Spatial Analysis of Fungicide Resistance Mutations in Botrytis Spp

Spatial analysis of fungicide resistance mutations in Botrytis spp.

populations

Hervé Van der Heyden

Master of Science

Department of Plant science,

McGill University,

Ste-Anne de Bellevue, Québec

August 2013

A thesis submitted to McGill University in partial fulfillment of the

requirements for the degree of Master of Science

TABLE OF CONTENTS 2

LIST OF TABLES 5

LIST OF FIGURES 6

ABSTRACT 9

RÉSUMÉ 11

ACKNOWLEDGMENT 13

CONTRIBUTION OF AUTHORS 14

1 GENERAL INTRODUCTION 16 1.1 INTRODUCTION 16

1.2 OBJECTIVES 17

1.3 HYPOTHESES 18

2 LITERATURE REVIEW 19 2.1 SPATIAL STATISTICS 19

2.1.1 DISTANCE-BASED APPROACH 20

2.1.2 QUADRAT-BASED APPROACH 22

2.1.3 GEOSTATISTICAL APPROACH 24

2.2 BOTRYTIS SPP. 25

2.2.1 BOTRYTIS 25

2.2.2 BOTRYTIS CINEREA 25

2.2.3 BOTRYTIS SQUAMOSA 28

2.3 FUNGICIDE RESISTANCES 32

2.3.1 RESISTANCE TO ANTI-MICROTUBULE FUNGICIDES 32

2.3.2 RESISTANCE TO DICARBOXIMIDES AND PHENYLPYRROLES 32

2.3.3 RESISTANCE TO ANILINOPYRIMIDINES 33

2 2.3.4 RESISTANCE TO STEROL BIOSYNTHESIS INHIBITORS 33

2.3.5 RESISTANCE TO QUINONE OUTSIDE INHIBITORS 34

2.3.6 RESISTANCE TO SUCCINATE DEHYDROGENASE INHIBITORS 34

2.4 CONNECTING TEXT FOR CHAPTER 3 35

3 COMBINED APPLICATION OF SPATIAL STATISTICS AND MOLECULAR GENETICS FOR CHARACTERIZATION OF SMALL-SCALE SPATIAL RELATIONSHIPS BETWEEN POLYMORPHISMS RELATED TO FUNGICIDE RESISTANCE IN BOTRYTIS CINEREA POPULATIONS 36 3.1 ABSTRACT 36

3.2 INTRODUCTION 37

3.3 MATERIAL AND METHODS 42

3.3.1 INVENTORY OF PHENOTYPIC FUNGICIDE RESISTANCE 42

3.3.2 FUNGICIDES AND CONIDIAL GERMINATION ASSAY 42

3.3.3 INCIDENCE OF SNPS RELATED TO FUNGICIDE RESISTANCE 43

3.3.4 DNA EXTRACTION AND PCR ANALYSIS 43

3.3.5 UNIVARIATE SPATIAL ANALYSIS 44

3.3.6 BIVARIATE SPATIAL ANALYSIS 47

3.4 RESULTS 49

3.5 DISCUSSION 51

3.6 ACKNOWLEDGEMENTS 56

3.7 CONNECTING TEXT FOR CHAPTER 4 67

4 A NOVEL PCR-RFLP ASSAY FOR THE DETECTION OF A SINGLE NUCLEOTIDE POLYMORPHISM RELATED TO DICARBOXIMIDE RESISTANCE IN BOTRYTIS SQUAMOSA FIELD ISOLATES 68 4.1 ABSTRACT 68

4.2 INTRODUCTION 68

4.3 MATERIAL AND METHODS 70

3 4.4 RESULTS 71

4.5 DISCUSSION 72

4.6 ACKNOWLEDGEMENTS 72

4.1 CONNECTING TEXT FOR CHAPTER 5 76

5 SPATIAL DISTRIBUTION OF SINGLE NUCLEOTIDE POLYMORPHISMS RELATED TO FUNGICIDE RESISTANCE AND IMPLICATIONS FOR SAMPLING 77 5.1 ABSTRACT 77

5.2 INTRODUCTION 78

5.3 MATERIALS AND METHODS 82

5.3.1 SAMPLING PROTOCOLS 82

5.3.2 DNA EXTRACTION AND RFLP-PCR 83

5.3.3 GEOSTATISTICAL ANALYSES 85

5.3.4 DISTRIBUTIONAL ANALYSIS 87

5.3.5 SAMPLING CURVES 89

5.4 RESULTS 90

5.4.1 SNP INCIDENCE 90

5.4.2 GEOSTATISTICAL ANALYSES 91

5.4.3 SPATIAL STATISTICS 91

5.4.4 SAMPLING CURVES 92

5.5 DISCUSSION 93

5.6 ACKNOWLEDGEMENTS 98

6 GENERAL CONCLUSION AND CONSIDERATION FOR FUTURE RESEARCH 108 6.1 GENERAL CONCLUSION 108

6.2 CONSIDERATION FOR FUTURE RESEARCH 110

7 LITERATURE CITED 111

4 List of tables

TABLE 1: LIST OF FUNGICIDE GROUPS AND ACTIVE INGREDIENTS AND

DISCRIMINANT DOSE USED FOR THE PHENOTYPIC ASSAY...... 57

TABLE 2: CHEMICAL GROUPS, ACTIVE INGREDIENTS AND SEQUENCES OF THE

PRIMERS AND PROBES USED FOR THE DETECTION OF THE MUTANT ALLELES

ASSOCIATED WITH FUNGICIDE RESISTANCE...... 58

TABLE 3: INDEX OF DISPERSION D AND PARAMETERS ESTIMATE OF THE BETA-

BINOMIAL AND BINOMIAL DISTRIBUTION...... 59

TABLE 4: AMINO ACID SUBSTITUTIONS OF BCOS-1 HOMOLOGOUS IN 18 SINGLE-

SPORE ISOLATES OF BOTRYTIS SQUAMOSA USED FOR DEVELOPMENT...... 73

TABLE 5: RESULTS OF CRAD ANALYSIS (COREGIONALIZATION ANALYSIS WITH A

DRIFT) AND VARIOGRAM MODEL PARAMETER ESTIMATES (RANGE, NUGGET

EFFECT, SILL OF SPHERICAL STRUCTURE) FOR SNP DISTRIBUTIONS WITHIN B.

CINEREA AND B. SQUAMOSA POPULATIONS IN TWO VINEYARDS AND TWO

ONION FIELDS SAMPLED IN 2009-2011...... 99

TABLE 6: OBSERVED VALUES OF THE INDEX OF DISPERSION AND PARAMETER

ESTIMATES FOR THE BETA-BINOMIAL AND BINOMIAL DISTRIBUTIONS ...... 102

TABLE 7: MINIMUM SAMPLE SIZE (N) REQUIRED TO ESTIMATE THE MEAN SNP

INCIDENCE (P) FOR LOW, INTERMEDIATE AND HIGH VALUES OF THE LOCAL

AGGREGATION INDEX (Θ) AND INCREASING VALUES OF THE CV (EXCERPTS

FROM FIGURE 3; SEE TEXT FOR DETAILS) ...... 104

List of figures

FIGURE 1: SCHEMATIC REPRESENTATION OF THE THREE POSSIBLE OUTCOMES IN

THE ANALYSIS OF INTER-SNP DISTANCES A) ABSENCE OF A SPATIAL

RELATIONSHIP, B) EXCLUSIVE RELATIONSHIP AND C) INCLUSIVE

RELATIONSHIP. THE CONTINUOUS LINE REPRESENTS THE OBSERVED

CUMULATIVE RELATIVE FREQUENCY DISTRIBUTION, WHILE THE DASHED

LINES REPRESENT THE 2.5TH- AND 97.5TH-PERCENTILE ENVELOPES...... 60

FIGURE 2: PROPORTION OF RESISTANT ISOLATES TO SIX FUNGICIDES OBTAINED IN

A CONIDIAL GERMINATION ASSAY FOR 232 BOTRYTIS CINEREA SAMPLES,

COLLECTED FROM 23 PLOTS IN 18 VINEYARDS FROM DIFFERENT

PRODUCTION AREAS IN THE PROVINCE OF QUÉBEC...... 61

FIGURE 3: A) FREQUENCY DISTRIBUTION OF THE SINGLE NUCLEOTIDE

POLYMORPHISM (SNPS) FOUND IN TWO VINEYARDS IN 2011 AND B) NUMBER

OF SNPS PER ISOLATES. GREY BARS: FIRST FIELD; BLACK BARS: SECOND

FIELD...... 62

FIGURE 4: UNIVARIATE 2-D SPATIAL POINT PATTERNS OBSERVED ON THE FIRST

FIELD. A POINT INDICATES THE PRESENCE OF A SNP FOR MUTATION G143A

(A), H272R (B), DAF1 (C), H272Y (D), N230I (E), F412I (F), H272L (G), G143A-

INTRON (H) AND F412S (I), RESPECTIVELY...... 63

FIGURE 5: UNIVARIATE 2-D SPATIAL POINT PATTERNS OBSERVED ON THE SECOND

FIELD. A POINT INDICATES THE PRESENCE OF A SNP FOR MUTATION G143A

(A), H272R (B), DAF1 (C), H272Y (D), H272L (E) AND F412S (F), RESPECTIVELY. . 64

FIGURE 6: BIVARIATE ANALYSIS OF THE 2-D SPATIAL POINT PATTERNS OF FIGURE

4. THE CONTINUOUS LINES REPRESENT THE OBSERVED CUMULATIVE

RELATIVE FREQUENCY DISTRIBUTIONS OF INTER-SNP DISTANCES, WHILE

THE DASHED LINES REPRESENT THE 2.5TH- AND 97.5TH-PERCENTILE

ENVELOPES...... 65

6 FIGURE 7: BIVARIATE ANALYSIS OF THE 2-D SPATIAL POINT PATTERNS OF FIGURE

5. THE CONTINUOUS LINES REPRESENT THE OBSERVED CUMULATIVE

RELATIVE FREQUENCY DISTRIBUTIONS OF INTER-SNP DISTANCES, WHILE

THE DASHED LINES REPRESENT THE 2.5TH- AND 97.5TH-PERCENTILE

ENVELOPES...... 66

FIGURE 8: PREDICTED AMINO-ACID PARTIAL SEQUENCE OF B. SQUAMOSA BCOS-1

HOMOLOGOUS, ALIGNED WITH BCOS-1 OF B. CINEREA. THE ASTERISK (*) AT

POSITION 210 INDICATES THE I86S SUBSTITUTION SITE AND THE DOTS (.)

REPRESENT GAPS INTRODUCED TO MAXIMIZE ALIGNMENT...... 74

FIGURE 9: VALIDATION RESULTS OF THE PCR-RFLP ASSAY USING 24 SINGLE-

COLONY ISOLATES. THE PRESENCE OF TWO FRAGMENTS OF 400 AND 160 BP

FOLLOWING DIGESTION WITH TAQΑ1 INDICATES THE ABSENCE OF THE I86S

MUTATION. THE PRESENCE OF AN ENTIRE 560-BP FRAGMENT IMPLIES THE

PRESENCE OF THE I86S MUTATION. WILD-TYPE STRAINS SENSITIVE TO

IPRODIONE ARE IN LANES 1 TO 12, 13, 15 TO 17, AND 24, WHILE MUTANT

STRAINS RESISTANT TO IPRODIONE ARE IN LANES 14 AND 18 TO 23. THE PCR

NEGATIVE CONTROL IS IN LANE N...... 75

FIGURE 10: RAW COUNT DATA MAPS FOR B. CINEREA IN VINEYARD OR FOR A)

H272R (OBSERVED MEAN INCIDENCE, 0.68, ± 0.023, STANDARD ERROR), B)

H272Y (0.21 ± 0.021), AND C) I86S (0.55 ± 0.024), AND IN VINEYARD TE FOR E)

H272R (0.61 ± 0.021) F) H272Y (0.20 ± 0.018), AND G) I86S (0.59 ± 0.020), AND FOR B.

SQUAMOSA FOR I86S HOMOLOGOUS IN D) 2009 (0.16 ± 0.020) AND H) 2010 (0.42 ±

0.025)...... 105

FIGURE 11: FREQUENCY DISTRIBUTION OF THE OBSERVED NUMBER OF BOTRYTIS

ISOLATES CARRYING A GIVEN SNP PER QUADRAT (EMPTY BARS), TOGETHER

WITH THE EXPECTED NUMBERS OF BOTRYTIS ISOLATES CARRYING THE SNP

PER QUADRAT UNDER THE BETA-BINOMIAL DISTRIBUTION MODEL (FILLED

BARS) AND THE BINOMIAL DISTRIBUTION MODEL (DASHED BARS), FOR B.

CINEREA IN VINEYARD OR FOR A) H272R, B) H272Y AND C) I86S, IN VINEYARD

TE FOR E) H272R, F) H272Y AND G) I86S, AND FOR B. SQUAMOSA FOR I86S IN D)

2009 AND H) 2010...... 106

FIGURE 12: SAMPLING CURVES CALCULATED UNDER THE ASSUMPTION OF A

BETA-BINOMIAL DISTRIBUTION, USING LOW, INTERMEDIATE AND HIGHER

VALUES OF THE PARAMETER Θ THAT CORRESPOND TO THE MINIMUM,

MEDIAN AND MAXIMUM VALUES OF Θ -ESTIMATES OBTAINED IN OUR

STUDY, FOR THREE LEVELS OF PRECISION GIVEN BY CV VALUES OF 10%, 20%

AND 30%. THE CONTINUOUS LINE REPRESENTS THE MINIMUM SAMPLE SIZE

REQUIRED TO ESTIMATE MEAN SNP INCIDENCE WITH A CV OF 10%; THE

DOTTED LINE, WITH A CV OF 20%; AND THE DASHED LINE, WITH A CV OF 30%.

...... 107

Abstract

The objectives of this project were: 1) to study the spatial interactions of single nucleotide polymorphisms (SNPs) related to fungicide resistance within Botrytis cinerea populations isolated from grapes; 2) to study the spatial distribution patterns of SNPs related to fungicide resistance within B. cinerea populations in grape and within B. squamosa populations in onion; and 3) to compute sampling curves relative to mean SNP incidence estimation. In a first experiment, B. cinerea isolates were collected following a quadrat-based design (100 10x10m quadrats) in two commercial vineyards. The presence of 9 SNPs related to resistance to iprodione, boscalid, azoxystrobin and fenhexamid were detected using PCR-RFLP, PIRA-PCR and RT-qPCR assays. These data were spatially referenced and considered as a multivariate point pattern in a given vineyard.

