simons, 1994; Penet et al., 2014) which limits dispersal to several meters from the source (Ntahimpera et al., 1999).

Sexual reproduction can nevertheless occur at the end of epidemics, on dead leaves or stems, which—when left on the soil surface—may contribute to the primary infection of yam (Ripoche et al., 2007). Ascospores (sexual spores) are produced in dried-perithecia that can eventually be wind-dispersed on a long distance (Abang, 2003). Besides, C. gloeosporioides generates quiescent infections in yam tuber epidermis (Simons and Green, 1994); As farmers traditionally save tuber portions from a previous harvest for the new crop propagation, mycelia can pass from one crop season to the next. Thus, commercial exchanges of yam tubers and quiescent infections confer to C. gloeosporioides a capacity for long-distance migration and for clonal propagation (Green, 1994).

Dioscorea alata foliar anthracnose has been reported in Oceania and India (Winch et al., 1984; McDonald et al., 1998), in the West and Central Africa (Asiedu et al., 1998; Mignouna et al., 2001) and in the Caribbeans (Coursey, 1967; Green and Simons, 1994; Cowger and Mundt, 2002). In the West Indies, anthracnose disease became the most damaging disease on water yam at the end of the 1970s (McDonald et al., 1998) and was associated with serious yield losses reaching 80–100% for the common cultivars (cv) of this time, i.e., cv Pacala, in Guadeloupe (Green and Simons, 1994). In the 1970s, the cultivar Plimbite was introduced in the Guadeloupean islands and has been extensively cultivated until it became as susceptible to anthracnose as cv Pacala. After the late 1990s, farmers started to use more resistant cultivars such as cv Kabusah and cv Tahiti, but kept planting previous cultivars (Penet et al., 2016).

Genetic determinants of the interaction between C. gloeosporioides and D. alata are poorly understood. For a long time, this resistance was thought to be polygenic (Green and Simons, 1994; Abang et al., 2001). More recently, Petro et al. (2011) work based on the Boutou and Pyramide D. alata cultivars provided strong evidence for the quantitative inheritance of the components associated with the D. alata resistance to C. gloeosporioides. Next to this quantitative resistance, a gene-for-gene-like component (Flor, 1971) of the interaction between C. gloeosporioides and D. alata cultivars has been unraveled (Mignouna et al., 2002; Petro et al., 2011). For instance, the single dominant locus (Cdg1) explains the specific resistance of cv TDa-95/00328 to moderately virulent C. gloeosporioides strains.

The genetic structure of C. gloeosporioides populations, causing yam anthracnose, has been mostly studied in Nigeria (Abang et al., 2001, 2004, 2005, 2006), which is the first yam producer in the world (71.5% of the world production, FAOstat 2007). There, C. gloeosporioides was shown to exhibit high genetic and phenotypic diversity from the lesion to the region scales (morphotypes, Vegetative Compatibility Groups, RAPD, MP-PCR) and a weak population structure according to the host or to the agroecological areas (Abang et al., 2004, 2006).

Yam breeding and the management of yam resistances in the French West Indies require knowledge about the influence of the migration, the reproductive mode and the yam quantitative resistance on C. gloeosporioides evolution. Moreover, the impact of the host quantitative resistance on the evolution of the pathogen aggressiveness has never been investigated. Aiming to document a part or all of these aspects, the objectives of this work were: (i) to estimate genetic diversity and its distribution among fields with neutral markers (AFLP); (ii) to infer the influence of recombination at the field scale; (iii) to study how the degree of host quantitative resistance to anthracnose influences the distribution of fungal aggressiveness diversity.

### MATERIALS AND METHODS

### Fungal Sampling and Isolation

A total of 222 C. gloeosporioides isolates were collected from December 2001 to January 2002, from six yam fields within a 5-km-diameter area of the Guadeloupe (FWI, between Morne à l'eau and Le Moule, 16◦ 15N, 61◦ 35O, **Figure 1** and (**Supplementary Table S1**). Sampling was done on three yam (D. alata) cultivars with different levels of resistance to anthracnose: cv Pacala (highly susceptible), cv Kabusah (moderately susceptible) and cv Tahiti (resistant). The study was carried out in six production fields of approximately 0.5 ha, with identical soil composition, topology, crop management and climate (**Supplementary Table S1**). One month before harvest, leaves with typical anthracnose lesions were collected from 50 plots per field, each distant from 2 to 5 m. Fungus isolation was performed from freshly harvested infected yam leaves (on the same day). Circa 1 mm<sup>2</sup> of typical anthracnose lesions were plated on potato dextrose agar medium (PDA). C. gloeosporioides isolates were identified according to morphological characteristics (Sutton, 1992), then monoconidial cultures were obtained. The strains were grown on PDA-medium under photoperiodical conditions of 12 h day at 25◦C and 12 h night at 23◦C. Spores of 7 day-old cultures were suspended in distilled water to reach concentrations of 5.10<sup>5</sup> conidia per mL. Mycelia were dried, thrown in liquid nitrogen and stored at −80◦C for DNA extraction.

### DNA Extraction and AFLP Markers Genotyping

One milliliter of a suspension at 5.10<sup>5</sup> conidia per mL was mixed with 30 mL of potato dextrose broth medium (PDB). After 72 h of incubation at 27◦C, the mycelium was harvested, dried on sterile filter paper and stored at −80◦C. Total genomic DNA was extracted from frozen mycelium with a phenol/chloroform protocol as described in Villaréal et al. (2002). DNA quality was checked by electrophoresis on 0.8% agarose gel and stored at −20◦C in TE 0.1 (10 mM trisHCL, pH8; 0.1 mM EDTA). DNA digestion (MseI and EcoRI), ligation, pre-amplification, selective amplification and electrophoresis were conducted as described in Villaréal et al. (2002). The primer combinations used for selective amplifications were MseI-0 +AT/EcoRI-0 +AT, MseI-0+ AG/EcoRI +AG and MseI-0 +TA/EcoRI-0 +CAA. The 6% bisacrylamide-urea gel was stained with silver nitrate as described in Chalhoub et al. (1997). Polymorphic bands between 100 and 500 pb were manually referenced as present ("1") or absent ("0"), which gave a 0–1 combination for each isolate. The AFLP patterns were checked for reproducibility with the two DNA replicates extracted and only clearly repeatable bands were used for further analyses. Co-migrating bands were assumed to be the same locus, and isolates having the same AFLP-pattern were assumed to be one single haplotype.

### Plant Material, Inoculation and in vitro Aggressiveness Assessment

In order to study the adaptive patterns of C. gloeosporioides on its host cv of origin, cross-infection experiments were performed. The diversity for aggressiveness was investigated on a 45-strain subsample composed of six to nine strains per field, randomly chosen in a cloned-corrected sample (i.e., sample containing only single-copy AFLP patterns).

### Host-Plant Material

Colletotrichum gloeosporioides strain aggressiveness was tested against the three host-plant cultivars from which they were isolated, i.e., water yam cv Pacala (high susceptibility),cv Kabusah (moderate susceptibility), cv Tahiti (low susceptibility), and also tested on one cultivar of reference for susceptibility to anthracnose, cv Plimbite (high susceptibility). Plants were grown from tuber pieces. All the yam tubers were produced in the INRA-URPV experimental center (Guadeloupe, French West Indies) which ensured the origin of the tubers as well as the sanitary control of their production. A total of 15-g tuber pieces were dipped for 24 h into fungicides (Chlorothalonil 550 g L−<sup>1</sup> and Carbendazim 100 g L−<sup>1</sup> supplied by Bayer CropScience), airdried and covered with wood ashes. Tuber pieces were planted into humidified sterile compost in individual pots and put in a greenhouse under the natural conditions of Guadeloupe at the beginning of the wet season (25–30◦C, >75% humidity, 12 hphotoperiod). To prevent any leaf infection on the plants used in the in vitro tests, we applied pesticides (oxythioquinox at 10 mg/mL supplied by Bayer CropScience) to the compost every 3 weeks until the first leaf emergence. As advised by Simons and Green (1994), fully expanded young leaves were harvested on 3-month old plants. Leaves were then washed with sterilized water and their adaxial surfaces were placed on moistened filterpaper disks in Petri dishes (one leaf per petri dish). Leaves were inoculated on the entire abaxial surface, as described by Toribio and Jacqua (1978).

Inoculation Strains were grown on PDA-medium in 12 h day at 25◦C and 12 h night at 23◦C conditions. Spores of 7-day old cultures were suspended in distilled water to raise concentrations of 5.10<sup>5</sup> conidia mL−<sup>1</sup> . For each strain, 22.5 mL of the spore-suspension were sprayed on freshly cut leaves of each cultivar (five leaves per cultivar) using ecospray <sup>R</sup> atomizer (LCF labo chimie France, Aix en Provence). At the same time, non-inoculated leaves (sprayed with water) of each cultivar were used as controls. Leaves were incubated at 100% humidity and in photoperiodical conditions of 12 h day at 27◦C, 12 h night at 22◦C. To reproduce the natural conditions of infection of D. alata with C. gloeosporioides, inoculations were carried out just before night. For all host-strain combination and condition (control and inoculation), five independent repeats were performed simultaneously.

Disease assessment two different symptoms are commonly observed when inoculating D. alata leaves with C. gloeosporioides: the typical lesion (large necroses, dark brown, coalescent, high expansion rate) and the pinpoint lesion [punctual necroses of less than 2 mm in diameter, black, not coalescent;

Winch et al. (1984)]. Percentages of healthy, typical lesion and pinpoint lesion areas were visually estimated from the first to the 11th day after inoculation. In order to estimate strain aggressiveness, we use the analytical framework detailed in Frézal et al. (2012), which transforms the evolution of disease severity over time into one index, Ag. The construction of the Ag index is based the analysis of the symptom dynamics using a linear mixed-effects regression (PROC MIXED procedure; Laird and Ware, 1982) with two-knots, using SAS software (SAS Institute, 1987). The model is defined with a starting point at day 0 and two changes in slope (at day 6 and day 9). The typical lesion (Na) and total diseased areas (Da) were then defined as:

$$\text{Da}\left(t\right) = \alpha\_{\text{Da}} \cdot t + \beta\_{\text{Da}}\left(t - 6\right) + \gamma\_{\text{Da}}\left(t - 9\right)$$

$$\text{Na}\,(t) = \alpha\_{\text{Na}} \cdot t + \beta\_{\text{Na}}\,(t-6) + \gamma\_{\text{Na}}\,(t-9)$$

where, (t−6) = (t−6) if (t−6) > 0, and 0 otherwise; where (t−9) = (t−9) if (t−9) > 0, and 0 otherwise.

Strain aggressiveness index on each cultivar was defined as:

$$\text{Ag} = \frac{3}{10} \cdot 100 \left( \left( \alpha\_{\text{Da}} \frac{6}{100} \right)^{\frac{1}{4}} \cdot \frac{\text{Da} \left( \text{day} \, 11 \right)}{100} \right) + \frac{7}{10} \cdot 100$$

$$\left( \left( \alpha\_{\text{Na}} \cdot \frac{6}{100} \right)^{\frac{1}{4}} \cdot \frac{\text{Na} \left( \text{day} \, 11 \right)}{100} \right)$$

For each strain, four indices were calculated: Ag-K on yam cv Kabusah, Ag-P on cv Pacala, Ag-T on cv Tahiti and Ag-Pl on cv Plimbite. We also focused on the initial lesion expansion rates (i.e., αNa) and the typical lesion area on day 11 (Na) as this provides detailed insight on strain fitness and helps to imagine hypothetic outcomes of the competition between strains.

### Data Analysis

### Population Genetics Analyses

Diversity at each locus was estimated within subset of isolates grouped by field or cultivar. Unbiased gene diversity (Hd; Nei, 1987) was computed using GENETIX 4.05 (Belkhir et al., 1996/2004) on data sets containing only one copy of each haplotype per subset (i.e., clone-corrected data sets) (Chen and McDonald, 1996). Frequencies at each locus were estimated from each subset and used to calculate standard population statistics. In order to see whether the number of AFLP markers used in this study was sufficient to recover the maximal genotypic diversity, we plotted the genotypic diversity calculated from the multilocus genotype frequencies (Nei, 1987) versus the number of loci: our 23 polymorphic loci were discriminating and powerful enough (**Supplementary Figure S1**).

A minimum spanning tree based on the 23 ALFP markers was constructed for the full sample (222 isolates) using Bionumerics 5.1 with 100,000 resampling and partitioning of two majority trees. To determine population subdivision without any a priori knowledge, a second analysis was performed on the clone-corrected dataset using the Bayesian clustering method implemented in the software Structure version 2.3.3 (Pritchard et al., 2000). We used admixture and non-admixture models to identify genetic groups, both with distinct allele frequencies (Falush et al., 2003), and we assumed uniform priors for the vector of proportion (qi) of the individual i's genome in each cluster. The scores of individuals in the genetic groups (i.e., the posterior estimates of the qi) correspond to the probability of ancestry in each one of them. We varied K from 1 to 10, with 10 replicates for each K, and with each simulation consisting in 800,000 Monte-Carlo Markov Chain (MCMC) iterations preceded by a burn-in period of 200,000 iterations. The most probable structure was determined by computing the posterior probability for each K using the distribution of maximum likelihoods and 1K distribution (Evanno et al., 2005; **Supplementary Figure S2**), which is a quantity, related to the second order rate of change of the log probability of data with respect to the number of genetic clusters. Finally, to determine population subdivision without any a priori knowledge and without making any assumptions regarding the population genetics model, we described the genetic clusters from the clone-corrected dataset using the discriminant analysis of principal components (DAPC, Jombart et al., 2010) implemented in the R package adegenet 2.1.1 (Jombart and Ahmed, 2011; R Development Core Team, 2015).

Pairwise Weir and Cockerham (1984) values were calculated between all pairs of populations using GENETIX 4.05 (Belkhir et al., 1996/2004). Populations were defined as field samples and as subsets of individuals collected on the same yam cultivar, i.e., Kabusah, Tahiti and Pacala. Then, the significance of the pairwise-Fst was tested by 10,000 random permutations of the haplotypes (i.e., individuals) between the populations using Multilocus 1.3b software (Agapow and Burt, 2001).

The number of genotypes (Gs) was calculated in the whole sample and per filed from the clone-corrected dataset. As recommended by Arnaud-Haond et al. (2007) we calculated the modified index of clonal diversity, R from Dorken and Eckert (2001) as

R=(G−1)/(N−1) , with G, the number of identical multilocus genotypes and N, the number of individuals.

The degree of association between loci was estimated on the clone-corrected dataset by using the index of association, which is the ratio between the observed variance of the number of differences between pairs of strains and the expected variance under the hypothesis of absence of linkage disequilibrium (LD) (Agapow and Burt, 2001). We used a modified version of this index (rD), corrected for the dependence to the number of loci used. r<sup>D</sup> has an expected value of 0 if there is no association of alleles at unlinked loci, as is expected in a randomly mating population. The significance of r<sup>D</sup> was tested by a randomization procedure (1,000 times) by comparing the observed value to that expected under the null hypothesis of complete random mating (Agapow and Burt, 2001). Value calculation and tests were performed by Multilocus 1.3b software (Agapow and Burt, 2001). To prevent any effect of the population structure or of the LD between markers, we tested the estimated and tested (LD) between markers and retained 16 loci with no residual LD. We estimated and tested the significance of r<sup>D</sup> within each genetic cluster using the 16 loci clone-corrected dataset.

### Aggressiveness Analysis

fpls-09-01554 October 29, 2018 Time: 18:17 # 5

Analyses of variance were conducted for the factors host cultivar, field of origin, genetic group using the General Linear Models Procedure of SAS software (SAS Institute, Inc., Cary, NC, United States). A standard analysis of variance (ANOVA) of the aggressiveness indices on each of the four hosts was performed using the SAS software (SAS Institute, Inc., Cary, NC, United States) to test the effect of the host cultivar, the field and the genetic group on strain aggressiveness. Moreover, the mean-values of aggressiveness indices (i.e., Ag-K, Ag-T, Ag-P, Ag-Pl), of initial lesion expansion rate (αNa) and of final lesion area (Na) were compared using the Student's t-test. Means values were considered as different for a probability under the threshold of 0.01.

### RESULTS

### Genetic Diversity

A total of 222 monoconidial strains of C. gloeosporioides were isolated from three host cultivars (**Figure 1**). The AFLP analysis generated 89 clearly reproducible bands. Among the 89 clearly reproducible markers, 23 were polymorphic, independent and had a frequency between 0.05 and 0.95, i.e., non-rare alleles. All the markers were amplified from several isolates originating from at least three different fields. Among the 222 isolates, 197 haplotypes were identified, which correlated with high R ratios (i.e., modified G/N ratio, where the number of identical multilocus genotypes, G, is divided by the number of individuals, N) recorded for the entire dataset, 0.89, and in each fieldpopulation, from 0.83 to 0.93 (**Table 1**). Sixteen haplotypes were found in two, three or four copies, but the few isolates sharing the same haplotype were always recovered from a unique field (**Supplementary Figure S3**). Nei (1987) gene diversities were all superior to 0.20 whether they were calculated on field sample or on the entire dataset (**Table 1**).

### Population Structure

The population structure analysis with no a priori, displayed in the Minimum Spanning Tree (**Figure 2A**) was concordant with the structure analysis using both the Structure software (**Figure 2B**) and the DAPC—adegene package (**Supplementary Figures S4**–**S6**). According to Evanno's method, the clone-corrected sample (197 isolates) was most likely divided into three genetic groups (**Supplementary Figure S2**). The group 1 (blue in **Figure 2B**, k = 3) was composed of more than 98% of strains isolated from yam cv Tahiti. This "Tahiti-exclusive" genetic group contained 78.8% of the strains collected on cv Tahiti, including 97% of the field Tah<sup>A</sup> sample and 61% of the field Tah<sup>B</sup> sample. For the groups 2 and 3 (green and orange in **Figure 2B**, k = 3) the clustering did not correspond to one host cv nor to one geographical location. The group 2 (green in **Figure 2B**, k = 3) was composed of isolates from the fields Pac<sup>C</sup> (50%), Kab<sup>A</sup> (33%), and Tah<sup>B</sup> (13.5%). The group 3 (orange in **Figure 2B**, k = 3) contained the entire Pac<sup>D</sup> sample and 91.6% of Kab<sup>B</sup> sample, which represented 93% of the total of strains contained in this genetic group. The genetic grouping was more complex for the group 2. Pac<sup>C</sup> strains belonged to one genetic group only, but Kab<sup>A</sup> strains belonged to two genetic groups, one common with strains originated from Tah<sup>B</sup> and one with strains from PacC.

The computed pairwise-Fst proved that populations from different fields were significantly differentiated, whatever the host cultivar, cv, and the field (**Table 2**). The pairwise-Fst values were clearly consistent with the genetic clustering given by the analysis with K = 6. Indeed, the "Tahiti-private" group and the group 3, respectively, split into two genetic groups corresponding to distinct field of origin (i.e., Tah<sup>A</sup> vs TahB; Pac<sup>D</sup> vs KabB). The lowest Fst-value (0.161) was observed between the fields Kab<sup>A</sup> (host cv Kabusah) and Pac<sup>C</sup> (host cv Pacala) distant of 1.3 km, whereas higher Fst-values (0.406 and 0.404) were observed between two contiguous fields (Kab<sup>A</sup> and TahA) and between the two cv Pacala fields (Pac<sup>C</sup> and PacD).

Altogether, the populations sampled on the cultivar Tahiti tend to form a "Tahiti-specific' group, the populations from the fields Pac-D and Kab-B form two clearly distinct genetic groups. The sample from the Kab-A field does not form a uniform genetic group, as to a lesser extent, the sample from the fields Pac-C and Tah-B.

### Indices of Association

All indices of association values (rD) calculated were low (rDvalues ranging from 0.003 to 0.103) but with significant departure to 0 for all populations except cluster 1 (**Table 3**).


(a)Number of polymorphic loci out of the 86 loci. (b)Non-biased Nei's (1987) diversity; standard errors are specified in brackets. (c)R, the debiased G/N index calculated as (G−1)/(N−1) with G, the number of identical multilocus genotypes and N, the number of individuals.

FIGURE 2 | Genetic structure of the 222 Colletotrichum gloeosporioides strains sampled on six fields of three different cultivars. (A) The minimum-spanning tree based on ALFP markers was constructed using Bionumerics 5.1. Haplotypes collected on cv Tahiti are coded in blue, on cv Kabusah in green and on cv Pacala in Red (bottom right). The size of the circles corresponds to the number of replicates of the same haplotype (bottom right). Numbers of differences between haplotypes are coded by the type of line joining two circles (bottom right). (B) Structure software output based on 23 AFLP markers with the number of genetic groups (K) ranging from 2 to 6. Kab, Tah, Pac, respectively, stands for the host, Dioscorea alata cultivars Kabusah, Tahiti and Pacala. The letters A, B, C, D stand for the four geographical origins (see Figure 1). Gray arrows indicate the strains randomly chosen for the pathogenicity tests.

### Strain Aggressiveness Distribution

The diversity for strain aggressiveness was investigated on a 45 strain subsample. These strains belong to the main genetic groups described above (**Figure 3** and **Supplementary Figure S7**). No symptom nor chlorosis appeared on control leaves 11 days after inoculation. We detected an effect of the genetic group (cluster) on strain aggressiveness on the cultivar Tahiti but none on the cultivars Kabusah, Pacala and Plimbite (**Table 4**). Besides, strain aggressiveness on yam cv Plimbite (taken as the reference for susceptibility) and on yam cv Kabusah (moderate susceptibility) did not depend on the host nor on the field of origin (**Table 4**). Nevertheless, the Student's t-test showed that strains collected from cv Kabusah were more aggressive on their host cv Kabusah (Ag-index mean-value of 47.8) than the strains sampled from cv Tahiti (Ag-index mean-value of 20.3). More strikingly, strain aggressiveness, respectively, on the yam cv Tahiti (low susceptibility) and cv Pacala (high susceptibility), significantly depended on the host cultivar and the field of origin

TABLE 2 | Pairwise-Fst among the six Colletotrichum gloeosporioides populations (222 isolates) collected from Yam (Dioscorea alata) fields in Morne-à-l'eau (French West Indies, Guadeloupe), estimated with 23 polymorphic AFLP-markers.


Below the diagonal: Pairwise-Fst were calculated following the Weir and Cockerham (1984) indications. <sup>∗</sup> indicates significant Fst (p < 0.001) after 1,000 permutations. Above the diagonal: Exact p-values of the exact test of sample differentiation based on haplotype frequencies with Markov chain length of 10,000 steps (Raymond and Rousset, 1995; Goudet et al., 1996).

TABLE 3 | Standardized index of association within Colletotrichum gloeosporioides populations.


(a) rD: standardized index of association calculated on the clone-corrected data set using 23 ploymorphic markers, i.e., with the haplotypes (Haubold et al., 1998; Agapow and Burt, 2001). (b) rD<sup>∗</sup> : standardized index of association calculated on the clone-corrected data set using 16 polymorphic markers where loci had no significant LD with each others; ∗∗indicates significant values with p-values below 0.01 after 1,000 random resampling, ∗∗∗indicates significant values with p-values below 0.001 after 1,000 random resampling.

(**Table 4**). The mean aggressiveness on cv Tahiti was higher for the strains collected on this cultivar (mean Ag-index value of 44.7) than the strains from other host yam cv (mean Ag-index values of 16.6 and 11.7). In the same way, the strains collected on cv Pacala displayed the highest level of aggressiveness on cv Pacala than the other strains (**Table 4**). Finally, the effect of the host cultivar on the values for the initial lesion expansion rates (αNa; **Figure 3**) and the typical lesion area on day 11 (Na; **Figure 3**) was similar to the effect observed on the Ag-index (aggressiveness). The initial lesion expansion rate seemed critical to explain the adaptive evolution of the aggressiveness.

### Patterns of Strain Adaptation to Its Host Cultivar

The pairwise comparison of strain aggressiveness (Ag-indices) on the three host cv populations on the same inoculated cv (**Table 4** and **Figure 3**) revealed clear pattern of local adaptation on cv Tahiti and cv Kabusah with high differences in mean aggressiveness values. Local adaptation to the host cv Pacala (highly susceptible) was also shown, however, the mean aggressiveness values were all higher on cv Pacala than on the other inoculated cultivars which could be seen as a general adaptation to this cultivar. The patterns of local adaptation to the host cultivar were reinforced by the absence of any pattern of differential adaptation of the strains to the cultivar of reference, cv Plimbite. Finally, the pattern of C. gloeosporioides adaptation to its host cultivar, especially to the cultivar Tahiti, was similar to the pattern expected for locally adapted pathogen populations.

### DISCUSSION

Our purpose was to better understand how and in which proportions the migration, recombination and yam quantitative resistance influence the C. gloeosporioides evolution, and more particularly the evolution of C. gloeosporioides pathogenicity. Does the adaptive pattern of C. gloeosporioides correspond to a general adaptation to the most common host cultivar, or to a highly fragmented pathogen population resulting from the adaptation to each cultivar subpopulation (i.e., local adaptation; Gandon et al., 1996; Gandon and Van Zandt, 1998)? Furthermore, can the history of cultivars' succession in Guadeloupe be reflected by the recorded fungal pathogenicity distribution?

### Colletotrichum gloeosporioides Diversity and Sexual Reproduction Imprint

The AFLP analysis revealed high genetic diversity. Such levels for neutral molecular diversity were previously recorded for C. gloeosporioides collected from different yam hosts in three agro-ecological areas of Nigeria (Abang et al., 2006). The indirect evaluation of the reproductive mode, that is the standardized index of multilocus LD, was in accordance with a significant departure from total panmixia, but not reaching complete clonality. Our results thus revealed significant the impact of clonal reproduction on C. gloeosporioides's evolution, which is in concordance with the role of conidia dispersal in the epidemic expansion and the presence of quiescent mycelia in tubers. However, the influence of recombination (sexuality) was not entirely excluded. Besides previous studies reported the presence of sexual reproductive structures of C. gloeosporioides on dead yam leaves at the end of the crop season (Toribio and Jacqua, 1978; Toribio et al., 1980; Degras et al., 1984; Abang et al., 2001, 2006). Therefore, it seems parsimonious to consider that recombination explains the high haplotype diversities observed in every subset. Altogether, the influence of sexual reproduction and the high levels of genetic diversity seem to be general traits in C. gloeosporioides's biology as it was previously suggested for C. gloeosporioides populations damaging yams (Mignouna et al., 2002; Abang et al., 2005, 2006) and damaging strawberry (Xiao et al., 2004). To conclude, even if C. gloeosporioides, pathogenic on D. alata, undergoes weak recombination between two crop seasons, sexuality likely confers this fungus a great opportunity to evolve.

Kabusah (K, moderate susceptibility), Pacala (P, high susceptibility) and Tahiti (T, low susceptibility) were cross-inoculated on these cultivars and on a cultivar of reference (Plimbite, Ref.). (A) Boxplot for the necrosis area, Na. (B) Boxplot for the initial (i.e., day 0–day 6) necrosis expansion rate, αNA were obtained for each cross-inoculation using R-software. (C) Pairwise comparison of the strain aggressiveness inoculated on their local and foreign hosts. Aggressiveness of the strains isolated from cv Kabusah are, respectively, represented with black circles (green dotted lines), from cv Pacala in black square (red dashed lines) and from cv Tahiti in gray triangles (blue lines). Arrows underline the loss or the gain of mean aggressiveness in comparison with the one of strains isolated on cv Pacala, the oldest and susceptible cultivar.

### Colletotrichum gloeosporioides Migration Rate

The haplotype distribution within one population gives insights into its dispersal potential (McDonald, 1997). In the present study, the AFLP genotyping of the haploid fungus C. gloeosporioides revealed a strong population structure: each field sample can be considered as one fungal population. The number of genetic groups revealed with the no a priori analyses and the high Fst-values recorded between field populations, whatever their location or host cultivar, suggest a weak influence of C. gloeosporioides migration between the fields studied during the crop season. Indeed, no haplotypes were shared between fields. Moreover, the strains from adjoining fields (KabA-Tah<sup>A</sup> and KabB-TahB; **Figure 1** and (**Supplementary Figure S3**) belonged to different genetic groups, whereas individuals from the field Pac<sup>C</sup> (Pacala) were genetically close to some of the individuals from the field Kab<sup>A</sup> (Kabusah). Furthermore, out of the 18 multicopy haplotypes, only seven were sampled from neighboring sampling positions whereas 11 were sampled from dispersed sampling positions within the same field (**Supplementary Figure S3**). In other words, for the fungal populations studied here, the dispersal by splashing during the crop season occurred but had a minor impact at a local scale (i.e., not beyond few meters).