Spatial point patterns were analyzed by pairs, using an extension of Diggle’s procedure for the analysis of nearest-neighbor distances. In this randomization testing procedure, the cumulative relative frequency distribution of the inter-SNP distances was used to characterize the spatial relationship between SNPs related to fungicide resistance. In the second experiment, two SNPs known to be responsible for boscalid resistance and one SNP known to be responsible for dicarboximide resistance in B. cinerea on grape were studied, in addition to one SNP responsible for dicarboximide resistance in B. squamosa on onion. One onion field was sampled in 2009 and another one was sampled in 2010 for B. squamosa, and two vineyards were sampled in 2011 for B. cinerea, for a total of four sampled sites.

Sampling was carried following the same design as in the first experiment, except

10 samples were collected in each quadrat. Samples were analyzed by RFLP-

PCR. The characterization of spatial distribution patterns was made through the fitting of discrete probability distributions. The level of mutations obtained in the first experiment was 90%, 64%, 67%, 33% and 1% for G143A, I86S, H272R,

H272Y and N230I, respectively. Our results show that three spatial relationships can arise when spatial point patterns representing the presence of SNPs related to fungicide resistance are compared by pairs: spatial exclusiveness (12%), spatial co-existence (31%) and absence of a spatial relationship (56%). Despite the fact that more than one half of the pairs of SNPs tested showed no spatial relationship, the presence of about a third of inclusive spatial relationships supports the models of co-existence between sensitive and resistant strains postulated in the literature, but suggests a higher level of complexity in the resistant-sensitive interactions. In the second experiment, the beta-binomial distribution was found to fit the data better than the binomial distribution for all data sets. This indicates local SNP aggregation among sampling units, as supported by estimates of the parameter θ of the beta-binomial distribution ranging from 0.09 to 0.23, with an overall median value of 0.20. On the basis of the spatial distribution patterns of SNP incidence that we found in Botrytis populations, sampling curves were developed for various levels of precision, emphasizing the importance of sampling for early detection of fungicide resistance in plant disease epidemiology.

10 Résumé

Les objectifs de ce projet étaient: 1) d’étudier les interactions spatiales entre polymorphismes nucléotidiques simples (PNS) associés à la résistance aux fongicides dans les populations de Botrytis cinerea provenant de raisins infectés;

2) d’étudier les patrons de distribution spatiale des PNS associés à la résistance aux fongicides au sein de populations de B. cinerea provenant de vignobles et de populations de Botrytis squamosa dans des champs d’oignions; et 3) de développer des courbes d’échantillonnages associées à l’estimation de l’incidence moyenne des PNS. Dans une première expérience, des isolats de B. cinerea ont

été récoltés dans deux vignobles commerciaux en suivant une grille d’échantillonnage de 100 quadrats de 10x10m. La présence de 9 PNS associés à la résistance à l’iprodione, au boscalid, à l’azoxystrobine et au fenhexamid a été détectée par PCR-RFLP, PIRA-PCR et RT-qPCR. Les données ont été référencées spatialement et pour chaque vignoble, considérées comme un patron ponctuel multivarié. Les analyses ont été réalisées par paire, à l’aide d’une extension de la méthode de Diggle pour l’analyse des distances aux plus proches voisins. Dans cette procédure de test par randomisation, la distribution des fréquences relatives cumulatives des distances inter-PNS est utilisée afin de caractériser les patrons de relation spatiale entre PNS associés à la résistance aux fongicides. Dans une deuxième expérience, deux PNS associés à la résistance de

B. cinerea au boscalid et un PNS associé à la résistance de B. cinerea aux dicarboximides ont été étudiés dans la vigne, en plus d’un PNS associé à la résistance de B. squamosa aux dicarboximides dans l’oignion. Pour B. squamosa,

deux champs ont été échantillonnés, un en 2009 et un en 2010, et pour B. cinerea, deux champs ont été échantillonnés en 2011, pour un total de quatre sites d’échantillonnage. L’échantillonnage a été réalisé en suivant le même dispositif expérimental que pour la première expérience, à la différence que 10 échantillons ont été prélevés dans chaque quadrat. Les échantillons ont été analysés par PCR-

RFLP, et les patrons de distribution spatiale ont été caractérisés sur base de l’ajustement des lois de distributions. Dans la première expérience, les proportions de PNS étaient de 90%, 64%, 67%, 33% et 1% pour G143A, I86S,

H272R, H272Y et N230I, respectivement. Ces résultats démontrent que, lorsque les PNS associés à la résistance aux fongicides sont comparés par paires, trois types de relation spatiale peuvent survenir: l’absence de relation spatiale (56%), l’inclusion spatiale (31%) et l’exclusion spatiale (12%). En dépit du fait que plus de la moitié des paires de PNS testées ne montraient aucune relation spatiale, la présence de relation spatiale inclusive (31%) supporte les modèles de coexistence entre phénotypes sensibles et résistants, mais suggère un niveau de complexité supérieur. Pour la seconde expérience, la distribution bêta-binomiale s’ajustait mieux aux données que la distribution binomiale pour tous les jeux de données.

Les valeurs estimées de l’indice d’agrégation θ étaient comprises entre 0.09 et

0.23 (valeur médiane de 0.20), indiquant une agrégation locale des PNS au sein d’une même unité d’échantillonnage. Finalement, en se basant sur les niveaux d’agrégation observés, des courbes d’échantillonnages ont été calculées pour différentes incidences de PNS et différents niveaux de précision. Ces résultats mettent ainsi l’emphase sur l’importance de l’échantillonnage pour une détection rapide de la résistance aux fongicides en épidémiologie.

Acknowledgment

Un grand nombre de personnes doivent être remerciés (dans l’ordre ou dans le désordre), sans qui tout ceci n’aurait par été possible. Tout d’abord, j’aimerais remercier Luc Brodeur, avant tout pour m’avoir fait confiance et pour m’avoir laissé aller jusqu’au bout; merci Luc. Je voudrais également remercier

Odile Carisse, elle aussi pour m’avoir fait confiance, et la première à avoir suggéré que j’en étais capable. Merci aussi pour ton temps, ton savoir et pour ton appui. Merci à Pierre Dutilleul pour tout votre temps et votre enseignement, pour vos statistiques spatiales et pour vos questions à 5 Francs Belge. Merci aussi à

Jean-Benoît Charron pour ton aide et ton soutien.

Merci aussi aux membres du labo, Mathieu, Annie, Audrey et Mamadou.

Merci à Mélanie pour son travail de fourmis. Merci à Ameur et Liwen pour leur aide avec SAS et MatLab. Merci à l’équipe de PRISME et Phytodata pour leur aide et leur compréhension.

J’aimerais aussi te remercier pour ton soutien, ton appui et ton amour; merci Véro. Merci pour nos filles (Camille, Laurence et Mathilde) pour votre présence dans ma vie.

Enfin, j’aimerais remercier Agriculture et Agroalimentaire Canada ainsi que la Compagnie de Recherche Phytodata pour avoir fourni le financement pour que ce projet puisse être réalisé.

13 Contribution of authors

This thesis has been written in the form of manuscripts. The Faculty of

Graduate and Postdoctoral Studies at McGill University have approved this format, as described in the “Guidelines for Thesis Preparation and Submission”

(www.mcgill.ca/gps/sites/mcgill.ca.gps/files/theses_guidelines_2013.pdf). This research was planned with Dr. Odile Carisse in collaboration with Dr. Pierre

Dutilleul for the statistical aspects. This thesis is composed of seven chapters.

Chapters 1 and 2 are the general introduction and the literature review, respectively. Chapter 3 is a theoretical study of spatial co-occurrence between mutations related to fungicide resistance in B. cinerea populations. In Chapter 4, we present a new PCR-RFLP assay developed for the detection of a mutation related to fungicide resistance in B. squamosa, and Chapter 5 reports an original study of the spatial distribution pattern of mutations related to fungicides resistance in B. cinerea and squamosa populations. Chapters 6 and 7 are made of the general conclusions and the literature cited, respectively. The manuscripts derived from this research were or will shortly be submitted for publication.

The first manuscript (Chapter 3) is co-authored by the candidate, Dr.

Pierre Dutilleul, Dr. Odile Carisse, Dr. Jean-Benoît Charron and Mr. Luc Brodeur.

The candidate performed the sampling, PCR and statistical analyses and wrote drafts of the manuscript, all of this under the guidance of Dr. Carisse and Dr.

Dutilleul. Dr. Carisse and Mr. Luc Brodeur provided the funding to conduct this research. All the authors were involved in the manuscript edition and revision.

The second manuscript (Chapter 4) is co-authored by the candidate, Mr.

Mathieu Tremblay, Dr. Jean-Benoît Charron, Dr. Dutilleul and Dr. Odile Carisse.

Mr. Tremblay designed the PCR assay, and the candidate validated the assay, performed the data analyses and wrote drafts of the manuscript. Dr. Carisse provided the funds to conduct this research and Dr. Charron provided useful insight in the early stage of preparation of the manuscript. All the authors were involved in the manuscript edition and revision.

The third manuscript (Chapter 5) is co-authored by the candidate, Dr.

Pierre Dutilleul, Dr. Odile Carisse and Mr. Luc Brodeur. The candidate performed the sampling, PCR and statistical analyses and wrote drafts of the manuscript, all of this under the guidance of Dr. Carisse and Dr. Dutilleul. Dr. Carisse and Mr.

Luc Brodeur provided the funding to conduct this research. All the authors were involved in the manuscript edition and revision.

1 General introduction

1.1 Introduction

Botrytis species are among the most ubiquitous and versatile plant pathogens (Jarvis, 1977). They are the causing agents of grey mould diseases on more than 235 hosts worldwide (Williamson, Tudzynski, Tudzynski, & Van Kan,

2007). They are particularly important in grapes, small fruits, vegetables and bulbous monocotyledons (Jarvis, 1977). Considerable efforts are spent every year for protecting crops against botrytis-induced diseases. Thus, the market size for fungicides specifically targeting Botrytis sp. represents 10% of the world fungicide market, with annual cost of about CA $700 million (Dean et al., 2012).

The wine and table grapes segment singly, represent 50% of the total botryticides market (Dean et al., 2012). Control losses incurred by the presence of fungicide(s) resistant pathogens is of increasing concern for growers, crop specialists and for the agro-chemical industry as well.

One of the most interesting characteristics of Botrytis sp. is the capacity of the pathogen to remain quiescent for various periods of time (Elad et al., 2007).

This particularity of Botrytis makes the assessment of fungicide efficacy difficult and can results in disease control failure. Among the conditions that can lead to disease control failure, fungicide resistance is of considerable importance.

Consequently, there is an increasing need for monitoring fungicide resistance.

Conventional methods for fungicide resistance testing based on the use of discriminant doses are tedious and time consuming, limiting overall sampling

capacity. Two approaches are conventionally followed and are both based on the culture of the target fungus in presence of a given fungicide. The first test consists in evaluating the germination of a spore suspension, and the second one, in measuring the radial growth of the target fungus on a growth medium supplemented with fungicide.

Recent development in the identification of mutations responsible for the resistant phenotypes and in new DNA based tools aiming at detecting these target mutations, have increased the potential for fungicide resistance monitoring.

Despite these tremendous outbreaks, changes in monitoring scale (from phenotypes to genotypes), suggest that the sampling approach should be adapted.

In plant pathology, like in entomology, sound sampling strategies are largely based on the identification of spatial distribution patterns of the targeted pathogen

(Madden, Hughes, & Munkvold, 1996). The spatial distribution pattern of mutations related to fungicide resistance has not been studied yet.

1.2 Objectives

The overall objective of this study is to determine the minimum sampling number of quadrat required for the estimation of mean incidence of mutations related to fungicide resistance in Botrytis populations. More specifically:

1. To develop a reliable PCR-RFLP assay for the identification of

dicarboximide-resistant strains of B. squamosa.

2. To investigate the spatial relationships between SNPs related to fungicide

resistance.

3. To formally characterize the spatial distribution pattern of three SNPs

related to fungicide resistance to boscalid and iprodione within Botrytis

cinerea populations in vineyards and for one SNP related to fungicide

resistance to iprodione within Botrytis squamosa populations in onions

fields.

4. To compute sampling curves relative to the estimation of SNP frequency

for various levels of spatial aggregation.

1.3 Hypotheses

1. H0: Mutation relative to dicarboximide resistance in B. squamosa is at the

same position than the mutation relative to dicarboximide resistance in B.

cinerea.

2. H0: There is no spatial relationship between mutations related to fungicide

resistance

3. H0: Mutations related to fungicide resistance follow a completely random

spatial distribution pattern;

4. H0: The minimum sample size required to estimate mean SNP incidence

with a given level of precision is inversely proportional to the frequency of

mutations.

18 2 Literature review

2.1 Spatial statistics

The study of spatial distribution patterns is thought to be the first step in the study of ecological and epidemiological processes (Karrandinos, 1976; Madden,

Hughes and van den Bosch, 2007). For instance, when looking at the spatial pattern of a plant pathogen, it is the complex interaction between dispersal processes and environmental factors that is being observed (Madden, Hughes and van den Bosch, 2007). In other words, spatial patterns are considered to be partial realizations of an underlying dispersal process. The characterization of spatial distribution patterns is helpful to understand disease dynamics and for the development of sampling strategies.

Spatial pattern characterization is largely dependent on the nature of the data collected. Simply put, the data can take two forms: point pattern versus surface pattern (Dutilleul, 2011, Chapters 3-5 and 6-8). A 2-D spatial point pattern can be seen as a set of locations (s1, s2,…, sn) in a defined study area; at these locations, a binary response takes the value 1 because of the occurrence of an event of interest or the presence of an individual of interest, and elsewhere, the value 0. Examples include the presence/absence of a given plant species in a wild environment and of a given pathogen in an agricultural field. In point patterns in general, the observations are represented as points that can be locations or times.

Conversely, if the response is continuous quantitative, meaning that it can take an infinitely large number of possible values, and is defined everywhere in the study

area, the corresponding 2-D spatial pattern observed at a finite number of locations is a surface pattern, made of the values taken by the response at those locations. Examples of surface patterns are provided by the measurement of soil pH, soil or foliar nitrogen content and airborne spore concentration. Like point patterns, surface patterns can be collected in space, time and space-time.

The nature of the data collected directs the choice of the statistical methods to be used for their analysis. For point patterns, the statistical analysis can follow the distance-based approach or the quadrat-based approach, and geostatistical methods are very popular and relevant for the analysis of spatial surface patterns.

2.1.1 Distance-based approach

A classical objective in the distance-based approach to point pattern analysis is to detect departure from complete, spatial or temporal, randomness (Diggle,

2003; Dutilleul, 2011; Upton & Fingleton, 1985). This approach relies on the information contained in the spacing between points of interest to characterize the spatial pattern (Clark & Evans, 1954; Diggle, 2003; Dutilleul, 2011; Gatrell,

Bailey, Diggle, & Rowlingson, 1996).

Typically, Euclidean distances between points are used in the statistical analysis. The Euclidean distance between two locations, denoted s and s’, is || s’ – s ||. In 2-D space, the coordinate vectors of s and s’ are (x, y) and (x’, y’), which are used to calculate the Euclidean distance (�! − �)! + (�! − �)! . When analyzing a spatial point pattern, three options are therefore possible: working

20 with all the distances between pairs of points; using distances between a number of sampling locations and the nearest point for each of them; or analyzing only the nearest-neighbour distances from the distances between points (Dutilleul, 2011;

Gatrell et al., 1996). In this chapter, the emphasis will be given to the third option because completely random and regular point patterns are hardily distinguishable when the frequency distribution of all the distances between points is used to characterize spatial heterogeneity (Dutilleul, 2011). Only the left-hand side of the frequency distribution of distances between points is studied when nearest- neighbour distances are used for analysis. The frequency distribution of nearest- neighbour distances is characterized by a relative constancy of frequencies for a completely random point pattern, a bell-shaped curve for a regular point pattern and an excess of small distances for an aggregated point pattern.

The presence of heterogeneity in the distribution of nearest-neighbour distances can be assessed with the randomization testing procedure developed by

Peter J. Diggle and initially presented in Diggle (1983); see also Dutilleul (2011,

Chapter 3). This randomization testing procedure is based on the construction of an upper and a lower envelope (similar to confidence intervals) by simulating independent partial realizations of the point process under the null hypothesis and the comparison of some statistical function evaluated for the observed point pattern with the two envelopes. More specifically, to test the departure from a completely random point pattern at an approximate significance level of 5%, the

2.5 and 97.5-th percentile envelopes obtained from 999 independent partial realizations of a completely random point process is plotted against the

cumulative relative frequency distribution of nearest-neighbour distances of the observed point pattern. When the observed cumulative relative frequency distribution lies outside the envelopes, the null hypothesis of complete randomness for the underlying point process is rejected, in favour of aggregation if above or regularity if under. Interesting, though not plant-oriented examples of the application of this randomization testing procedure are given in Dutilleul et al.

(2009) and Haltigin et al. (2010). In Chapter 3 of this thesis, the application is extended to bivariate point patterns.

2.1.2 Quadrat-based approach

In the quadrat-based approach to point pattern analysis in 2-D space, the studied area is divided into subareas of equal size (quadrats), points of interest are counted in each subarea, and the resulting counts are transformed into frequencies. The statistical methods available for the analysis of quadrat counts are based on the characteristics of their frequency distribution (Figs 10 and 11,

Chap. 5). Basically, the analysis relies on the relation and even the equality between the theoretical mean (E(X)) and the theoretical variance (Var(X)) of quadrat counts for a simple Poisson process (Dutilleul, 2011). Accordingly, the sample mean (�) and the sample variance (�!) of quadrat counts are used to calculate a ratio �! � , which is used to measure departure from complete spatial randomness. Thus, the value of �! � is expected to be close to 1.0 for a completely random spatial point pattern, well below 1.0 for a regular spatial point pattern and well above 1.0 for an aggregated spatial point pattern. However, observed values of the variance-to-mean ratio should be interpreted with due care,

since a simple Poisson process implies a variance-to-mean ratio of 1.0, but the reciprocal is not true (Dutilleul, 2011).

Therefore, the variance-to-mean ratio can only be used as an exploratory tool in the quadrat-based approach and the results obtained with it should be confirmed by other means. In plant pathology, disease incidence can be considered as a Bernoulli random variable, a plant being diseased with a probability p and the same plant being healthy with the probability 1 – p (Madden,

Hughes and van den Bosch, 2007). If p is the same for all plants in a given sampling unit (quadrat), the number of diseased plants in this sampling unit follows a binomial distribution with parameters n and p, which has a mean and a variance equal to np and np(1 – p), respectively (Madden & Hughes, 1995). If the plants in the same sampling unit tend to have the same health status and this status can change drastically from quadrat to quadrat (almost all plants in a quadrat are diseased or almost all are healthy), then p is not constant and can be considered, itself, as a random variable with the beta probability density function with parameters (α, β) (Dutilleul, 2011; Madden, Hughes and van den Bosch, 2007).

Thus, the number of diseased plants per sampling unit follows a β-binomial distribution with mean np and variance np(1 – p)(1 + nθ), where � = � (� + �) is the average probability of a plant being diseased and � = 1 � + � is an index of local aggregation (Dutilleul, 2011; Hughes & Madden, 1993, 1995). This approach has been introduced in plant pathology by Madden and Hughes (1993), and has been used to describe spatial patterns of several plant diseases since

23 (Hughes & Madden, 1993; Madden & Hughes, 1995; Madden, Hughes, & Ellis,

1995; Turechek & Madden, 1999; Xiao, Hao, & Subbarao, 1997).

2.1.3 Geostatistical approach

Statistical methods for the study of heterogeneity of spatial surface pattern have been first developed for mining and the earth sciences, and are now known as geostatistics (Krige, 1953; Matheron, 1962). These methods apply to all types of spatially referenced variables (incidence or severity in plant pathology). In plant pathology, the use of geostatistics has been introduced by Chellemi (1988).

Geostatistical methods are based on the analysis and estimation of the autocovariance ( �(ℎ) ) and semivariance ( � ℎ ) functions for pairs of observations separated by distance h (Dutilleul, 2011, Isaaks, 1989). Under second-order stationarity, the theoretical semivariance and the autocovariance function are related by � ℎ = � 0 − �(ℎ) (Dutilleul, 2011). Usually, the results are presented graphically in a (semi-) variogram, by plotting semivariance theoretical values or estimates against distances (classes or lags). In the analysis of a variogram, three characteristics are of interest: the range, the nugget and the sill (Dutilleul, 2011; Madden, Hughes and van den Bosch, 2007). The range represents the distance above which pairs of observations in space or time are no longer autocorrelated. The nugget (or nugget effect) indicates the presence of spatial variability at a scale smaller than the minimum distance between two sampling locations, and includes the experimental error. The sill is the maximum value of ℎ , which corresponds to the theoretical variance of the spatial

stochastic process. For positive spatial autocorrelation, the semivariance increases with increasing distance.

2.2 Botrytis spp.

2.2.1 Botrytis

Botrytis species are amongst the most important ubiquitous and saprophytic plant pathogens worldwide. The genus Botrytis belongs to the ascomycota phylum, the Leotiomycetes class, the Heliotiales order and the Sclerotiniacea family. The genus botrytis includes 22 species and one hybrid, reported to infect more than 235 hosts (Staats, van Baarlen, & van Kan, 2005). Fungi of the genus

Botrytis are important pathogens of many crops such as grapes, tomatoes, bulb and ornamentals crops. Among them some species have a broad range of host plants (i.e. Botrytis cinerea) while others are host specific (i.e. Botrytis squamosa).

2.2.2 Botrytis cinerea

Botrytis cinerea Pers. Fr. (telomorph Botryotinia fuckelinia (de Bary)

Whetzel) is probably the most known member of the genus Botrytis. It is a haploid, filamentous, heterothallic pathogen reported to infect over 200 commercial crops worldwide (Williamson et al., 2007). B. cinerea is responsible of grey mould in many crops including vegetables (i.e. lettuce, cucumber, tomato), ornamentals (i.e. roses, saintpaulia), bulbs (i.e. tulips, onion), stone fruits

(i.e. cherry) and small fruits (i.e. strawberry, raspberry). Crop losses related to B.

cinerea-induced diseases, represent 20% of the world harvest (Dean et al., 2012).

In grapes, B. cinerea is responsible for Botrytis bunch rot (BBR).

2.2.2.1 Epidemiology

B. cinerea overwinters as sclerotia or mycelium on canes, rachis or tendrils

(Elad, 2007). In the spring, sclerotia can either produce asexual conidiophores or sexual apothecia. The latter produce haploid ascospores that germinate directly into mycelium (De Istvanffi, 1903). Conversely, conidiophores produce airborne conidia that are multinucleated and heterocaryotic. Conidia are dispersed in vineyards through air (Jarvis, 1962a), insects, and were even reported to be transported through rain droplets (Jarvis, 1962b). Conidia germination requires the presence of a thin water film or of high relative humidity and the presence of plant debris (i.e. leaf, flower cap, petal) as nutritive substrate (Blakeman, 1980).

B. cinerea can infect at every stage of the plant development, but the flowering stage is of particular interest. If the infection occurs before the flowering stage, it causes the flower buds drop and if the infection occurs at flowering, the pathogen can remain quiescent until berries begin to ripen (Nair et al., 1995). Flower infection was reported to be a key stage in B. cinerea epidemiology (Keller, Viret, & Cole, 2003). Injury to berries by physical damage from insect, hail, machinery and wind prone cluster to B. cinerea infection and, together with wet conditions, lead to bunch rot expression (Keller et al., 2003).

2.2.2.2 Management

Management of BBR includes cultural (prophylactic), chemicals and biological control methods. Removal of crop debris on the vineyard floor can

26 induce an important reduction in B. cinerea inoculum (Savage, 1983). Pruning and shoot and leaf removal are common operations aiming at favouring air circulation hence reducing the risk of B. cinerea infections. Among cultural practices, measures such as reduction of nitrogen fertilization to reduce extreme vigour (P. Leroux & Clerjeau, 1985), effective control of powdery and downy mildew with active ingredients also effective against BBR, a good control of the grape berry moth and other insects and a good weeding around the vines also help managing B. cinerea.

The modes of action of biological control agents are various.

Microorganisms can modify leaf surface and interfere with attachment and growth of the pathogen on plant surface (Bunster, 1986). They can act by competition for nutrient (nitrogen, carbon, macro and micronutrients) or space. Interspecific competition results in reduced rate of spore germination and mycelial growth

(Elad, 2007). Parasitism of fungi by microorganisms is another approach generally based on cell wall-degrading enzymes (Elad, 2007). An important mode of bio-control is the induction of systemic plant resistance by microorganisms. In

Canada, only few bio-control agents are available for growers: Trichoderma hazarium (Rootshield), Bacillus subtilis (Serenade) and Streptomyces griseoviridis (Mycostop).

Several chemicals are available for the control of BBR, including multi- site inhibitors (e.g Chlorothalonil, dithiocarbamate), which generally act by blocking enzymes related to spore's respiration (Leroux, Gredt, Leroch, &

Walker, 2010). Even if they are considered as weak botryticides, multi-site fungicide often offers a good protection if used prior to infection (Van der

Heyden, Carisse, & Brodeur, 2012). Development of single-site fungicides has improved disease control but lead to fungicide resistance selection (Brent &

Hollomon, 2007). Among single-site fungicide registered against B. cinerea, strobilurins and carboxamide (e.g. azoxistrobine, boscalide) are affecting the mitochondrial respiration by inhibiting the activity of the mitochondrial complex

ІІІ and ІІ, respectively (Leroux et al., 2010). The dicarboximide fungicides (e.g.

Iprodione) are affecting osmoregulation by inducing lipid peroxidation, membrane destruction and overproduction of reactive oxygen forms. Another important family is the sterol biosynthesis inhibiting fungicide; they include the hydroxianilide group (e.g. Fenhexamid), which affects the C-4 demethylation during the ergosterol biosynthesis (Leroux et al., 2010).

2.2.3 Botrytis squamosa

2.2.3.1 Onion production

Dry bulb onion (Allium Cepa L.) is an important crop worldwide; it occupies the fourth place in the world production of vegetables, with a volume of 57.9 million tons annually (FAOSTAT, 2005). In North America, 166,238 ha were seeded with onion in 2009. Onion is one of the three main vegetable commodities produced in the USA. In Canada, onions are mainly produced in the Ontario and

Québec provinces. In 2007, 6,000 ha of dry bulb onion were produced in Canada for an estimated farm gate value of 57.3 million CDN ("Statistique Canada,

Division de l’Agriculture, Section des cultures. Productions de fruits et légumes.

June 2008. Statistique Canada, Ottawa, Canada. Catalogue No 22-003-X, vol 77, no 1,"). In the province of Québec, almost all onions (> 90%) are produced in a muck land area located in the southwest of Montréal (BPR, 2005).

2.2.3.2 Onion leaf blights

Several diseases can cause onion leaf blights, an induced premature defoliation resulting in reduced bulb diameter and consequent yield losses. The most destructive disease affecting onion leaves is downy mildew, caused by

Peronospora destructor (Berk.). Even if downy mildew can destroy a field rapidly, the disease is sporadic. Purple blotch, caused by Alternaria porri, is another common disease affecting onion leaves. Unlike downy mildew, A. porri is considered as an opportunist fungus infecting plant subjected to stress conditions.

Botrytis Leaf Blight (BLB) of onion is caused by the pathogen Botrytis squamosa

Walker (Walker, 1925). In the Montréal Muck Lands as well as in the Province of Ontario and the States of New York and Michigan, BLB is endemic and several fungicide sprays are needed every year to avoid economic losses, especially during wet growing season.

2.2.3.3 Importance of Botrytis leaf blight

In dry bulb onion, BLB symptoms are characterized by the development of small spots on green leaves, followed by blighting of the leaf tips. The disease can be detected in the field as soon as the first leaf has appeared, but most of the yield losses are induced by defoliation, which occurs at bulb initiation when bulb filling begins. At this growth stage, onions are more susceptible to BLB infection

(Carisse, Tremblay, McDonald, Brodeur, & McRoberts, 2011) Shoemaker and

29 Lorbeer (1977) have shown that BLB can reduce yield up to 30%. In the

Netherlands, De Visser (1996) has shown a yield reduction of 26% in untreated plots. In Québec, BLB can reduce yield up to 47% (Carisse et al., 2011).

2.2.3.4 Botrytis leaf blight Biology, symptoms and epidemiology

B. squamosa is a member of the Sclerotiniaceae family. It is characterized by straight conidiophores splitting into several branches at the end. These conidiophores carry ellipsoid conidia that are 10-to-16 X 16-to-24 µm (Carisse et al., 2011). B. squamosa sclerotia are round and flat, disk shaped, ranging from 0.5 to 4.0 mm in diameter, changing from white to black as they age.

The disease is characterized by yellow spots on green leaves, followed by leaf tip dieback and leaf blighting, reducing photosynthesis and consequently bulb diameter. Botrytis Leaf Blight is a polycyclic disease and epidemics are initiated with conidia produced on over-wintering sclerotia. The main sources of inoculum for initial infection are sclerotia formed on crop debris that over-wintered on soil or in culled piles (Ellerbrock and Lorbeer, 1977). The first cohort of conidia infects onion leaves and cause small whitish lesions surrounded by silver halos.