In fact, the population structure recorded here (high diversity and strong structure) corresponds to the one expected for a primary inoculum. It is also consistent with the correlation found between the presence of C. gloeosporioides mycelia in the yam tubers planted and the subsequent leaf anthracnose


TABLE 4 | Aggressiveness indices (Ag) of Colletotrichum gloeosporioides populations from six D. alata yam fields tested on four yam cultivars: Kabusah, Pacala, Tahiti and Plimbite (the cultivar of reference for the susceptibility to anthracnose).

Fields KabA/TahA, and KabB/Tah<sup>B</sup> were contiguous (see Figure 1). a, b stands for the groups of significant differences of aggressiveness tested using Student's t-test with a p-value p < 0.01. Tests were performed considering separately each of the four inoculated cultivars and each of the fungal population grouping levels, host cv, field and genetic cluster. <sup>∗</sup> , ∗∗∗, respectively, indicates significant values with p-values below 0.05 and 0.001. Standard errors are specified in brackets.

severity (Green and Simons, 1994). Ultimately, as farmers usually divide yam tubers into pieces (seeds) and save them for the next crop season, when a tuber is infected with one isolate, this isolate will be randomly dispersed within a field where seeds are planted. This fully explains the genetic diversity we observed within and between fungal field-populations. More importantly, the spatial distribution of the haplotypes seemed more likely related to seeds distribution than to fungus dispersal by splashing.

Finally, if dead stems and leaves left on the soil after the previous crop season are a potential source of primary inoculum (Ripoche et al., 2007), the impact of such event cannot be detected in this study, as for all the fields the previous crop was sugar cane. Although no strong influence of migration was detected in this study, the pathogen migration can occur between fields of severely damaged plants (Degras et al., 1984; Green, 1994).

### Colletotrichum Gloeosporioides Local Adaptive Response to Yam Resistances

Previous works already reported the ability of C. gloeosporioides to overcome host resistance (Miles and Lascano, 1997; Chakraborty et al., 1999; Abang et al., 2001; Abang, 2003) or to develop resistance to fungicides (Bayart and Pallas, 1994). The last point of our investigations focused on the adaptation of the field-structured populations to their host quantitative resistance. Are C. gloeosporioides field populations adapted to every host cultivar (general adaptation) or are they specifically adapted to their host cultivar?

We can also draw a parallel between the patterns of adaptation recorded and the "age" and intensity of each cultivar's production in Guadeloupe. As a matter of fact, the cultivar Pacala was the most widespread in Guadeloupe until the end of the 1960s, when its resistance to anthracnose was overcome. Despite its susceptibility to anthracnose, cv Pacala is still one of the most planted cultivars in Guadeloupe (Penet et al., 2016). The cv Plimbite, introduced during the 1970s, was widely cultivated in Guadeloupe but its prevalence decayed when its resistance was overcome in the early 1990s (Ano et al., 2002). The water yam cv Kabusah was the last cultivar introduced at the beginning of the 1990s and has been widely cultivated ever since. Finally, cv Tahiti has been cultivated from the 1960s but its prevalence increased in the 1990s.

The patterns of adaptation described in the present study are concordant with a general adaptation to the formerly highly prevalent hosts (cv Pacala and cv Plimbite) and a local adaptation (i.e., diversifying selection) to the cultivars which became abundant more recently (cv Kabusah and cv Tahiti; **Table 4** and **Figure 3**). Surprisingly a pattern of local adaptation to the old but still widely planted cv Pacala was observed despite a strong general adaptation pattern, revealing an ongoing selective pressure of cv Pacala on C. gloeosporioides populations.

In our sample, strain virulence (not null aggressiveness) on the four yam cultivars prevailed. It is unclear whether this corresponds to an accumulation of virulences within C. gloeosporioides populations or to a core virulence component fixed in all populations. The first hypothesis seems more likely as a former survey on Brazilian C. gloeosporioides populations infecting Stylosanthes showed an increase of fungal race complexity within few years (Chakraborty et al., 2002). The overall virulence of the studied strains reveals that the recorded adaptation for aggressiveness does not depend on the virulence combination. The adaption of C. gloeosporioides to its yam host is quantitative. Besides, we did not find any correlation between strain aggressiveness (Ag-values) against the Kabusah, Tahiti

or Pacala cultivars (**Supplementary Figure S8**). This indicates a certain level of specificity in the C. gloeosporioides–D. alata interaction, even if it seems minor compared to the quantitative aspect. The evolutionary potential of the C. gloeosporioides populations pathogenic on D. alata reinforces the necessity to better understand the aggressiveness evolution in time and space in order to efficiently predict the durability of D. alata resistance to anthracnose. Further investigations of the in-field population dynamics are needed to evaluate the potential of aggressiveness evolution during one crop season. Finally, as previously mentioned, the yam tuber infection can be crucial for the parasite long-distance dispersal and its survival between crop seasons. The capacity of C. gloeosporioides to infect and survive in its host tuber is clearly a determinant aspect of its own fitness thus its evolutionary potential. The pattern of genetic diversity revealed by our study underlines the need to improve the control for seed quality at all cost.

### AUTHOR CONTRIBUTIONS

LF, GJ, and CN contributed to conception and design of the study. LF analyzed the results and wrote the first draft of the manuscript. CN supervised the work. All authors contributed to manuscript revision, read and approved the submitted version.

### FUNDING

This work was funded by the French National Institute of Agronomical Research (INRA) and the Antilles-Guyane region.

### ACKNOWLEDGMENTS

We particularly thank Bruno Le Cam, Bernard Tivoli, Jean Carlier, and Virginie Ravigné for their advices.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018.01554/ full#supplementary-material

FIGURE S1 | Power of the number of markers used to detect genotypes and measure diversity.

### REFERENCES


FIGURE S2 | Evanno's Delta K test.

FIGURE S3 | Detailed distribution of haplotypes at the field scale. The relative positions where the haplotypes where sampled are specified. The genetic groups obtained in the Structure analysis with K = 3, are color coded in orange, green, and blue. Strains belonging to several genetic groups are specified using dashed color code. The letters T, A, K, M, P, F correspond to the strains, respectively, originating from the fields specified in Figure 1, Tah-A, Tah-B, Kab-A, Kab-B, Pac-C, Pac-D. The strains found to be the same haplotype are specified using black boxes, the names of the duplicates are given with the letters a, b, c. No haplotypes were shared between fields.

FIGURE S4 | Principal Component Analysis of the 199 haplotypes. PCA was performed using the adegenet-DAPC software (Jombart and Ahmed, 2011). (A) The PCA shows the dispersion of the 197 haplotypes (23 independent AFLP markers). Colors represent the cultivar of origin: Tahiti, Kabusah, Pacala.

FIGURE S5 | Principal Component Analysis of the 199 haplotypes. PCA was performed using the adegenet software (Jombart and Ahmed, 2011). The PCA shows the dispersion of the 197 haplotypes (23 independent AFLP markers) when the number of groups is 6 or 3. For each group defined by the DAPC analysis, the number of haplotypes sampled in each field specified in Figure 1 (i.e., Tah-A, Tah-B, Kab-A, Kab-B, Pac-C, Pac-D) are specified in the tables.

FIGURE S6 | Principal Component Analysis of the 199 haplotypes. DPCA was performed using the adegenet software (Jombart and Ahmed, 2011). The DPCA shows the dispersion of the 197 haplotypes (23 independent AFLP markers) when the number of groups is, respectively, 4, 5, 7, and 8.

FIGURE S7 | Structure groups for the strains used in the pathogenicity tests. Structure software output based on 23 AFLP markers with the number of genetic groups (K) ranging from 2 to 6. Kab, Tah, Pac, respectively, stands for the host, Dioscorea alata cultivars Kabusah, Tahiti, and Pacala. The letters A, B, C, D stand for the four geographical origins (see Figure 1).

FIGURE S8 | Pairwise correlation between C. gloeosporioides aggressiveness (Ag-Index values) against four D. alata cultivars. The psych package (Revelle, 2018) was used to plot and test the pairwise regressions of the 45 C. gloeosporioides strains' aggressiveness recorded on four different D. alata cultivars, cv Kabusah, Pacala, Tahiti, and Plimbite. Below the diagonal: pairwise scatter plots of Ag-indexes, respectively, assessed on cv Kabusah (KAB), Pacala (PAC), Tahiti (TAH), and Plimbite (PLI). Diagonal: the histogram and lowess locally fit regression line of the Ag-indexes assessed on each cultivar, respectively. The x axis in each scatter plot represents the column Ag-indexes, the y axis the row Ag-indexes. Above the diagonal: the Pearson correlation values. The significances of single correlation were tested with multiple comparisons adjustment using the Holm correction and are specified as ∗∗∗ for p ≤ 0.001 and ∗∗ for p ≤ 0.01.

TABLE S1 | Main characteristics of the six Dioscorea alata fields located in the center of Guadeloupe (Morne-à-l'eau). (A) Sugar-cane (sc) and pasture (p); TahA/B and KabA/B refers to the adjoining fields described in Figure 1; (B) Farm indicates that the tuberpieces (i.e., seeds) planted were harvested during the previous crop-season in the same farm. Imported, means that tuber-pieces planted were obtained from other farms or, more likely imported from Costa Rica.

anthracnose disease of yam (Dioscorea spp.). J. Phytopathol. 153, 137–142. doi: 10.1111/j.1439-0434.2005.00942.x




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

Copyright © 2018 Frézal, Jacqua and Neema. 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.

# Populations of the Beet Cyst Nematode Heterodera schachtii Exhibit Strong Differences in Their Life-History Traits Across Changing Thermal Conditions

Sylvain Fournet\*, Lucile Pellan, Catherine Porte, Christophe Piriou, Eric Grenier and Josselin Montarry

IGEPP, INRA, Agrocampus Ouest, Université de Rennes 1, Le Rheu, France

It is widely accepted that climate has an essential influence on the distribution of species and that temperature is the major abiotic factor that affects their life-history traits. Species with very restricted active dispersal abilities and a wide geographical distribution are thus expected to encompass distinct populations adapted to contrasted local conditions. The beet cyst nematode Heterodera schachtii is a good biological model to study temperature adaptation in populations collected from different environments. Here, we tested the effect of temperature on H. schachtii life-history traits using seven field populations from Morocco, Spain, France, Germany, Austria, Poland and Ukraine. We tested hatching and multiplication rates of each population at different temperatures, as well as hatching rates of each population after storage at different temperatures – simulating survival conditions during the inter-cropping period. Results showed a strong temperature effect on the life-history traits explored. Temperature impact on hatching (at different temperatures and after storage at different temperatures) depended on the origin of populations, separating southern from northern ones. Surprisingly, low temperatures influenced hatching less in southern populations. However, for these populations, a storage period at low temperatures strongly reduce subsequent hatching. Conversely, nematode multiplication was not differentially affected by temperatures, as favorable conditions for the host are also favorable for the parasite. Finally, a significant correlation between the genetic diversity and the level of specialization showed that the less diverse populations were more specialized than the more diverse ones.

Keywords: adaptation, genetic diversity, hatching, specialist/generalist, survival

### INTRODUCTION

Citation:

Edited by: Brigitte Mauch-Mani,

> Reviewed by: Anna-Liisa Laine,

> > Pablo Castillo,

University of Neuchâtel, Switzerland

Consejo Superior de Investigaciones

University of Helsinki, Finland

Científicas (CSIC), Spain \*Correspondence: Sylvain Fournet sylvain.fournet@inra.fr

Specialty section: This article was submitted to Plant Microbe Interactions, a section of the journal Frontiers in Microbiology Received: 02 July 2018 Accepted: 31 October 2018 Published: 16 November 2018

Fournet S, Pellan L, Porte C, Piriou C, Grenier E and Montarry J (2018) Populations of the Beet Cyst Nematode Heterodera schachtii Exhibit Strong Differences in Their Life-History Traits Across Changing Thermal Conditions. Front. Microbiol. 9:2801. doi: 10.3389/fmicb.2018.02801

There is no doubt that climate has a major influence on the natural distribution of species (Pearson and Dawson, 2003) and that temperature is the major abiotic factor that affects species phenology (Scranton and Amarasekare, 2017). In a context of ongoing climate change, predicting the temperature impact on life-history traits, such as fecundity, development and survival, and thus on the distribution of species is a very challenging scientific question with implications for conservation biology, ecology, agronomy and pest management.

Species characterized by restricted active dispersal ability and a wide geographical distribution are expected to comprise distinct populations adapted to contrasted local conditions

(Davis and Shaw, 2001). Plant pathogens, being often distributed across a wide range of climates, are good biological models to study the temperature adaptation of populations from different environments. For instance, patterns of local adaptation to temperature have been highlighted in natural (Laine, 2008) and agricultural host-pathogen systems (Zhan and McDonald, 2011; Mboup et al., 2012; Mariette et al., 2016).

The life cycle of many plant pathogens can be split into two alternating phases: the epidemic period, onto or into the host, and the inter-epidemic period, often outside the host. Several investigations have shown the necessity to take into account the inter-epidemic period in order to better describe the evolutionary trajectories of plant pathogens (e.g., Pfender and Vollmer, 1999; Abang et al., 2006; Montarry et al., 2007; Sommerhalder et al., 2011; Suffert et al., 2015; Pasco et al., 2016). Interestingly, temperature regimes are likely to differ across locations much more during the inter-epidemic periods than during the epidemic periods, because optimal temperatures for host growth is quite similar whatever the geographical locality.

The beet cyst nematode Heterodera schachtii, which is regarded as the most important pest in sugar beet production worldwide (Müller, 1999; Amiri et al., 2002), is globally distributed with populations living in highly contrasted environments (see Subbotin et al., 2010 for a review). In European wild beets, H. schachtii is relatively rare in south of Spain and Portugal, but more frequent from the northern coast of Spain to Denmark (Gracianne et al., 2014). In European sugar beet fields, H. schachtii is widely present from east to west, from Portugal to Ukraine, and from south to north, from Morocco to Finland (Subbotin et al., 2010). Also, as for all plant parasitic nematodes, H. schachtii has only very limited active dispersal capabilities in the soil (Wallace, 1968), and temperature is one of the most influential environmental factor affecting key steps of its biological cycle (Trudgill et al., 2005; Kakaire et al., 2012; Kaczmarek et al., 2014; Vandenbossche et al., 2015). Indeed, temperature can act on all or part of life-history traits involved in embryogenesis, hatching and development after infection and survival (Wallace, 1955; Thomason and Fife, 1962; Jones, 1975; Griffin, 1988). For H. schachtii, such influence is well known and extensively described in many previous studies, but on a limited number of field populations coming from north European fields (e.g., Vandenbossche et al., 2015). Therefore, H. schachtii populations may be thus assumed to be differentially adapted to local temperatures, both for traits involved in development within their host plant (sugar beet), and for traits involved in their survival in the soil when plants are not present.

Here, we explored the effect of temperature on H. schachtii life-history traits using seven field populations from Morocco, Spain, France, Germany, Austria, Poland, and Ukraine. We tested the hatching and the multiplication rates of each population at different temperatures, as well as the hatching rate of each population after storage at different temperatures, simulating survival conditions during the inter-cropping period. Finally, to test the hypothesis that greater genetic diversity is associated with less ecological specialization, we calculated the level of specialization to the temperature and assessed the genetic diversity using microsatellite markers for each population.

## MATERIALS AND METHODS

### Nematode Populations

The seven H. schachtii populations were sampled between 2014 and 2016 in a single field in seven different countries and localities representing contrasted climatic zone : Morocco (Mor; Al Aroui ), Spain (Spa; Lebridja), France (Fra; Prunay), Germany (Ger; Schlanstedt), Austria (Aus; Tadten), Poland (Pol; Radlowek), and Ukraine (Ukr; Teofipol). Each population corresponded to a single soil sample constituted of a minimum amount of 2 kg and a minimum of 10 elementary samples collected randomly in the field between 0 and 20 centimeters deep. The cysts extracted from each of these seven soil samples were multiplied one time in standardized conditions, (20◦C, room chamber, 16:8) for two to four generations on the cultivar Ardan. Newly formed cysts were then stored at 5◦C in moistened sandy soil.

### Genetic Diversity

The seven H. schachtii populations were genotyped using eight microsatellite markers developed by Montarry et al. (2015): Hs33, Hs36, Hs55, Hs56, Hs68, Hs84, Hs111, and Hs114. Between 21 and 39 individuals per population were successfully genotyped using only one larva (i.e., 1 s-stage juvenile) per cyst, in order to limit biases due to sibling relationships. DNA of each larva was extracted with a procedure using sodium hydroxide and proteinase K, following Boucher et al. (2013).

Three multiplex panels were defined to genotype the 207 individuals at the eight loci. PCR was performed in a 5 µL volume containing 1X of Type-it Microsatellite PCR kit, 0.4 µM of primer mix and 1 µL of template DNA. Cycling conditions included an initial denaturation at 95◦C for 5 min, followed by 30 cycles of denaturation at 95◦C for 30 s, annealing at 58◦C for 90 s and extension at 72◦C for 30 s, followed by a final extension at 60◦C for 30 min. PCR products were then diluted to 1:10 in sterile water and 3 µL of this dilution were mixed with 0.05 µL of GeneScan 500 LIZ Size Standard (Applied Biosystems) and 5 µL of formamide (Applied Biosystems). Analyses of PCR products were conducted on a ABI Prism <sup>R</sup> 3130xl sequencer (Applied Biosystems).

Allele sizes were determined with the automatic calling and binning module of GeneMapper v4.1 (Applied Biosystems), and complemented by a manual examination of irregular results. Samples showing dubious genotypes were tested again.

Genetic diversity in each H. schachtii population was estimated using GENETIX 4.05.2 (Belkhir et al., 2004) as the unbiased gene diversity (Hnb) (Nei, 1978). Hnb corresponds to the average probability across loci to draw at random different alleles in the same population.

### Hatching at Different Temperatures

For the seven H. schachtii populations, larvae emergence was investigated at three temperatures: 11, 17, and 23◦C, which cover the range of temperatures encounter by European H. schachtii populations, from the North to the South, during the cropping period. For each population and each temperature, six

independent replicates were performed. Each replicate consisted of a pool of four cysts, calibrated in size between 450 and 500 µm. Cysts were placed on a sieve of 250 µm, allowing the active passage of larvae. Each sieve was put in one well of a 24-well plate containing root exudates of sugar beet (cv. Ardan). Plates were placed in the dark in three different climate chambers set at the respective temperatures. Temperature within each climate chamber was recorded every hour using thermo-tracers (ThermoTracer, Oceasoft, Montpellier, France – **Supplementary Figure S1**). The number of hatched larvae was counted after 1, 2, 3, 4, 7, 11, 15, 19, 23, 28, and 33 days, the exudate being changed at each date. The number of unhatched eggs was assessed at the end of the experiment in order to calculate the percentage of hatching.

### Multiplication at Different Temperatures

For the seven H. schachtii populations, H. schachtii multiplication was investigated at three temperatures: 11, 17, and 23◦C. For each population and each temperature, 27 independent replicates were performed, i.e., 27 sugar beet plants of the cultivar Ardan. Plants were grown in a 4:1 sand-kaolin mixture watered with a nutrient solution (Hakaphos: NPK 15/10/15). Kaolin was used to aerate the sand and thus to promote the mobility of H. schahctii larvae. Each plant was inoculated at the two-leaf stage with 1 mL of a suspension of larvae concentrated at 500 larvae.mL−<sup>1</sup> . Plants were placed in three different climate chambers set at the respective temperatures with a photoperiod of 16 h. Temperature within each climate chamber was recorded every hour using thermo-tracers (ThermoTracer, Oceasoft, Montpellier, France – **Supplementary Figure S1**). Plants were cut at the end of the cycle, when white females were well formed and contained eggs, i.e., after 4 weeks at 23◦C, after 5 weeks at 17◦C, and after 9 weeks at 11◦C. Plants were kept several weeks in their climatic chamber until the maturation of cysts, i.e., three, four and six additional weeks at 23, 17, and 11◦C, respectively. Newly formed cysts were extracted by filtering the soil through two sieves (800 and 250 µm) and counted under a stereomicroscope.

### Hatching Following Storage at Different Temperatures

Each H. schachtii population was multiplied one generation on the sugar beet cultivar Ardan and newly formed cysts were then stored during 2 months at four temperatures: −3, 4, 10, and 25◦C, which cover the range of temperatures encounter by European H. schachtii populations, from the North to the South, during the inter-cropping period. For each population and each temperature, six independent replicates were performed. Each replicate consisted of a pool of four cysts, calibrated for their size between 450 and 500 µm, and placed in boxes containing 33 g of sand and 7 mL of water. Boxes were stored in the dark in incubators set at the respective temperatures. Temperature within each incubator was recorded every hour using thermo-tracers (ThermoTracer, Oceasoft, Montpellier, France – **Supplementary Figure S1**). After the storage period, each pool of four cysts was placed on a sieve of 250 µm, allowing the active passage of larvae, and each sieve was put in a 24-well plate containing root exudates of sugar beet (cv. Ardan). Plates were placed in the dark at 20◦C ( ± 1) in an incubator. The number of hatched larvae was counted after 1, 2, 3, 4, 7, 11, 15, 19, 23, and 28 days, the exudate being changed at each date. The number of unhatched eggs was assessed at the end of the experiment in order to calculate the percentage of hatching.

### Statistical Analysis

In order to take into account the hatching dynamic in its entirety, and not only the final point of the hatching curve, the Area Under the Hatching Curve (AUHC) was calculated for each replicate. For each population, and for each measured traits (i.e., AUHC, number of produced cysts and AUHC after storage), the temperature effect was tested through a one-way ANOVA and multiple comparisons of means were done with the Tukey contrasts test (α = 0.05). Normality and homogeneity of variances were checked using the Shapiro–Wilk and the Leven tests, respectively.

Standardized niche breadth (BA) – a parameter estimating the level of ecological specialization of a species (or a population) – was calculated for each H. schachtii population using the standardized index developed by Levins (1968):

$$B\_A = \frac{1}{n-1} \left[ \frac{1}{\sum p\_i^2} - 1 \right]$$

where p<sup>i</sup> is the proportion of the measured fitness trait into the environment i and n the total number of tested environments. In this study, n = 3 for the hatching and the multiplication at different temperatures (11, 17, and 23◦C) and n = 4 four for the hatching following storage at different temperatures (−3, 4, 10, and 25◦C). The standardized niche breadth varies from 0 to 1. A low niche breadth points to a high level of specialization (i.e., a specialist species) and an high niche breadth to a low level of specialization (i.e., a generalist species).

The relationships between the level of specialization (BA) and the genetic diversity (Hnb) were tested with the Pearson's product-moment correlation. All statistical analyses were performed using the R software version 3.3.3 (R Development Core Team, 2017).

## RESULTS

### Hatching and Multiplication at Different Temperatures

ANOVAs revealed a very strong temperature effect for five H. schachtii populations, Fra, Ger, Aus, Pol, and Ukr (**Table 1** and **Figure 1**), showing that the hatching pattern was very different at 11◦C: larvae emergence was lower at 11◦C than at 17 or 23◦C, in terms of AUHC (**Figure 1**) but also in terms of cumulative hatching at the end of the experiment (**Supplementary Figure S2**). Conversely, the temperature effect was very small or even non-significant for the two remaining populations, Mor and Spa (**Table 1** and **Figure 1**), for which larvae emergence (i.e., AUHC – **Figure 1** – and cumulative hatching at the end of the experiment – **Supplementary Figure S2**) was similar whatever the temperature.


TABLE 1 | Results from the analyses of variance (ANOVAs) assessing the temperature effect on the three explored life-history traits (AUHC at different temperatures, on the number of cysts produced at different temperatures and on AUHC after storage at different temperatures) for each Heterodera schachtii populations.

The statistically significant effects are indicated as <sup>∗</sup>P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001.

Regarding the number cysts produced on sugar beet plants at different temperatures, ANOVAs revealed a temperature effect for all populations, except for the French one, showing that H. schachtii populations produced more cysts at the intermediary temperature, i.e., 17◦C (**Table 1** and **Figure 2**).

### Hatching Following Storage at Different Temperatures

After a prior exposure for 2 months to different storage temperatures, the hatching experiment was performed at 20◦C (±1) because at this temperature larvae of all populations hatched very well, as showed by the previous experiment (**Figure 1**). Larvae emergence was reduced after storage at −3 ◦C for all populations except the German one (**Table 1** and **Figure 3**). The difference between the storage step at −3 ◦C and the three other temperatures (4, 10, and 25◦C) was much higher for the two Mediterranean populations (Mor and Spa) than for the five others (Fra, Ger, Aus, Pol, and Ukr), both for the AUHC (**Figure 3**) or the cumulative hatching at the end of the experiment (**Supplementary Figure S3**).

### Relationship Between Genetic Diversity and Ecological Specialization

Regarding the hatching (AUHC) at different temperatures, the standardized niche breadth B<sup>A</sup> – a parameter to measure how specialized a population is within a given environment – highlighted the generalist status of H. schachtii populations from Morocco and Spain (wide niche breath, 0.98 < B<sup>A</sup> < 0.99) and

FIGURE 2 | Number of cysts obtained from a multiplication at different temperatures (11, 17, and 23◦C) for each H. schachtii population (mean values ± standard deviations). Letters represent the homogenous groups identified by the Tukey contrasts test at the 5% threshold.

FIGURE 3 | Area Under the Hatching Curve (AUHC) after storage at different temperatures (–3, 4, 10, and 25◦C) for each H. schachtii population (mean values ± standard deviations). Letters represent the homogenous groups identified by the Tukey contrasts test at the 5% threshold.

TABLE 2 | Genetic diversity (Hnb) and standardized niche breadth (BA) calculated on AUHC at different temperatures, on the number of cysts produced at different temperatures and on AUHC after storage at different temperatures for each H. schachtii populations (Pop).


n indicates the number of genotyped individuals.

showed a lower niche breadth for the five other populations (0.58 < B<sup>A</sup> < 0.89), when taking temperatures as different environments (**Table 2**). Moreover, there was a significant correlation between the genetic diversity and the level of specialization (cor = 0.758; P = 0.048), showing that the less diverse populations were more specialized than the more diverse ones, and vice versa (**Figure 4**).

Regarding the multiplication (number of cysts) at different temperatures and the hatching (AUHC) after storage at different temperatures, the standardized niche breadth B<sup>A</sup> highlighted the generalist status of all H. schachtii populations (0.97 < B<sup>A</sup> < 0.99 for the number of cysts and 0.92 < B<sup>A</sup> < 0.99 for the AUHC, **Table 2**) and no correlation with the genetic diversity (cor = −0.373; P = 0.410 for the number of cysts and cor = −0.464; P = 0.294 for the AUHC).

### DISCUSSION

Although the hatching and development of the beet cyst nematode H. schachtii have been already studied (e.g., Griffin, 1988), far less is known about the survival abilities of this species under different climate conditions that can be encountered during the inter-cropping season. The main objectives of the present work were thus to test the effect of temperature and to compare the level of specialization of different H. schachtii populations for key traits involved in survival and reproduction: hatching, development in the host plant and hatching after a survival period.

Low temperatures strongly reduced both the hatching pattern and the cumulative hatching rate of five populations (France, Germany, Austria, Poland, and Ukraine): at 11◦C, hatching was slowly and was incomplete. Previous studies already showed the same behavior in populations from the same geographical areas (e.g., Vandenbossche et al., 2015). Conversely, temperature had only a small effect on the two southern populations, from Spain and Morocco. This was a surprising result: one could imagine that northern populations should be less constrained by lower temperatures than southern ones, which faced much higher temperature levels over the year. In an unexpected way, temperature appeared thus to act differently according to the geographical origin of populations in the sequence leading to the hatch of larvae. Perry and Gaur (1996) showed that differences related to the host, for instance the age of the host plant, on which the nematodes multiplied could strongly affect their subsequent hatching rate. In our study, cysts all came from a single multiplication cycle on the same sugar beet cultivar, and experiments were performed in controlled conditions. The differences observed in hatching behavior could thus not be assigned to the conditions for obtaining cysts used in this study, but would rather be a clear answer of each population to the thermal conditions applied in the experiment. Indeed, cyst nematode species are all obligate parasites, and their reproductive success mainly depends on the synchronization between hatching and host presence. Consequently, it seems that natural selection has optimized the hatching response to enhance fitness in a different way for northern and southern populations (Spain and Morocco). For these last populations, the most important factor driving hatching may rather be the occurrence of favorable conditions (moisture and presence of root exudates), rather than temperature only.

Regarding the development in the host plant, thermal conditions did not affect each population differentially. In all cases, the number of newly formed cysts matched the usual shape of a thermal performance curve (Schulte et al., 2011), and all populations exhibited higher performances at 17◦C. This is consistent with what was already described in the literature for field populations (e.g., Griffin, 1988) and for wild populations of H. schachtii (Gracianne et al., 2014). This optimum corresponds also to the optimal temperature for sugar beet cropping. Indeed,

even if sugar beet is grown throughout Europe, cropping is always realized under conditions that ensure optimal development and yields. These conditions correspond to the period from April to September in the majority of European countries, but from November to May in Morocco and in the south of Spain.