The symptoms can be visible as early as 24 hours after infection. As the lesion ages, the center becomes necrotic and dry and may develop a characteristic vertical slit. The number of lesions on leaves increase as the epidemic progresses, causing tip and leaf dieback. Because B. squamosa sporulates mostly on dead tissues (Hancock, and Lorbeer, 1963), secondary inocula increase as the dead leaf surface area becomes more important. Therefore, the disease progression rate is

considerably influenced by the rate of leaf dieback. The main part of B. squamosa life cycle is driven by asexual spores (conidia). The number of sporulation- infection cycle’s increases as the season progresses and it is only towards the end of the season that sexual reproduction occurs.

2.2.3.5 Management

Traditional recommendations to manage BLB include long rotation (3-4 years) with non-susceptible crops, removal of volunteer and cull piles, as well as use of tolerant cultivars. In the Montréal Muck lands, most growers use short rotations (1-2 years) and susceptible cultivars, due to lack of performing tolerant ones. Hence, in practice, BLB management relies mostly on fungicide applications.

Several fungicides are registered for the control of BLB and their mode of actions is either single- or multi-sites. Fungicides that belong to the multi-site group, such as dithiocarbamates, are usually protective and inhibit, by contact, spore germination and penetration into leaves (Brent & Hollomon, 2007). Onion growers largely use them because they are not expensive and when they are used under low disease pressure, they offer a relatively good protection. The impact factor on health and environment of multi-site fungicides is high because of the low specificity and activity of the active molecules. Alternatively, a few single- site fungicides such as Iprodione (dicarboximide group) are available. They are more efficient under high disease pressure because of their specific activity and most of them have a lower impact on the environment. For the control of BLB, fungicide spray programs usually consist in applying fungicides every 7 to 10

days from the 4-leaf growth stage until onion lodging. A common strategy is to begin the spray program with a preventive multi-site fungicide such as dithiocarbamate (mancozeb) or Chloronitril (Chlorotalonil) (Carisse & Tremblay,

2007). As the season progresses and disease pressure increases, iprodione is often used alternately with or in tank mixed with a dithiocarbamate. The direct consequence of this control strategy is a large number (10 to 14) of fungicide applications during the growing season. Hence, populations of B. squamosa are exposed to repeated applications of fungicides and fungicide resistance to certain fungicides has already been detected (Tremblay, Talbot, & Carisse, 2003).

2.3 Fungicide resistances

2.3.1 Resistance to anti-microtubule fungicides

The binzimidazoles fungicides (benomyl, carbendazim and thiophanate-methyl) prevent the microtubule assembly by binding to β-tubulin, inhibiting the mycelial growth and germ-tube elongation. The first resistant strains of Botrytis cinerea were found rapidly after the fungicide introduction (Leroux & Clerjeau, 1985).

The resistant phenotype was associated with a mutation of the Mbc1 gene (E198E and F200Y) encoding for the β-tubulin (Yarden & Katan, 1993). In Canada, the registration for the benzimidazoles fungicides was withdrawn in 2003.

2.3.2 Resistance to dicarboximides and phenylpyrroles

Dicarboximides (ipridione, vinclozolin and procymidone) and phenypyrroles

(fludioxinil and pyrrolnitrin) fungicides are both affecting osmoregulations. The precise target site of these fungicides has not been determined yet but they are

32 interfering with a two-component histidine kinase encoded by the OS1 gene (Cui,

Beever, Parkes, & Templeton, 2004). These fungicides affect osmotic signal transduction pathway leading to a lethal accumulation of glycerol within the fungus cells (Ochiai et al., 2001). Resistance to dicarboximides was associated with mutations (I86R/N/S) on the second repeat of the BcOS1 gene in field and laboratory mutants of Botrytis cinerea (Cui et al., 2004). However, the I86S substitution si thought to be the more frequent among field strains of B. cinerea

(Oshima et al., 2006; Oshima et al., 2002).

2.3.3 Resistance to anilinopyrimidines

The anylinopyrimidines fungicides (pyrimiethanil, mepanipyrim and cyprodinil) are methionine biosynthesis inhibitors. The target site has not been identified yet, and despite the possible candidates cystathionine β-lyase and the cystathionine γ- synthase, no mutation in the corresponding genes (BcmetC and BcmetB) could be associated with the resistant phenotype (Leroch, Kretschmer, & Hahn, 2011;

Leroux et al., 2002). However, recent developments in the identification of multidrug resistant strains have showed that MDR1h phenotypes lead to a strong resistance to cyprodynil and fludioxinil (Leroch et al., 2013).

2.3.4 Resistance to sterol biosynthesis inhibitors

In the SBI group, the only fungicide registered for the control of Botrytis sp. is fenhexamid. This fungicide inhibits the 3-ketoreducatse enzyme involved in the

C-4 sterols demethylation (Leroux et al., 2002). Resistance to fenhexamid was detected in field very quickly (only four years) after fungicide introduction

(Walker et al., 2013). The resistant phenotype was recently associated with mutation in the erg27 gene and classify into 3 groups: Hyd1, Hyd3- and Hyd3+.

The Hyd1 phenotypes are associated with detoxification of fenhexamid and belong to a new cryptic species (B. pseudocinerea) and were thought to be present in wild type populations before the introduction of fenhexamid (Walker et al.,

2012). This phenotype is considered to be of little importance compared to the

Hyd3 phenotypes. The Hyd3- phenotype is caused by many possible mutations in codons 195, 309, 314, 336, 369 or 400 but they only affect the efficacy of fenhexamid during the germination stage (Fillinger et al., 2008). The Hyd3+ phenotype is by far the most important one. It is caused by mutations in the codon

412 only (F412I/S/V) and leads to resistance at the germination and mycelial growth stage (Fillinger et al., 2008).

2.3.5 Resistance to Quinone Outside Inhibitors

Strobilurins or QoIs inhibits the cellular respiration by targeting the mitochondrial complex III; more specifically by targeting the active site of the cytochrome b

(Leroux, 2010). They are not targeting Botrytis spp. directly, but they are often used in mixture with SDHIs fungicides. Resistance to strobilurins is common in pathogenic fungi worldwide, and is caused by the G143A substitution in the cytochrome b protein (Walker et al., 2013).

2.3.6 Resistance to succinate dehydrogenase inhibitors

Boscalid was the first succinate dehydrogenase inhibitors registered against

Botrytis. More recently, another member of the SDHI group, fluopyram was

registered for the control of Botrytis. The SDHIs are affecting mitochondrial respiration by interfering with the subunits B, C and D of the complex II (Leroux et al., 2010). In the case of boscalid, resistance has been detected only few years after the fungicide introduction. The resistant phenotype was associated with several mutations in SDHB gene (H272Y/R/L, N230I or P225L/F/T) or in the

SDHD gene (H132R) (Leroux et al., 2010). Recently, Veloukas et al. (2011) have showed that mutations H272L, N230I and P225L/F/T were also related to resistance to fluopyram. The mutations H272R/Y are by far the most frequently observed (Veloukas, Leroch, Hahn, & Karaoglanidis, 2011; Yin, Kim, & Xiao,

2011).

2.4 Connecting text for chapter 3

The following chapter examine the spatial co-occurrence between different mutations related to resistance to different fungicides among Botrytis cinerea population in vineyards. This exploratory research was extended from the spatial coexistence model proposed by Parnell et al. (2005; 2006) who included, for the first time, spatial coexistence between resistant and sensitive phenotype in his model for evolution of fungicide resistance. However, coexistence between resistances to different fungicides was not yet considered. In this chapter the authors intend to determine if there is relationships between mutations related to fungicide resistance and their spatial location. The results of this research will be submitted for publication in an international journal in a near future.

35 3 Combined application of spatial statistics and molecular genetics for characterization of small-scale spatial relationships between polymorphisms related to fungicide resistance in Botrytis cinerea populations

H. Van der Heyden, P. Dutilleul, J. -B. Charron, L. Brodeur, O. Carisse

Keywords: Fungicide resistance monitoring, nearest-inter distances, SNP, spatial point pattern analysis

3.1 Abstract

The objectives of this project were to study the occurrence and distributions of

SNPs related to fungicide resistance and explore their spatial interactions within

Botrytis cinerea populations isolated from grapes. B. cinerea isolates were collected following a quadrat-based design (100 10x10m quadrats) in two commercial vineyards. The presence of 9 SNPs related to resistance to iprodione, boscalid, azoxystrobin and fenhexamid was detected using PCR-RFLP, PIRA-

PCR and RT-qPCR assays. These data were spatially referenced and considered as a multivariate point pattern in a given vineyard. Spatial point patterns were analyzed by pairs, using an extension of Diggle’s procedure for the analysis of nearest-neighbor distances. In this randomization testing procedure the cumulative relative frequency distribution of the inter-SNP distances was used to characterise the spatial relationship between pairs of SNPs related to fungicide resistance.

Using discriminant doses, we found that 95%, 66%, 56% and 3% of the B. cinerea isolates were resistant to azoxystrobin, iprodione, boscalid and fenhexamid, respectively. When using molecular methods based on SNP detection, the proportions obtained indicate resistance to azoxystrobin (90%) and

36 iprodione (64%), low resistance (67%), and moderate resistance (33%) to boscalid, and resistance to fenhexamid (1%). Our results show that three spatial relationships can arise when spatial point patterns representing the presence of

SNPs related to fungicide resistance are compared by pairs: spatial exclusiveness

(12%), spatial co-occurrence (31%) and absence of a spatial relationship (56%).

Despite de fact that the majority of the tested pairs of single nucleotide polymorphism (SNP) show no spatial relationship, the presence of inclusive spatial relationship support models of co-existence between sensitive and resistant strains, but suggest a higher level of complexity in the resistant-sensitive interactions.

3.2 Introduction

Pers. Fr. (telomorph Botryotinia fuckelinia (de Bary) Whetzel) is probably the most known member of the genus Botrytis. This haploid, filamentous and heterothallic pathogen has been reported to infect over 200 commercial crops

(Williamson et al., 2007). B. cinerea is responsible of grey mould in many crops, including vegetables (i.e. lettuce, cucumber, tomato), ornamentals (i.e. roses, saintpaulia), bulbs (i.e. tulips, onion), stone fruits (i.e. cherry) and small fruits (i.e. strawberry, raspberry). Crop losses related to B. cinerea-induced diseases

represent 20% of the world harvest (Dean et al., 2012). In grapes, B. cinerea is responsible for Botrytis bunch rot (BBR).

Considerable efforts are spent every year on protecting crops against

Botrytis-induced diseases. Consequently, the fungicides specifically targeting

Botrytis sp. represents 10% of the world fungicide market (Dean et al., 2012). The segment for wine and table grapes solely represents 50% of the total botryticides market (Dean et al., 2012). Control losses incurred by the presence of fungicide- resistant pathogens are of increasing concern to growers, crop specialists and the agro-chemical industry as well. Nowadays, fungicide resistance is becoming a major concern in several crops, including those affected by B. cinerea (Brent and

Hollomon, 2007). Research efforts made on fungicide resistance have been targeted towards developing new modes of action, understanding molecular mechanisms and defining anti-resistance strategies (van den Bosch and Gilligan,

2008). The elaboration of anti-resistance strategies requires a good understanding of field resistance, including the spatial distribution patterns of resistant isolates.

Currently, very little is known about small-scale spatial distribution and field dynamics of fungicide resistance. Models of fungicide resistance evolution have assumed that resistant strains necessarily outcompete sensitive strains over time

(Chin, 1987; Gubbins and Gilligan, 1999; Milgroom, 1990) and that spatial co- occurrence is unlikely to occur. In those models, only two outcomes are expected at the field level: complete invasion of the resistant strains or non-invasion.

However, co-existence has been included as a third possible outcome in more recent models. For example, models developed by Parnell et al. suggest that co- existence of resistant and sensitive phenotypes is possible and that co-existence is

38 probable over an extended period of time (Parnell et al., 2005; Parnell et al.,

2006).

While most of the work on fungicide resistance so far has been aimed at understanding the mechanisms underlying the development of resistant phenotypes at the individual level, understanding fungicide resistance at the population level has proven to be challenging. The classical phenotypic methods are tedious and time-consuming, and often represent an obstacle to the processing of large numbers of samples. Molecular tools recently developed for rapid detection and characterization of fungicide resistance in fungal populations have increased sampling capacity and effectiveness, while allowing research work to be conducted at the population scale. Moreover, moving from classical phenotypic assays to molecular assays enables breaking down phenotypic resistance into single nucleotide polymorphisms (SNPs) among other genetic mechanisms, allowing the study of sub-population dynamics (Ma and Michailides, 2005). In fact, acquiring knowledge on the diversity and frequency of isolates carrying different SNPs in a fungal population is of high practical importance because each of them has a particular weight in terms of fitness and resistance factor (Billard et al., 2012b; Veloukas et al., 2011). Accordingly, the presence and abundance of isolates carrying a given SNP may have an impact on the presence and abundance of isolates carrying another SNP, and because of the resistance factor associated with each SNP, this may influence the efficacy of fungicide applications at the field level. However, the possible occurrence of fungicide resistance polymorphisms suggests a supplementary level of complexity in the interactions between resistant and sensitive strains of fungal plant pathogens.

39 During the last decade, a large number of studies were conducted to identify mutations related to fungicide resistance in B. cinerea in both laboratory and field mutants. For iprodione resistance, a polymorphic major gene (bc-Os1) has been reported as a histidine kinase (HK) gene (Cui et al., 2002; Oshima et al.,

2002), while a major mutation (G-143) was identified in the cytochrome b gene for azoxystrobin resistance (Leroux et al., 2010). A recent study (Billard et al.,

2012a) demonstrated that three possible mutations on the sterol 3-ketoreductase encoded by the erg27 gene were encountered for the fenhexamid-resistant phenotypes (erg27-412I, erg27-412S, erg27-412V). For boscalid resistance, several mutations (H272Y-R-L, P225F-L-T, N230I, H132R) have been identified on the B and D subunits of the succinate deshydrogenase (SDH) gene (Veloukas et al., 2011; Yin et al., 2011). With the availability of these DNA-based tools for the assessment of fungicide resistance, B. cinerea can be used to study small-scale spatial co-occurrence of SNPs related to fungicide resistance.

The spatial locations where a given polymorphism is present can be seen as points forming a pattern (i.e. a point pattern), which constitutes one partial realization of an underlying stochastic process (i.e. a point process). The analysis of such a univariate spatial point pattern consists in assessing departures from complete spatial randomness, following the distance-based approach or the quadrat-based approach or both these approaches (Diggle, 2003; Dutilleul, 2011;

Hughes and Madden, 1993). From an epidemiological perspective, the research hypotheses are whether the spatial point pattern is aggregated or regular, instead of completely random (i.e. the null hypothesis). To perform this assessment, the approach based on the smallest spatial distance between a point and all other

40 points in the same pattern (i.e. the nearest-neighbor distance) is often followed

(Diggle, 2003; Dutilleul et al., 2009; Gottwald et al., 2002; Haltigin et al., 2010).

Expanding the analysis from univariate to bivariate, the research and null hypotheses can be expressed in terms of spatial relationship. A spatial relationship between polymorphisms can be either present or absent, and in the presence of a spatial relationship, this can be inclusive or exclusive. A variant of the nearest- neighbor distance (i.e. the smallest spatial distance between a point from one pattern to all points of the other pattern) can be used for bivariate point pattern analysis (Dutilleul, 2011). A popular analytical procedure used in the quadrat- based approach to assess small-scale spatial aggregation is based on the beta- binomial distribution (Dutilleul, 2011; Hughes and Madden, 1993; Madden et al.,

2007), an extension of the binomial distribution which counts the number of

‘successes’ among n ‘trials’ with a probability of success, p.

Despite the importance of fungicide resistance and the availability of detection tools, the spatial relationships between isolates carrying different SNPs associated with fungicide resistance are not yet fully understood. In this study, we evaluated the abundance and spatial distribution of fungicide-resistant strains in B. cinerea populations in different grape production areas of Québec, and the SNPs related to phenotypic resistance to four fungicides were studied in two vineyards.

In these two vineyards, the resistant fraction of the B. cinerea population was considered as a sub-population, and was used to investigate spatial relationships between isolates carrying different SNPs related to fungicide resistance.

41 3.3 Material and methods

3.3.1 Inventory of phenotypic fungicide resistance

The first data collection done here was conducted at a regional scale in different growing areas of the Province of Québec. In 2010 and 2011, a total of 230 samples were taken in 23 fields in 18 vineyards located in Québec, Montréal,

Dunham, Frelighsburg, Vaudreuil-Dorion, Havelock, Rougemont, Napierville,

Brigham, St-Paul-d’Abbotsford, Mont-Saint-Hilaire, Farnham, Saint-Rémi, Saint-

Basile-le-Grand, Sutton, Ste-Etienne-de Beauharnois, St-Hyacinthe and Saint-

Charles-de-Bellechasse. In each vineyard, spores from 10 infected berries bearing single sporulating colonies were randomly collected on different bunches, using a dry BBL culture swab (Fisher Scientific, Ottawa, Ontario). Each swab was placed on a potato dextrose agar (PDA) plate supplemented with novobiocin (100µg ml-

1) and tetracycline (50µg ml-1) for purification and conservation.

3.3.2 Fungicides and conidial germination assay

All the fungicides used were analytical standard grades purchased from

Sigma-Aldrich (Sigma-Aldrich, St. Louis, Missouri). Each active ingredient was dissolved in ethanol or acetone (50ml L-1), and added to the liquid media at 50°C temperature after autoclaving. For each fungicide, a discriminant dose was applied following Leroux et al. (1999) (Table 1). All assays, except for boscalid, were conducted on solid PG agar media containing 10 g of glucose, 2 g of

K2HPO4, 2 g of KH2PO4 and 10 g of agar. For boscalid, the assays were conducted on solid PS agar media containing 4 g of sodium succinate, 2 g of

42 K2HPO4, 2 g of KH2PO4 and 12.5 g of agar. The agar media (PG and PS) were transferred into a 100-mm Petri dish. Spore suspension was adjusted to 200,000 spores ml-1, and 0.3 ml was spread over the surface of the agar upon medium solidification. Agar plates were incubated at 19°C in the dark for 24 h. After incubation, the percentage of germinated spores was evaluated from 100-200 spores for each treatment. Only the spores having a germ tube at least twice the average length of non-germinated spores were considered germinated.

3.3.3 Incidence of SNPs related to fungicide resistance

This second data collection was conducted in commercial vineyards in the Eastern

Townships region of the Province of Québec during the fall of 2011. Sampling was conducted in two vineyards planted with the Seyval cultivar. In each vineyard, a 100x100m-sampling grid, divided into 100 10x10m quadrats, was used. At the center of each sampling quadrat, spores from a sporulating colony on infected berries were randomly collected in duplicate, using a dry BBL culture swab (Fisher Scientific, Ottawa, Ontario). Each swab was plated on PDA amended with novobiocin (100µg ml-1) and tetracycline (50µg ml-1) for purification and conservation.

3.3.4 DNA extraction and PCR analysis

DNA extraction was performed by using a commercial isolation Kit (MP

Biomedicals, Solon, Ohio). The mutations related to distinct fungicide resistance were identified in RFLP-PCR and PIRA-PCR assays, using published primers

(Table 2). The PCR reactions were conducted in a 25-µl reaction volume

43 containing 3 µl of extracted DNA, 12.5 µl of PCR buffer (2X), 0.55 µl of each primer (10 µM) and 0.55 µl of Terra hot start polymerase (Clontech Laboratories,

Mountain View, California). The amplification was carried in a SureCycler 8800

(Agilent, La Jolla, California), with conditions set as follows: initialization at

98°C for 2 min; then, 30 cycles of 10 sec at 98°C and 40 sec at 68°C; and a final elongation step of 5 min at 68°C. The PCR products amplified by the different primer pairs were digested with a restriction enzyme, as recommended by the manufacturer. Digested products were separated by electrophoresis in 1.7-% agarose gel in 1X TAE buffer, and visualized after ethidium bromide staining under UV exposure. The detection of mutations F412wt-I-S-V related to fenhexamid resistance was performed using an ASPPAA assay as described by

Billard et al. (Billard, et al. 2012a).

3.3.5 Univariate spatial analysis

The presence of each SNP in the two Eastern Townships vineyards during the fall of 2011 was first mapped. Basically, the data were binary, the presence of a SNP in a sampling quadrat being coded 1 (i.e. a point) while its absence was coded 0 (i.e. no point). The resulting set of points (constrained to be located at the center of sampling quadrats) provided the observed univariate spatial point pattern for a given SNP.

In a preliminary step, the index of dispersion, D, was used to detect departure from complete spatial randomness. This index is defined as the ratio of the observed sample variance to the variance estimated under the assumption of a

44 completely random distribution in space (Carisse, et al. 2011; Hughes and

Madden 1993; Madden 2007):

s2 D = (1) pˆ(1− pˆ) / n where � is the estimated probability (empirical frequency) of presence of a SNP.

A value of 1.0 for D is suggestive of a completely random pattern over sampling quadrats, whereas values greater than 1.0 indicate departures due to heterogeneity. Under complete randomness, D follows, up to a multiplicative constant, a χ2-distribution (Madden 2007).

Then, beta-binomial and binomial distributions (Dutilleul 2011; Hughes and Madden 1993; Madden 2007) were fitted to the observed SNP frequency distributions as follows. The 100 initial sampling quadrats in a given vineyard were grouped into 25 larger units, made of 2x2 initial sampling quadrats each (i.e. the initial 10x10 sampling grid was reduced to a 5x5 grid, with 4 samples per unit). Thus, the number N of sampling units was 25, and the number n of samples per unit was 4. In these conditions, if the spatial location of an individual (i.e. a sporulating colony) possessing a given polymorphism is independent of the spatial location of another individual possessing the same polymorphism, the number X of individuals possessing the same polymorphism among the n samples across the N sampling units follows the binomial distribution:

n! P(X = x) = px (1− p)n−x x = 0, 1, 2, …, n (2) x!(n − x)! where p denotes the probability of success, a ‘success’ being an individual possessing the polymorphism of interest.

45 By comparison, the beta-binomial distribution (Y) has one more parameter:

θ > 0, which is a measure of the small-scale spatial aggregation of individuals possessing the polymorphism of interest, above what is expected (θ = 0) for the binomial distribution (i.e. complete spatial randomness):

y−1 n−y−1 ∏( p + jθ) ∏ (1− p + jθ) n! j 0 j 0 P(Y = y) = = = y = 0, 1, 2, …, n (3) y!(n − y)! n−1 ∏(1+ jθ) j=0

While a good fit by the binomial distribution is suggestive of a completely random spatial pattern of polymorphism incidence within sampling units, a good fit by the beta-binomial distribution is indicative of an aggregated spatial pattern at that level or scale. For a beta-binomial distribution with parameters n, p and θ,

np(11−p)( + nθ ) the population mean and variance are E(Y) = np and VRA (Y ) = . 1+θ

Accordingly, the population variance is factored as the binomial variance (i.e. np(1 – p)) times a heterogeneity component (1 + nθ)/(1 + θ) (Hughes and

Madden 1993). Since n is greater than 1 and θ is restricted to take non-negative values, the beta-binomial variance, together with the heterogeneity component, increases with θ, and as θ approaches 0, the beta-binomial variance approaches the binomial variance. Thus, the parameter θ is an index of spatial aggregation within the sampling unit (i.e. among samples within a unit). A likelihood ratio test and a C(α) test were used to determine whether the beta-binomial distribution fits the observed frequencies better than the binomial distribution. When possible, χ2- test statistics of goodness-of-fit were computed for both distributions. These

46 analyses were performed with a SAS routine obtained from http://www.oardc.ohio-state.edu/pp702/default.htm.

3.3.6 Bivariate spatial analysis

The statistical method used to analyze the spatial relationships between pairs of SNPs is a bivariate extension of Diggle’s randomization testing procedure originally defined for univariate spatial point patterns and using nearest-neighbor distances (Diggle 2003; Dutilleul 2011). This extension is based on the cumulative relative frequency distribution of n1 inter-SNP spatial distances, where n1 denotes the number of points (i.e. individuals, or sporulating colonies) of the first of the two spatial patterns compared. More specifically, for each of the n1 individuals with a given SNP, the nearest individual from the n2 sporulating colonies with another SNP is located from their respective spatial coordinates; the corresponding inter-SNP spatial distance is then the Euclidean distance between these two individuals.

The observed cumulative frequencies of inter-SNP distances were first obtained by ranking in ascending order the n1 distances between an individual with a given SNP and the nearest individual with the other SNP; dividing then the cumulative frequencies by n1 provided the cumulative relative frequencies.

Thereafter, the observed cumulative relative frequency distribution could be plotted as a step function, each cumulative relative frequency corresponding to a given observed inter-SNP distance. Prior to being superimposed on the observed cumulative relative frequency distribution, the 2.5th- and 97.5th-percentile

‘envelopes’ to be used in Diggle’s randomization testing procedure were

47 generated under the null hypothesis (i.e. absence of a spatial relationship). As recommended by Upton and Fingleton and others (Dutilleul 2011; Thébaud, et al.

2005; Upton and Fingleton 1985), these envelopes were constructed from 999

‘toroidal shifts’ of the spatial point pattern observed for the second SNP of the pair, while keeping the observed spatial point pattern of the first SNP of the pair fixed, under the assumption that there is no spatial relationship between the two.

In the torus model, points located near a corner of the vineyard are at small distance from points located near the opposite corner, so that the upper and lower edges are connected and so are the right and left edges. The method of toroidal shifts provides a nonparametric way to test for spatial independence when the study area is rectangular. These analyses were performed using a customized

SAS/IML code modified from “figure3.1.sas” on the CD-ROM distributed with

Dutilleul (2011).

For later interpretation of our results, an observed cumulative relative frequency distribution of inter-SNP distances falling within the 2.5th- and 97.5th- envelopes indicates the absence of a spatial relationship between the two SNPs considered; in this case, the observed curve lies inside the 95% ‘acceptance region’ of the null hypothesis (Fig. 1A). By contrast, the presence of one SNP at a given location may tend to exclude the presence of another SNP at the same location. Their spatial relationship is then mutually exclusive, and the observed cumulative relative frequency distribution lies below the lower envelope (Fig. 1B) because no zero inter-SNP distance is observed; in our study, the smallest, strictly positive value of inter-SNP distance to be observed is 10 m, the distance between centers of two contiguous 10x10m sampling quadrats. The third possible scenario

48 is that of an inclusive spatial relationship between two SNPs, the presence of one

SNP at a location implying the presence of the other SNP at that location.

Accordingly, the observed cumulative relative frequency distribution lies above the upper envelope, due to an accumulation of zero inter-SNP distances (Fig. 1C).

3.4 Results

From our Québec phenotypic survey, the incidence of fungicide resistance appears to be relatively constant over the two sampling years (Fig. 2). Based on the resistance percentage values plotted in Figure 2, the differences between fungicides, averaged over 2010 and 2011, can be summarized as follows. In decreasing order, about 91%, 89%, 61%, 60%, 49% and 5% of the collected strains were resistant to azoxystrobin, benlate, iprodione, pyrimethanil, boscalid and fenhexamid, respectively. Thus, using a discriminant dose, the level of fungicide resistance amongst B. cinerea populations in Québec vineyards varies from one fungicide to another, but is greater than 40% for most of the fungicides that we tested.

The genotypic survey conducted in two Québec vineyards in 2011 showed that the relative abundance of each SNP tested was similar in both fields (Fig.

3A). The most abundant SNP was G143A, which is related to azoxystrobin resistance and was detected in 89% (field 1) vs. 90% (field 2) of the collected strains. Mutations H272R, H272Y, H272L, N230I and P225F, which are related to boscalid resistance, were respectively detected in 65%, 29%, 2%, 4% and 0%

(field 1) vs. 67%, 24%, 2%, 0% and 0% (field 2) of the collected isolates. The

49 I86S mutation related to iprodione resistance was detected in 63% (field 1) vs.

65% (field 2) of the collected strains. Finally, the mutations F412I, F412S and

F412V related to fenhexamid resistance were detected in 3%, 0% and 0% (field 1) vs. 0%, 1% and 0% (field 2) of the collected strains.

In the first field, the number of mutations per isolate varied from 0 to 4, with the majority (58%) the collected strains carried 3 mutations (Fig. 3B). In the second field, the numbers of mutations per isolate varied from 1 to 4, with a majority (52%) of the collected strains were carrying 3 mutations (Fig. 3B).

Figures 4 and 5 present the observed univariate 2-D spatial point patterns for all the SNP positions identified in the sampled B. cinerea populations of the two Québec vineyards in the genotypic survey. Very few locations were tested positive for some of the SNPs studied, so little information could be extracted from the corresponding point patterns. Therefore, five of the univariate 2-D spatial point patterns from field 1 and four from field 2 were kept for further analyses. In the first vineyard, the index of dispersion, D, was significantly different from 1.0 for the mutations G143A and H272R, while it was different from 1.0 only for the mutation G143A in the second vineyard (Table 3). For all the other mutations tested, D was not different from 1.0 according to the χ2-test.

For both fields, the beta-binomial distribution fitted the observed frequencies of the mutations G143A and H272Y better than the binomial distribution (Table 3), indicating aggregation in their spatial patterns at small scale. Frequency distributions of the SNPs I86S and N230I in field 1 were better described by the binomial distribution, while the incidence of H272R in that vineyard was almost equally well described by both distributions. In the latter

50 case, estimates of the index of spatial aggregation, θ, were very close to zero, indicating a pattern close to complete spatial randomness (Table 3). Similarly, the incidence of the SNPs I86S and H272R in field 2 was better described by the binomial distribution (Table 3).