Most studies only considered the effect of temperature on the multiplication of plant parasites during the epidemic phase, i.e., during the interaction with the host. Here, we also investigated its impact on the survival of the progeny inside the cysts, because the ability to survive when host plants were absent is also an important fitness component submitted to selection (Roff, 1992). In our case study, temperature during the inter-epidemic phases are much more variable than those encountered during the cropping period: they correspond either to very hot summer conditions for the southern populations and to cold winter conditions for the populations of northern and central Europe. In both cases, these periods are characterized by unfavorable conditions for the growth of host plants. We thus exposed cysts of each H. schachtii population during 2 months to temperatures from −3 to 25◦C, which correspond to the range encountered in natural conditions. The results confirmed the existence of two distinct groups of populations: one (Morocco, Spain) for which survival at a low temperature (−3 ◦C) reduced strongly the subsequent hatching rate, and a second one (all others populations) where survival temperatures had only a limited impact on subsequent hatching. At this point we could therefore conclude, rightly as these populations never encountered such low temperatures, that both southern populations were thus clearly maladapted to survival under cold conditions, while northern populations seemed adapted to survive over a larger range of temperatures, including cold ones.

However, closer examination of the mean standard deviations leads to a more nuanced conclusion. Indeed, southern populations both exhibited greater variation for this trait than northern ones (**Figure 3** and **Supplementary Figure S3**) and seemed to contain some individuals that could survive cold conditions and others that could not (see **Supplementary Figure S4** showing the hatching curves for each independent replicate). This observation makes sense in an evolutionary context, where northern populations can be seen as the result of a gradual northward (re)colonization process of Europe after the Last Glacial Maximum (LGM), with the selection among southern individuals of those able to survive the coldest temperatures. Several studies make this hypothesis relevant. First, the south of the Iberian Peninsula is well known as an historical refuge, climatically stable during geological times (Rodrıguez-Sanchez et al., 2008) and therefore during the LGM which corresponded to the latest coldest conditions encountered in Europe. Second, two recent studies tend to prove that wild populations of H. schachtii and of its wild hosts Beta maritima recolonized the north of Europe following a northward process of migration from the south of the Iberian Peninsula (Leys et al., 2014; Gracianne, 2015; respectively). After the LGM, the following huge increase of temperature pushes northward favorable conditions to the development of host plants and nematodes and make possible the recolonization of northern Europe. While our study was performed with field populations that share only part of their evolutionary history with wild ones, our results matched this scenario rather well. Indeed, we also observed a northward decrease of genetic diversity in our field populations, as expected in the case of sequential founder events associated to recolonization processes and limited gene flow among population (Austerlitz et al., 1997).

As a last challenge, we explored the link between genetic diversity in populations and their ecological specialization, testing the hypothesis that greater genetic diversity is associated with less ecological specialization. Such a relationship was only demonstrated so far between species: in 11 of 14 studies identified by Li et al. (2014), the specialist species showed lower genetic diversity than related generalists. Here, a similar trend can be observed between populations: Spanish and Morrocan populations were both generalists, with a high niche breath and a higher genetic diversity, while Northern populations appeared to be more specialized and less diverse. The generalist character fitted the independence of the hatching and development according to thermal conditions. The association of this generalist trait with higher genetic diversity fitted also well with Gracianne's scenario, which assumes that Spain and more southern regions kept favorable climate conditions during the LGM, promoting the maintenance of a huge genetic diversity in these areas. The less diversity and the higher specialization to cold climate conditions we observed in northern populations fitted again well with this scenario. The northward recolonization of Europe was associated with the selection for a better capability to survive and a less susceptibility to hatch at low temperatures, maximizing the synchronization between hatching and host presence.

Further investigations are now needed, using wild H. schachtii populations, to determine whether the counter-selection of the capability to hatch at low temperatures and the selection of the capability to survive at low temperatures are the result of a fast evolution process in the cultivated compartment (i.e., an adaptation to crop practices in northern Europe), or of a longer process in the wild, associated to the northward recolonization of Europe. Anyway, the high tolerance to contrasted survival conditions observed in northern Europe field populations explains well the presence of H. schachtii all around the world where sugar beet is grown (Subbotin et al., 2010). Furthermore, in the framework of global climate change, our data imply that northern populations of H. schachtii should not be strongly affected for both survival and development by global warming.

### DATA AVAILABILITY

All data used in this article are available at data.inra.fr (doi: 10.15454/ZKGTAH).

### AUTHOR CONTRIBUTIONS

LP, CaP, and ChP performed the experiments according to a protocol elaborated jointly by SF, EG, and JM. SF and JM analyzed the data. SF, EG, and JM wrote the text and prepared the figures. All authors edited the paper and have approved the current version.

### ACKNOWLEDGMENTS

fmicb-09-02801 November 16, 2018 Time: 12:44 # 8

We gratefully acknowledge ITB, SES, KWS, Strube, and SYNGENTA who kindly provided the different Heterodera

### REFERENCES


schachtii populations. Dr. D. Andrivon is acknowledged for comments on earlier draft of this manuscript.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.02801/full#supplementary-material



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

Copyright © 2018 Fournet, Pellan, Porte, Piriou, Grenier and Montarry. 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.

# Role of Temperature and Coinfection in Mediating Pathogen Life-History Traits

Elise Vaumourin\* and Anna-Liisa Laine

Research Centre for Ecological Change, University of Helsinki, Helsinki, Finland

Understanding processes maintaining variation in pathogen life-history traits is a key challenge in disease biology, and of importance for predicting when and where risks of disease emergence are highest. Pathogens are expected to encounter tremendous levels of variation in their environment – both abiotic and biotic – and this variation may promote maintenance of variation in pathogen populations through space and time. Here, we measure life-history traits of an obligate fungal pathogen at both asexual and sexual stages under both single infection and coinfection along a temperature gradient. We find that temperature had a significant effect on all measured life-history traits while coinfection only had a significant effect on the number of sexual resting structures produced. The effect of temperature on life-history traits was both direct as well as mediated through a genotype-by-temperature interaction. We conclude that pathogen life-history traits vary in their sensitivity to abiotic and biotic variation in the environment.

Keywords: abiotic interactions, biotic interactions, life-history evolution, overwintering success, pathogen evolution, Plantago lanceolata, Podosphaera plantaginis

### INTRODUCTION

Pathogens are a major threat to human health and food security (Strange and Scott, 2005; Jones et al., 2008). The success of pathogens may be attributed to their evolutionary potential which allows them to emerge, adapt, and persist in space and time within their host populations (Mundt, 2002; Friesen et al., 2006). Hence, understanding processes maintaining variation in pathogen life-history traits is a key question in evolutionary ecology and a major challenge for the design of disease control programs (Galvani, 2003; Grenfell et al., 2004). Environmental variation may be a powerful process maintaining variation in pathogen populations if infection outcomes depend on conditions directly or through genotype-by-environment interactions, and if variation is spatially and/or temporally structured (e.g., heterogeneous landscapes and seasonality). To understand the role of environmental variation in maintaining variation in pathogen populations, to date most studies have focused on the abiotic environment (e.g., climatic conditions, Harvell et al., 2002). Recently it has become evident that pathogens often occur simultaneously with other strains or species of pathogens colonizing the same host (Tollenaere et al., 2016). Thus, biotic variation in the co-occurring pathogen community may be an overlooked component of the processes maintaining variation in pathogen populations.

#### Edited by:

Cindy E. Morris, Genetics and Improvement of Fruit and Vegetables (INRA), France

#### Reviewed by:

Remco Stam, Technische Universität München, Germany Frédéric Suffert, Institut National de la Recherche Agronomique (INRA), France

> \*Correspondence: Elise Vaumourin elise.vaumourin@helsinki.fi

#### Specialty section:

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

Received: 15 June 2018 Accepted: 26 October 2018 Published: 20 November 2018

#### Citation:

Vaumourin E and Laine A-L (2018) Role of Temperature and Coinfection in Mediating Pathogen Life-History Traits. Front. Plant Sci. 9:1670. doi: 10.3389/fpls.2018.01670

The abiotic environment has been shown to affect the evolutionary trajectories of pathogen life-histories (Blanford et al., 2003; Fels and Kaltz, 2006). Variation in temperature represents one of the most ubiquitous sources of environmental variation, and is known to greatly affect biochemical, physiological and behavioral processes of organisms. Pathogens with a free transmission stage are considered particularly vulnerable to variation in temperature (Truscott and Gilligan, 2003). In many pathosystems, temperature has been shown to affect pathogen ability to establish or maintain infection, its latency as well as its aggressiveness (e.g., Burdon, 1987; Thomas and Blanford, 2003; Fels and Kaltz, 2006). There is increasing evidence that the effect of temperature on pathogen fitness may be mediated through genotype-by-environment (G × E) interactions, suggesting that adaptation to biotic and abiotic habitats may be strongly linked (Ferguson and Read, 2002; Thomas and Blanford, 2003; Price et al., 2004; Mitchell et al., 2005; Fels and Kaltz, 2006). For example, it has been shown that the diversity Arcobacter populations is modulated by a temperature gradient (Fisher et al., 2014).

More recently, in addition to the surrounding abiotic environment, the biotic environment has been also highlighted to impact the maintenance of the variation in pathogen populations. For example, epidemic outcomes have been shown to change when multiple pathogens are present on a same host (Susi et al., 2015a). Strains under coinfection are generally expected to compete for the same limited resources provided by the host. If this process was acting alone, it could favor higher host exploitation rates (i.e., higher virulence) under coinfection than under single infection (Alizon, 2013). For many pathogens coinfection is also the pre-requisite for sexual reproduction, and thus changes in pathogen growth rates under coinfection may also represent facilitation to increase the probability of generating of new allelic combinations between the coinfecting strains (Otto, 2009; Carter et al., 2013). Thus, coinfection may largely contribute to maintain the genetic variation in pathogen life-history traits and can speed up the adaptation to environmental change (Becks and Agrawal, 2012).

Both abiotic and biotic environments may have strong impacts on infection outcomes, but little is known about their relative importance, and whether outcomes of coinfection are mediated by temperature (but see Marçais et al., 2017). In this study, we investigate how pathogen life-history traits that constitute both asexual and sexual stages are affected by the biotic environment (i.e., coinfection) and abiotic environment (i.e., temperature). Despite its ecological importance, the sexual stage is difficult to study for many pathosystems. Our study is focused on the interaction between Podosphaera plantaginis – Plantago lanceolata (powdery mildew – ribwort plantain). Given that this pathogen completes its entire life-cycle on the surface of the host plant, we are able to monitor the life-history traits visually in a non-destructive manner. Another advantage of this pathosystem is that P. plantaginis is a homothallic fungi (i.e., self-fertile, Tollenaere and Laine, 2013) so the production of the sexual structures (i.e., chasmothecia) can be achieved under both single strain infections as well as in coinfections. Thus, we can evaluate the impact of coinfection and temperature on this critical life-history trait affecting overwinter survival.

### MATERIALS AND METHODS

### The Pathosystem: Podosphaera plantaginis – Plantago lanceolata

We focused our study on the powdery mildew, Podosphaera plantaginis (Erysiphales, Ascomycota), which is an obligate fungal pathogen naturally infecting host plant Plantago lanceolata, the ribwort plantain. The epidemic cycle of the powdery mildew starts with the germination of a spore on a susceptible host resulting in lesion where clonal spores which are winddispersed are produced. To survive the winter, resting structures (chasmothecia) are produced. Each resting structure contains eight sexually generated ascospores, which initiate new infections in the spring upon their release. In P. plantaginis chasmothecia production can be achieved via haploid selfing of pure strains as well as outcrossing between two different strains infecting the same host plant (Tollenaere and Laine, 2013).

The dynamics of P. plantaginis in its fragmented host population network have been intensively studied in the Åland Islands (50 km × 70 km area) in southwest Finland since 2001 (Ojanen et al., 2013). The study system consists of approximately 4 000 host populations that are surveyed annually for the presence of P. plantaginis. The powdery mildew infection is visually conspicuous as whitish mycelia on the leaves. Previous studies have shown that prevalence of the pathogen is low in this system, with 2–17% of the host populations being infected each year in the metapopulation. Podosphaera plantaginis persists as highly dynamic metapopulation with frequent extinctions and (re)colonizations (Laine and Hanski, 2006; Jousimo et al., 2014). Genotyping has revealed considerable genetic diversity in this pathogen metapopulation, with most of the strains found in only one or few localities with a small subset of strains being common across multiple host populations (Tollenaere et al., 2012). The three strains chosen for our experiment – S1 (genotype i1), S2 (genotype 4000) and S3 (genotype 876-1) – are temporally and spatially widely distributed in the Åland metapopulation (**Supplementary Figure 1**).

Prior to the experiment, we carried out repeated cycles of inoculations to obtain adequate stocks of sporulating fungal material for the inoculation trials. The inoculations were done on petri dishes on detached host leaves of a P. lanceolata genotype (490-15) that has been characterized as broadly susceptible during previous experimental and maintenance work. For the experiment we used conidial chains from lesions that were 16–18 days old. To have sufficient leaf material for the experiment, we produced multiple clones by placing mother plants growing in pots with a perforated bottom in pots filled with vermiculite. After 9 weeks the mother plant was cut from the roots it had developed in the pot of vermiculite. Roots in the vermiculite yielded new rosettes, which were planted into pots with sand-rich humus. The plants were grown under glasshouse conditions, with 16 h of light and a temperature of +22◦C.

### Measuring Asexual and Sexual Pathogen Life-History Traits Under Single Versus Coinfection Along a Temperature Gradient

In order to investigate how pathogen life-history traits are impacted by the biotic environment (i.e., coinfection) and by the abiotic environment (i.e., temperature), we conducted an inoculation experiment where we recorded life-history traits of strains on their own (single infection) or with another strain (coinfection). We measured pathogen life-history traits under three different temperatures, +17◦C, +20◦C and +23◦C. This range represent typical growing season temperatures that P. plantaginis experiences in the Åland archipelago (Laine, 2008). We measured two life-history traits that constitute the asexual stage (i.e., affecting pathogen growth): time from inoculation to germination and time from germination to sporulation. We also quantified three life-history traits of the sexual stage (i.e., affecting pathogen overwinter survival): time from sporulation to mature chasmothecia, total number of chasmothecia and proportion of viable chasmothecia. Given that this pathogen completes its entire life-cycle on the surface of the host plant, we were able to monitor life-history traits visually in a non-destructive manner. All inoculations were performed on the broadly susceptible host genotype (490-15).

To provide single vs. coinfection comparisons under different temperature conditions, each of the three strains was inoculated both individually as well as with another strain in a fully crossed design under the three different temperatures (+17◦C, +20◦C and +23◦C). A cross-inoculation of a strain with itself is considered as a single infection and a cross-inoculation of a strain with another strain is considered as a coinfection. Plantago lanceolata leaves were placed on a moist filter paper in a Ø 9 cm Petri dish. Three conidial spore chains of both strains (i.e., six conidial chains at each inoculation site) were carefully placed on the same spot of a P. lanceolata leaf by single hair inoculation technique (Tollenaere and Laine, 2013). On each leaf we inoculated three distinct spots resulting in 54 sites of inoculation for each strain pairing per temperature. All inoculated leaves inside Petri dishes were kept at their respective temperatures with a 16L/8D photoperiod in growth chambers. We included non-inoculated leaves as negative controls. All P. lanceolata leaves used in the experiment were harvested from same-aged clones that were maintained in the greenhouse.

For each inoculated site, using a dissecting microscope, we measured daily the life-history traits related to the pathogen growth. Then, 14 days after the appearance of the first mature chasmothecia, the number of total chasmothecia produced were counted under a dissecting microscope and the viability of approximately 10 mature randomly collected chasmothecia were measured. To measure their viability, we have used the vital stain fluorescein diacetate (FDA) method (Firstencel et al., 1990; Ficke et al., 2002; Vági et al., 2016). This method consists of gently crushing chasmothecia on microscope slides in a droplet of a solution containing FDA (10 µg/mL). Slides were then scored for fluorescing ascospores under an epifluorescence microscope. FDA is lipophilic, membrane-permeable, non-fluorescent and it is hydrolyzed onto polar fluorescent molecules in the cytoplasm

of living cells. These molecules remain in the cytoplasm of viable ascospores with intact cell membranes and exhibit intense green fluorescence when examined under blue light (465–495 nm) thereby signaling chasmothecia viability. A chasmothecium was scored as being viable if it contained any ascospores that exhibited intense green fluorescence.

### Statistical Analyses

fpls-09-01670 November 20, 2018 Time: 12:50 # 4

To understand how temperature and coinfection affected variation in the asexual life-history traits - time from inoculation to germination, time from germination to sporulation and time from sporulation to mature chasmothecia – we used survival models as implemented in R-package survival (Therneau, 2015) with Cox Proportional Hazards model (Cox, 1972). All survival models had temperature, first strain identity, coinfection and their respective interactions with the temperature as fixed explanatory effects. The identity of the second coinfecting strain and the leaf were considered as random factors.

To understand how temperature and coinfection affected variation in the sexual life-history traits – total number of chasmothecia and proportion of viable produced chasmothecia – we fitted generalized linear mixed models. For the model estimating total number of produced chasmothecia we defined a normal error structure and an identity link function, and for the model analyzing proportion of viable chasmothecia we defined a binomial error structure and a logit link function. Then for each model, the temperature, first strain identity, coinfection and their respective interactions with the temperature were fixed explanatory effects and the second coinfecting strain and leaf were considered as random factors.

### RESULTS

We found that all measured life-history traits are significantly affected by temperature (**Figures 1**, **2** and **Table 1**). The average timing of each life-history event at each temperature and strain

TABLE 1 | Results of the GLMMs analyzing infection development of Podosphaera plantaginis measured under single vs. coinfection under a temperature gradient.


Statistically significant results are shown in bold.

is summarized in **Table 2** and in **Supplementary Figure 2**. At the lowest temperature, +17◦C, we find all sexual life-history traits to perform better than at higher temperatures (**Table 2**). For example, for strain S1 the time from first sporulation to first mature chasmothecia is 2.1 times faster at +17◦C than at +23◦C, 2.1 times faster for strain S2 and 2.0 times faster for the strain S3. The total number of produced chasmothecia is 73.3 times higher at +17◦C than at +23◦C for strain S1, 13.8 times higher for strain S2, and 10.0 times higher for strain S3. The proportion of viable chasmothecia is 9.4 times higher at +17◦C than at 23◦C for the strain S1, 5.3 times higher for the strain S2 and the same for the strain S3. For the asexual life-history traits studied – time from inoculation to germination and time form germination to sporulation – the effect of temperature tended to be the opposite. Strains were typically faster to germinate and launch spore production at +20◦C and +23◦C than at +17◦C (**Table 2** and **Figure 1**).

Coinfection did not have a significant effect on the asexual life history traits measured. However, we found that under coinfection significantly more chasmothecia are produced (**Figure 2** and **Tables 1**, **3**). For the strain S1 2.0% chasmothecia are produced more under coinfection, 45.6% more for the strain S2 and 63.4% more for the strain S3. The proportion of viable chasmothecia was not significantly different between single infection and coinfection (**Figure 2** and **Table 1**). The interaction between temperature and coinfection was not significant for any of the measured life-history traits (**Table 1**).

We found significant variation among strains in the timing of chasmothecia production and their viability (**Figure 2** and **Table 1**). We also found that for all traits measured, no strain can outperform all others in all temperatures (**Figures 1**, **2**, **Table 2**, and **Supplementary Figure 2**). This trend was most pronounced when measuring the timing and total number of chasmothecia, as well as proportion of viable chasmothecia, as evidenced by the statistically significant strain identity × temperature interaction (**Figure 2** and **Table 1**).

### DISCUSSION

Understanding processes maintaining variation in pathogen lifehistory traits is a long-standing challenge in disease ecology as this variation underlies risks of pathogen emergence and risks of infection (Galvani, 2003; Grenfell et al., 2004). Here, we have measured life-history traits at both asexual and sexual stages of an obligate fungal pathogen. We find that when we simultaneously investigate the effect of abiotic and biotic environmental variation, only temperature had a consistent and significant impact on all measured pathogen life-history. For several of the measured traits, the pathogen strains also differed significantly in how they responded to variation in temperature. Coinfection had a significant direct effect on the number of produced chasmothecia. Our finding suggests that abiotic variation may be more important in maintaining variation in pathogen populations than biotic variation.

In line with a previous study (Laine, 2007), our results demonstrate that the asexual life-history traits leading to infection are significantly affected by temperature. For the asexual life-history traits the optimal temperature depended on the lifehistory trait as well as strain identity. In general, although the specific effects of temperature vary among pathosystems, it has been shown that temperature may impact all key pathogen life-history traits including the ability to establish infection, subsequent development as well as transmission (Burdon, 1987; Thomas and Blanford, 2003; Fels and Kaltz, 2006). Here, we also find that temperature significantly affects traits linked with the sexual stage. Our inoculation study revealed that at +17◦C, the number and viability of produced chasmothecia was the highest for all three strains. This effect of the temperature is in line with the biology of the powdery mildew, as the production of the sexually produced resting structures may be partly triggered by lower temperatures toward the end of the growing season (Tollenaere and Laine, 2013).

Genotype-by-environment interactions are considered an important mechanisms maintaining variation. In support of this, here we find that for nearly all measured life-history

#### TABLE 2 | The means and standard deviations of the measured life-history traits for each temperature and strain under single infection.


The time is indicated in days. The best performing strain at each temperature is indicated in bold.

TABLE 3 | The means and standard deviations of the measured life-history for each temperature and strain under coinfection.


The time is indicated in days. The best performing strain at each temperature is indicated in bold.

traits, the strains switch ranks along the temperature gradient. Hence, no strain outperforms all others across all temperatures. This trend was statistically significant when measuring all traits linked with sexual reproduction: the timing and number of produced chasmothecia, as well as their viability. These represent core fitness traits in P. plantaginis, as chasmothecia production is strongly linked with the ability to survive over winter (Tack and Laine, 2014). Although the experiment only included three strains of P. plantaginis, it is in line with previous work demonstrating significant temperature and genotype x temperature effects on asexual life-history traits using a diverse range of P. plantaginis strains (Laine, 2004, 2008). An interesting avenue of future work would be to test whether the results we find here – both direct and strain mediated sensitivity to temperature– also holds for a larger number of strains and under more extreme temperature variation.

For pathogens the biotic environment may vary depending on the community of coinfecting pathogens sharing the same host. Coinfection is a common phenomenon across plant pathosystems (Tollenaere et al., 2016). In the P. plantaginis metapopulation, approximately half of the local pathogen populations support coinfection (Susi et al., 2015a). Theoretically coinfection has been proposed to change pathogen growth rates as strains compete for the same limited resources of the host, and there is some empirical support for this (Clement et al., 2012; Tollenaere et al., 2016; Suffert et al., 2018). Although previous studies on P. plantaginis have revealed coinfection to change both within- and between host dynamics (Susi et al., 2015a,b), here coinfection only had a significant effect on the number of chasmothecia produced. Higher production of resting spores under coinfection is expected to have profound fitness consequences as local pathogen populations go through a severe decline each winter (Tack and Laine, 2014), often resulting in local extinction (Jousimo et al., 2014). As coinfection varies spatially (Susi et al., 2015a), it can be a powerful mechanism maintaining diversity in this natural pathosystem.

Interestingly, we also did not detect a significant coinfectionby-temperature interaction for any of the measured life-history traits. It is likely that the effect of temperature in our experiment was so strong, that most of variation in these data is explained

by the temperature gradient. However, it should also be noted that previous studies have shown strains to vary in how they respond to coinfection (Laine and Mäkinen, 2018). Hence, it is possible that the strains included in this study do not represent those that respond strongly to coinfection. Moreover, the effects of coinfection – either direct or mediated through an interaction with temperature – may not become apparent in the development of a single lesion but accumulate during the epidemic season (Susi et al., 2015a,b). Hence, a future study using more strains and allowing for several cycles of auto- and allo-infection would help elucidate whether coinfection changes infections outcomes also under pronounced abiotic variation.

### CONCLUSION

Here, we find both abiotic and biotic environments to impact pathogen life-history traits directly as well as through genotype specific responses. Moreover, we find that the effect of abiotic and biotic variation depends on pathogen life-history traits. Temperature variation affected all studied pathogen lifehistory traits and hence, we may expect it to be a powerful mechanism maintaining diversity during the epidemic season. We found coinfection to significantly affect the number of produced chasmothecia which is expected to have far-reaching consequences for pathogen survival from one epidemic season to the next (Tack and Laine, 2014). For an epiphytic pathogen such as P. plantaginis, it is not surprising to find high sensitivity to ambient temperature. Temperature fluctuations may be gradual (e.g., global warming), seasonal (e.g., summer vs. winter resource availability), or continuously fluctuating (e.g., daily temperatures). Our results suggest that P. plantaginis can cope with such variation in the environment, as we documented plastic responses in many life-history traits under temperature variation. We conclude that environmental variation is expected

### REFERENCES


to be an important mechanism maintaining variation in pathogen populations, given that many studies have reported life-history trait plasticity in response to temperature variation (Blanford et al., 2003; Wolinska and King, 2009).

### DATA AVAILABILITY STATEMENT

All data associated with this study is available on the Dryad Digital Repository: doi: 10.5061/dryad.fg25kk0.

### AUTHOR CONTRIBUTIONS

EV and A-LL conceived and designed the experiments, analyzed the data, and wrote the paper. EV performed the experiments.

### FUNDING

This work was funded by grants from the Academy of Finland (296686), and the European Research Council (Consolidator Grant RESISTANCE 724508) to A-LL.

### ACKNOWLEDGMENTS

Krista Raveala is acknowledged for assistance in the greenhouse and laboratory.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018.01670/ full#supplementary-material



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

Copyright © 2018 Vaumourin and Laine. 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.

# Striking Similarities Between Botrytis cinerea From Non-agricultural and From Agricultural Habitats

Marc Bardin\*, Christel Leyronas, Claire Troulet and Cindy E. Morris

Pathologie Végétale, INRA, Montfavet, France

Investigations into life history of microorganisms that cause plant diseases have been limited mostly to contexts where they are in interaction with plants, and with cropped or otherwise managed vegetation. Therefore, knowledge about the diversity of plant pathogens, about potential reservoirs of inoculum and about the processes that contribute to their survival and adaptation is limited to these contexts. The agro-centric perspective of plant pathogen life histories is incoherent with respect to the capacity of many of them to persist as saprophytes on various substrates. In this context we have investigated the ubiquity of the broad host range necrotrophic fungus Botrytis cinerea, outside of agricultural settings and have determined if the populations in these natural habitats can be distinguished phenotypically and phylogenetically from populations isolated from diseased crops. Over a period of 5 years, we isolated B. cinerea from 235 samples of various substrates collected in France including rainfall, snowpack, river, and lake water, epilithic biofilms in mountain streams, leaf litter and plant debris, rock surfaces, bird feathers and healthy wild plants from outside of agricultural fields. All substrates except rock surfaces harbored B. cinerea leading us to establish a collection of purified strains that were compared to B. cinerea from diseased tomato, grapes and various other crops in France. Phylogenetic comparisons of 321 strains from crop plants and 100 strains from environmental substrates based on sequences of 9 microsatellite markers revealed that strains from crops and the environment could not be distinguished. Furthermore, the genetic diversity of strains outside of agriculture was just as broad as within agriculture. In tests to determine the aggressiveness of strains on tomato stems, the mean disease severity caused by strains from environmental substrates was statistically identical to the severity of disease caused by strains from tomato, but was significantly greater than the severity caused by strains from grape or other crops. Our results suggest that highly diverse populations of this plant pathogen persist outside of agriculture in association with substrates other than plants and that this part of their life history is compatible with its capacity to maintain its potential as plant pathogen.

Edited by:

Jeremy Astier, INRA UMR1347 Agroécologie, France

#### Reviewed by:

Renaud Travadon, University of California, Davis, United States Michelina Ruocco, Istituto per la Protezione Sostenibile delle Piante (IPSP), Italy

> \*Correspondence: Marc Bardin marc.bardin@inra.fr

#### Specialty section:

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

Received: 28 August 2018 Accepted: 22 November 2018 Published: 05 December 2018

#### Citation:

Bardin M, Leyronas C, Troulet C and Morris CE (2018) Striking Similarities Between Botrytis cinerea From Non-agricultural and From Agricultural Habitats. Front. Plant Sci. 9:1820. doi: 10.3389/fpls.2018.01820

Keywords: microbial ecology, epidemiology, gray mold, polyphagous, diversity, aggressiveness, fitness, selection

## INTRODUCTION

fpls-09-01820 December 3, 2018 Time: 11:7 # 2

Investigations into the life history of microorganisms that can cause disease to plants have been limited mostly to contexts where they are in interaction with plants, and even more frequently with cropped or otherwise managed vegetation (Morris et al., 2009). Therefore, knowledge about the diversity of plant pathogens, about potential reservoirs of inoculum and about the processes that contribute to survival and adaptation of plant pathogens is limited to these contexts. By contrast, there are many examples of human pathogens for which habitats and biological activities are known outside the strictly medical context (Morris et al., 2009). The agro-centric perspective of plant pathogen life histories is incoherent with respect to the capacity of many plant pathogens to persist as saprophytes on various substrates. Thus, various plant pathogens, including bacteria, fungi and viruses have been detected in non-agricultural habitats or in the absence of host plants (Morris et al., 2009). For instance, recently the fungal species Fusarium oxysporum, a major fungal plant pathogen, has been isolated in tap water (Edel-Hermann et al., 2016). However, apart from the bacterial species Pseudomonas syringae (Morris et al., 2013), and the oomycete genera Phytophthora (Hansen et al., 2012), the ecology and evolutionary history of plant pathogens at spatiotemporal scales wider than their strict interaction with plants (either cultivated or wild) has not been extensively studied.