In the bivariate analyses, the three possible types of spatial relationship were observed on the pairs of SNPs tested (Figs. 6 and 7). The spatial relationships between the pairs G143A-H272R, G143A-I86S and H272R-I86S are essentially inclusive in fields 1 and 2 because the observed cumulative relative frequency curve for inter-SNP distances is mostly above the upper envelope, while being close, touching or even crossing it sometimes. For the pair H272R-

H272Y, the spatial relationship is exclusive in both fields as expected, since both mutations are on the same codon. For the remaining pairs (i.e. G143A-H272Y,

I86S-H272Y and G143A-N230I in fields 1 and 2, and I86S-N230I and H272R-

N230I in field 1), there is no evidence for a particular type of spatial relationship, as the observed curve is comprised between the 2.5th- and 97.5th-percentile envelopes.

3.5 DISCUSSION

Despite the relatively young history of grape production in Québec, the fungicide resistance levels found in the vineyards surveyed in our study, with the exception of fenhexamid, were comparable to those found in countries with longer history in grape production. The level of resistance to azoxystrobin (90%) found in Québec vineyards was higher than results obtained recently in Greece,

California and Germany, with respective resistance levels of 49%, 66% and 76%

51 for this class of fungicide (Fernández-Ortuño et al., 2012; Samuel et al., 2011;

Weber, 2011). In our study, the resistance level for benzimidazol was still around

89% even if the registration for this class of fungicide has been withdrawn in

2003. This result is in agreement with those reported for Monilinia fructicola

(Zhonghua et al., 2005), suggesting that this resistant phenotype is persistent over time. In Greece, New Zealand and Spain, resistance levels for benzimidazol were found to vary from 41 to 64% (Beever et al., 1989; Moyano et al., 2003;

Myresiotis et al., 2008). For dicarboximide, the resistance level in Québec (61%) compares to those of New Zealand, Japan, Germany and France (51 to 65%)

(Beever et al., 1989; Leroux and Clerjeau, 1985; Oshima et al., 2006; Oshima et al., 2002; Weber, 2011). The levels of resistance in Québec for boscalid (49%) and anilinopyrimidine (60%) are consistent or tend to be consistent with those in other regions of the world, respectively, 20-62% and 12-42% (Fernández-Ortuño et al., 2012; Latorre and Torres, 2012; Moyano et al., 2003; Moyano et al., 2004;

Veloukas et al., 2011; Weber, 2011; Yin et al., 2011). Conversely, resistance to fenhexamid appears to be very low in Québec (5%), whereas other studies showed levels of resistance ranging from 17 to 45% (Esterio et al., 2011; Grabke et al.,

2012; Latorre and Torres, 2012; Weber, 2011). A recent study by Billard et al.

(2011) showed that resistance to fenhexamid has a high fitness cost and that the most important consequence of acquiring that resistance is a low production of sclerotia at temperatures below 11°C (Billard et al., 2012b). Since the daily mean temperature between mid-September and mid-October (i.e. when sclerotia production takes place) is below 10˚C in southern Québec, this could explain the

52 very low proportion of fenhexamid-resistant isolates, compared to other production areas.

Recent developments in molecular detection of fungicide resistance have enhanced the surveillance and monitoring capacities of researchers and crop advisors. In our study, resistance was broken down into SNPs, and the resistance levels provided by classical phenotypic assays were generally similar to those obtained with PCR-based assays. The benefit of using molecular tools to test for fungicide resistance is not only to provide insight on the specificity of a polymorphism towards a particular resistance phenotype, but it could also lead to a more precise characterization in terms of fitness and practical resistance

(resistance factor). The case of boscalid is the most relevant in this regard. In

Québec, we found four mutations out of the five tested on 200 single-colony isolates. Interestingly, the mutation H272R (67% frequency) was by far the most dominant one, followed by the H272Y mutation (24% frequency). The value of the resistance factor associated with these two SNPs varies from low to intermediate, while the other possible SNPs related to boscalid resistance confer a much higher resistance factor (Veloukas et al., 2011; Yin et al., 2011). Hence, interpreting results about fungicide resistance on the basis of each SNP frequency may give rise to a more accurate estimation of practical resistance. In other words, because field resistance is influenced by the proportion of resistant individuals carrying SNPs associated with a low, intermediate or high resistance factor, the mixture of SNPs has a very important practical significance for field resistance.

Moreover, the large number of sampling quadrats with two or three SNPs observed in this study is strongly suggestive of SNP co-existence.

53 The statistical analysis of nearest-neighbor distances in the univariate case has seldom been used in plant pathology (Dallot et al., 2003; Gottwald et al.,

2002), while the procedure is more commonly applied in plant ecology (Dutilleul,

2011; Perry et al., 2006; Souris and Bichaud, 2011). Here, we performed bivariate spatial point pattern analyses based on inter-SNP distances to characterize spatial relationships between SNPs related to fungicide resistance in the field. The null hypothesis of absence of a spatial relationship was accepted for several pairs of

SNPs. Among the pairs for which the null hypothesis was rejected (i.e. there was evidence for a spatial relationship), the three SNPs related to resistance to azoxystrobin (G143A), dicarboximides (I86S) and boscalid (H272R) were found to have an inclusive spatial relationship. From the results obtained by fitting beta- binomial and binomial distributions to observed frequencies in the univariate case, spatial exclusiveness could be expected for pairs of SNPs with different spatial distributions (i.e. completely random vs. aggregated), and spatial inclusiveness might be expected for pairs of SNPs with aggregated spatial distributions or for

SNPs with high incidence. The SNPs in B. cinerea populations can be seen as genetic niches where one or several SNPs establish depending on their fitness cost or position in the organism’s genome. From a strictly phenotypic point of view, invasion by the resistant phenotype, co-existence between resistant and sensitive phenotypes and extinction of the resistant phenotype are the only three possible outcomes (Parnell et al., 2005; Parnell et al., 2006). When resistance is expressed in terms of SNPs, the number of possible outcomes increases, and a additional level of complexity is achieved.

54 Our results suggest that a genetic niche occupied by a specific polymorphism may either prevent invasion by other SNPs or co-exist with them in the field. This is well illustrated by the two following examples. First, among boscalid resistant strains, the abundant polymorphism H272R, which has a low- to-intermediate resistance factor, occupies the genetic niche and prevents invasion by the polymorphism H272L, which has a high resistance factor. Second, over two fungicide classes, the H272R and I86S pair of SNPs showed an inclusive spatial relationship, which is suggestive of the co-existence of these two SNPs.

Such results are supportive of a model of co-existence between sensitive and resistant strains, and demonstrate not only that co-existence is possible but also that heterogeneity in population phenotypes can slow down resistance development (Parnell et al., 2005; Parnell et al., 2006). The processes underlying these interactions are not yet fully understood, but the inclusion of spatial relationships in day-to-day monitoring or in modeling will facilitate our understanding of fungicide resistance at the population level. The monitoring of polymorphisms related to fungicide resistance should allow a better assessment of practical resistance. However, moving from conventional resistance monitoring with a binary outcome (resistant or sensitive) to molecular monitoring requires appropriate new protocols for sampling. Therefore, further research on temporal and spatial distributions is needed, in particular to include SNPs with low incidence because these may be important indicators of shifts in fungal populations.

55 3.6 Acknowledgements

The authors are grateful to Mathieu Tremblay for his constant support and

Annie Lefebvre and Audrey Levasseur for their essential technical assistance. We also thank Anne-Sophie Walker for her help during the experiment. This work was financially supported in part by Agriculture and Agri-Food Canada. The contributions of H. Van der Heyden and L. Brodeur to this work were partially supported by the Compagnie de Recherche Phytodata inc.

Table 1: List of fungicide groups and active ingredients and discriminant dose used for the phenotypic assay.

Active Discriminant dose Group ingredient (ppm)

Dicarboximides iprodione 2.5

Benzimidazoles carbendazim 1.0

Anilinopyrimidines pyrimethanil 1.0

Hydroxianilides fenhexamid 0.4 and 4.0

QoI azoxystrobin 10.0

Carboxanilides boscalid 3.3 and 7.0

57 Table 2: Chemical groups, active ingredients and sequences of the primers and probes used for the detection of the mutant alleles associated with fungicide resistance.

Chemical Active Restriction Mutation Primer Sequence Reference group ingredient enzyme Bgl-II (Veloukas et Anilide boscalid H272L H272L 5'-GGCAGCTTTGGATAACAGCATGAGTTTGTACAGAGATC-3' -FOR al., 2011) Hha-I Anilide boscalid H272R H272R-FOR 5'-GGCAGCTTTGGATAACAGCATGAGTTTGTACAGATGGC-3' EcoR-V Anilide boscalid H272Y H272Y-FOR 5'-GGCAGCTTTGGATAACAGCATGAGTTTGTACAGATAT-3'

Anilide boscalid H272rev H272-REV 5'-GCCATTTCCTTCTTAATCTCCGC-3'

Anilide boscalid N230I N230I-FOR 5'-GACCCAGCACCAGAAGGAAAA-3' BamH-I

Anilide boscalid N230I N230I-REV 5'-GATAGCTGGTCCCAAGTACTCCTCACGG-3' Hind-III Anilide boscalid P225F P225F-FOR 5'-GTATTCTCTGCGCATGCTGCTCGACATCAAGC3'

Anilide boscalid P225F P225-REV 5'-AAGCTGCCTTACGTTCTTCC-3' Fnu4h-I (Leroux et al., Qoi azoxystrobin G143A Qoi2ext 5'-GGTATAACCCGACGGGGTTATAGAATAG-3' -FOR 2010)

Qoi azoxystrobin G143A Qoi4ext-REV 5'-AACCATCTCCATCCACCATACCTACAAA-3' Taqα-I (Oshima et Dicarbopximide iprodione I86S Bcos5- 5’-GAGGCTTTCCAAAAAGCTCT-3’ FOR al., 2002)

Dicarboximide iprodione I86S Bcos10-REV 5’-TCTTGGTCAAATCTCCTCTGGCGACA-3’ (Billard et al., Hydroxyanilide fenhexamid F412 erg27- 5'-TGTTTCGGAGATCATGCCC-3' FOR 2012)

Hydroxyanilides fenhexamid F412wt erg27WT-REV 5'- CCATCCATCTTACAAGGTCGAAG-3'

Hydroxyanilide fenhexamid F412wt erg27WT-PRO 5'- FAM-TTATCTACAGATTGATCTTC-MGB-NFQ-3'

Hydroxyanilide fenhexamid F412I erg27F412S-REV 5'- CCATCCATCTTACAAGGTCGG-3'

Hydroxyanilide fenhexamid F412I erg27F412S-PRO 5'- FAM-TTTATCTACAGATTGATCTCC-MGB-NFQ-3'

Hydroxyanilide fenhexamid F412S erg27F412I-REV 5'- CCATCCATCTTACAAGGTCGATG-3'

Hydroxyanilide fenhexamid F412S erg27F412I-PRO 5'- FAM-TTATCTACAGATTGATCATC-MGB-NFQ-3'

Hydroxyanilide fenhexamid F412V erg27F412V-REV 5'- CATCCATCTTACAAGGTCGACG-3'

Hydroxyanilide fenhexamid F412V erg27F412V-PRO 5'- FAM-TTTATCTACAGATTGATCGTC-MGB-NFQ-3'

58 Table 3: Index of dispersion D and parameters estimate of the beta-binomial and binomial distribution.

v w x y z SNP D θ(StdE) ρ(StdE) χ2(BBD) χ2(BD)

Field 1 G143A 1.51* 0.1617 (0.1662) 0.8898 (0.0360) 0.050 0.514

Daf-1 1.17 0.0447 (0.0975) 0.6300 (0.0512) 0.899 0.528

H272R 1.38 0.1203 (0.1211) 0.6603 (0.0544) 2.138 2.888

H272Y 1.48* 0.1687 (0.1387) 0.2791 (0.0533) 0.004 1.331

N230I 0.91 <0.001(<0.0001) 0.0400 (0.0197) - -

Field 2 G143A 1.85* 0.3932 (0.3016) 0.9019 (0.0379) 0.003 2.812

daf1 0.92 <0.001 (<0.0001) 0.6190 (0.2097) - 0.008

H272R 0.92 <0.001 (<0.0001) 0.6633 (0.2111) - 0.950

H272Y 1.08 0.0149 (0.0909) 0.2400 (0.0435) 0.270 0.401

v D is the index of dispersion, values followed by * are significant at the 0.05 level. w Parameter estimated of the beta-binomial distribution (θ) and standard error in parenthesis. x Parameter estimated of the binomial distribution (p) and standard error in parenthesis. y χ2 values of the goodness of fit test for the beta-binomial distribution. z χ2 values of the goodness of fit test for the binomial distribution.

59 1.0 A Inclusive 0.8 Random

0.6

0.4

Exclusive 0.2

0.0 0 20 40 60 80

1.0 B

0.8 Inclusive

0.6

0.4

0.2 Random Exclusive Cumulative relative frequency relative Cumulative

0.0 0 10 20 30 40 50

1.0 C Inclusive 0.8

0.6 Random 0.4

0.2 Exclusive

0.0 0 2 4 6 8 10 12 14 Distance (m)

Figure 1: Schematic representation of the three possible outcomes in the analysis of inter-SNP distances A) absence of a spatial relationship, B) exclusive relationship and C) inclusive relationship. The continuous line represents the observed cumulative relative frequency distribution, while the dashed lines represent the 2.5th- and 97.5th-percentile envelopes.

60 Figure 2: Proportion of resistant isolates to six fungicides obtained in a conidial germination assay for 232 Botrytis cinerea samples, collected from 23 plots in 18 vineyards from different production areas in the Province of Québec.

61 Figure 3: A) Frequency distribution of the single nucleotide polymorphism

(SNPs) found in two vineyards in 2011 and B) number of SNPs per isolates. Grey bars: first field; black bars: second field.

62 A (G143A) B (H272R) C (daf-1) 100

0 D (H272Y) E (N230I) F (F412I) 100

40 Distance (m)

0 G (H272L) H I (F412S) 100 (QoI Intron)

0 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 Distance (m)

Figure 4: Univariate 2-D spatial point patterns observed on the first field. A point indicates the presence of a SNP for mutation G143A (A), H272R (B), daf1 (C),

H272Y (D), N230I (E), F412I (F), H272L (G), G143A-Intron (H) and F412S (I), respectively.

63 A (G143A) B (H272R) C (daf-1) 100

0 D (H272Y) E (H272L) F (F412S) 100 Distance (m) 80

0 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 Distance (m)

Figure 5: Univariate 2-D spatial point patterns observed on the second field. A point indicates the presence of a SNP for mutation G143A (A), H272R (B), daf1

(C), H272Y (D), H272L (E) and F412S (F), respectively.

64 1.0 A B C

0.8

0.6

0.4

G143A-H272R G143A-I86S G143A-H272Y 0.2 L L L U U U 0.0 0 10 20 30 40 A) 0 5 10 15 20 25 0 5 10 15 20 25 30 1.0 D E F

0.8

0.6

0.4

H272R-I86S H272R-H272Y I86S-H272Y 0.2 L L L U U Cumulative relative frequency relative Cumulative U 0.0 0 5 10 15 20 25 0 5 10 15 20 25 30 35 0 10 20 30 40

1.0 G H I

0.8

0.6

0.4

G143A-N230I I86S-N230I H272R-N230I 0.2 L L L U U U 0.0

0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 inter-distance (m)

Figure 6: Bivariate analysis of the 2-D spatial point patterns of Figure 4 (first field). The continuous lines represent the observed cumulative relative frequency distributions of inter-SNP distances, while the dashed lines represent the 2.5th- and

97.5th-percentile envelopes.

65 1.0 A B C

0.8

0.6

0.4

G143A-H272R G143A-I86S G143A-H272Y 0.2 L L L U U U 0.0 0 5 10 15 20 25 0 5 10 15 20 25 0 10 20 30 40

1.0 D E NN Distance (m) F

0.8 Cumulative relative frequency relative Cumulative

0.6

0.4

0.2 H272R-I86S H272R-H272Y I86S-272Y L L L U U U 0.0

0 5 10 15 20 25 0 10 20 30 40 0 10 20 30 40 Inter-distance (m)

Figure 7: Bivariate analysis of the 2-D spatial point patterns of Figure 5 (second field). The continuous lines represent the observed cumulative relative frequency distributions of inter-SNP distances, while the dashed lines represent the 2.5th- and

97.5th-percentile envelopes.

66 3.7 Connecting text for chapter 4

In the previous chapter, we identify three types of possible spatial relationships between mutations related to different fungicide resistance: coexistence, exclusion and no spatial relationship. Among the interesting mutations, the I86S (dicarboximides resistance) substitution was of particular interest because of its coexistence relationship with G143A and H272R mutations.

Furthermore, resistance to dicarboximide have been found in B. squamosa populations. In order to deepen our knowledge on spatial distribution pattern of mutations related to fungicide resistance, and to find out if similarity can be found in two different pathosystem (in chapter 5), we developed in chapter 4 a new

PCR-RFLP assay to detect the presence of I86S mutation related to dicarboximide resistance in B. squamosa.

The manuscript derived from chapter 4 was submitted to Crop Protection.

The manuscript was written by the candidate and co-authored with Dr. Odile

Carisse, Dr. Jean-Benoit Charron, Mr. Mathieu Tremblay and Dr. Pierre Dutilleul.

67 4 A novel PCR-RFLP assay for the detection of a single nucleotide polymorphism related to dicarboximide resistance in Botrytis squamosa field isolates

H. Van der Heyden, D.-M. Tremblay, J.-B. Charron, P. Dutilleul, O. Carisse

Keywords: Botrytis leaf blight, fungicide resistance, Histidine kinase, Iprodione.

4.1 Abstract

A PCR-RFLP assay was developed to accurately detect dicarboximide-resistant strains of Botrytis squamosa, responsible for Botrytis leaf blight of onion. This assay is aimed at the detection of mutation I86S, known to confer Botrytis cinerea resistance to dicarboximide, and is first described for Botrytis squamosa in our study.

4.2 Introduction

Botrytis leaf blight of onion (BLB) is caused by Botrytis squamosa J.C.

Walker (Walker, 1925). The pathogen is known to be highly specific to alliums, and has been reported on onion crops worldwide. In the eastern part of North

America, B. squamosa is endemic. The biology of B. squamosa and the epidemiology of Botrytis leaf blight have been under the scope of many researchers, and despite the availability of integrated disease management programs, disease control still relies mainly on fungicide applications (Carisse et al., 2012, Carisse et al., 2011, Van der Heyden et al., 2012). Several classes of fungicides against B. squamosa are registered, including multi-site and single-site inhibitors. Among single-site inhibitor fungicides, the dicarboximide family is

68 widely used for the control of Botrytis-induced diseases such as BLB. The occurrence and incidence of resistant isolates in targeted populations are key factors when making a disease management decision. For Botrytis cinerea, the mechanisms of resistance to several fungicides are well documented, and are mainly caused by single nucleotide polymorphisms (SNP). For the dicarboximide family, the I86S SNP, at position 86 of the second repeat of the osmosensing histidine kinase encoded by the BcOS1 gene, is primarily responsible for the resistant phenotype (Cui et al., 2002, Oshima et al., 2002). DNA-based detection assays are available for dicarboximide resistance and these assays have been used to draw up inventories of resistance in B. cinerea populations (Cui et al., 2004,

Oshima et al., 2006). However, such knowledge is unavailable for B. squamosa, even if dicarboximide resistance is a real problem for the control of BLB. Till now, dicarboximide resistance has only been studied with limited sample size and classical phenotypic assays (Carisse and Tremblay, 2007, Currah and Maude,

1984, Presly and Maude, 1982, Tremblay et al., 2003). To understand field resistance, it is essential to develop reliable tools based on the detection of mutations associated with resistant phenotypes.

In this study, we describe a reliable PCR-RFLP assay devised for the identification of dicarboximide-resistant strains of B. squamosa. The test was designed to detect the I86S SNP that is responsible for the dicarboximide-resistant phenotype.

69 4.3 Material and methods

A total of 18 single-spore isolates (Tremblay et al., 2003) were used for the development of the assay and 24 single-colony isolates were used for validation.

These 42 isolates were collected from onion fields in the Montréal muck lands located in the western-southern part of the Province of Québec. As a first step, a portion of a gene homologous to BcOs-1 was isolated from B. squamosa. Single- spore and single-colony isolates were maintained on Potato Dextrose Agar medium (Difco Laboratories, Detroit, MI), and DNA extraction was carried out with a fastDNA extraction kit (MP Biomedical, Solon, OH). The PCR primers

BcOS5 (5’-GAGGCTTTCCAAAAAGCTCT-3’) and BcOS10R (5’-

TCTTGGTCAAATCTCCTCTGGCGACA-3’) (Oshima et al., 2006) were used to amplify a 1030-bp fragment corresponding to nucleotides 22 to 1051 of the B. cinerea BcOS1 gene. Each PCR reaction was conducted in a 25-µl reaction volume containing 45 ηg of template DNA, 1X PCR buffer, 0.150 µM of each primer, 1.6 µg of BSA and 1X Titanium taq DNA polymerase (Clontech

Laboratories, Mountain View, CA). The amplification was carried in a Biometra

T-gradient (Montréal Biotech Inc., Montréal, Canada), with conditions set as follows: initial denaturation at 95°C for 5 min; then 35 cycles of 30 sec at 95°C,

45 sec at 62°C and 60 sec at 68°C; final elongation step of 5 min at 68°C.

Amplicons were cloned using a TOPO TA cloning kit (Life technologies,

Burlington, ON) and sequenced. The sequences were aligned using Megalign

(DNASTAR inc., Madisson, WI), to confirm the presence of the Iso86 → Ser

(I86S) substitution among resistant isolates (Fig.8). Based on the consensus, a

70 sequence-specific primer pairs was designed for the amplification of a shorter fragment containing the single restriction site necessary to the detection of the

ATC x AGC mutation underlying the I86S SNP. Primers BsqdxF (5’-

CAGATATTCTCGAGTGAGGTCTCC-3’) and BsqdxR (5’-

CTCTGCCTGAACCTTCTGTGTCAA-3’) were developed to amplify a 560-bp fragment. After amplification of the wild-type isolates, the PCR products were digested into two fragments of 160 and 400 bp, using the restriction enzyme

TaqαI (NEB, Whitby, ON), and were separated by electrophoresis on 1.7% agarose gel.

4.4 Results

Using the newly devised PCR-RFLP assay, a fragment of size 1030 bp was obtained for each of the 18 single-spore isolates. Following sequencing, the 7 single-spore isolates resistant to iprodione were found to carry the Iso86 → Ser

(ATC x AGC) substitution (Table 4). The PCR-RFLP assay developed in our study was further validated using 24 single-colony isolates. A 560-bp fragment was obtained in all cases, but only the PCR products of wild-type isolates could be digested into two fragments of 400 and 160 bp by the restriction enzyme

Taqα1, whereas PCR products from resistant isolates remained undigested (Fig.

9). Among the 24 single-colony isolates used for validation, 7 had the I86S mutations, indicated by a single 560-bp fragment (Fig. 9). In all cases, the results obtained with the developed PCR-RFLP assay were confirmed by standard phenotypic assays (Tremblay et al., 2003).

71 4.5 Discussion

Genetic resistance mechanisms have been widely studied for B. cinerea, and various techniques are available for the detection of isolates resistant to several fungicide classes. In particular, Samuel et al. (2011) recently developed a Taqman qPCR assay allowing the detection and quantification of B. cinerea strains resistant to strobilurin. Contrastingly, genetic aspects of fungicide resistance have not been addressed so far for B. squamosa, thus delaying progress in the understanding of many facets of disease control such as spatial and spatiotemporal distributions of fungicide resistance. The detection of field resistance with large sample sizes is crucial for the understanding of resistance dynamics and the development of anti-resistance management strategies. The newly developed PCR-RFLP assay that we devised could further be used to draw up field inventories of dicarboximide-resistant strains in B. squamosa populations.

4.6 Acknowledgements

The authors are grateful to Annie Lefebvre and Audrey Levasseur for their valuable technical assistance. This work was financially supported in part by

Agriculture and Agri-Food Canada through a DIAP project. The contribution of

H. Van der Heyden to this work was partially supported by the Compagnie de

Recherche Phytodata Inc.

72 Table 4: Amino acid substitutions of BcOS-1 homologous in 18 single-spore isolates of Botrytis squamosa used for development.

Codon at position 86 of Amino acid at position Strain the second repeat 86 of the second repeat

Sensitive strains 12R11 ATC ILE 13R13 ATC ILE 18R01 ATC ILE 1R10 ATC ILE 1R04 ATC ILE 21R07 ATC ILE 23R02 ATC ILE 31R1 ATC ILE 5R01 ATC ILE 5R02 ATC ILE 7R15 ATC ILE Resistant strains 13R9 AGC SER 34R3 AGC SER 23 AGC SER 26R8 AGC SER 29R1 AGC SER 29R2 AGC SER 34R3 AGC SER

73 Figure 8: Predicted amino-acid partial sequence of B. squamosa BcOS-1 homologous, aligned with BcOS-1 of B. cinerea. The asterisk (*) at position 210 indicates the I86S substitution site and the dots (.) represent gaps introduced to maximize alignment.

74 600-bp

400-bp Sensitive isolate

100-bp Resistant I86S 600-bp

400-bp

100-bp

Figure 9: Validation results of the PCR-RFLP assay using 24 single-colony isolates. The presence of two fragments of 400 and 160 bp following digestion with Taqα1 indicates the absence of the I86S mutation. The presence of an entire

560-bp fragment implies the presence of the I86S mutation. Wild-type strains sensitive to iprodione are in lanes 1 to 12, 13, 15 to 17, and 24, while mutant strains resistant to iprodione are in lanes 14 and 18 to 23. The PCR negative control is in lane N.

75 4.1 Connecting text for chapter 5

In chapter 3 and 4, we identify three types of possible spatial relationships between mutations related to different fungicide resistance in B. cinerea populations and developed a PCR-RFLP assay to detect the presence of I86S mutation related to dicarboximide resistance in B. squamosa. In order to deepen our knowledge on spatial distribution pattern of mutations related to fungicide resistance, and to find out if similarity can be found in two different pathosystem, in chapter 5 we choose three mutations with various level of incidence in B. cinerea (I86S, H272R and H272Y) and the mutation developed in chapter 4 for B. squamosa, to formally investigate the spatial distribution pattern of these mutations related to fungicide resistance. Furthermore, we also study the impact of the spatial distribution pattern on the sample size required to estimate mean mutation incidence.

The manuscript derived from chapter 5 is considered for publication in

Phytopathology. The manuscript was written by the candidate and co-authored with, Dr. Pierre Dutilleul, Mr. Luc Brodeur and Dr. Odile Carisse.

76 5 Spatial Distribution of Single Nucleotide Polymorphisms Related to

Fungicide Resistance and Implications for Sampling

H. Van der Heyden, P. Dutilleul, L. Brodeur, and O. Carisse

Keywords: Local aggregation, spatial heterogeneity.

5.1 Abstract

Spatial distribution of single nucleotide polymorphisms (SNPs) related to fungicide resistance was studied for Botrytis cinerea populations in vineyards and for Botrytis squamosa populations in onion fields. Heterogeneity in this distribution was characterized by performing geostatistical analyses based on

(semi-)variograms and through the fitting of discrete probability distributions.

Two SNPs known to be responsible for boscalid resistance (H272R and H272Y), both located on the B subunit of the succinate dehydrogenase gene, and one SNP known to be responsible for dicarboximide resistance (I86S) were chosen for B. cinerea in grape. For B. squamosa in onion, one SNP responsible for dicarboximide resistance (I86S homologous) was chosen. One onion field was sampled in 2009 and another one was sampled in 2010 for B. squamosa, and two vineyards were sampled in 2011 for B. cinerea, for a total of four sampled sites.

Cluster sampling was carried on a 10x10 grid, each of the 100 nodes being the center of a 10mx10m quadrat. In each quadrat, 10 samples were collected and analyzed by RFLP-PCR or AS-PCR. Mean SNP incidence varied from 16 to 68%, with an overall mean incidence of 43%. In the geostatistical analyses, omni- directional variograms showed spatial autocorrelation characterized by ranges varying from 21 to 31m. Various levels of anisotropy were detected, however,

77 with variograms computed in four directions (at 0°, 45°, 90°, and 135° from the within-row direction used as reference) indicating that spatial autocorrelation was prevalent or characterized by a longer range in one direction. For all eight data sets, the beta-binomial distribution was found to fit the data better than the binomial distribution. This indicates local aggregation of fungicide resistance among sampling units, as supported by estimates of the parameter θ of the beta- binomial distribution ranging from 0.09 to 0.23 (overall median value: 0.20). On the basis of the observed spatial distribution patterns of SNP incidence, sampling curves were computed for different levels of reliability, emphasizing the importance of sample size for the detection of mutation incidence below the risk threshold for control failure.