The aerially disseminated fungus Botrytis cinerea is considered to be a broad-host range necrotrophic pathogen able to attack about 586 plant genera, mostly dicots (Elad et al., 2016). It is also capable of growing as a saprophyte on dead and decaying plant matter (Williamson et al., 2007) and it probably has an underestimated facultative endophytic behavior, i.e., it survives in host plant tissue without generating disease symptoms (Van Kan et al., 2014). Additionally this fungus has been detected in non-agricultural contexts such as outdoor or indoor air in agricultural or urban settings (Bartlett et al., 2004; Lugauskas and Krikstaponis, 2004; Leyronas and Nicot, 2013), in clouds (Amato et al., 2007), in precipitation (Monteil et al., 2014), in agricultural or non-agricultural soils (Azmi and Seppelt, 1998; Leyronas et al., 2015c), on rocks and monuments (Gaylarde and Morton, 1999), in stocks of hydrocarbons (Gaylarde et al., 1999), on external or in internal parts of insects (Fermaud and Lemenn, 1989; Louis et al., 1996), on pelage of small mammals (Shchipanov et al., 2006) and human hairs (Ali-Shtayeh et al., 2001). However, to our knowledge, the genetic diversity and the pathogenic potential of saprophytic B. cinerea strains isolated from non-agricultural habitats have never been studied.

Recent research is revealing that natural environments harbor strains of microorganisms that have likely given rise to strains that have emerged in plant disease epidemics (Hansen et al., 2012; Monteil et al., 2016). In this context, we have assessed the ubiquity of the fungus B. cinerea outside of agricultural settings and have determined if the populations in these natural habitats can be distinguished phenotypically and phylogenetically from populations isolated from diseased crops. The objectives of the present study were (i) to determine the presence and the abundance of B. cinerea in various non-agricultural habitats, (ii) to evaluate the genetic diversity and the phenotypic diversity of the strains collected, and (iii) to compare this diversity with strains collected from crops.

### MATERIALS AND METHODS

### Isolation of B. cinerea From Non-agricultural Habitats

Over a period of 5 years (2005–2010), we collected 235 samples of various substrates, including rainfall, snowpack, fresh water from rivers and lakes, epilithic biofilms in mountain streams, rocks, leaf litter and plant debris, healthy wild plants from outside of agricultural fields and feathers from the great tit (Parus major) (**Table 1**). Most of samples originated from the Provence-Alpes-Côte d'Azur region in France. Some of the strains obtained from rainfall and snowfall were part of a previous study (Monteil et al., 2014).

Each sample was plated on Petri plates containing the semiselective Botrytis Spore Trap Medium (BSTM) (Edwards and Seddon, 2001). The Petri plates were sealed with parafilm and incubated at 20◦C in daylight for 14 days to allow the development of fungal colonies. The colonies with mycelium resembling that of B. cinerea were individually sub-cultured on Potato Dextrose Agar (PDA) and their identity was confirmed 7 days after transplanting by observing characteristic asexual sporulation of B. cinerea (Barnett, 1998). Dry sterile cotton swabs were rubbed on sporulating plates to collect B. cinerea spores. The swabs were stored at −20◦C until isolate purification.

All isolates were purified and single-spored in a classical way (Leyronas et al., 2012) prior to their genotypic and phenotypic characterization. Hereafter these characterized single spore isolates will be referred to as "environmental strains."

### Strain Genotyping

A subset of 109 strains collected from non-agricultural habitats were genotyped (**Table 1**). They were compared to 327 agricultural strains sampled from lettuce and tomato plants grown in several greenhouses in the South of France (Leyronas et al., 2015a,b,c).

Genomic DNA was extracted from aliquots of 15 mg lyophilized fungal material (harvested from two-week old cultures on Potato Dextrose Agar), following the DNeasy Plant extraction Kit protocole (Qiagen). The nine microsatellite markers designed for B. cinerea by Fournier et al. (2002) were amplified following the protocol described by Leyronas et al. (2015b). To determine the size of the microsatellites, the PCR products were scanned with the help of an ABI 3730 sequencer (Applied Biosystems). GeneMapper software version 4.1 (Applied Biosystems) was then used for the microsatellite size analysis. Complete microsatellite size profiles (referred to as "haplotypes" hereafter) were obtained for 109 environmental strains and 327 agricultural strains.

### Genetic Characteristics of Strains

In a first step, the strains with the private allele at microsatellite locus BC6 associated with the cryptic species B. pseudocinerea

TABLE 1 | Quantification of the number of Botrytis cinerea strains collected in non-agricultural habitats (environmental strains) and number of strains tested for different traits.


(Walker et al., 2011) were removed from further analyses. This private allele is found only in B. pseudocinerea strains and never in B. cinerea strains. Then, in order to compare the B. cinerea strains sampled from the different reservoirs, several indices of genetic diversity were used. The software FSTAT version 2.9.3 (Goudet, 1995) was used to compute allelic richness (corrected for the smallest sample size) and unbiased gene diversity per locus. Unbiased gene diversity (Hnb) and allelic richness (average over the nine loci) were computed separately for the strains collected in the different reservoirs with the Genetix software (Belkhir et al., 1996–2004). The number of different multilocus haplotypes (MLH) was computed with GenClone 1.0 software (Arnaud-Haond and Belkhir, 2007). We used the index of haplotypic diversity (based on the number of individuals and the number of distinct MLH), which estimates the proportion of haplotypes present in a population and takes a value of 1 when a population is composed exclusively of unique haplotypes (Arnaud-Haond et al., 2007).

### Phylogenetic Relationships Between Strains From Non-agricultural Habitats and From Crops

To assess the relationships between the different strains, we computed a neighbor-joining (NJ) tree with the program POPULATIONS (version 1.2.32 provided by Olivier Langella, Quantitative Genetics and Evolution, Gif-sur-Yvette, France). The NJ-tree was based on the distances of Cavalli-Sforza and Edwards (1967), computed from our microsatellite loci. The tree was visualized and edited with TREEVIEW (Page, 1996). All analyses were conducted on data sets excluding clone replicates.

In addition, Arlequin version 3.5 (Excoffier et al., 2005) was used to assess genetic differentiation between strains collected from non-agricultural habitats (NH) and from crops by computing RST values as suggested by Slatkin (1995) for microsatellite data. The clonally corrected data set was used.

### Aggressiveness of Strains on Tomato Plants

The aggressiveness of a subset of 128 environmental strains (**Table 1**) was assessed on 8-week old tomato plants cv. Monalbo (INRA). Tomato plants were grown in a greenhouse and watered daily with a nutrient solution as described previously (Decognet et al., 2009). Each strain was inoculated on three plants. On each plant, three leaves were removed, leaving 1 cm petiole stubs on the stems and the wounds were inoculated with 10 µL aliquots of spore suspension. The spore suspensions were prepared from two-week old cultures on PDA and were adjusted to 10<sup>6</sup> spores mL−<sup>1</sup> . All plants were incubated in a growth chamber with a photoperiod of 14 h, with a light intensity of 162 µmol m−<sup>2</sup> s −1 , maintained at 21◦C with a relative humidity above 90%. The length of resulting stem lesions was monitored daily from the 3rd to the 7th day after inoculation and these data were used to compute the area under the disease progress curves (AUDPC). Due to a limited place in the growth chamber, strains were tested in series of 10–15 strains, and in each series, strain BC1 was used as a reference to calculate an index of aggressiveness (IA) for each strain, relative to that of this strain (Decognet et al., 2009), as follows:

$$\text{IA} = 100^\* \text{ (AUUDPC}\_{\text{strain}} / \text{AUUDPC}\_{\text{BC1}})^\*$$

with AUDPCstrain being the value of AUDPC computed for the tested strain and AUDPCBC1 the value of AUDPC obtained for the reference strain BC1. Two to three independent repetitions of the test were realized for each strain.

The aggressiveness of environmental strains was compared to that of 156 single-spored agricultural strains having a wide diversity regarding year, region of isolation and host plants. All these strains were collected from 1988 to 2008 with a majority collected after 2000 (108 strains). Regarding their geographical origin, 149 strains were sampled from France, of which 122 were from the southern part of the country. In addition, four strains came from Italy, two from Syria and one from Portugal. They were collected from various diseased plants: 101 from tomato, 33 from grape, and 22 from other plants including rose, cucumber, strawberry, artichoke, pepper, carrot, onion, asparagus, peach fruit, cherry fruit, kiwifruit, hydrangea, gerbera, cyclamen, and poinsettia.

### Traits of B. cinerea in vitro

fpls-09-01820 December 3, 2018 Time: 11:7 # 4

Mycelial growth, sporulation and sclerotia production were evaluated for a sub-sample of 31 agricultural strains and 24 environmental strains of B. cinerea (**Table 1**), which represent the full range of aggressiveness of the tested strains. Mycelial growth was evaluated by inoculating PDA with a 5 mm-diameter mycelial plug of each B. cinerea strain and incubating in a growth chamber (21◦C, photoperiod 14 h, 114 µmol m−<sup>2</sup> s −1 ). Two perpendicular measurements of the diameter of the mycelial colony were performed every day for 3 days and the rate of mycelial growth between the first and the third day of incubation was calculated for each strain. Three plates were inoculated for each strain and the whole experiment was repeated two times. The spore production was determined 14 days after inoculation on PDA medium (21◦C, photoperiod 14 h, 114 µmol m−<sup>2</sup> s −1 ). Spores were scrapped from the media, suspended in water and spore concentration was determined using a hemacytometer. The number of sclerotia produced by each strain of B. cinerea on PDA was recorded after five weeks of incubation at 21◦C (1 week with a photoperiod of 14 h at 114 µmol m−<sup>2</sup> s −1 and 4 weeks in the dark) and 5 weeks of incubation at 4◦C in the dark. For the sporulation and sclerotia production experiments, three replicates were realized for each strain.

### Statistical Analyses

Statistical analyses were performed with Statistica (version 12, Statsoft). Non-parametric tests (Mann and Whitney) were used to determine significant differences between gene diversity and allelic richness of environmental and agricultural strains. To compare the different phenotypic traits evaluated, analysis of variance (ANOVA) were performed. In the case of a significant effect of the tested factor, a comparison of mean was realized using the Newman–Keuls test. The relationship between the various phenotypic traits tested (mycelial growth, sporulation, sclerotia production) and aggressiveness was realized using linear regressions. Statistical inferences were made at the 5% level of significance, unless indicated otherwise.

### RESULTS

### B. cinerea Is Ubiquitous in Non-agricultural Habitats

Botrytis cinerea was recovered in 56 samples out of the 235 samples collected (**Table 1**). From these 56 samples, we were able to purify 141 strains of B. cinerea. All substrates except rock surfaces harbored B. cinerea. The fungus was detected at a rate of 51% in precipitation, 50% from feathers of Great Tits, 21% in fresh water and 20% in snowpack. Detection rates were between 11 and 17% in the other substrates including epilithic biofilm, litter, plant debris and asymptomatic plants.

### Environmental and Agricultural Strains Cannot Be Significantly Differentiated Based on Their Genetic Diversity

Among the 109 genotyped environmental strains and 327 genotyped agricultural strains, 9 and 6 strains, respectively, carried the private allele at microsatellite locus BC6 associated with the cryptic species B. pseudocinerea (Walker et al., 2011) and were removed from further analyses. We thus compared the indices of genetic diversity based on the haplotypes obtained for the 100 and 321 remaining strains considered to be B. cinerea (**Table 2**). There was no significant difference between the number of alleles in each microsatellite locus and the gene diversity per locus of environmental strains and agricultural strains (respectively, PMann and Whitney = 0.60, PMann and Whitney = 0.79) (**Table 3**). The global gene diversity and the mean number of alleles per locus were in the same range for both population of strains (**Table 2**). However, environmental strains had higher haplotypic diversity than agricultural strains (**Table 2**). Two hundred and seventy different multilocus haplotypes (MLH) were found among the 423 strains analyzed but none of these was shared by environmental strains and agricultural strains.

The distribution of the strains collected from crops and those collected from non-agricultural habitats in the NJ tree shows that they were widely intermixed in the tree regardless of their origin (**Figure 1**). This result is supported by the low level of genetic differentiation (RST = 0.021, P = 0.019) between these two groups of strains.

### Environmental Strains Represent a Greater Variability in Aggressiveness and an Overall Greater Mean Aggressiveness Than Agricultural Strains

Based on the lesion expansion on the stem of the tomato plants, different levels of aggressiveness were observed within the 128 environmental strains and the 156 agricultural strains of B. cinerea tested (**Figure 2**). For each of the population of strains (agricultural strains and environmental strains), significant differences were observed among strains (ANOVA, P < 0.0001). MOP7-4 isolated in rainfall in 2010, and H6 and BC25 isolated on diseased tomato plants in 1991 were not able to generate symptoms on the stem of the plant, suggesting that they are hypo-aggressive on tomato stems. However, these strains were able to infect the tomato petiole stubs (data not shown). The index of aggressiveness for the 284 strains of B. cinerea ranged from 0 to 144% on potted tomato plants with the wider range of aggressiveness for environmental strains compared to agricultural strains (**Table 4**). The mean disease severity caused by strains from environmental substrates was statistically identical to the severity of disease caused by strains from tomato, but was significantly greater than the severity caused by strains from grape or other crops. Globally, environmental strains are significantly more aggressive than agricultural strains on plants (58.9% vs. 50.4%, P = 0.009).


TABLE 2 | Indices of genetic diversity among strains collected from non-agricultural habitats (environmental strains) and from crops (agricultural strains).

TABLE 3 | Allelic richness (AR) and unbiased gene diversity (Hnb) of environmental and agricultural strains of B. cinerea, at each of the nine microsatellite loci (corrected for the smallest sample size: 100).


### Environmental Strains Grow Faster and Sporulate Less Abundantly but Do Not Differ in Sclerotial Production Compared to Agricultural Strains

Among the 31 agricultural and 24 environmental strains tested, the mycelial growth between the first and the third day after the plating of mycelial plug varied widely among strains (**Figure 3**; ANOVA, P < 0.0001 for each set of strains). Four strains (BC83, 06C12, 06C163, and S381) had a mycelial growth rate significantly lower than the other strains. These four strains were all isolated from diseased plants (rose, grape and tomato). Globally, mycelial growth rate was significantly higher for environmental strains compared to agricultural strains (29.8 ± 0.4 vs. 25.7 ± 0.4; P < 0.0001). A significant correlation was detected between aggressiveness of strains on tomato plants and their mycelial growth on PDA medium (r <sup>2</sup> = 0.12, P = 0.009 for all the strains). However, this correlation is not significant for each set of strains taken independently (r <sup>2</sup> = 0.09, P = 0.17 for environmental strains and r <sup>2</sup> = 0.11, P = 0.08 for agricultural strains).

All strains of B. cinerea tested were able to sporulate on PDA medium but the number of spores produced varied widely among strains, both for environmental and for agricultural strains (**Figure 4**; ANOVA, P < 0.0001 for each population of strains). Globally, sporulation was significantly lower for environmental strains compared to agricultural strains (12.9 ± 1.1 vs. 17.4 ± 1.1 million spores/mL; P = 0.005). There was no significant correlation between aggressiveness of strains on tomato plants and sporulation (r <sup>2</sup> = 0.00001, P = 0.98 for all strains).

The number of sclerotia produced on PDA medium varied greatly among strains (**Figure 5**; ANOVA, P < 0.0001 for both environmental and agricultural strains) and the average number of sclerotia produced was not different between the two sets of strains (41.1 ± 5.5 and 65.5 ± 12.6 for environmental and agricultural strains, respectively, P = 0.11). Twelve strains of B. cinerea (7 from diseased plants and 5 from non-agricultural context) were not able to produce sclerotia under the conditions used here and one strain (BC96 isolated from hydrangea) produced a significantly higher number of sclerotia than all the other strains. We found a significant negative relationship for all strains between the aggressiveness and the number of sclerotia produced (r <sup>2</sup> = 0.09, P = 0.031). This correlation is highly significant for the 24 non-agricultural strains (r <sup>2</sup> = 0.47, P = 0.0002) while it was not significant for the 31 agricultural strains (r <sup>2</sup> = 0.03, P = 0.33), suggesting that the relationship between aggressiveness and sclerotial production differ among these two populations of strains.

### DISCUSSION

This study is the first systematic census of the fungal plant pathogen B. cinerea in diverse non-agricultural environments that could come into contact with agriculture such as snowpack, fresh water from river and lake, epilithic biofilms in mountain streams, leaf litter, plant debris and bird feathers. These results complement the prior detection of B. cinerea in the diverse substrates described in the introduction. It also contributes to information about sources of B. cinerea in precipitation (Monteil et al., 2014) and corroborates reports about its capacity to

TABLE 4 | Mean and range of aggressiveness of strains of B. cinerea collected on various substrates.


∗ Ia (index of aggressiveness) is expressed in % relative to the reference strain BC1. Values with the same uppercase letter are not significantly different according to Newman-Keuls multiple range test (P < 0.05).

colonize and survive internally in plants without causing any symptoms (Van Kan et al., 2014), traits that are essential for an endophytic phase. Our results illustrate that B. cinerea can survive in environmental reservoirs outside of agriculture sensu stricto and possibly in the absence of susceptible plants. This knowledge contributes to the growing notion that the life history of plant pathogens should be viewed from a wider angle than their direct interaction with plants in contexts that are favorable for disease. Furthermore, the relative ease with which we found strains of B. cinerea in environmental reservoirs highlights how defining this fungus as a plant pathogen skews research questions on its life history and ecology in spite of the numerous previous reports of its presence in a wide range of diverse habitats.

Our results also suggest that the environmental reservoirs are probable sources of inoculum. Previous reports of the genetic diversity of B. cinerea from wild plants (e.g., primrose, bramble, dandelion, Hypochaeris radicata, Plantago lanceolata, and Sonchus asper), illustrated the potential reservoirs of inoculum for crops on wild plants and weeds in agricultural systems (Fournier and Giraud, 2008; Rajaguru and Shaw, 2010; Walker et al., 2015; Wessels et al., 2016), but there have been no studies on non-agricultural sources. Phylogenetic comparisons of strains from crop plants and from environmental substrates presented in this study revealed that strains from crops and from the environment could not be distinguished, thereby supporting the hypothesis that strains from these different sources mix. Furthermore, the genetic diversity of strains outside of agriculture was just as broad as within agriculture (**Table 5**; Bardin et al., 2014), suggesting that there are no strong selective pressures that remarkably distinguish these two categories of habitats. These results are consistent with those previously published for B. cinerea in different agricultural contexts, using the same genotyping tools developed by Fournier et al. (2002). In such studies, the haplotypic diversity (ratio of the number of different haplotypes over the total number of characterized strains) varied from 0.37 to 1.00. In the present study, this index was

high, with a value of 0.82, comparable to what is observed in populations of B. cinerea collected on diseased plant in open field situations. It reflects the scarcity of strains with identical genotypic profiles in the non-agricultural context. Although we have demonstrated that strains from a non-agricultural context cannot be genetically distinguished from strains from diseased plants, various questions remain about the rate of flux of strains between non-agricultural environment and crops and on the contribution of environmental reservoirs to epidemics.

In tests to determine the severity of disease that strains could induce on tomato stems, the mean disease severity caused by strains from environmental substrates was statistically

TABLE 5 | Genetic diversity reported for strains of B. cinerea collected from different origins and characterized using the same microsatellite markers developed by Fournier et al. (2002), as in this study.

value represents the mean of three replicates. The vertical bar associated to each histogram correspond to SE.

s


The References presented here range from 2015 to 2018. For information on previous references, see Bardin et al. (2014). a Indication between brackets indicates the type of crop. OF, open field; G, greenhouse. <sup>b</sup>Haplotypic diversity index directly available or computed (on the basis of reported information) as the ratio of (number of distinct MLG – 1) over (sample size – 1). <sup>c</sup>na, data not available.

greater than the severity caused by strains from diseased crops (P = 0.009). It suggests that there are no selective pressures for strains of B. cinerea to lose pathogenicity while they are in contexts where they do not cause plant disease or are not even colonizing plants. On the other hand, the environment seems to select for more rapid mycelial growth whereas cropping contexts select for more abundant sporulation. This suggests that there are fitness or survival benefits for these traits in each of the respective environments. This hypothesis is reinforced by the fact that mycelial growth is positively correlated with aggressiveness. Therefore, it suggests that the non-agricultural environment could select strains for factors that would facilitate their pathogenicity on plants. It would be interesting to determine how much time is needed for mycelial growth and sporulation rates to change under the selection pressures of each environment and if in fact these traits are fixed or due to epigenetic processes. This could give clues to the amount of time that strains from these habitats have been localized in one habitat or another. Interestingly, a trade-off between the aggressiveness and the number of sclerotia produced was observed for the environmental strains of B. cinerea. This suggests that the least aggressive environmental strains compensate this low level of aggressiveness by their high ability to survive in the environment in the absence of host plants.

Our results are consistent with those concerning the life history of the bacterial plant pathogen P. syringae (Morris et al., 2013) in that highly diverse populations of strains that maintain their pathogenic potential can persist outside of agriculture in association with substrates other than plants. These findings set the stage for new perspectives on factors that foster to the evolution and emergence of pathogenic potential in microorganisms, on disease epidemiology and on the factors that contribute to the durability of disease control methods. They also beg for the development of systematic approaches to widen our understanding of the life history of plant pathogens in general beyond the context of agriculture.

### AUTHOR CONTRIBUTIONS

fpls-09-01820 December 3, 2018 Time: 11:7 # 10

MB, CT, and CM achieved the sampling of environmental strains. MB and CT realized the phenotypic experiments and the associated data analysis. CL carried out microsatellite genotyping and the genetic data analysis. MB coordinated the writing of the manuscript. MB, CL, and CM wrote the manuscript. All authors reviewed the manuscript.

### REFERENCES


### ACKNOWLEDGMENTS

The research leading to these results has received funding from INRA department of Plant Health and Environment. We are grateful to Dr. Philippe Nicot for extremely useful discussions throughout this research project and to Dr. Caroline Monteil, Jean-Charles Bouvier, Caroline Guilbaud, Charlotte Chandeysson, and Magali Duffaud for their valuable help in samples collection. Part of this work was carried out by using the resources of the Molecular Biology Platform of INRA-PACA center. Research on plant biology depends heavily on the cultivation of plants under experimental conditions that are controlled, monitored, and repeatable. For ensuring a perfect plant production, the contribution of the Plant Production and Environmental Management Service of the Plant Pathology Research Unit was greatly appreciated.



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

Copyright © 2018 Bardin, Leyronas, Troulet and Morris. 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.

# Pathogenicity Traits Correlate With the Susceptible Vitis vinifera Leaf Physiology Transition in the Biotroph Fungus Erysiphe necator: An Adaptation to Plant Ontogenic Resistance

#### Edited by:

Omer Frenkel, Agricultural Research Organization (ARO), Israel

#### Reviewed by:

Andreia Figueiredo, Universidade de Lisboa, Portugal Courtney Coleman, University of Missouri, United States

#### \*Correspondence:

Agnes Calonnec agnes.calonnec@inra.fr

#### †Present Address:

Sylvain Schnee, Station de Recherche Agroscope Changins-Wädenswil ACW, Nyon, Switzerland

#### Specialty section:

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

Received: 15 June 2018 Accepted: 20 November 2018 Published: 11 December 2018

#### Citation:

Calonnec A, Jolivet J, Vivin P and Schnee S (2018) Pathogenicity Traits Correlate With the Susceptible Vitis vinifera Leaf Physiology Transition in the Biotroph Fungus Erysiphe necator: An Adaptation to Plant Ontogenic Resistance. Front. Plant Sci. 9:1808. doi: 10.3389/fpls.2018.01808

#### Agnes Calonnec<sup>1</sup> \*, Jerome Jolivet <sup>1</sup> , Philippe Vivin<sup>2</sup> and Sylvain Schnee1†

<sup>1</sup> UMR1065 SAVE Santé et Agroecologie du Vignoble, INRA, ISVV, Villenave d'Ornon, France, <sup>2</sup> EGFV, Bordeaux Sciences Agro, INRA, University of Bordeaux, ISVV, Villenave d'Ornon, France

How and when the pathogen cycle is disrupted during plant development is crucial for harnessing ontogenic resistance in sustainable agriculture. Ontogenic resistance against powdery mildew (Erysiphe necator) was quantified on Vitis vinifera. Shoots were sampled in the vineyard at several dates during seasonal growth and processed in the laboratory under controlled conditions. Experiments were conducted on two susceptible Vitis vinifera Cabernet Sauvignon and Merlot. The process of leaf ontogenic resistance was investigated by measuring three quantitative traits of pathogenicity: the infection efficiency, sporulation and mycelium growth. Morphological and physiological plant indicators were used to identify leaf changes that resulted in ontogenic resistance and to predict pathogen variations that were linked to pathogenicity traits. The process of ontogenic resistance was established early in correspondence with the physiological transition of the leaf from sink to source status and was characterized by its increase in sugar content. The three traits of pathogenicity that we measured were affected, and their variation was strongly correlated with leaf age. Using leaf age, we were able to accurately predict the susceptibility of the leaf: a leaf aged, on average, 13.3 days had a very high probability (0.8) of being susceptible, while this probability decreased to 0.5 one week later. Sporulation was more closely correlated with variations in sugar and the infection efficiency in leaf water. The results for both cultivars were consistent. Ontogenic resistance on grapevine leaves is thus interpreted to be a strong, immutable physiological process that E. necator is able to circumvent by restricting its development to sink tissue. Future research should explore how this native plant resistance can be incorporated into grape management strategies to better control powdery mildew (PM) epidemics with reduced amounts of fungicides.

Keywords: powdery mildew, ontogenic resistance, sink-source transition, traits of pathogenicity, grapevine

## INTRODUCTION

Biotrophs totally depend on their hosts to complete their life cycles, deriving their nutrients from living host cells through the development of specialized infection structures (O'Connell and Panstruga, 2006). Erysiphe necator Schwein, the powdery mildew of grapevines, is very efficient and can adopt highly sophisticated mechanisms to invade living plant cells successfully, such as upregulating the expression of cell wall invertase and plant hexose transporter genes or downregulating vacuolar invertase genes (Hayes et al., 2010).

Powdery mildew of grapevine is a ubiquitous disease that affects common cultivars of Vitis vinifera L., and it is responsible for significant damage (Calonnec et al., 2004) if no protective measures are used. While fungicide-based strategies are generally effective in reducing disease (Delière et al., 2010; Deliere et al., 2015; Valdés-Gómez et al., 2017), finding alternative methods that consider practitioners and the environment and that are economically feasible remains a challenge for sustainable viticulture. One solution relies on the exploitation of resistant grape varieties, in which sources of resistance coming from other Vitis of American or Asian origin are introgressed (Calonnec et al., 2013b). The difficulty for breeders in associating traits of agronomical and oenological values and genes of pathogen resistance leads also to exploration of the native potential of plant defenses in cases of a susceptible V. vinifera host. Innovative strategies to counteract pathogen efficiency could rely on integration into the management of the disease of the timecourse evolution of the plant (i.e., leaves and berries) resistance during the season (Burie et al., 2011; Valdes-Gomez et al., 2011; Gadoury et al., 2012; Calonnec et al., 2013a).

In many plant-pathogen interactions, a susceptible host can acquire resistance to a pathogen at a certain developmental stage (Develey-Rivière and Galiana, 2007). This age-related resistance, commonly called ontogenic resistance (when young tissues are susceptible) or receptivity (when old tissues are susceptible), could be defined as the dynamic modification of tissue receptivity during organ development, triggering resistance/tolerance to pathogenic micro-organisms. Ontogenic resistance has been described for many plant-pathogen systems [e.g., strawberrypowdery mildew (Carisse and Bouchard, 2010), cucurbit fruit— Phytophthora capsici (Ando et al., 2009), tobacco—Phytophthora parasitica (Hugot et al., 1999), apple-apple scab (Li and Xu, 2002), cocoa—Phytophthora megakarya (Takam Soh et al., 2012), pea-powdery mildew (Fondevilla et al., 2006), pea-aschochyta blight (Richard et al., 2012)]. They are therefore common traits but remain underexploited in disease management, principally because of a lack of understanding of the underlying mechanisms and of the potential variability linked to the physiological responses of the host to environmental factors. The phenomenon of ontogenic resistance or receptivity to the main pathogens of grapevines has been highlighted and assessed on bunches for Plasmopara viticola (Kennelly et al., 2005), Guignardia bidwellii (Molitor and Berkelmann-Loehnertz, 2011), and Botrytis cinerea (Deytieux-Belleau et al., 2009). Botrytis susceptibility is known to increase with fruit maturity. Although the mechanisms involved remain speculative, they might differ from those involved in response to infection in the classical defense system (Develey-Rivière and Galiana, 2007; Gee et al., 2008). For powdery mildew, the process has been underscored for berries, on which infection stops when berry sugar content reaches 8% (Delp, 1954), and it is characterized by a low penetration rate of infectious structures following potential modification of the structure and chemical composition of the cuticle (Ficke et al., 2002). The level of irradiance received before inoculation can strengthen ontogenic resistance through modifications of berry physiology, such as decreases in pH and K concentration and increases in polyphenol and anthocyanin concentrations (Zahavi and Reuveni, 2012). Ontogenic resistance also exists in mature leaves, which suffer less infection (Doster and Schnathorst, 1985). Experiments on cuttings grown in a glasshouse under a controlled temperature showed that infection is maximal for leaves that have just ceased importing assimilates (Merry et al., 2013). Simulations have shown that, depending on its parameter value, ontogenic resistance could significantly reduce the severity of late epidemics (Calonnec et al., 2008, 2011). However, little information is available regarding the behavior of the pathogen at each step of its sequential development for a broad range of crop development, and no information is available for field-grown material.

In our study, leaf ontogenic resistance to powdery mildew in V. vinifera was investigated using the susceptible cultivars Cabernet Sauvignon and Merlot. They are the two most commonly grown cultivars in the world for wine grapes (OIV, 2018). Shoots were characterized and sampled in the field, and leaves were processed in the laboratory. Our questions were: When is ontogenic resistance established? How does it vary with variety, crop development and climatic environment? Which pathogenicity traits are the most affected, and how variable is the effect? We used leaf age, leaf surface, sugars and LWC to identify changes in leaf physiology linked to ontogenic resistance and to determine whether they could be used to predict variations in pathogenicity.

### MATERIALS AND METHODS

### Plants

Vitis vinifera Cabernet cv. Sauvignon and Merlot were grown on two INRA experimental vineyards near Bordeaux, France, at Cadaujac and Latresne, respectively. Plantation density was 5,050 vines/10,000 m²; the vinestocks were between 7 and 12 years old. The experimental plots were regularly protected with fungicides from natural infection by downy ("Remiltine or Dauphin" (cymoxanil, mancozeb) and powdery mildews (powder or wettable sulfur). Treatments were applied between 6 and 20 days before sampling. Plant phenological stages were monitored from bud break until ripening (Baillod and Baggiolini, 1993). Each new fully expanded primary leaf (width from 3 cm) was tagged by a set of plastic color markers to indicate the date of leaf appearance and subsequently to calculate leaf age. These color markers allowed leaf age (LA) and the sum of temperatures (above 10◦C) accumulated by the leaf during its development (LAdev) to be determined.

### Bioassays

Shoots were sampled at different dates during the vegetative growth of grapevines (May 12, May 26 and June 10 for Cabernet Sauvignon) in 2005; June 4, 2009, and May 27 and June 24, 2010, for Merlot). Selected shoots were cut early in the morning and immediately brought to the laboratory, with the base of the shoots kept in water. No more than one shoot per vine. Before leaf sampling, petioles were marked with a color code to specify their respective positions on the shoot and to maintain traceability throughout the whole experiment. An approximation of the leaf extension rate (LER) per day was calculated by measuring leaf dimensions 3 days (t-3) and the day before (t-1) sampling with. LER = (LSt-3−LSt-1)/2 (LSt-3−LSt-1)/2 The maximum leaf width (W) and maximum leaf length (L) of each leaf were measured to calculate the leaf surface (LS) in accordance with Montero et al. (2000) (LS = L × W × 0.6). For each shoot, the leaves were detached, abundantly washed under permuted water and then dried on filter paper. Under axenic conditions, the leaves were disinfected in a 5% (w/v) aqueous calcium hypochlorite at 65% for 10 min, rinsed in sterile water and dried in sterile filter paper. Three leaf discs (Ø 22 mm) were cut into each leaf and distributed in three different Petri dishes prepared for the pathogenicity tests. The leaf discs were disposed abaxial face down on an agar medium (20 g L−<sup>1</sup> ) supplemented with benzimidazole (30 mg L −1 ). Each Petri dish contained six discs from an identical leaf level (not necessarily the same LA, as shoots are not synchronous in their growth) from six different shoots (replicates). Petri dishes corresponding to the infection efficiency test and to the total production of spores per cm² were randomly placed in a settling tower and then artificially inoculated by blowing conidia from a leaf infected 14 days earlier (4–5 conidia on each conidiophore), in accordance with Cartolaro and Steva (1990). All leaf levels were inoculated together in one tower. Two towers were necessary to inoculate all of the replicates for each experiment. The average density of spores inoculated was 120, 388, and 610 spores/cm² for 2005, 2009, and 2010, respectively. The dishes were placed in a growth chamber at 22◦C (12:12 h light:dark photoperiod) for pathogen development. One different monospore isolate was used each year. It was sampled from vinestock leaves from a vineyard in the Bordeaux area or in the greenhouse before each annual experiment and was bulked on Cabernet Sauvignon leaves from greenhouse-grown cuttings. Infection efficiency was assessed at 72 hpi (hours post-infection) by counting the conidia that reached the branching hyphae stage (stage 5–6 according to Leinhos et al., 1997) after removing tape from leaves and then applying a cotton blue staining procedure (Cartolaro and Steva, 1990). The total production of spores per infected disc was measured at 14–15 dpi. For colony growth and spores per colony tests, the leaf discs were inoculated by the deposit, with a needle, of a few conidia (from 1 to 10) in their centers. Colony growth was assessed at ∼4, 7, 10, 14, and 18 dpi (days postinfection) and the production of spores per colony at 14 or 18 dpi. Sporulating disks were placed in a vial filled with 20 mL of isotone 2 and one drop of non-ionic dispersant (Nacconol 90F), shaken, and the spore production was assessed by counting the number of particles between 15 and 38µm diameter on a sample of 500 µl using a particle counter (Coulter Counter <sup>R</sup> MultisizerTM 3—Beckman Coulter©, U.S). The disease variables calculated were: Inf (% of infectious spores as defined above), Spo (number of spores per cm² of infected leaf disc), Spoc (number of spores per isolated colony), and Diam (final colony diameter in mm). A total of 12–20 shoots (120–220 leaves) were tested depending on the experiments. The experimental conditions and variables measured are summarized in **Tables 1**, **2**, respectively and experimental design in **Supplementary Figure S1**.

### Physiological Indicators

For each sampled leaf, six discs (Ø 6 mm) were collected into a 2-ml Eppendorf tube, fresh weighed, dried in a hood (60–70◦C for 72 h) and then weighed again. Leaf water content (LWC) was calculated from fresh (FW) and dry mass measurements (DW) (LWC = (FW − DW)/FW). The total non-structural carbohydrates were extracted and dosed as described by Gomez et al. (2007). Briefly, two metallic balls (Ø 5 mm) were added to each disc leaf samples, and the Eppendorf tube was submitted to intense shaking (3 min at 30 swings sec-1, Retsch Mixer Mill MM200). An ethanolic solution (80%, v/v) was added to the finely ground tissue, sonicated for 10 min at 60◦C and centrifuged (3,700 trs min−<sup>1</sup> for 5 min). The pellet containing soluble starch (ST) was rinsed three times with nanopure water, autoclaved and then hydrolysed in the presence of 200 µl of amyloglucosidase at 55◦C for 1 h. The supernatant representing soluble sugars (including glucose, fructose and sucrose) (SS) was dried and then hydrolysed by addition of 200 µl of invertase (Sigma 19253) at 55◦C for 1 h. After drying, 800 µl of nanopure water were added to the two types of extracts and vigorously shaken by vortex agitation and sonication to ensure total carbohydrate solubilisation. Soluble sugars and starch solutions were prepared (kit LISA 200C, CETIM, France) and measured spectrophotometrically at 340 nm, with an ELx800UV automated micro-plate reader (Biotek Instruments Inc., Vermont, USA). Glucose (Sigma G8270) was used for the standard range, and all of the obtained data were expressed in glucose equivalents.

### Environmental Growing Conditions

The sum of degree-days above 10◦C (growing degrees) was used to determine grapevine organogenesis (Schultz, 1992). Five variables were calculated to characterize the environmental conditions of vine growth: the sum of GD at the day of sampling, which is also an indicator of the time of the experiment; the sum of GD during the 10 days before sampling; the average evapotranspiration rate, according to the Penman formula for this period (ETPP in mm.day−<sup>1</sup> ); and the sum of rainfall for the 10 days before the experiments. The environmental conditions are summarized in **Table 1**.

### Data and Statistical Analyses

For infection efficiency (Inf), only data based on more than 30 counted spores were analyzed, and for all variables, only shoots with more than 3 validated samples were considered. To lessen the variation between shoots and to emphasize the variation between leaves of different ages, relative infection (RInf) and sporulation (RSpo) rates were also calculated by dividing the value obtained for a given leaf by the maximum value on its own shoot.

#### TABLE 1 | Conditions of the experimentations.


Nb, number; σ, standard deviation; ETPP, Evapotranspiration rate; GD, growing degrees, sum of degrees above 10◦C.

TABLE 2 | Measured variables for the experiments.


<sup>a</sup>No physiological indicators available.

Variance analyses (general linear modeling procedure in SAS) were performed on different variables, such as infection efficiency (Inf), sporulation (Spo), maximum diameter (Diam), and the number of spores produced per colony (Spoc), with leaf age as a factor. For Diam and Spoc, some of the age classes were grouped to allow for statistical comparison when the sample numbers were too low (<5) due to a failure of the infection process related to ontogenic resistance (only a few spores are deposited) (**Table 2**). The effect of leaf age on the independent variables was measured by performing Tukey-Kramer's multiple comparisons test adapted for unbalanced designs. The mean number of spores produced per area of a colony (Spoc/[(Diam/2)².π]), a proxy of the sporulation capacity according to Lannou (2012), was compared by permutation testing (5000 permutations) for two classes of leaf age.

The disease data functions of leaf age were fitted to models using ordinary least squares methods. The choice of one specific model was based on Akaike's information criterion, which selects the model most likely to have generated the data. Goodness of fit was measured by R ² and the adequacy of the model by replicate testing. A fitting procedure was performed using Prism software (version 5.04).

Relationships among experimental conditions, leaf growth characteristics, leaf physiological indicators and disease were studied by a partial least squares path model analysis (PLS-PM) (Tenenhaus et al., 2005). The PLS-path model was described by six unobservable or latent variables (LVs): "Variety," "Climate," "Leaf growth," "Leaf physiology," "Inoculum," and "Disease." This analysis provides how much variation of "Disease" can be predicted by the other latent variables. Each LV was constructed by a set of observable variables called manifest variables (MVs). "Variety" and "Inoculum" were constructed in a formative way by qualitative variables (CS and M for variety, 2005–2009–2010 for inoculum), corresponding either to the varieties Cabernet Sauvignon and Merlot or to the year using different isolates. "Climate" was described in a reflexive way by the average temperature over the 10 days before the experimentation (STJ-10) and the sum of rain (SRJ-10). "Leaf growth" was described in a reflexive way by leaf age (LA), the sum of GD temperature during leaf development (LAdev) and the leaf surface (LS). It is worth noting that, for one experiment, leaves sharing the same leaf age encountered the same climatic conditions, but they do not necessarily have the same plastochron index, which is a function of the rate of growth of the shoot, nor do they have the same LS, which is dependent on the variety, vigor and environmental conditions. An identical leaf age yields various LAdev depending on experiments. "Leaf physiology" was described by the leaf soluble carbohydrates (SS), starch (ST) concentrations and LWC. The "Disease" variable was described by the variables characterizing the two basic pathogen processes: the infection efficiency (Inf) and the sporulation per colony (Spoc). The standardized latent variables were estimated as linear combinations of their centered MVs. Collinearity between manifest variables was allowed. The PLS-path model was described by the measurement model, which related the different MVs to their own LVs, and the structural model, which linked the endogenous LV "Disease" to the other LVs: "Leaf growth," "Leaf physiology," and "Year." The inner estimate was performed using a path weighting scheme algorithm. Confidence intervals on regression coefficients are estimated by bootstrap methods (500 random sampling of 280 individuals of the observed data set). Two criteria of the goodness of fit of the model (GoF) are provided to evaluate the external model (relationship between MVs and LVs) and the internal model (relationship between LVs). Altogether, 321 observations (leaves) were considered (49 from 2005\_exp1, 64 from 2005\_exp2, 117 from 2009\_exp4, 50 from 2010\_exp5, and 41 from 2010\_exp6). Observations on leaves older than 31 days were not considered to avoid introducing bias between the early and late experiments. Missing values were estimated using NIPALS procedures. The analysis was performed using the XLstat PLS-PM module (version 2012.2.02).

Finally, logistic regression was performed on the whole data set, including old leaves. The Spo variable was transformed into binary form with a value of 1 for Spo >3000 (significant peak with the particle counter) and 0 for less. Then, a model was fitted to describe how the chance of the event of a "diseased" leaf occurring depends on a maximum of one or two covariables, among which are "LA," "LAdev," "LS," "SS," "ST," and "1/LWC." The regression assumes that P(Y = 1/X = exp(β<sup>0</sup> + β1X<sup>1</sup> + β2X2)/exp(β0+β1X<sup>1</sup> + β2X2) with Y the outcome and X the covariates. The model parameters β are estimated by the method of maximum likelihood. The analysis was performed on 497 individuals from the 5 experiments [261 leaves classified as "healthy" = "non-sporulating"; and 236 classified as "diseased"; null hypothesis P(D) = 0.47]. Among these leaves, 30 were randomly sampled for model validation. The ability of the model to distinguish between low and high risk for a leaf to be diseased was tested by ROC (receiver operating curve) curve analysis. The ROC curve plots the sensitivity (proportion of truly "positive," here "diseased," observations classified as such by the model) function of the 1-specificity (the specificity is the probability of the model predicting the sample "negative," given that the observation is "negative," here "healthy"). The area under the curve yields the ability of the model to correctly distinguish between healthy and diseased leaves. When the AUC is high (>0.8), the model has high discrimination ability.

### RESULTS

### Environmental Growing Conditions

The first leaves were marked on April 29 for Cabernet Sauvignon (2005, exp. 1–2–3), corresponding to a sum of 138.7◦C growing degrees (GDs) and on April 14 and April 19 for Merlot, corresponding to 52.1◦C GD (2009 exp. 4), and 66.2◦C (2010 exp. 5–6), respectively. For each experiment, there was a linear relationship between leaf age (LA) and the sum of GD for a leaf during its development (LAdev), with an increased difference for older leaves. The GD globally explained the number of leaves per shoot for each experiment, with ∼21◦C for each newly expanded leaf. For Cabernet Sauvignon, the leaves sampled for experiment 1 (early May 2005) were those developed at the coldest temperature: a 13-day-old leaf developed under a similar GD as a 10-day-old leaf from experiment 2 (**Figure 1**). In comparison, for Merlot, the leaves sampled for exp. 6 (late June) encountered the coldest period: 13-day-old leaves developed under a similar GD as 10-day-old leaves from exp. 5 (end of May).

### Infection Rate Function of Leaf Age

In 2005, for the first sampling date (exp. 1), no significant difference in infection efficiency was observed between the different leaf ages of Cabernet Sauvignon (**Table 3**), whereas for the second (exp. 2), and third experiments (exp. 3), which occurred later in the season, the youngest leaves up to 6 days old displayed significantly higher infection rates. For experiments 2 and 3, we could adjust the data of infectious efficiency to the same exponential decrease model (**Figure 2A**, **Table 4**-M1). In such a model, 10-day-old leaves had an infection efficiency that decreased to 0.25. When the data were standardized (data divided

by the maximum % of infection on the shoot), the whole set of data could be adjusted to a decreasing exponential model with an R ² of 0.59 (**Figure 2B**, **Table 4**-M1). Thus, a 10-day-old leaf lost approximately half of its infectivity, compared to the most susceptible ones. For the Merlot cultivar, the rate of infection efficiency also showed an exponential decrease with leaf age. Very young leaves were excluded from the analysis since we could not exclude that the tape test was not disturbed on particularly hairy young leaves (**Figure 5A**, **Table 5**-M1). For experiment 4 (4 June 2009) and 5 (27 May 2010), with similar environmental conditions, the decrease with leaf age was quite similar, despite a higher rate of infection for very young leaves (I0) in experiment 4. For experiment 6 (24 June 2010), leaves up to 16 days old that developed at cooler temperatures showed a high rate of infection, and then the rate decreased very quickly.

## Colony Growth Function of Leaf Age

The maximum diameter of a colony produced by single conidia was, on average, smaller on leaves older than 10 days for Cabernet Sauvignon (**Table 3**). A model with a decreased diameter at the inflection point of 8.35 days (a50), a maximum diameter (M) on young leaves of 6.8 mm and a minimum diameter (m) of 4.6 mm for older leaves, fit the data with an R <sup>2</sup> = 0.41 (**Table 4**- M2, **Figure 3A**). On average, the sporulation of a colony was exponentially correlated with the maximal diameter attained by the mycelium network (**Table 4**-M4, **Figure 3B**), but at <5 mm, the level of sporulation was low and constant, indicating that a colony with only a two-millimeter difference in diameter (e.g., between 5 and 7 mm) will generate twice as many more spores. For Merlot, the trends and models were the same but with different parameters depending on the experimentation. In experiments 4 and 6, the maximum diameters were greater than in experiment 5. Similar to infection efficiency, slower growth appeared for older leaves in experiment 6 (a50 = 16.93 vs. 12.6 or 12.17 for exp. 5 and 4) (**Table 5**-M4, **Figure 5B**). Sporulation per colony was much higher for experiment 4 (**Figure 5D**).

### Sporulation Function of Leaf Age

The variable sporulation (Spo) measured on inoculated leaf discs results from three successive processes: infection efficiency, mycelium growth and sporulation. From this point of view, it could be considered an integrative variable of susceptibility. The Spo variable followed a clearly bell-shaped curve function of leaf age for experiments 1 and 2, with higher levels of spores produced on 3- to 6-day-old leaves and a decrease for 8-dayold leaves (**Figure 4A**, **Tables 3**, **4**-M3). For experiment 3, on average, the youngest leaves had the highest level of sporulation. When the three experiments were pooled and standardized to homogenize the shoots, the same bell-shaped curve was obtained (**Figure 4B**, **Table 4**-M3) with a high level of sporulation for leaves aged between 3 and 8 days, corresponding to a plastochron index (LPI) between 1.3 and 1.7 (first to second leaf expended). Sporulation per colony (Spoc), which is independent of infection efficiency, also had this bell-shaped curve (**Figure 4C**, **Table 4**- M3). For experiment 2, for which the data were most numerous, the number of spores produced per area of sporulating colony (the sporulation capacity) was significantly higher (P = 1) for very young, susceptible leaves (1–6 days), compared to older leaves (≥10 days) (295 sp.mm² against 167 sp.mm²). This finding is consistent with the hypothesis that, for very young leaves, the higher sporulation results from an increase rate in the hypha branching combined with an increased number of spores produced per conidiophore. For Merlot, the sporulation curve with leaf age was similar to that observed on Cabernet Sauvignon, but the stage of high sporulation (Spo and Spoc) lasted longer, with increased sporulation for 5- to 16-day-old leaves with a peak at 6, 9, or 14 days, depending on the experiment (**Figure 5C**, **Table 5**-M3). This outcome matched an LPI of 2.5 (exp. 6) to 3.3 (exp. 4).

### Leaf Physiological Evolution

Total non-structural carbohydrate (TNC) concentrations tended to follow sigmoid curves regarding leaf age. The soluble sugars concentration (SS) displayed a strong and rapid increase for young leaves, followed by a plateau for older ones. For Cabernet Sauvignon, this change appeared on 8- to 12-day-old leaves (LPI



Inf, percentage of infectious spores; Spo, sporulation of inoculated leaf disks; Spoc, sporulation of the colony; Diam, maximal colony diameter. Different letters near the average indicate significant difference at P < 0.05 according to the Tukey-Kramer multiple comparison test.

= 2.7 to 3.4), while for Merlot, it appeared later: 10- to 14-day-old leaves (exp. 4 and 5, LPI = 3.9) to 14-to 16-day-old leaves (exp. 6, LPI = 3.7) (**Figures 6A,C**). The level reached for old leaves depended on the experiments. The leaf starch concentration was also highly variable with leaf age, but no starch was detected within the youngest leaves (<10 days) (**Figures 6B,D**). Leaves showing an increase in SS reached ∼50% of their final size (∼150 and 270 cm² for, respectively, CS and M), consistent with the time for transition between sink-to-source tissue (**Figures 7A,C**). The increases in the leaf soluble sugar content and leaf age were associated with a decrease in leaf water content (**Figures 7B,D**).

Thus, the processes of infection efficiency, sporulation and colony growth associated with leaf age showed similarities between the experiments on Cabernet Sauvignon and Merlot: a lower rate of mycelium growth and the end of the peak of a high level of sporulation matching an increase of sugar (SS); in comparison, the decrease in infection efficiency was more gradual with leaf age, as was the decrease in leaf water content. Infection efficiency is more closely correlated with leaf age (−0.68), leaf water content (−0.41), and starch (−0.32) than is sporulation (−0.54, −0.32, −0.30). Sporulation per colony was also significantly correlated with soluble sugar (−0.25), although it was more influenced by the experiments on Cabernet Sauvignon, for which we had more colonies for a wider range of leaf ages. For Cabernet Sauvignon, the proxy of the sporulation capacity (number of spores produced per unit area of sporulating colony) was significantly correlated with leaf age (−0.52) and with the amount of sugar (−0.64), indicating that colonies on



TABLE 5 | Results of model fitting for comparison

 of infection rate (Inf), colony diameter (Diam), and sporulation

 rate (Spo) for experiment

 on Merlot.

colony; RSpo, relative rate of spores generated per cm2 of inoculated leaf.

young leaves with smaller amounts of sugar are significantly able to produce more spores per colony. When a period of cold is encountered, processes are delayed (increased sugar, leaf development, decreased leaf water content), and the duration of the sensitivity period is lengthened (exp. 6). Very early in the growing season (exp. 1), all of the leaves are as susceptible as each other to infection.

### Disease Severity Prediction

Partial least-squares path modeling allowed for highlighting the relative contributions of the studied composite variables ("Leaf growth" and "Leaf physiology") in the global explanation of the global "Disease" variable, depending on the experimental conditions ("Variety," "Climate," "Year"). The "Disease" variable was explained at R ² = 0.61 by "Leaf growth," "Leaf physiology," and "Year," with relative contributions of 57.3, 19.5, and 23.2%, respectively (**Figure 8**). Variations in leaf physiology were explained at R ² = 0.50, mainly by leaf growth (87%) (leaf age, leaf temperature for development and leaf surface). "Variety" and "Climate" only poorly contributed to leaf physiology level. When considering total effects (indirect plus direct effects) on the disease, "Leaf growth" had a higher effect on "Disease" (−0.68) compared with "Leaf Physiology" (−0.27), due to the correlation between variables. Leaf water content (1/LWC) and simple sugars (SS) made the greatest contributions to "Leaf physiology" (w = 0.62 and 0.38, respectively), indicating that, regardless of whether sugar (SS) increases or water content decreases, the "Disease" variable decreases. The "Year" effect certainly reflects the variation of isolate aggressiveness that was obvious when considering the level of sporulation per colony diameter. The relative goodness of fit of the model was high (GoFrel = 0.85) with moderate absolute goodness of fit (GoFabs = 0.55), indicating that the predictors used only partly explained disease variation.

Logistic regression showed that, with a maximum of one covariable, leaf age was the best predictor (Pr > LR <0.0001, β<sup>0</sup> = 4.12, β<sup>1</sup> = −0.20), with sensitivity of 86.02% (73.3% for the validation sample) and very high discrimination (AUC = 0.927) of "diseased" vs. "not diseased" leaves. The individuals not well assigned were those at the transition state or very young leaves that were not sufficiently sporulating but that were considered susceptible by the model. If we analyzed the probability after model fitting, there was a probability for a leaf to be susceptible of 0.8 for an average leaf age of 13.3 days [11.4; 14.9] and a probability decrease at 0.5 for a leaf age of 20 days [18.7; 21.4]. With two co-variables, leaf age and starch content were the best predictors, but the model was only slightly improved (AUC = 0.936) with slightly higher specificity (prediction that the leaf is healthy given that it is observed to be healthy).

## DISCUSSION

Understanding and quantifying the process of ontogenic resistance of the grapevine leaf of the two susceptible, first widely cultivated wine cultivars Cabernet Sauvignon and Merlot to one of the major fungal pathogens, E. necator is a prerequisite to promote more sustainable pest management strategies in viticulture. Although ontogenic resistance on grapevine bunches has been well documented (Ficke et al., 2002; Gee et al., 2008), little evidence is available of similar processes for leaves, although leaves are the primary targets for the onset of epidemics. In this study, we attempted to analyse and quantify how the life traits of the pathogen were impacted by ontogenetic resistance and find predictors on the plant side associated with them. To obtain a generic view of the phenomenon, we used shoots coming from vinestocks grown in vineyards under different environmental conditions, and we used several predictable variables that were more or less easy to measure and relevant to analyse leaf state change.

Pathogen behavior was analyzed in relation to leaf age and was characterized by: (i) highly infectious colony growth and sporulation rates for young leaves (<10 days for Cabernet Sauvignon and <16 days for Merlot), with significantly higher sporulation on very young leaves; and (ii) lower colony growth, with a strong decrease in the sporulation rate and in the relative infection rate for older leaves (>10–16 days). Then, all of the chronologic infectious steps of the pathogen—infection efficiency, mycelium growth and sporulation—are affected by the leaf age. The sporulation capacity is even increased on very

disk) (Spo) (A), relative sporulation (RSpo) (B) and number of spores generated per colony (Spoc) (C) as a function of leaf age for experiments on Cabernet Sauvignon. Plain lines correspond to the fitted model M3 in Table 4. Dashed lines correspond to the 95% confidence bands on the models, and bars indicate standard deviation.

young leaves, at least for Cabernet Sauvignon. This increase in sporulation was correlated to low sugar content, whereas the decrease in infection efficiency was more closely correlated with a decrease in leaf water content and an increase of starch. If we consider the leaf plastochron index, instead of leaf age, compared with other studies, the peak of sporulation was 1.4 for Cabernet Sauvignon and 2.7 for Merlot. These values of LPI varied by less than one point for the average temperature encountered during the experiments (15.9–18.7◦C), indicating that, at each time during the growing season, very few leaves are potentially highly susceptible (∼2 to 3 per shoot), and the landscape of susceptibility should be higher for Merlot than for Cabernet Sauvignon. In a previous study (Merry et al., 2013), Cabernet Sauvignon leaves showing the maximum severity were those at LPI 3.7 at 18◦C (constant temperature in the glasshouse), and the range of maximum severity was for leaf positions 2.1- 4.8. At 25◦C, for a higher rate of leaf emergence, the mean modal leaf position for maximum severity was 4.4. Comparatively, in our study with leaves coming from field shoots under natural irradiance and fluctuating temperatures, the maximum severity (corresponding more or less to Spo) was earlier (1.4), with the decrease of leaf susceptibility even sharper. In another study performed on cuttings in growth chambers, colony hyphal length and infection efficiency decreased on both resistant and susceptible cultivars as leaves matured (LPI 1 to LPI 8), with, however, high variations between cultivars (Doster and Schnathorst, 1985).

The high infection rate and increased colony growth on young leaves resulted in an increase in penetration sites and haustoria formation favoring pathogen nutrient uptake and, consequently, a higher sporulation rate. In our study, the very young, susceptible leaves were characterized by a small amount of glucose content, higher water content and a high leaf extension rate. These expanding leaves reached only ∼1/3 of their final size, consistent with the status of sink tissues for grapevines (Lebon et al., 2008; Scott, 2008). Sink leaves depend on other leaves for their energy supply, and the carbohydrates that they receive from source plant parts are immediately used for respiration, growth and the synthesis of new cell components and are therefore not stored. Sink leaves correspond to leaf plastochron indices of 1–3, showing maximal photosynthetic activity (30% for LPI 3) (Zufferey et al., 2000) and higher day respiratory rates (Rd) over the whole growing period. Rd values in young leaves (LPI 3-5) on primary shoots were two times greater than those of mature leaves (Zufferey, 2016). The leaves in transition from sink to source showed a high increase in sugar levels and were able to export sugars when the supply in carbohydrate exceeded the needs for metabolism. In our study, this transition matched decreases in mycelium growth, sporulation and infection efficiency when the number of source leaves exceeded that of sink leaves. When it was not the case, early in the growing season (like for experiment 1), the infection efficiency was not affected by leaf age, and only sporulation decreased for the two basal leaves. The correspondence between the sink to source transition and susceptibility to powdery mildew were once identified on leaves from Cabernet Sauvignon cuttings by autoradiography after <sup>14</sup>CO<sup>2</sup> labeling (Merry et al., 2013). Our study provided some clues about the ontogenic resistance process by decomposing the sequential development of the pathogen (infection rate, colony growth and sporulation), which converges toward a specific age-related transition in the receptivity of the leaf tissues. The complete life cycle of the fungus (up to sporulation) is restricted to sink tissue (small amount of sugar), while the first steps (infection, colony growth) can extend later. Based on these experiments for different environmental conditions and two varieties, we can accurately predict the susceptibility of the leaf with its age as a predictor: a leaf aged on average 13.3 days had a very high probability (0.8) of being susceptible, while this probability dropped to 0.5 one week later.

FIGURE 5 | Infection efficiency rate (Inf) (A), colony diameter (Diam) (B), and sporulation (number of spores produced per cm² of infected leaf disc) (Spo) (C) as a function of leaf age and sporulation per colony (Spoc) as function of colony diameter (D) for Merlot experiments. Plain lines correspond to the fitted models M1 (A), M2 (B), M3 (C), and M4 (D) in Table 5. Dashed lines correspond to the 95% confidence bands on the models, and bars correspond to the standard deviation.

Decreases in incidence and severity of powdery mildew with leaf age or development stage were also identified on strawberry (Carisse and Bouchard, 2010) and were reported to be a robust and biologically consistent process with a response consistent across cultivars and isolates employed and between the various sites tested (Asalf et al., 2014). Regarding pea powdery mildew, a resistance gene involved in post-penetration cell death and a reduction in percentage penetration success in mature leaves confers complete resistance in the field under high temperatures or on mature leaves (Fondevilla et al., 2006). On Eucalyptus grandis, all of the steps of the development of the rust Puccinia psidii—germination, appressorium formation, penetration and sporulation—were altered with increased leaf age (Xavier et al., 2015). For other pathosystems, the susceptibility can increase with leaf age, as is the case for another rust, Melampsora larici populina, which is more aggressive on mature leaves of poplar (Maupetit et al., 2018). The authors observed an increase in uredia (colony) size and a higher sporulation rate with increased LPI and lower sporulation capacity. In this case, pathogen traits were not correlated with leaf sugar, but there was a negative correlation with phenolic content, which was higher in younger leaves. On pea, detached stipules were more receptive to Mycosphaerella pinodes as soon as visual senescence was observed (Richard et al., 2012).

### Which Factors Are Involved in the Variation of the Ontogenic Resistance Process?

Our results attest to the repeatability of ontogenic resistance on leaves during vegetative growth for two susceptible cultivars. However, our results only explained 61% of disease variation. These variations can be linked to sampling shoots with differences in the implementation age of the ontogenic resistance process or between years/cultivars with variations in disease level. To explain variation between years, one parsimonious hypothesis is that it is based on differences in isolate aggressiveness. Indeed, a strong difference in sporulation for the same colony diameter was observed for different years. Variations between shoots and experiments depend on factors with effects on the sinkto-source status of the leaf, such as exposure to light before field sampling, crop load or even leaf damage. The transition to leaf resistance is, however, clearly related to the timing of the sink-to-source status rather than to the final level of glucose attained. Variations were also observed when leaves encountered a cold temperature, which slowed the physiological processes and delayed the timing of ontogenic resistance establishment (such as in experiment 6). We also observed that infection efficiency was not significantly different for the first leaves that emerged early in the season, which could emphasize the particular status of these first expanding leaves for which the source of assimilates used for growth first comes from retranslocation from the parent vine (Keller, 2015). Using <sup>14</sup>C tracer, it has been shown that, at the 5-leaf stage shoot, only the first basal leaf is able to export assimilates to the youngest one, and at the 6-leaf stage, two to three basal leaves behave as sources (Yang and Hori, 1980). The experiments were conducted on a hybrid of Labrusca, but this outcome was consistent with what we observed. Indeed in our first experiment on Cabernet Sauvignon, only the two basal leaves being in the transition from sink to source showing significantly less sporulation. This should enhance infection success when epidemics start very early in the season and surely contributes to their particularly damaging effects.

Although infection efficiency is very low on old leaves, it is not always null. We do not know the impact of these rare infections considering that they were not sporulating at the time

of assessment. In the pathosystems of apple/apple scab, diffuse colonies and mycelium that formed on old leaves continued to grow (Li and Xu, 2002), and these authors wondered whether these diffuse colonies could sporulate later in the season and become a source of mycelium for the formation of pseudothecia. On grapevines, we did not exclude that the translocation of carbohydrates from old leaves to berries during grape ripening would not modify leaf susceptibility or allow rare infected spores to continue their development and affect late colonies. This phenomenon has already been observed for other pathosystems, as especially demonstrated in the case of hermaphrodite plants, such as papaya (Carica papaya L.). Female fruit-bearing plants are susceptible to Oidium caricae on old leaves and fruit, whereas male plants are not attacked (Jarvis et al., 2002). These authors have advanced hypotheses of the shunting of assimilates from leaf to fruit to explain the differences. However, we could not observe

inner latent variable. Blue lines indicate negative correlations, and red lines indicate positive correlations.

any modification of ontogenic resistance processes after bunch closure (late July) on Merlot.

### What Hypotheses Can Be Advanced to Explain the Process of Ontogenic Resistance Observed on Old Leaves? Role of Sugar

The increase in carbohydrate concentration in old leaves could be involved in the activation of secondary metabolism. Soluble carbohydrates are known to control the expression of various metabolic and plant defense-related genes (Koch, 1996; Rolland et al., 2006), and high sugar content can also suppress fungus degradation enzymes, such as grape chitinase (Saito et al., 2011). Powdery-mildew diseases have already been classified as "high-sugar resistance" pathogens (Vanderplank, 1984), indicating reversal of susceptibility to high levels of sugars. The relationship between sugar amounts and osmotic pressure might also be involved (Koroleva et al., 2002). Increased osmotic pressure can challenge the pathogen during the first step of infection, slowing tissue colonization by limiting the process of penetration of secondary haustoria. This scenario is consistent with our observed reduction of colony diameter on old leaves. Increased osmotic pressure was hypothesized to explain the greater resistance of mature leaf epidermal cells to other powdery mildews on apple (Cimanowski and Millikan, 1970) and peach (Schnathorst and Weinhold, 1957; Weinhold and English, 1964).

### Activation of the PTI

Transcriptome and proteome analyses showed that the susceptible variety Cabernet Sauvignon is able to initiate basal defense mechanisms (Fung et al., 2008; Fekete et al., 2009; Marsh et al., 2010). Transcriptome analyses of leaves (Dufour et al., 2016) also revealed overexpression of defense genes (phenylalanine ammonialyase, stilbene synthase, anthocyanidin synthase, chitinase, lipase) on Cabernet Sauvignon leaves at ages of 15 and 20 days, compare to leaves at ages of 7 and 10 days, only after powdery mildew inoculation (unpublished data). For young expanding leaves, a high rate of cellular reactions of plant defense might be energetic and, for the carbohydrates, its consumption too costly (Bethany et al., 2014). MLO genes, known as "powdery mildew susceptibility genes," might be involved in this regulation. It was hypothesized that adapted powdery mildew species release effectors that, by combining with specific MLO proteins, could suppress the host PTI (plant triggered immunity) in the young leaves (Dry et al., 2010). One hypothesis explaining ontogenetic resistance is that changes occur in the old leaves in the expression or activity of the MLO genes with the consequence of restoring the PTI (Qiu et al., 2015).

### Constitutive Defenses

The transition in the trophic statute of the leaf bound up with development can also trigger the establishment of constitutive defenses to infection: cuticle thickness in Lactuca sativa, (Schnathorst, 1959) and Prunus persica (Weinhold and English, 1964); or the synthesis of antimicrobial compounds (Tattersall et al., 1997; Salzman et al., 1998). Precocious events of infection progress, such as germ tube initiation and appressorium differentiation, require physical or chemical cues (Muller and Riederer, 2005). Variations in cuticular waxes (quality and amount) during ontogenic development could be an inhibiting factor, consistent with the lower rate of conidium germination on old leaves that we observed. Constitutive defenses and PTI can be both active in source leaves.

### How Can Ontogenic Resistance Be Used in Epidemic Management?

Given the repeatability of the results, it is unlikely that the susceptibility of young leaves can be reduced without significant cost for the plant. Our results clearly demonstrated that powdery mildew takes advantage of the compromise that the plant must make between growth and defense (Bethany et al., 2014). Even if it is tempting to think that new genome editing techniques will allow for the easy knockout of genes, such as MLO, the effect of these knockouts on the physiology of the plant remain unknown (Acevedo-Garcia et al., 2014; Malnoy et al., 2016). The rate of both primary and secondary leaf appearance can, however, be slowed by the winegrower, hence reducing the proportion of available susceptible young tissues per unit of time. Indeed, agronomical practices, such as cover-cropping and the use of a vigor-controlling rootstock, can decrease the rate of leaf appearance and consequently disease level (Calonnec et al., 2009; Schnee et al., 2011; Valdes-Gomez et al., 2011). Simulations have shown that a reduction in grapevine vigor, through a decrease in both the numbers and the rates of secondary leaf appearance, can slow epidemics (Burie et al., 2011; Mammeri et al., 2014). Other practices, such as thinning of old leaves, will decrease the photosynthesis of younger leaves, thereby potentially delaying their ontogenic resistance, whereas shoot topping of younger leaves does not impact leaf photosynthesis per unit of leaf area (Petrie et al., 2003). Consequently, thinning might be less favorable to increasing leaf resistance than shoot topping, without considering the potentially positive effect of light on tissue receptivity (Zahavi and Reuveni, 2012). Other cultural practices, such as minimal pruning, which results in increased shoot numbers, reduce leaf and shoot size, and more synchronous leaves emergence (Poni et al., 2000), could allow the plant to escape disease by simultaneously causing all leaves to behave as a source. Such cultural practices could also be used to enhance the resistance of varieties with specific resistance genes. It would be interesting to revisit the theory of grapevine balance using tools of genetics and physiology in a general context of dieback and climate change, in which yields are sought, while maintaining control of pathogens that are very deleterious, such as mildews, is a prerequisite. Coupling functional structural grapevine physiology (Weinhold and English, 1964) and pathogen development models (Calonnec et al., 2008) would help to better understand crop susceptibility under various environmental conditions and complex canopies, while early protection of the foliage, which allows for desynchronizing of the disease cycle and the crop cycle toward a higher percentage of resistant leaves, should be preferred.

## CONCLUSIONS

Our results revealed a clear correlation between the appearance of grapevine leaf ontogenic resistance and the leaf transition state from sink to source. The soluble and complex sugars, leaf water content and leaves' extension rate constitute indicators of a change in the metabolic processes and are indirect markers of tissue age-related resistance. From a practical point of view, the additional effects of ontogenic resistance modification through crop management and resistant cultivars should be explored to obtain better control of powdery mildew epidemics in sustainable agriculture (Calonnec et al., 2013a). Moreover, the integration of this developmental process into a grapevine disease management program requires reconsideration of fungicide spraying with regard to the formatting of organ susceptibility windows.

### AUTHOR CONTRIBUTIONS

AC: head of the research programme, data analyses, writer. JJ: experimental technician. PV: collaborator for physiology data analysis and experimentation. SS: post-doctoral student, experimenter, writer.

### REFERENCES


### FUNDING

This study was supported by a grant from the Agence Nationale de la Recherche program SYSTERRA (ANR-08-STRA-04) and by the Aquitaine Regional Council.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018. 01808/full#supplementary-material

Supplementary Figure S1 | Scheme of the experimental design.


in Eucalyptus grandis leaves of different ages. Trop. Plant Pathol. 40, 318–325. doi: 10.1007/s40858-015-0043-7


the field: I. Interaction of age, light and temperature. Vitis 39, 19–26. Available online at: https://ojs.openagrar.de/index.php/VITIS/article/view/4599

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

Copyright © 2018 Calonnec, Jolivet, Vivin and Schnee. 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.

# A Trade-Off Between Sporangia Size and Number Exists in the Potato Late Blight Pathogen Phytophthora infestans, and Is Not Altered by Biotic and Abiotic Factors

Nicolas Mariette, Alexander Kröner, Romain Mabon, Josselin Montarry, Bruno Marquer, Roselyne Corbière, Annabelle Androdias and Didier Andrivon\*

INRA, UMR1349 Institute for Genetics, Environment and Plant Protection, Le Rheu, France

### Edited by:

Michele Perazzolli, Fondazione Edmund Mach, Italy

### Reviewed by:

Remco Stam, Technische Universität München, Germany Daolong Dou, Nanjing Agricultural University, China

> \*Correspondence: Didier Andrivon didier.andrivon@inra.fr

#### Specialty section:

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

Received: 13 July 2018 Accepted: 27 November 2018 Published: 19 December 2018

#### Citation:

Mariette N, Kröner A, Mabon R, Montarry J, Marquer B, Corbière R, Androdias A and Andrivon D (2018) A Trade-Off Between Sporangia Size and Number Exists in the Potato Late Blight Pathogen Phytophthora infestans, and Is Not Altered by Biotic and Abiotic Factors. Front. Plant Sci. 9:1841. doi: 10.3389/fpls.2018.01841 The negative relationship between offspring size and number is a classic example of trade-off between life-history traits, reported many times in animal and plant species. Here, we wanted to ascertain whether such a trade-off occurred in the oomycete Phytophthora infestans, and whether it was impacted by biotic and abiotic factors. We thus conducted three infection experiments under controlled conditions and measured the number and the size of sporangia (asexual propagules) produced on potato by different P. infestans isolates. In all experiments, we observed a negative relationship between sporangia size and number, demonstrating the existence of a trade-off. Moreover, although the potato host cultivar, temperature and host of origin (tomato or potato) all affected sporangia number, sporangia size or both, none of these biotic and abiotic factors did change the trade-off. Therefore, the trade-off between sporangia size and number could maintain the polyphenism for these traits in P. infestans populations, and favors the coexistence of distinct reproductive strategies within this species. Our results emphasize the relevance to focus on the relationship between offspring size and number in other fungal plant pathogens, as well as to study the impact of offspring size on fitness-linked traits (virulence and disease lesion development) in these organisms.

Keywords: fitness, life-history traits, plant pathogen, reproductive strategies, Solanum tuberosum

## INTRODUCTION

The widely accepted trade-off concept states that an organism cannot simultaneously maximize all traits involved in fitness, owing to dilemmas in resource allocation (Law, 1979). According to this theory, any beneficial change in one trait will unavoidably have a detrimental effect on another trait (Anderson and May, 1982; Stearns, 1989). The net result is a negative correlation between traits that mobilize the same resources. The trade-off concept has a prominent importance in evolutionary biology, as it can explain how the adaptation of populations to new environments can be constrained and how life-history strategies can evolve (Rausher, 1984; Roff, 1992). Tradeoffs between fitness components are of central importance in plant pathogens, as they could limit their evolutionary potential (Lannou, 2012; Laine and Barres, 2013) and may thus be exploited to engineer durable control strategies (Quenouille et al., 2013; Brown, 2015).

Since they are key to determining fitness, trade-offs involving reproductive traits, especially the one between offspring size and number, have received the most attention in the scientific literature (Stearns, 1989; Roff, 2002; Begon et al., 2006). The trade-off between offspring size and offspring number is based on the hypothesis that, for a given total reproductive investment, if most resources are devoted to the production of many offspring individuals, each of them will have limited lower share of the reproductive investment (Begon et al., 2006). If this trade off exists (that is, if resources available for reproduction are constrained), reproductive strategies should range along a continuum, from the production of few large offsprings to the production of many small ones (Timi et al., 2005). This size-number trade-off was first advocated by Lack (1947) with experiments on clutch-size in birds, and has been subsequently observed in many animals (Charnov and Ernest, 2006; Walker et al., 2008) and plants (Werner and Platt, 1976; Jakobsson and Eriksson, 2000; Linkies et al., 2010).

The size-number trade-off could also be found in plant parasites, especially oomycetes and fungi. For such sporulating organisms (with spores as asexual propagules), producing large spores should indeed be more energetically costly than producing small spores, but may increase spore viability or infectious ability. Although trade-offs involving spore production on one hand and other life-history traits such as latent period (Pariaud et al., 2013), virulence (Thrall and Burdon, 2003; Montarry et al., 2010) or transmission success rate (Pasco et al., 2016) on the other hand, were observed in various plant pathogens, very few studies have investigated the existence of a trade-off between spore size and number (Delmotte et al., 2014). Fortunately, the advent of new techniques and equipments for counting and measuring thin particles like spores now make the exploration of such a trade-off possible.

As in all pathogenic interactions, the development of a disease caused by a plant pathogen is strongly influenced by both abiotic environment and host genotype (Laine, 2007; Wolinska and King, 2009; Clément et al., 2010; Delmotte et al., 2014). As a consequence, these factors can also have an impact on the shape and even the direction of the correlation between life-history traits in pathogens. Hence, fitness costs can differ among environments (e.g., temperature) as observed in the fungal pathogen Podosphaera plantaginis (Laine, 2008). Likewise, some studies reported shifts of life-history correlations in plant pathogens, depending on the cultivars tested (Huang et al., 2010; Susi and Laine, 2013). These data thus stress that it is crucial to take into account sufficient variation in both the host and the parasite, by measuring life-history traits in different environments, to draw conclusions on the status of a life-history trade-off in a plant pathogen.

Here, we aimed to investigate the existence of a trade-off between offspring number and size in the oomycete Phytophthora infestans, and the impact of some biotic and abiotic factors on this eventual trade-off. Infamous for having triggered the Irish Great Famine in the 1840s, P. infestans is the causing agent of late blight in solanaceous species and is responsible for significant losses in both potato and tomato crops (Fry et al., 2015). This pathogen is particularly suitable for exploring trade-offs, as it can be isolated and maintained as axenic cultures on artificial media whereas miniature biotests performed in controlled conditions can be used for the assessment of life-history traits of the pathogen, particularly those linked to spore formation (Montarry et al., 2010). We thus conducted three experiments under controlled conditions in which we measured the size and number of sporangia (i.e., asexual spores) produced by P. infestans isolates after artificial inoculations on potato leaflets. In each experiment, we changed one factor [host resistance level, temperature, or the host of origin (tomato or potato)] which could impact the trade-off. We measured the effect of these factors on both sporangia size and number, analyzed the relationships between these two traits and compared them within each set of conditions.

### MATERIALS AND METHODS

We conducted three inoculation experiments of P. infestans isolates on detached potato leaflets in which we measured the size and the number of sporangia produced by the pathogen. Different biotic and abiotic parameters – host resistance level, temperature and host of origin (tomato or potato) – were modified in these experiments. The experimental protocols used for these experiments were mostly similar but contained some slight differences (**Table 1**).

### Isolate Collections

Single-lesion isolates were established and maintained as axenic cultures on pea agar medium as previously described (e.g., Montarry et al., 2010). Briefly, each single-lesion isolate was obtained by placing a fragment of infected leaf tissue on tuber slices of the susceptible potato cultivar Bintje. After incubation at 15◦C in growth chambers for four to eight days, pure cultures were established by transferring the hyphal tips growing through the slices to sterile pea agar and subsequently maintained at 15◦C in darkness by serial transfers on fresh pea agar.

In the first experiment, we used 119 isolates sampled in commercial potato fields of western France during late blight

TABLE 1 | Protocol specificities applied for each experiment lead during the study.


epidemics of 2014 (Mariette, 2016). All isolates were tested on three potato cultivars: Bintje, Robijn, and Möwe. These cultivars do not bear race-specific resistance genes but differ in levels and components of partial resistance to late blight: Bintje is the reference susceptible cultivar to P. infestans commonly used in European laboratories, whereas the partially resistant Robijn and Möwe notably reduce spore production of the pathogen (Clément et al., 2010). In the second experiment, 16 western French isolates sampled in 2013 were tested on the potato cultivar Bintje under three thermal regimes: 10, 14, and 18◦C. This temperature range was chosen because it covers the biological activity of P. infestans (Maziero et al., 2009). The third experiment included 21 isolates sampled either on potato (11) or tomato (10), in French and Algerian crops from 2013 to 2015, and were tested on the potato cv. Bintje.

### Plant Material

Three potato cultivars (Bintje, Robijn, and Möwe) and one tomato cultivar (Marmande) were used. Certified seed tubers of the potato cultivars were obtained from the INRA Biological Resource Center BrACySol (Ploudaniel, France), whereas commercial standard seed of the tomato cultivar was used.

Plants were grown in pots filled with 1:1:1 sand-peat-compost mixture placed in a glasshouse regulated at 15–20◦C (night/day temperatures) with 16 h of photoperiod. Once a week, plants were watered with a nutrient solution (Hakaphos; NPK 15/10/15). For the inoculum preparation and life-history traits measurements, leaflets of similar size were picked from the median area of 6- to 8-week-old plants.

### Inoculum Preparation

Prior the assessment of their life-history traits and in order to restore pathogenicity possibly lost in axenic cultures (Jinks and Grindle, 1963), P. infestans isolates were first inoculated onto detached leaflets of the original host species: the potato cv. Bintje and the tomato cv. Marmande for potato and tomato isolates, respectively. For this purpose, droplets of sporangial suspensions, prepared from 3- to 4-week-old pea agar cultures by flooding with sterile water and scrapping the colony surface, were deposited on the lower (abaxial) side of detached leaflets kept on the empty lids of inverted Petri dishes containing 10 g L−<sup>1</sup> water agar and acting as humid chambers. After 7 days of incubation at 18/15◦C (day/night temperature) and 16-h day length, newly formed sporangia were collected from infected leaflets by gently shaking in sterile water. Sporangia were then counted using a hemocytometer, adjusted to a final concentration of 3 × 10<sup>4</sup> or 5 × 10<sup>4</sup> sporangia mL−<sup>1</sup> , depending on the experiment (**Table 1**), and then kept at 4◦C for 2 h to promote zoospore release.

### Inoculation

Sporangial suspensions were inoculated onto a potato leaflet, except in the second experiment where three leaflets of potato cv. Bintje were inoculated due to the three thermal regimes tested (**Table 1**). To this end, leaflets were placed, abaxial face up, onto the empty lids of inverted Petri dishes containing 10 g L−<sup>1</sup> water agar. A 20-µL droplet of the prepared sporangial suspension was then deposited on the center of the leaflet. The Petri dishes were kept in clear boxes and incubated in climate chambers regulated at thermal conditions desired (**Table 1**) with a 16 h photoperiod.

### Life-History Traits Measurements

Three days after the formation of the first sporangia – checked by daily observations under a magnifying glass – or 5 days after inoculation (depending on the experiment, **Table 1**), newly formed sporangia were washed from each leaflet in 10 mL Isoton II (saline buffer; Beckman Coulter, Villepinte, France). Suspensions were kept in glass tubes at −20◦C until the counting of sporangia number (SN) produced using a Coulter Z2 counter (Beckman Coulter) equipped with a 100-µm aperture tube. Sporangia size (SS) was determined under a microscope fitted with a camera and using the image analysis software Histolab <sup>R</sup> v8.1.0 (Microvision Instruments, Evry, France). For this purpose, drops of the sporangial suspension were placed under the microscope and we used the camera measurement software to determine length and width of 100 randomly chosen sporangia. These measures allowed to calculate sporangia volumes, assuming a revolution ellipsoid shape for each sporangium (Philibert et al., 2011) by applying the standard equation <sup>4</sup><sup>π</sup> 3 ∗ length 2 ∗ width 2 2 .

### Statistical Analyses

All analyses were performed using the statistical software R v. 3.4.4 (R Core Team, 2017) and the significance threshold was fixed at α = 0.05. We first compared SS and SN of P. infestans isolates tested on different potato cultivars (experiment 1) and at different temperatures (experiment 2) using analysis of variance (ANOVA) followed by Tukey's honestly significant difference post hoc comparisons. For the comparisons between isolates sampled on potato and tomato (experiment 3), we applied a Student t-test. When necessary, SS and SN were log-transformed to achieve assumptions of homoscedasticity and normality. For the three tested modalities (potato cultivar, temperature and host of origin), each P. infestans isolate was regarded as a biological repetition. Then, we investigated the relationship between SS and SN for each modality tested in the three experiments using Pearson correlations and linear regressions (where SS and SN were log-transformed). To test whether slopes of regression lines differed between modalities within each experiment, variation in SS was analyzed by mean of analysis of covariance (ANCOVA) with SN as the covariate and host cultivar, temperature or host of origin as the factor. In the first experiment implying 119 isolates, phenotypic measurements were made only once on each of them as the large number of tested individuals allowed a good assessment of significance. In the case of experiments 2 and 3, we compared data obtained from the two independent replicates and, as results showing no significant differences (i.e., the criterion was statistically fulfilled or rejected), data from both experiments were pooled, reanalyzed and represented graphically.

### RESULTS

### Impact of Biotic and Abiotic Factors on P. infestans Life-History Traits

In experiment 1, potato cultivar had a significant effect on SS of P. infestans isolates (ANOVA, F2,<sup>352</sup> = 3.58, P = 0.029), with bigger sporangia produced on Bintje than on Robijn as revealed by Tukey post hoc tests (**Table 2**). Likewise, a significant effect of potato cultivar on SN was detected (ANOVA, F2,<sup>352</sup> = 158.48, P < 0.001), with the highest number of sporangia produced on Bintje, then on Robijn, when the lowest number of sporangia was observed on Möwe (**Table 2**). In the second experiment, SS was significantly impacted by temperature (ANOVA, F2,<sup>91</sup> = 54.804, P < 0.001): the lower the temperature, the bigger the sporangia (**Table 2**). In this experiment, temperature also had a significant effect on SN (ANOVA, F2,<sup>91</sup> = 106.31, P < 0.001) with in this case, an increase of SN with increasing temperatures (**Table 2**). In experiment 3, no significant effect of isolate origin was detected on SS [t(40) = 1.63, P = 0.111; **Table 2**], nor on SN [t(40) = 1.81, P = 0.079; **Table 2**].

### Relationship Between Life-History Traits

In experiment 1, significant negative relationships between SS and SN of P. infestans isolates were detected in the tests conducted on potato cultivars Bintje (r = −0.36, P < 0.001; **Figure 1**), Robijn (r = −0.49, P < 0.001; **Figure 1**), and Möwe (r = −0.48, P < 0.001; **Figure 1**). ANCOVA revealed no significant interaction between host cultivar and SN (**Table 3**), indicating that the relationship between SS and SN did not significantly differ among potato cultivars. Likewise, in the second experiment, the relationship between SS and SN was significantly negative at 10◦C (r = −0.70, P < 0.001; **Figure 2**), 14◦C (r = −0.62, P < 0.001; **Figure 2**), and 18◦C (r = −0.68, P < 0.001; **Figure 2**). Again, this relationship did not significantly differ among temperatures, since no significant interaction between temperature and SN was detected by ANCOVA (**Table 3**). Finally, in experiment 3, SS and SN were also negatively correlated and this relationship was significant for isolates sampled on both potato (r = −0.57,

TABLE 2 | Mean values (SE) of sporangia size and sporangia number measured in the three experiments for each tested factor (potato cultivar, temperature, and host of origin).


Different letters in a column indicates significant differences between means within an experiment at a 0.05 threshold (Tukey HSD or Student-t-tests).

FIGURE 1 | Relationship between size and number of sporangia produced by 119 P. infestans isolates after inoculations on potato cultivars Bintje, Möwe, and Robijn (experiment 1). All variables were LOG-transformed. For sporangia produced on Bintje, the linear regression is represented by a continuous gray line (y = 10.89–0.14×, R <sup>2</sup> = 0.13, P < 0.001), for sporangia produced on Möwe, the linear regression is represented by a continuous black line (y = 10.58–0.13×, R <sup>2</sup> = 0.23, P < 0.001), and for isolates inoculated on Robijn, the linear regression is represented by a dotted black line (y = 10.22–0.10×, R <sup>2</sup> = 0.24, P < 0.001).

TABLE 3 | Summary of analyze of covariance (ANCOVA) for testing the effects of biotic and abiotic factors on the relationship between size and number of sporangia produced by P. infestans isolates.


In the models, potato cultivar (experiment 1), temperature (experiment 2), and host of origin (experiment 3) were rated as factors, sporangia production (SP) as the covariate and sporangia size as the response variable. Statistical significances are indicated as follow (∗∗P < 0.01 and ∗∗∗P < 0.001). Sporangia size and sporangia number were LOG-transformed.

P = 0.006; **Figure 3**) and tomato (r = −0.58, P = 0.007; **Figure 3**). As in the other two experiments, no significant interaction was revealed by ANCOVA between 'origin' and SN (**Table 3**), indicating that the SS-SN relationship was not significantly different between tomato and potato isolates.

FIGURE 2 | Relationship between size and number of sporangia produced by 16 P. infestans isolates after inoculations on potato cultivar Bintje at 10, 14, and 18◦C (experiment 2). All variables were LOG-transformed. For sporangia produced at 10◦C, the linear regression is represented by a continuous gray line (y = 11.25–0.21×, R <sup>2</sup> = 0.49, P < 0.001), for sporangia produced at 14◦C, the linear regression is represented by a dotted black line (y = 11.58–0.23×, R <sup>2</sup> = 0.39, P < 0.001) and for isolates inoculated at 18◦C, the linear regression is represented by a continuous black line (y = 11.14–0.21×, R <sup>2</sup> = 0.46, P < 0.001).

FIGURE 3 | Relationship between size and number of sporangia produced by 21 P. infestans isolates sampled on potato or tomato after inoculations on potato cultivar Bintje (experiment 3). All variables were LOG-transformed. For sporangia produced by potato isolates, the linear regression is represented by a continuous black line (y = 10.23–0.10×, R <sup>2</sup> = 0.32, P = 0.006) and for sporangia produced by tomato isolates, the linear regression is represented by a continuous gray line (y = 10.20–0.11×, R <sup>2</sup> = 0.34, P = 0.007).

### DISCUSSION

The trade-off between offspring number and offspring size has been intensely studied and well characterized in animals and plant species (Werner and Platt, 1976; Charnov and Ernest, 2006; Walker et al., 2008). Here, we report that it also applies to a microorganism pathogenic to plants, potato late blight oomycete pathogen P. infestans: higher offspring numbers (in this case, asexual sporangia) correlates in this species with smaller offspring size. This trade-off has been seldom investigated in plant pathogens, even though a similar finding was reported in another plant pathogenic oomycete, the grapevine downy mildew pathogen Plasmopara viticola (Delmas et al., 2014). Such a trade-off is thus probably present in other sporulating parasites.

Interestingly, we have found significantly negative relationships between SS and number for all the modalities tested in our study, and with similar slopes within each experiment. Since host cultivar, temperature and isolate origin (potato or tomato) are important factors for the fitness of an almost obligate pathogen such as P. infestans (Andrivon et al., 2013), we could have supposed that these biotic and abiotic environmental variables could have altered the shape of the trade-off. Indeed, previous studies focused on plant pathogens reported for instance that host could modulate trade-offs, like those implying infectivity and transmission (Susi and Laine, 2013) or virulence and lesion size (Huang et al., 2010). Therefore, the stability of the trade-off relationship between SS and SN in P. infestans across a range of biotic and abiotic environments suggest that this trade-off is constitutive in this species, and might be exploited in control strategies of late blight epidemics.

The stability of the trade-off relationship over the range of environments tested is all the more remarkable that the environmental factors investigated strongly impacted one or both of the life history traits involved in the trade off. In fact, we showed that sporangia production of P. infestans was strongly host-dependent, confirming earlier observations made in this species (Clément et al., 2010) and more generally in a range of fungal and oomycete parasites (Hardham and Hyde, 1997; Delmotte et al., 2014). Interestingly, we also showed a clear effect of potato cultivars on SS, with smaller sporangia on Robijn – especially compared to Bintje. A similar observation was reported in P. viticola (Delmotte et al., 2014) and Venturia inaequalis (De Gracia et al., 2015), suggesting a possible selection by host on this trait among sporulating plant pathogens. Regarding temperature, the thermal optimum for P. infestans development is around 18–20◦C (Mizubuti and Fry, 1998; Mariette et al., 2016), and spore production quickly decreases beyond this. In our second experiment, sporangia were much less numerous at the lowest temperatures, but they were bigger. This phenomenon could be due to the longer latent period of the pathogen at low temperatures (Maziero et al., 2009; Mariette et al., 2016), leading to more time for sporangia formation. Finally, we observed that isolates sampled on tomato produced less sporangia on the potato cultivar Bintje than those sampled on potato, confirming the host specialization previously highlighted in P. infestans (Oyarzun et al., 1998; Kröner et al., 2017).

By preventing an organism to simultaneously maximize all traits involved in fitness, trade-offs can equalize fitness across individuals, genotypes, or species with different reproductive strategies, leading to the maintenance of phenotypic diversity

(Roff, 1992). For instance, Héraudet et al. (2008) suggested that the trade-off between latent period and spore production in H. arabidopsidis may explain why the latent period remains variable in nature and did not achieve a uniformly minimal value. The constraint of total reproductive effort imposing a trade-off in the size and number of sporangia in P. infestans, as observed in our study, could therefore maintain variability for these life-history traits within populations of this pathogens. This tends to be confirmed with the large range of sporangia production values generally reported amongst isolates of single P. infestans populations (e.g., Montarry et al., 2008b, 2010; Mariette et al., 2016). Moreover, the polyphenism of life-history traits values imposed by a trade-off can lead to the emergence of various life-history strategies (Roff, 1992). This can be crucial in pathogens highly dependent on their environmental variations, as different life-history strategies can explain the coexistence of genetically differentiated clades, in complete or partial reproductive isolation, within single plant pathogen species (Montarry et al., 2008a).

Here, we report that a trade-off between number and size of asexual spores occurred in an important oomycete plant pathogen, P. infestans, and that its shape was not affected by different biotic and abiotic factors. It would be interesting to know the relationship between asexual spore number and size in other fungal plant pathogens as trade-offs (and their shape) can play a key role in evolutionary outcomes of these organisms (Kamo et al., 2007). Nevertheless, it is far from being the case at present, because if the production of the asexual propagules (spore or sporangia) is a trait extensively studied in fungal plant pathogens (Hardham and Hyde, 1997; Montarry et al., 2010), their size is still often ignored. Yet, the size of fungal propagules may be an important trait for different aspects of their fitness. The infection efficiency should be higher for large spores as they are supposed to have more resources available for hyphal growth (Bässler et al., 2015). Likewise, SS in oomycetes could also impact their indirect germination (formation of zoospores) as suggested by investigations on P. viticola showing that larger sporangia produced more zoospores (Delmas et al., 2014). Conversely, small spores might have higher aerial dispersal abilities and could therefore tend to deposit further to the source than larger ones (Norros et al., 2014). Nevertheless, owing to the lack of empirical evidences, the link between asexual spore size and such fitness aspects remain uncertain (De Gracia et al., 2015).

### REFERENCES


Further research efforts on this topic are thus needed and it would be particularly crucial to study the impact of offspring size on virulence or disease lesion development in fungal plant pathogens. Besides, a future way of research in plant pathogens could be to not only limit on the trade-off between number and size of offspring but, as recommended by Begon et al. (2006), to look for a trade-off between the number of offspring and their individual fitness.

### DATA AVAILABILITY STATEMENT

The datasets generated for this study will be found on data.inra.fr (doi: 10.15454/BB0WJN).

### AUTHOR CONTRIBUTIONS

RM, AA, BM, NM, and AK performed the experiments according to a protocol elaborated jointly by RC and DA. NM and AK analyzed the data. NM, JM, RC, and DA wrote the text and prepared the figures. All authors edited the paper and have approved the current version.

### ACKNOWLEDGMENTS

A part of this work presented in this paper first appeared in the NM thesis, which can be accessed online (see the reference list). We gratefully acknowledge the financial support by INRA ACCAF Metaprogramme (CLIF project) and ARIMnet (PoH-MED project; KBBE 219262/ANR). We also warmly thank Prof. Zouaoui Bouznad and Sihem Belkhiter (ENSA, Algiers) for providing Algerian isolates and Mathilde Bodiou for help with the sampling of P. infestans isolates. We express our thanks to E. Rolland and C. Piriou for their help with isolate maintenance and biological experiments. We are grateful to the INRA Genetic Resources Center BrACySol of Ploudaniel for providing the plant material used in the tests. NM was supported by a Ph.D. project cofunded by INRA and ACVNPT (Association des Créateurs de Variétés Nouvelles de Pomme de Terre) through a CIFRE grant (Conventions Industrielles de Formation par la Recherche). AK was supported by a post-doctoral grant funded under the ARIMnet PoH-MED project.



mildew epidemics. Appl. Environ. Microbiol. 74, 6327–6332. doi: 10.1128/AEM. 01200-08


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

Copyright © 2018 Mariette, Kröner, Mabon, Montarry, Marquer, Corbière, Androdias and Andrivon. 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.

# Responses to Hydric Stress in the Seed-Borne Necrotrophic Fungus Alternaria brassicicola

Guillaume Quang N'Guyen<sup>1</sup>† , Roxane Raulo<sup>2</sup> , Muriel Marchi<sup>1</sup> , Carlos Agustí-Brisach<sup>3</sup> , Beatrice Iacomi<sup>4</sup> , Sandra Pelletier<sup>1</sup> , Jean-Pierre Renou<sup>1</sup> , Nelly Bataillé-Simoneau<sup>1</sup> , Claire Campion<sup>1</sup> , Franck Bastide<sup>1</sup> , Bruno Hamon<sup>1</sup> , Chloé Mouchès<sup>1</sup> , Benoit Porcheron<sup>5</sup> , Remi Lemoine<sup>5</sup> , Anthony Kwasiborski<sup>1</sup> , Philippe Simoneau<sup>1</sup> and Thomas Guillemette<sup>1</sup> \*

#### Edited by:

Christophe Le May, Agrocampus Ouest, France

#### Reviewed by:

Syama Chatterton, Agriculture and Agri-Food Canada (AAFC), Canada Antoine Gravot, University of Rennes 1, France

#### \*Correspondence:

Thomas Guillemette thomas.guillemette@univ-angers.fr

#### †Present address:

Guillaume Quang N'Guyen, Département de Biologie, PROTEO, Institut de Biologie Intégrative et des Systèmes, Université Laval, Pavillon Charles-Eugène-Marchand, Québec, QC, Canada

#### Specialty section:

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

Received: 18 September 2018 Accepted: 09 August 2019 Published: 30 August 2019

#### Citation:

N'Guyen GQ, Raulo R, Marchi M, Agustí-Brisach C, Iacomi B, Pelletier S, Renou J-P, Bataillé-Simoneau N, Campion C, Bastide F, Hamon B, Mouchès C, Porcheron B, Lemoine R, Kwasiborski A, Simoneau P and Guillemette T (2019) Responses to Hydric Stress in the Seed-Borne Necrotrophic Fungus Alternaria brassicicola. Front. Microbiol. 10:1969. doi: 10.3389/fmicb.2019.01969 1 Institut de Recherche en Horticulture et Semences – UMR 1345, INRA, Université d'Angers, Agrocampus-Ouest, SFR 4207 QUASAV, Angers, France, <sup>2</sup> Université de Lille, INRA, ISA, Université d'Artois, Université du Littoral Côte d'Opale, EA 7394 - ICV - Institut Charles Viollette, Lille, France, <sup>3</sup> Departamento de Agronomía, ETSIAM, Universidad de Córdoba, Córdoba, Spain, <sup>4</sup> Department of Plant Sciences, University of Agronomic Sciences and Veterinary Medicine of Bucharest, Bucharest, Romania, <sup>5</sup> Equipe "Sucres & Echanges Végétaux-Environnement," UMR CNRS 7267 EBI Ecologie et Biologie des Interactions, Université de Poitiers, Poitiers, France

Alternaria brassicicola is a necrotrophic fungus causing black spot disease and is an economically important seed-borne pathogen of cultivated brassicas. Seed transmission is a crucial component of its parasitic cycle as it promotes long-term survival and dispersal. Recent studies, conducted with the Arabidopsis thaliana/A. brassicicola pathosystem, showed that the level of susceptibility of the fungus to water stress strongly influenced its seed transmission ability. In this study, we gained further insights into the mechanisms involved in the seed infection process by analyzing the transcriptomic and metabolomic responses of germinated spores of A. brassicicola exposed to water stress. Then, the repertoire of putative hydrophilins, a group of proteins that are assumed to be involved in cellular dehydration tolerance, was established in A. brassicicola based on the expression data and additional structural and biochemical criteria. Phenotyping of single deletion mutants deficient for fungal hydrophilin-like proteins showed that they were affected in their transmission to A. thaliana seeds, although their aggressiveness on host vegetative tissues remained intact.

Keywords: hydrophilins, plant pathogenic fungus, seed transmission, dehydration, Alternaria brassicicola

### INTRODUCTION

The fungus Alternaria brassicicola causes black spot disease and is an economically important seed-borne pathogen of Brassicaceae species. This necrotrophic fungus strongly depends on seed transmission process for its long-term survival and dispersal (van den Bosch et al., 2010). However, fungal and plant factors that impact seed transmission are still poorly described. Such knowledge is crucial to propose strategies for improving the seed health, which remains a major issue for seed companies. Recent studies conducted with the Arabidopsis thaliana/A. brassicicola pathosystem showed that the level of susceptibility of the fungus to water stress strongly influenced its seed transmission ability. For instance, two osmosensitive fungal mutants, defective for the class III Histidine kinase (HK) AbNik1 (Pochon et al., 2013) and the MAP kinase AbHog1 (unpublished result), respectively, were highly jeopardized in their ability to colonize

seeds. Consistently, Iacomi-Vasilescu et al. (2008) had previously reported that A. brassicicola spontaneous phenylpyrrole resistant mutants with non-functional class III HK were found to be strongly impaired in their ability to infect radish seeds in field conditions, indicating that a functional high osmolarity pathway is required for efficient infection of seeds. Pochon et al. (2013) also showed that two dehydrin-like proteins were required for effective seed colonization by A. brassicicola. During colonization of maturing seeds, the fungi are exposed to severe water and osmotic constraints, and their ability to cope with such stresses is probably a key factor that determines their transmission to seeds.

The strategies to withstand low water availability have been mainly investigated in some organisms called anhydrobiotes belonging to kingdoms of bacteria, fungi and plants, which have the ability to resist desiccation, i.e., to survive to nearly total dehydration (Calahan et al., 2011; Dupont et al., 2014). Various mechanical, structural and oxidative constraints are induced during dehydration, which may lead to damage to membranes, proteins and DNA. Studies in the yeast Saccharomyces cerevisiae have shown that the protective system is based on a set of constitutive and inducible mechanisms. For instance, the synthesis of polyols occurs to maintain the osmotic balance. In yeast, the major polyol concerned with osmotic adjustment is glycerol and its synthesis, and hence its intracellular accumulation, is under control of the high osmolarity glycerol (HOG) pathway mediated by phosphorylation of the HOG1 MAP kinase. Some other molecules, such as heat shock proteins, hydrophilins or the non-reducing disaccharides trehalose and sucrose have been identified to be synthesized during desiccation and may be involved in preserving native protein structures or stabilizing membranes (Gadd et al., 1987; Garay-Arroyo et al., 2000; Sales et al., 2000; Tapia and Koshland, 2014).

Hydrophilins are a group of proteins that are present in organisms such as plants, fungi and bacteria, and that are biochemically defined as glycine-rich and highly hydrophilic disordered proteins. They are assumed to be involved in cellular dehydration tolerance, even if their specific role is still unclear. In Sa. cerevisiae, 12 genes, whose expression were induced in response to water deficit, encode proteins with specific hydrophilin characteristics (Garay-Arroyo et al., 2000). The deletion (or disruption) and overexpression of these genes showed that only some of them contributed to the desiccation tolerance of yeast (Dang and Hincha, 2011; López-Martínez et al., 2012; Rodríguez-Porrata et al., 2012). A first beneficial effect of hydrophilins on yeast viability during water deficit may be due to their antioxidant capacity which minimizes the accumulation of cellular ROS (López-Martínez et al., 2012). Another role of hydrophilins would be the protection of enzymes and the stabilization of membranes during drying (Sales et al., 2000; Goyal et al., 2005; Reyes et al., 2005). In virtue of their structural plasticity, which is due in part to the presence of disordered regions, hydrophilins may act as molecular shields or directly interact with various molecular partners (Tunnacliffe and Wise, 2007).

Except for the yeast hydrophilins, very little information is available on the presence and role of this class of proteins in filamentous fungi. To date, most studies have focused on a particular class of fungal hydrophilins called dehydrins. Dehydrin-encoding genes were first identified in Tuber borchii during searches for genes controlling fruiting body maturation or conidial dormancy (Abba' et al., 2006). Then, Wong Sak Hoi et al. (2011, 2012) identified three dehydrin-encoding genes in Aspergillus fumigatus. Two of them were found to be involved in the protection against oxidative, osmotic and pH stress and the third dehydrin regulated freezing tolerance. A farnesolinduced dehydrin-like protein, that contributed to the high tolerance of resting conidia against oxidative and heat stress, was also identified in Aspergillus nidulans (Wartenberg et al., 2012). In A. brassicicola, three proteins harboring the typical asparagine-proline-arginine (DPR) signature pattern and sharing the characteristic features of fungal dehydrin-like proteins have been identified (Pochon et al., 2013). The expression of these genes was found to be regulated by the AbHog1 mitogenactivated protein kinase (MAPK) pathway. Phenotyping of single deletion mutants showed that dehydrin-like proteins have an impact mainly on oxidative stress tolerance and on conidial survival upon exposure to high and freezing temperatures. Moreover, a double deletion mutant was strongly affected with respect to conidiation and showed a highly compromised pathogenicity with a lower aggressiveness on Brassica oleracea leaves and a reduced capacity to be transmitted to Arabidopsis seeds via siliques. These results support the importance of dehydrin proteins and, more globally, of hydrophilins, with respect to the ability of A. brassicicola to efficiently accomplish key steps of its pathogen life cycle.

To identify new fungal molecular determinants that may be involved in the silique and seed infection process, the present study aimed at analyzing the transcriptomic and metabolomic responses of A. brassicicola exposed to water stress. Then, the repertoire of putative hydrophilins in this fungus was established based on expression data and structural and biochemical criteria. Finally, functional approaches were initiated to investigate the role of some of these putative hydrophilins.

## RESULTS

## Application of Water Stresses

Several preliminary tests were conducted to evaluate the level of stress for each treatment applied to fungal cells. Considering that the reference strain is naturally tolerant to water stress (Iacomi-Vasilescu et al., 2008), we additionally analyzed the effect of treatments on three mutant strains (1abnik1, 1abhog1 and 1absch9) derived from the wild-type Abra43 strain, for which we hypothesized a higher sensitivity to water stress. As expected, the Abra43 reference strain exhibited high sorbitol tolerance with only 5% inhibition at a concentration of 1.2 M. At this concentration, a higher susceptibility was observed for the mutant strains (**Table 1**), confirming that the application of such a concentration indeed constitutes a stress for the fungal cells. In the same way, when a drastic decrease in relative humidity (RH) was applied to the germlings, no significant difference in mortality between the treated and the silica gel-free conditions was observed for the reference strain. On the contrary, the results

TABLE 1 | Susceptibility of Alternaria brassicicola wild-type, 1Abnik1,1Abhog1 and 1Asch9 strains to sorbitol and desiccation stresses.


Each condition was tested in triplicate and the experiments were repeated twice. Values are means of biological repetitions and represent the percentage growth inhibition under stress conditions compared with standard growth conditions.

showed a very high susceptibility of 1abnik1 and 1abhog1 to the desiccation stress, while the AbSch9 mutant exhibited an intermediate susceptibility (**Table 1**).

Based on these results, we chose to analyze the responses of the fungus when exposed to 1.2 M sorbitol for 0.5 and 2 h and when exposed to desiccation stress for 1 and 4 h. Regarding the latter stress, the chosen times implied that germlings exposed to silica gel beads were subjected to a RH value of 1% while control germlings were harvested at RH values of 70% and 90%, respectively.

In addition, we analyzed other phenotypic characteristics of these mutants: the susceptibility to oxidative stress (exposure to 10 mM H2O<sup>2</sup> and 20 mM menadione), the ability to transmit to the A. thaliana seeds and the aggressiveness on cabbage leaves. 1abnik1 and 1abhog1 mutants were more susceptible to oxidative stress, and, consistently, had reduced ability to colonize A. thaliana seeds compared to the wild-type strain (**Table 2**). The 1absch9 mutant was not susceptible to oxidative stress and its seed transmission capacity was not altered. Compared to seed transmission capacity, the ability of leaf infection was not or less dependent on the sensitivity of fungal genotypes to oxidative stress, since 1abnik1 and 1absch9 strains did not have reduced aggressiveness on cabbage leaves.

### Modulation of Fungal Transcriptomes in Response to Water Stress

In order to identify potential effectors of the response of A. brassicicola to dehydration stress, we first focused on the analysis of the transcriptional response in germinating conidia exposed to previously defined water stress (i.e., exposure to sorbitol 1.2 M for 0.5 and 2 h and exposure to silica gel beads for 1 h and 4 h). Each treated sample was compared to control sample and, thus, we produced four transcriptome data sets (each set was obtained from three biological replicates). We used A. brassicicola Nimblegen microarrays bearing one probe for each of the 10 633 ORFs predicted in the A. brassicicola automatically annotated genome database (JGI Genome Portal<sup>1</sup> ). Only probes with a P-values ≤0.05 and a log2 ratio ≥0.7 or ≤−0.7 were considered as differentially expressed. The numbers of genes that were induced or repressed under each condition, together with a summary of those regulated in more than one condition, are shown in **Figure 1**. To validate our microarray results, the regulation of

<sup>1</sup>https://genome.jgi.doe.gov/Altbr1/Altbr1.home.html

several genes by sorbitol or silica gel exposure was confirmed by quantitative PCR (**Supplementary Figure S1**).

Our data showed that, for all the treatments, with the exception of the 4 h-silica gel (4 h-desiccation) treatment, the number of up-regulated and down-regulated genes was less than 500 and 300, respectively. Most of these genes were found to be regulated in more than one condition, 37 were found to be induced in all 4 stress conditions while none were found to be repressed. The 4h-desiccation stress led to more profound transcriptomic modifications since 578 and 917 were induced or repressed compared to control condition, respectively. Moreover, this transcriptomic response seemed rather specific since more than 800 genes (268 up-regulated and 601 down-regulated genes) were specifically regulated after the fungus was exposed to this constraint. It should be noted that 157 genes, that were induced by the sorbitol exposure, were repressed following the silica gel treatment.

The GO enrichment analyses showed that the enriched GO terms were not generally shared within the gene sets representative of each type of treatment (sorbitol or desiccation) (**Figure 2**). One exception was the GO categories corresponding to the response to oxidative and osmotic stresses, which were enriched in all the lists of induced genes. Another configuration was that enriched categories emerging from the 4h-desiccation up-regulated genes were also enriched categories emerging from the down-regulated gene lists after sorbitol exposure. These results seem to point out that the two types of treatment did not cause the same effects at the cellular level and induced quite different transcriptomic responses.

Enriched categories emerging from genes regulated following silica gel exposure were mainly related to the translation process, mitochondrial functions, and carboxylic acid and amino acids metabolic processes. Within the up-regulated genes, the main enriched GO terms identified in the sorbitol data sets were predominantly associated with a response to oxidative stress and to sugar metabolic process. Within the sorbitol repressed gene list, enriched GO terms referred to a wider range of cellular processes such as gene expression (**Figure 2**).

### Metabolomic Responses to Water Stress

As several enriched GO categories emerging from transcriptomic data were linked to amino acid metabolism and catabolism, we investigated whether the response to water stress in A. brassicicola resulted in an alteration of the cellular content in amino acids. Regardless of the hydric stress applied, the total amino acid concentrations were significantly lower for the treatments conditions compared to the control conditions (Welch test P-value = 5.955E−<sup>08</sup> , **Supplementary Figure S2**). Indeed, considering the 24 amino acids and derivatives that were quantified in this study, the mean ratio of amino acid concentration in the treated conditions compared to the respective controls decreased significantly for 17 and 12 amino acids (9 in common), regardless of time exposure for the sorbitol and the desiccation treatment, respectively (**Figure 3A**). Amino acid profiles that were differentially less accumulated compared to controls were quite similar after 0.5 and 2 h of sorbitol exposure with 9 shared amino acids (over 13). Glutamine,

TABLE 2 | Other phenotypic characteristics of kinase mutants considered in this study.


Values were obtained in this study and are means of three biological repetitions and represent the percentage growth inhibition under stress conditions compared with standard growth conditions. <sup>a</sup> indicates a significant difference between the mutant and the Abra43 WT parental isolate (Student's t-test, P < 0.01). The abilities of fungal strains to transmit to Arabidopsis thaliana seeds and to develop necrosis on leaves were evaluated in this study or in indicated references, depending on the mutant; 0: unchanged comparing to the wild-type; –: decrease of the ability compared to the wild-type.

glutamate, alpha alanine, lysine and arginine concentration explained 70% [187.8/270.0 µmol.g−<sup>1</sup> DW (dry weight)] and 78% (156.3/199.8 µmol.g−<sup>1</sup> DW) of the amino acid variations for the sorbitol treatment at 0.5 and 2 h, respectively. For the 4 h desiccation condition, glutamine, GABA, alpha alanine and lysine explained 70% (134.7/194.6 µmol.g−<sup>1</sup> DW) of the amino acid variations. No significant concentration differences were observed for the 1 h desiccation treatment.

Then, we focused on sugar content, since their intracellular accumulation is a well-known protective mechanism to cope with the water constraints (Dupont et al., 2014). Eight saccharides and sugar alcohols were quantified in treated and control conditions (**Figure 3B**). The total concentration of those compounds was found to be significantly different only in the case of comparison between sorbitol-treated and untreated conditions (Welch test P-value = 2.05E−<sup>5</sup> , **Supplementary Figure S2**) regardless of the time exposure. Unexpectedly, we observed a strong accumulation of some sugars (arabitol, trehalose, mannitol) in the control of the desiccation condition. This result could explain that no significant difference on the sugar cellular contents was observed between treated and untreated samples, except for an arabitol accumulation after a 4 h-treatment.

As expected, glycerol was accumulated in fungal cells exposed to sorbitol. Two other polyols, arabitol and dulcitol, were also strongly accumulated under these conditions. Conversely, the amounts of other polyols and the disaccharide trehalose were lower in cells treated with sorbitol than in untreated conditions.

## Hydrophilins Identification Through in silico Analysis

Hydrophilins could be defined as glycine-rich, highly hydrophilic disordered proteins. This criterion was used to search in A. brassicicola for putative hydrophilins among the whole protein set deduced from the recently published genome of A. brassicicola strain Abra43 (12,456 entries, Belmas et al., 2018). A total of 732 predicted proteins with an average hydrophilicity above one (GRAVY index lower than −1) were first selected and analyzed for their glycine content. The 210 predicted proteins with at least 8% glycine were retained and subjected to secondary structure prediction analysis. Almost three quarters of them (165) were predicted to adopt a coil configuration over more than 50% of their length. After removing mis-annotated and incomplete proteins and those that had an obvious match upon Blast analysis with proteins of known functions (except "stress response"), a final set of 107 candidate hydrophilins (c.a. 0.9% of the whole predicted proteome) was selected (**Table 3**). Interestingly, among proteins that had physicochemical features of hydrophilins but were discarded after Blast analysis, several displayed similarities with proteins involved in nucleic acids metabolic processes (**Supplementary Table S1**). Each of the 107 candidate hydrophilins was characterized by an index of significance, calculated based on the following formula: [- (gravy index −1) X (% G – 8) X (% rc – 50)] X 100, whose value varied from 3 to 36.512. Six putative homologs of fungal hydrophilins were found within the whole set of A. brassicicola hydrophilin-like proteins: HSP12 (AB01868/Abra01064 and AB10131/Abra12207), STF2 (AB03803/Abra04061) from Sa. cerevisiae, CON10 (AB02056/Abra00844 and AB01782/Abra01163) and CON6 (Abra02927) from Neurospora crassa, and the dehydrin-like proteins from A. brassicicola AbDhn1 (AB02513/Abra00274) and AbDhn3 (AB05365, AB05364/Abra08036). None of the other yeast hydrophilins had putative homolog in the A. brassicicola genome except NOP6 (AB08803/Abra04570) and WWM1 (AB05865/Abra11690) that were not selected due to a glycine content below 8% or a gravy index above −1, respectively. Similarly, due to an average hydrophilicity below the threshold value, the other A. brassicicola dehydrin-like protein AbDhn2 (AB08993/Abra04787) was not selected within the candidate hydrophilins set. This protein set also contains the putative homolog DNA-damage responsive DDR48 (AB03330/Abra05531) protein from yeast.

### Expression of Hydrophilins During Osmotic and Hydric Stress

Then, we explored the expression pattern of genes encoding putative hydrophilins from transcriptome data sets generated after exposing germinating conidia of A. brassicicola to 1.2 M sorbitol and to low relative humidity (silica gel beads). These data sets were obtained using microarrays designed based on the genome of the A. brassicicola ATCC 96836 strain available on the JGI Genome Portal (see footnote 1). This genomic sequence contained 10,514 predicted genes and the proportion of gaps between scaffolding boards in this sequence was relatively high (estimated to 5.2% by Dang et al., 2015). As described above, the in silico analysis was performed from the recently published genome of A. brassicicola strain Abra43 (Belmas et al., 2018<sup>2</sup>,<sup>3</sup> ), which is a higher quality sequence containing 12,456 predicted protein-coding genes. Considering these proteome differences, expression patterns were available only for 78 putative hydrophilins that matched with a predicted ORF from the ATCC 96836 strain proteome. Under these conditions, 17 of the 78 (21.8%) candidate hydrophilin genes were up-regulated by the sorbitol treatment (**Figure 4A**). For all of them, except AB03330 and AB00441, an induced expression was observed at 0.5 h (**Figure 4B**). Moreover, 13 candidate hydrophilin genes were up-regulated by the desiccation treatment, 10 of them being also up-regulated upon the sorbitol treatment. All of these proteins were named with the abbreviation "SIH" for "stressinduced hydrophilin-like" (**Table 3**).

To gain insight on the possible regulation of hydrophilin genes expression by different kinases known to regulate the osmostress response, the expression profiles of hydrophilin genes in the 1abhog1, 1abnik1 and 1absch9 mutants were compared with those in the wild-type parental strain under both control condition and after exposure to sorbitol for 0.5 h (**Figure 4**). This analysis revealed that, among the sorbitolinducible hydrophilin genes, 6 were dependent of at least two of the three kinases for their expression. Two were regulated by AbHog1 and AbNik1, two others by AbHog1- and AbSch9 and three others by AbHog1, AbNik1 and AbSch9. The gene encoding AbDhn2 was also induced by the treatment and was found Hog1-Nilk1-Sch9-dependent.

## Role of Selected Hydrophilins in A. brassicicola

Two putative hydrophilins (AbSih3/AB01868 and Absih15/AB05148) were selected for further studies aiming to investigate their function in A. brassicicola. Knockout mutants deficient for AbSih3 and Absih15 (called 1absih3 and 1absih15), respectively, were constructed by replacing the respective ORFs with a hygromycin B resistance cassette (**Supplementary Figure S3**). For each targeted gene, two replacement mutants were selected using a PCR screen and further purified by two rounds of single-spore isolation. In all further experiments, the phenotypic characters for transformants of the same genotype were not found to be significantly different and the phenotyping values that we displayed were calculated by taking into account both transformants.

Although the expression of these genes was induced in germinating conidia exposed to previously defined water stress (i.e., exposure to 1.2 M sorbitol and to silica gel beads), the knock-out approach showed that only the AbSih3 deletion had an impact on phenotype, mainly on oxidative stress tolerance. The corresponding deficient single mutants were indeed characterized by a slightly increased susceptibility toward oxidative stress generated by exposure to menadione and H2O<sup>2</sup> compared to the wild-type (**Figure 5**). No significant difference in susceptibility to other tested stresses (exposure to PEG −0.7 MPa,

<sup>2</sup>https://www.ncbi.nlm.nih.gov/nuccore/PHFN00000000

<sup>3</sup>https://bbric-pipelines.toulouse.inra.fr/myGenomeBrowser?public=1

TABLE 3 | Set of candidate hydrophilins identified from the Alternaria brassicicola genome sequence through in silico analysis.


(Continued)

#### TABLE 3 | Continued

fmicb-10-01969 August 29, 2019 Time: 17:45 # 8


(Continued)

#### TABLE 3 | Continued

fmicb-10-01969 August 29, 2019 Time: 17:45 # 9


350 mM NaCl, 1.2 M sorbitol) was found compared to the wild-type (data not shown).

Consistently, a similar higher susceptibility to H2O<sup>2</sup> of the Sa. cerevisiae hsp12 null mutant was observed (**Figure 6**). To confirm that AbSih3 correspond to true functional homolog of the yeast hsp12 gene, complementation of hsp12 mutant null mutant was carried out with relevant recombinant pYES plasmids. As shown in **Figure 6**, AbSih3 successfully complemented the Sa. cerevisiae hsp12 phenotypes, even if the growth of 1hsp12 cells expressing AbSih3 was not restored to a level comparable to that of the WT strain containing the empty vector pYES2 on media supplemented with 1 mM H2O2.

Then, the virulence of the A. brassicicola wild-type, 1absih3 and 1absih15 strains were compared on B. oleracea and A. thaliana host plants during host vegetative tissue infection or during the seed transmission process. Brassica oleracea leaves were inoculated with drops of conidia suspension (10<sup>5</sup> , 10<sup>4</sup> , or 10<sup>3</sup> conidia/mL). As shown in **Figure 7**, the wild-type and KO strains were all able to produce typical symptoms and their aggressiveness was not impaired. Regardless of the inoculated strain, first necrotic symptoms appeared on leaves at 3 days post-inoculation (dpi). Necrotic areas enlarged and surrounded themselves with chlorotic halos at 6 dpi. During late stages of infection, the fungus produced conidia on the surface of necrosis.

Using the model pathosystem previously described by Pochon et al. (2012) for investigating seed transmission in Arabidopsis plants, the ability of the hydrophilin mutants to transmit to seeds was compared with that of the wild-type. As a positive control, we also analyzed the behavior of the osmosensitive fungal mutants 1abnik1 and 1abhog1, which were previously described as being strongly altered in their seed transmission capacity (Pochon et al., 2012). Silique inoculation with the wildtype and hydrophilin mutant strains resulted in the development of mycelium and typical lesions covered with conidia on siliques within a few days after inoculation. Seeds were then individually harvested and plated on PDA medium for a microbiological examination. As expected, a strong decrease in seed transmission rate was observed for 1abnik1 and 1abhog1 compared to the wild-type parental strain (**Figure 8**). Interestingly, the seed transmission capacity of the 1absch9 mutant was not altered although this strain was found to be sensitive to sorbitol and desiccation. While the deletion of AbSih3 had a slight significant impact on seed transmission, a stronger decrease was observed for the 1absih15 mutant, for which the seed

transmission rate was reduced by a factor two compared to the wild-type strain.

## DISCUSSION

The ability of A. brassicicola to cope with water-stress constitutes a key factor for successful seed colonization and thus accomplishment of its entire infectious cycle (Iacomi-Vasilescu et al., 2008; Pochon et al., 2012, 2013). Highly hydrophilic and disordered glycin-rich proteins containing repeated DPR-motifs called dehydrins were identified in A. brassicicola and might correspond to effectors of the response to water stress (Pochon et al., 2013). Based on their physicochemical features, dehydrins are considered as members of a larger protein family called hydrophilins initially described in Sa. cerevisiae (Garay-Arroyo et al., 2000). In an attempt to identify additional effectors of the response of A. brassicicola to water stress, the proteome deduced from the recently published genome of A. brassicicola strain Abra43 (Belmas et al., 2018) was first screened for proteins with characteristic features of hydrophilins. Candidate hydrophilins accounted for 0.9% of the whole predicted proteome. In comparison, hydrophilins represented only 0.2% of the yeast proteome (Garay-Arroyo et al., 2000) although the selection criteria were less stringent, i.e., proteins with lower glycine content were included. This difference might reflect a family expansion associated with environmental adaptation and/or fungal pathogenesis as it was proposed for other sets of proteins, e.g., secreted proteins, enzymes for secondary metabolites production (Soanes et al., 2008; Ohm et al., 2012). However, this assumption remains to be demonstrated and would require a more extensive search for hydrophilin repertoires in fungi with various lifestyles.

The fact that two of the three previously described dehydrins and additional homologs of fungal hydrophilins were included in the repertoire of candidate hydrophilins validated our screening process. Garay-Arroyo et al. (2000) proposed that hydrophilins that were not induced by water-stress conditions should not be considered as true hydrophilins. Thus, we analyzed changes in transcript levels in A. brassicicola germlings exposed to two kinds of water-stress: hyperosmosis (exposure to 1.2 M sorbitol) and desiccation (exposure to silica gel beads). Using gene expression array, which was designed from a previous genome sequence with a lower quality, we defined the expression patterns of 78 genes (over 107) encoding these putative hydrophilins. Twenty candidate hydrophilin genes were up-regulated either by the sorbitol treatment or by the desiccation stress, and, among them, ten were induced by both treatments. So, according to the hydrophilin definition given by Garay-Arroyo et al. (2000), we identified at least 20 hydrophilins in A. brassicicola. However, we can reasonably assume that their number is higher since we did not have access to the gene expression of 29 other genes encoding putative hydrophilins.

In this study, strains deficient for AbSih3 exhibited a slightly lower susceptibility to oxidative stress compared to the wildtype strain. No significant difference in susceptibility to other tested stresses (exposure to PEG, NaCl and sorbitol) was found for 1absih3 and 1absih15. Given the large number of hydrophilins that were co-expressed in this organism, functional redundancy may occur between some of these proteins. This process would explain the scarcity of significant phenotypes that we obtained from single deficient mutants since the presence of other hydrophilin genes would compensates for the loss. For instance, AbSih3 displays high sequence similarity to AbSih18 (AB10131), and both may be considered as homologs of the Sa. cerevisiae HSP12 hydrophilin, even if AbSih3 only partially compensated the HSP12 loss in yeast. The existence of such a functional redundancy was also previously observed between two dehydrins AbDhn1 and AbDhn2 in A. brassicicola. The double mutant 11abdhn1-abdhn2 was severely compromised in its pathogenicity while the respective single mutants were not affected (Pochon et al., 2013).

Despite this probable functional compensation, we showed here that the hydrophilins Absih3 and Absih15 could be considered as pathogenic factors since the corresponding null strains were affected in its seed transmission ability (more strongly for the 1absih15 strain). This result confirmed our initial assumption that an efficient fungal adaptive response to severe hydric stress conditions or to oxidative stress strongly influenced

FIGURE 6 | Complementation of yeast 1hsp12 mutants with AbSih3 coding sequences. (A) Growth curves of Saccharomyces cerevisiae (parental WT strain and 1hsp12 mutants) cells transformed with empty vector (pYES2CT) or recombinant vectors (pYAbsih3) exposed to 1 mM H2O<sup>2</sup> or to water. (B) Comparisons of growth inhibition rates obtained for each yeast strain (Wilcoxon test, <sup>∗</sup>p < 0.001).

the seed transmission ability. This is consistent with the fact that, during seed colonization, fungi are exposed to a gradual decrease in the water potential occurring in maturing reproductive organs and that they have to cope with this particular stress to be efficiently transmitted to seeds and complete their infection cycle. In the same vein, we showed in this study or in previous works that the class III HK AbNik1 and the MAP kinase AbHog1, two components of the high osmolarity pathway, were required for an effective seed colonization by A. brassicicola (Pochon et al., 2012). The 1absch9 mutant was also found to be sensitive to sorbitol and desiccation but, unexpectedly, its seed transmission capacity was not altered. The fact that this mutant strain did not exhibit a high susceptibility to oxidants suggested that the fungal ability to overcome oxidative stress would be an additional or the main key mechanism for an efficient seed infection process. This hypothesis is consistent with the fact that oxidative damages produced by free radical species are considered to be one of the major causes of desiccation injuries (Dupont et al., 2014).

A final point raised from these results is that several A. brassicicola mutants (1abnik1, 1absih3 and 1absih15) were affected in their transmission to A. thaliana seeds, although their aggressiveness on host vegetative tissues remained intact. This fact confirms the specificity of some molecular mechanisms controlling transgenerational transmission of fungal pathogens (Pochon et al., 2012) and validates the strategy to gain further insight into these mechanisms for the development of specific seed disease control strategies.

In addition to the induction of genes encoding hydrophilins, we identified other metabolic clues related to the activation of an adaptive response to water stress in A. brassicicola. First, the cellular content in amino acids was strongly disturbed following the application of both stresses. The amino acid levels were indeed significantly reduced in treated cells compared to controls. In line with this result, several enriched GO categories emerging from transcriptomic data were linked to amino acid metabolism and catabolism. Moreover, amino acid synthesis was also found to be strongly down-regulated in response to the osmotic stress treatment in another Dothideomycete fungus, the wheat pathogen Stagonospora nodorum (Lowe et al., 2008). As reported in plants (Hildebrandt et al., 2015), the degradation of amino acids may lead to the production of an additional energetic resource, that could be useful to cope with adverse conditions, by

providing intermediates of the tricarboxylic acid cycle and finally contributing to ATP production. Moreover, the oxidation of some amino acids feeds electrons into the mitochondrial electron transport chain. Most reaction steps of the catabolic pathways occur either in the mitochondria or cytosol. Consistent with this hypothesis, several enriched GO terms referred to mitochondria organization, carboxylic acid metabolism and to oxidationreduction process. Regarding the response to desiccation, as a GO enrichment of functions related to the genesis and transport of ribosomes and, more generally, to translation was observed, we could assume that a strong induction of the translation process would trigger a depletion of the cellular amino acid pool. So, the mechanisms leading to the decrease of the amino acid could depend on the type of applied stress.

An intriguing result from our data was that no significant difference in the amino acid concentrations was observed for the 1 h-desiccation treatment compared to the respective control. A more precise inspection showed that, for this condition, both control and treated samples showed a similar low amino acid content (**Supplementary Figure S2**). In this assay, germlings grown on cellophane were taken off from the PDA plate and directly transferred to sealed boxes (without additional agar medium) with or without silica gel beads. In the control box (without silica gel), we suspect that germlings were also exposed to a strong constraint in the early time of exposure, as a period of 4 h was required to gradually reach a RH value of about 90%. Therefore, although this experimental device triggered a specific regulation at the transcriptome level (**Figures 1**, **2**), this does not appear to be sufficient to reveal significant differences in amino acid accumulation between control and treated samples at this analysis time.

In the sorbitol-exposed samples, a strong increase in global sugar contents was observed compared to sorbitol control. Interestingly, the accumulation of non-reducing disaccharides, such as trehalose, did not seem to be favored in A. brassicicola germlings under the sorbitol stress conditions. On the contrary, the contents of three polyols, glycerol, arabitol and dulcitol, were mainly impacted by this stress, suggesting that they function as intracellular osmolytes to counteract the effects of decreases in the cell volume and the loss of turgor pressure (Sánchez-Fresneda et al., 2013; Dupont et al., 2014). This result was consistent with the increased sensitivity of the 1abhog1 mutant to osmotic stress since the MAP kinase Hog1 is known to be required for endogenous glycerol production in several fungal species (Brewster and Gustin, 2014; Hohmann, 2015). However, its involvement in the synthesis of other protective compounds, such as arabitol or dulcitol, is still unclear (Kayingo and Wong, 2005; Sánchez-Fresneda et al., 2013). Arabitol has been described as part of a set of polyols which is developed to counteract environmental challenges in several yeasts and filamentous fungi (Pascual et al., 2003; Sánchez-Fresneda et al., 2013). For instance, arabitol, and to a lesser extent glycerol, were found to accumulate in response to the osmotic stress treatment in the wheat pathogen St. nodorum (Lowe et al., 2008) or in the rice pathogen Magnaporthe grisea (Dixon et al., 1999). Conversely, the role of dulcitol in stress responses remains highly confidential and has been reported on rare occasions, for instance in the marine fungus Cirrenalia pygmea (Ravishankar et al., 2006).

In addition to its involvement in the synthesis of compatible solutes, we showed here that the HOG pathway, mainly regulated by the MAP kinase AbHog1 and the upstream HK AbNik1, also acted as an expression regulator of some hydrophilin encoding genes. The expression profiles of hydrophilin genes in the 1abhog1, 1abnik1, and 1absch9 mutants were indeed compared with those in the wild-type parental strain after exposure to sorbitol for 0.5 h. Of the 17 proteins that were induced under these conditions, 11 and 5 were found to be under the control of AbHog1 and AbNik1, respectively. This result, associated with the weakness of 1abnik1 and 1abhog1 to survive on reproductive organs, highlighted that the HOG pathway is a major target for controlling both the synthesis of osmolytes and hydrophilins, and thus limiting the transmission of fungal pathogens to seeds. Unexpectedly, AbDhn1 and AbDhn3 were not part of this set of HOGregulated proteins although Pochon et al. (2013) previously reported that these two proteins (in addition to AbDhn2) were targets for the AbHog1 signaling cascade when the fungus was exposed to a saline (NaCl) stress. This suggests a possible differential regulation of hydrophilins depending on the applied constraint.

Following a 0.5 h-sorbitol exposure, six other hydrophilins were also found to be AbSch9-dependent. As a major downstream effector of the target of rapamycin (TOR) complex 1, the Sch9 protein kinase was proposed to play multiple roles in stress resistance and nutrient sensing in Sa. cerevisiae (Pascual-Ahuir and Proft, 2007; Huber et al., 2009). However, the functions of Sch9 orthologs in filamentous fungi have been poorly characterized to date. In Fusarium graminearum, Gu et al. (2015) showed that FgSch9 served as a mediator of the TOR and HOG pathways and played important roles in regulating vegetative differentiation, secondary metabolism and stress responses. We showed here that both hydrophilins AbSih3 and AbSih15 were necessary factors for an optimal A. brassicicola transmission to seeds but their expression was not found to be regulated by AbSch9 (regulation by AbHog1 and AbHog1-AbNik1, respectively). This result, associated with the fact that the seed transmission capacity of the 1absch9 mutant was not affected, suggested that this kinase (unlike AbHog1 and AbNik1), although involved in resistance to sorbitol and desiccation, is not a major player in transgenerational transmission of fungal pathogens.

### MATERIALS AND METHODS

### Strains and Culture Conditions

The A. brassicicola wild-type strain Abra43 used in this study was previously described (Sellam et al., 2007). For routine culture, A. brassicicola was grown and maintained on potato dextrose agar (PDA). The genome of Abra43 strain was recently published and is available following this link https://bbric-pipelines.toulouse. inra.fr/myGenomeBrowser? (Carrere and Gouzy, 2017).

For transcriptomic analyses, exposure to sorbitol (osmotic stress) or to low relative humidity (desiccation stress) were achieved as follows. Conidia were first incubated for 24 h on cellophane disks overlaid on solid PDA medium. Germlings and the underlying cellophanes were then taken off from the agar plate and transferred either to a second PDA medium supplemented or not with 1.2 M sorbitol for 0.5 or 2 h.

The desiccation stress was achieved by applying a drastic decrease in relative humidity (RH) to the germlings (obtained as described below) by exposure to silica gel beads in sealed boxes for 0.5 or 4 h. Immediately after deposition of germlings (and the underlying cellophanes) in boxes, a strong RH increase was observed. In the control box (without silica gel), a gradual increase up to a RH value of about 90% in 4 h was observed. With silica gel beads, the opening of the boxes resulted in a rapid RH increase up to 20% followed by a decrease to a RH value of 1% in less than 0.5 h.

To evaluate the level of stress for each treatment applied to fungal cells and considering that the reference strain is naturally tolerant to water stress (Iacomi-Vasilescu et al., 2008), we additionally analyzed the effect of treatments on three mutant strains (derived from the wild-type Abra43 strain) for which we hypothesized a higher sensitivity to water stress. Two mutant strains (1abnik1 and 1abhog1) were described in previous studies (Dongo et al., 2009; Joubert et al., 2011) and were deficient for genes encoding key regulators of the HOG pathway, the HK AbNik1 and the MAP kinase AbHog1, respectively. The third mutant was obtained in this study and was deficient for the kinase AbSch9, which is also known to be involved in the fungal response to osmotic constraints (Pascual-Ahuir and Proft, 2007). Under non-stress conditions, no significant differences in mycelium growth were detected in any of the tested mutants as compared to the wild-type (data not shown). To assess the impact of low RH exposure on A. brassicicola, viability tests were conducted on young hyphae of wild-type and mutant strains. About 200 conidia were first incubated for 24 h on cellophane disks overlaid on solid PDA medium. Then, germinated conidia were transferred for 4 days to sealed boxes containing or not silica gel beads. The cellophane membranes were then removed from the boxes and deposited again on PDA medium. The survival rate was calculated by counting the number of mycelial colonies (derived from treated or untreated germlings) after 3 days of incubation at 24◦C. As previouslys reported (de Lima Alves et al., 2015; Wang et al., 2017), exposure to these two treatments induces water stress in fungal cells. Exposure of wild-type and mutant strains to 1.2 M sorbitol were performed following the procedure described below. The radial growth was calculated after 6 days of incubation.

To study hyphal growth in liquid media, conidial suspensions (10<sup>5</sup> conidia/mL, final concentration) were inoculated onto microplate wells containing the considered substance (10 mM H2O<sup>2</sup> or 20 mM menadione) at the desired concentrations in potato dextrose broth (PDB, Difco) in a total volume of 300 µl. Fungal growth was monitored automatically over a 30 h period with a laser-based microplate nephelometer (NEPHELOstar, BMG Labtech) as described by Joubert et al. (2010). For each condition, the area under the growth curve representative of the lag phase was used to calculate the percentages of growth inhibition under stress conditions compared with standard growth conditions (PDB medium without additive), as described by Gaucher et al. (2013). At least three replicates were conducted per treatment.

### RNA Extraction and Microarray Analysis

Total RNA was extracted using the nucleospin RNAplant kit (Macherey Nagel, Düren, Germany) and 400 ng were amplified using the Ambion messageAmp II (Ambion, Austin, TX, United States) according to the manufacturer's instructions. Five micrograms of amplified RNA were retro-transcribed with 400 U of Superscript II reverse-transcriptase (Invitrogen Corp., Carlsbad, CA, United States) and labeled with 1.5 mmol of Cyanine-3 (Cy3) or Cyanine-5 (Cy5) (Interchim, France) and then purified using NucleoSpin Gel and PCR Cleanup column kits (Macherey-Nagel, GmbH and Co., KG, Germany). Purified and labeled cDNA were quantified using a NanoDrop ND-1000. Corresponding Cy3- and Cy5 labeled samples (30 pmol) were combined and cohybridized to the Abra \_v1.0 12 × 135K arrays on a NimbleGen Hybridization System 4 (mix mode B) at 42◦C overnight. The Abra\_v1.0 chip was in situ synthesized by Nimblegen (Madison, WI)

and contained 10,633 60-mer oligoprobes that were designed from the A. brassicicola automatically annotated genome database<sup>4</sup> , containing one probe per sequence. Slides were washed, dried, and scanned at 2 µm resolution and high sensitivity with a Roche-NimbleGen MS200. DEVA Software version 1.2.1 was used to extract pair data files from the scanned images. Three biological replicates were analyzed per comparison using the dye-switch method. Statistical analyses of the gene expression data were performed using the R language, version 2.14, and the Linear Models for Microarray Analysis package (Smyth et al., 2005) from the Bioconductor project. For the preprocessing step, data were normalized by the LOWESS method. Log ratio and log intensity were calculated before differential expression analyses and performed using the lmFit function and the Bayes-moderated Student's t-test in a linear model for microarray analysis. Probes with P-value <0.05 were considered as differentially expressed. Gene expression datasets were deposited in the Gene Expression Omnibus (GEO)<sup>5</sup> .

### Quantitative PCR

Total RNA was extracted as described above. Amplification experiments were conducted as previously described (Joubert et al., 2011). The relative quantification analysis was performed using the 11Ct method (Winer et al., 1999). All the primer sequences used in real-time quantitative PCR are presented in the **Supplementary Table S2**.

### Generation of Targeted Gene Knockout Mutants and Fusion Strains

We followed the procedure described by Pigné et al. (2017). Gene replacement cassettes were generated with the Hph gene cassette from pCB1636 (Sweigard et al., 1995) conferring resistance to hygromycin B. The sets of primers used to amplify the 5<sup>0</sup> and 3 <sup>0</sup> flanking regions of each targeted gene are presented in the **Supplementary Table S2**. The replacement cassettes were used to transform A. brassicicola protoplasts as described by Cho et al. (2006). Primer combinations used to confirm integration of the replacement cassette at the targeted locus were presented in **Supplementary Figure S3**.

### Yeast Complementation Assays

Complementation assays were performed in Sa. cerevisiae cells (strains: parental BY4743 Acc#Y20000, 1HSP12 Acc#Y27070, Euroscarf, Germany) as described by Joubert et al. (2011). Sequences encoding AbSih3 were amplified by PCR with primer pairs AbSih3-Bam (5<sup>0</sup> - **GGATCC**AAAAATGACCGACGCATTCCGC-3<sup>0</sup> ) – AbSih3-Eco (50 - **GAATTC**CCATTCCAAGGGCGTTTTTGG-3<sup>0</sup> ). Unique BamH1-EcoRI restriction sites (indicated in boldface) were thus introduced at the upstream start and downstream of the coding sequences, respectively and were used to clone the PCR product into pGEM-T (Promega) and then into the yeast expression vector pYES2CT (Invitrogen) to generate the recombinant plasmid pYAbhsp12. Sensitivity to 1 mM H2O<sup>2</sup> was analyzed based on microcultivation in 300 µl liquid medium. Growth in microplates was automatically recorded using a spectrophotometer (Spectrostar Nano, BMG Labtech) at 30◦C.

### Infection Assays

Leaf infection assays were performed on B. oleracea plants (var. Bartolo) as described by Joubert et al. (2011). The experiment was repeated twice and involved at least 10 leaves inoculated by each fungal genotype.

Seed infection assays and seed contamination assessments were performed as described by Pochon et al. (2012). At least five plants per fungal genotype were inoculated and the experiment was repeated twice.

### Metabolic Extraction and Amino Acids and Sugar Analysis

Metabolic extractions were made on 15 mg of dry powdered samples. Procedures for methanol–chloroform–water-based extractions as well as determination of amino acid and carbohydrate contents were described by Thouvenot et al. (2015).

### AUTHOR CONTRIBUTIONS

GN'G carried out the sample preparation and microarray analyses. RR, BI, and MM involved in the sample preparation for metabolic analysis and performed the quantitative PCR. NB-S carried out the yeast complementation assays. CM, MM, and CA-B involved in the construction and phenotyping of Alternaria brassicicola mutant strains. PS and TG conceived the study and participated in the design of the experiments as well as in the analysis of the results. RL and BP involved in sugar analyses from fungal samples. SP and J-PR involved in the gene expression analyses. CC, MM, FB, BH, AK, and BI involved in the pathological tests.

### FUNDING

This research was supported in part by grants awarded by the Region des Pays de la Loire (FUNHY project). This research was also supported by post-doctoral fellowships (CA-B) provided by the University of Angers.

### ACKNOWLEDGMENTS

The authors thank the Metabolic Profiling and Metabolomic (P2M2) Platform from the French public institute IGEPP for assistance during metabolic phenotyping.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2019. 01969/full#supplementary-material

<sup>4</sup>https://genome.jgi.doe.gov/Altbr1/Altbr1.home.html

<sup>5</sup>https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE133507

FIGURE S1 | Comparison of transcript levels obtained by microarray or RT-qPCR methods. The expression of 9 genes was monitored in response to sorbitol and desiccation stresses. The fold inductions (log2 values) are presented in (A) and the correlation plot in (B).

FIGURE S2 | Total organic contents in a 1 mg sample of dried fungal cultures sample exposed or not to sorbitol and desiccation treatments. (A) Total contents of amino acids. (B) Total content of saccharides and sugar alcohols. Colors display the treatments. Box plots show the median, 1st and 3rd quantiles (n = 3). <sup>∗</sup>denotes a significant difference (P-value <0.001, Welsh test).

FIGURE S3 | Generation of null mutants by homologous recombination. (A) Schematic representation of the deleted locus in the wild-type and the

### REFERENCES


replacement construct with the Hyg B resistance cassette (Hph gene). Arrows indicate the position of primers used for the cassette generation and for PCR screening of mutants. (B) Gel electrophoresis of PCR products obtained from template DNA of the wild-type, 1absch9, 1absih3 and 1absih15 strains with the indicated primer pairs. Molecular sizes (kb) were estimated based on a 1 kb ladder (lane M, NEW ENGLAND BioLabs <sup>R</sup> Inc.). A supplementary PCR was done with the primer pair ITS1/ITS4 as a positive control to assess the quality of the wild-type template DNA.

TABLE S1 | List of predicted sequences from the Alternaria brassicicola genome exhibiting physicochemical features of hydrophilins and an obvious match upon Blast analysis with proteins of known functions.

TABLE S2 | List of primers used in this study.



**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 N'Guyen, Raulo, Marchi, Agustí-Brisach, Iacomi, Pelletier, Renou, Bataillé-Simoneau, Campion, Bastide, Hamon, Mouchès, Porcheron, Lemoine, Kwasiborski, Simoneau and Guillemette. 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.