5.2 Introduction

The spatial distribution of plant pathogens is the result of multiple, complex interactions between dispersal processes and environmental factors (Madden et al., 2007). For several decades, the spatial patterns of plant diseases have been of great interest, resulting in the development of a whole range of improved tools to be used by plant disease epidemiologists. This is especially true for the understanding of disease dynamics and the design of sampling strategies in space, time and space-time (Dutilleul, 2011; Ferrandino, 2004). Thoughtful sampling is at the basis of sound decision making in plant pathology and plant health management. The development of effective sampling strategies is dependent on the type of data available as well as on the spatial aggregation patterns within

fields and sampling units. The assessment of disease status is usually based on the evaluation of plant parts, entire plants or the whole plant population. In plant pathology, the collected data can either be binary (a plant is diseased, or not), discrete (counts), or continuous (severity). Thus, because the number of samples is often limited because of spatial, temporal or economic constraints, count data are commonly pooled to measure disease intensity (Ferrandino, 2004).

In the early 1990’s, Gareth Hughes and Laurence Madden have started to use discrete probability distributions to characterize spatial heterogeneity in plant disease epidemiology (Hughes & Madden, 1993, 1995; Madden & Hughes, 1994;

Madden et al., 1995; Madden et al., 1996; Madden et al., 1995). In this approach, the number of diseased plants among n plants randomly sampled in a quadrat, for example, is assumed to follow a binomial distribution for a completely random spatial pattern and a beta-binomial distribution for an aggregated spatial pattern.

In fact, one of the parameters (θ) of the beta-binomial distribution is a direct measurement of local aggregation. These two discrete probability distributions have been widely used to characterize homogeneity vs. heterogeneity of plant disease epidemics, for soilborne pathogens and airborne pathogens and viruses.

For example, Xiao et al. (1997) found that Verticiulium dahlia microsclerotia in soil were aggregated; in strawberry fields, Turechek et al. (1999, 2000) showed that leaf blight and leaf spot diseases tended to be aggregated; and for viruses,

Dallot et al. (2003) recently discovered aggregation in the Shraka disease, caused by the plum pox virus. These results were used in part to develop effective

79 sampling strategies, in order to help determine disease thresholds and eventually develop management strategies.

Even though many integrated management strategies have been developed to delay plant disease epidemics, disease control continues to depend on fungicide applications. An important factor for an efficient fungicide application is the proportion of resistant individuals present in the target population (Brent &

Hollomon, 2007). Despite the importance of the spatial distribution of fungicide resistance (to be distinguished from the spatial distribution of a plant pathogen), most of the epidemiological studies of fungicide resistance have been targeted on temporal changes, such as the increase in resistance frequency following consecutive fungicide applications (Mavroeidi & Shaw, 2005; 2006; Myresiotis et al., 2007). In the late 1990’s, Elmer et al. (1998) have studied the spatial component of dicarboximide resistance in Molinia fructicola populations on stone fruit. Surprisingly, the results obtained for stone fruit suggested various levels of heterogeneity in the spatial distribution of dicarboximide-resistant strains of M. fructicola populations. Invasion of a resistant strain over a sensitive one mainly follows from the difference in fitness between the two strains as well as the resistance factor associated with the resistant phenotype. Recent studies by Parnell et al. (Parnell et al., 2005; Parnell et al., 2006) have demonstrated that invasion is not the only possible outcome and there exist various levels of co-existence between resistant and sensitive strains over a prolonged period of time in a mixed fungus population.

In the last decade, breakthroughs at the individual level have been made in the identification of molecular mechanisms related to fungicide resistance. For the model organism Botrytis cinerea, responsible for diseases of more than 225 hosts, single nucleotide polymorphisms (SNPs) have been identified in both field and lab mutants and related to resistant phenotypes for at least six fungicide groups

(Banno et al., 2008; Billard et al., 2012; Billard et al., 2012; Cui et al., 2004; Cui et al., 2002; Fillinger et al., 2008; Fournier et al., 2003; Leroux et al., 2002;

Leroux et al., 2010; Ziogas et al., 2009; Oshima et al., 2006; Oshima et al., 2002;

Saito et al., 2009; Samuel et al., 2011). Ranges of resistance factor and impact on pathogen fitness where also determined for each of these SNPs. In addition, the results of Van der Heyden et al. (submitted) suggest that patterns of spatial co- existence are also possible at the SNP scale and that a genetic niche occupied by a specific SNP might either prevent invasion by or co-existence with other SNPs within a given population.

Therefore, breaking down the phenotypic resistance into genotypic factors (e.g.

SNPs) is of great practical importance for monitoring fungicide resistance and foreseeing shifts in resistant populations. Considering the value of knowing the frequency of resistant strains carrying a particular SNP in a given population, the question of sampling naturally arises in this approach. Few studies on fungicide resistance have focused on spatial aspects, and we could not find any studies in which the spatial distribution patterns of SNPs related to fungicide resistance were investigated (Elmer et al., 1998). Accordingly, by using the geostatistical and distributional approaches, the objectives of this study were: i) to formally characterize the spatial distribution pattern of three SNPs related to fungicide

81 resistance to boscalid and iprodione within B. cinerea populations in vineyards; ii) to do the same characterization for one SNP related to fungicide resistance to iprodione within Botrytis squamosa populations in onion fields; and iii) to compute sampling curves relative to the estimation of SNP frequency for various levels of spatial aggregation.

5.3 Materials and methods

5.3.1 Sampling protocols

Sampling for B. cinerea was conducted during the fall of 2011 in two commercial vineyards, each one planted with the cultivar Seyval and located in the Eastern Townships region of the Province of Québec (Canada). In both vineyards, sampling was organized following a 100mx100m grid divided in 100

10mx10m quadrats. At the center of each quadrat, spores from 10 single sporulating colonies were individually sampled in duplicate, using a dry BBL culture swab (Fisher Scientific, Ottawa, ON, Canada). Once in the lab, culture swabs were placed into a 2-ml skirted screw cap micro-test tube, containing 380µl of isopropanol (Sigma-Aldrich Canada Ltd., Oakville, ON, Canada). Each sample was placed at -20°C until upstream PCR analysis.

Sampling for B. squamosa was conducted in commercial onion fields in a muck land area near Montréal (Québec, Canada) during the falls of 2009 and

2010. In 2009, onion cultivar Red Bull was seeded on April 23rd, at the rate of 30 seeds m-1 with row spacing of 0.45m; in 2010, onion cultivar Hamlet was seeded on April 26th, at the same rate with same row spacing. Sites were different in the

82 two years, but at each site, sampling was organized following a 100mx100m grid divided in 100 10mx10m quadrats. At the center of each quadrat, 10 infected leaves were randomly collected on different onion plants and placed on ice, prior to being brought to the lab by the end of the day. Each leaf was cut into 2-cm parts and placed in a 100mmx15mm Petri dish on a sheet of 90-mm Whatman filter paper amended with 500µl of miliQ water (Fisher Scientific, Ottawa, ON,

Canada). The Petri dishes (10 per quadrat) were placed at room temperature (22 to

25°C) for 2 to 5 days. In each Petri dish, spores from single sporulating lesions were harvested using a cotton swab drenched with isopropanol (Sigma-Aldrich

Canada, Oakville, ON, Canada), and placed into a 2-ml skirted screw cap micro- test tube containing 300µl of isopropanol (Sigma-Aldrich Canada, Oakville, ON,

Canada). Each sample was placed at -20°C until upstream PCR analysis.

5.3.2 DNA extraction and RFLP-PCR

To obtain total DNA, 80µl of spore suspension was placed in a sterile 2-ml screw cap micro-test tube containing 100mg of glass beads (425-600µm) (Sigma-

Aldrich Canada, Oakville, ON, Canada). Tubes were shaken in a FastPrep-24 instrument (MP Biomedicals, Irvine, CA) at 4m/s during 20s, to fragilize the conidia cell wall and promote membrane disruption. Conidia were precipitated by centrifugation for 5s at 10,000g, and isopropanol was evaporated under vacuum using a SpeedVac (Savant Instruments, Inc., Holbrook, NY) instrument for 20min at 55°C. Conidia were re-suspended in 300µl of DNA extraction buffer, consisting of nuclease-free water (Integrated DNA Technologies, Inc., Coralville, IA) and

5% chelex-100 molecular biology grade resin (BioRad Laboratories, Hercules,

83 CA). Following resuspension, tubes were placed in a dry bath at 105°C for 15min, to complete cell disruption and DNA extraction. Tubes were finally agitated for

5s, using a table vortex first and then a centrifugation at 15,000g for 5min at 4°C.

Supernatant containing genomic DNA was placed at 4°C until upstream PCR analysis for a maximum of 2h or placed at -20°C until future use.

Detection of the I86S allele in B. cinerea was performed using a PCR-RFLP assay, as described by Oshima et al. (2002), and that of the H272R and H272Y alleles was completed with the procedure described in Yin et al. (2011). Similarly, identification of the I86S allele in B. squamosa was performed in a PCR-RFLP assay. The primer pair BsqdxF (5’-CAGATATTCTCGAGTGAGGTCTCC-3’) and BsqdxR (5’-CTCTGCCTGAACCTTCTGTGTCAA-3’) was used for amplification of a 560-bp fragment. The PCR reactions were conducted in a 25-µl reaction volume containing 3µl of DNA extraction, 2.5µl of PCR buffer (10X),

2µl of BSA (10mg ml-1), 0.25µl of dNTPs (20mM), 0.38µl of each primer (10µM), and 0.5µl of Titanium Taq polymerase (Clontech Laboratories, Mountain View,

CA). The amplification was carried in a SureCycler 8800 (Agilent, La Jolla, CA), with conditions set as follows: 95°C for 5min first and then 35 cycles of 30s at

95°C, 45s at 62°C, and 30s at 72°C; finally, 10min at 72°C.

The PCR products amplified by the primer pairs BcOS10R/BcOS5a and

BsqdxF/R were digested with the enzyme TaqαI, as recommended by the manufacturer. Digested products were separated by electrophoresis in 1.7%- agarose gel in 1X TAE buffer, and visualized after ethidium bromide staining under UV light.

5.3.3 Geostatistical analyses

Initially, maps of SNP incidence were produced from the number (from 0 to 10) of samples tested positive for the presence of a given SNP at the 100 spatially referenced quadrats on a 100mx100m sampling grid; these count data were assigned to the centers of quadrats. Then, the first phase of the method of coregionalization analysis with a drift (CRAD) was performed, to detect and remove any drift (i.e. large-scale component of the spatial pattern) (Pelletier et al.,

2009). In this initial geostatistical analysis, the count data collected on a grid for a given SNP were considered as a partial realization of a random function (Z), and their spatial pattern was decomposed into an estimated drift (�) plus a residual component (�):

� �! = � �! + � �! (i = 1, ...., 100) (1) where �! represents quadrat i. The drift was estimated with the “estimated generalized least squares” (EGLS) method, performed in a local procedure (L1) including the selection of an optimal moving window size. Our CRAD analyses were carried out with the freeware available at http://environmetricslab.mcgill.ca/Programs.html.

Secondly, to assess the presence of spatial autocorrelation in the count data, omni-directional and directional experimental variograms were computed on residuals (�!) and analyzed using PROC VARIOGRAM of SAS Version 9.3 (SAS

Institute, Cary, NC). Calculations were made at distance classes (lags) up to 60m for all 8 data sets. There were at least 150 pairs of observations in each distance

85 class, which assured a good level of precision in the estimation of semivariance values. Matheron’s classical semivariance estimator (Matheron, 1962) was used, which for omni-directional variograms is given by:

! ! ! � ℎ = ! (�(� ) – �(� )) (2) !! ! !,! ! !

! where h refers to a certain distance class, �! and �! denote any two quadrats whose centers are separated by a distance belonging to the distance class considered, and A(h) is the number of pairs of such quadrats. In addition, the semivariance was also estimated as a function of distance in four directions: at 0°,

45°, 90°, and 135° from the within-row direction, 90° representing the across-row direction. In this case, only the pairs of observations (i.e. CRAD first phase residuals) in the direction considered were included in the computation of the semivariance estimator. Variogram ordinates that start low and increase with increasing distance lag are the sign of presence of spatial autocorrelation. The distance lag at which the semivariance reaches a plateau (i.e. a constant value) is called “range”, and indicates the distance beyond which observations are no longer spatially correlated. When the plateau can only be reached asymptotically, the range is said to be “practical”, and then corresponds to the distance at which the semivariance is equal to 95% of the plateau. Other important variogram model parameters are: the nugget effect (i.e. the variogram ordinate at infinitely small distance) and the partial sill (measured as the difference between the height of the plateau and the nugget effect).

86 Because the spherical variogram model provides a good compromise between the Gaussian and exponential models in terms of rate of decrease of spatial autocorrelation with distance (Dutilleul, 2011), the spherical model was chosen for fitting:

! � ℎ = � + � 1.5 ! − 0.5 ! for 0 < h < a (3) ! !! where c represents the nugget effect, b is the sill associated with the spherical variogram model, and a denotes the range of spatial autocorrelation. The fitting of spherical variogram models to experimental variograms was performed by EGLS

(Pelletier et, 2004). Anisotropy was assessed by comparing the variogram model parameter estimates (i.e. range, nugget effect, partial sill) obtained in the four directions and with the corresponding omni-directional variogram.

5.3.4 Distributional analysis

The index of dispersion, D, was used to assess departure from complete randomness on each count data set. This index is a ratio of two variances: the observed variance of SNP incidence in the numerator and the variance of SNP incidence, estimated under the assumption of a binomial distribution, in the denominator:

!! � = ! (4) ! !!! /! where � is the estimated probability (empirical frequency) of presence of a SNP.

The general rule of interpretation is that a value of D equal to 1.0 (i.e. the two variances are equal) indicates a completely random pattern within a sampling unit

(i.e. a quadrat). A value of D greater than 1.0 is indicative of a departure from a

87 completely random pattern, and is suggestive of heterogeneity. Under the assumption of a completely random spatial distribution, (N − 1) D follows a χ2- distribution with N − 1 degrees of freedom, where N is the sample size (i.e. number of quadrats) (Madden et al., 2007).

Binomial and beta-binomial distributions were fitted to the frequency distributions of each count data set, using a SAS code obtained from http://www.oardc.ohio-state.edu/pp702/Downloads.htm. When the location of an isolate carrying a given SNP is independent of the location of another isolate carrying the same SNP, the probability p of an isolate carrying the SNP is constant over the field and the number X of isolates carrying the SNP, out of n isolates randomly sampled, follows a binomial distribution characterized by the probability function:

n! P(X = x) = px (1− p)n−x x = 0, 1, 2, …, n. (5) x!(n − x)!

The theoretical mean and variance of X are equal to �� and ��(1 − �) , respectively. Compared to the binomial distribution (X), the beta-binomial distribution (Y) has a supplementary parameter, θ, used as a measure of local aggregation in space, and its probability function changes for:

y−1 n−y−1 ∏( p + jθ) ∏ (1− p + jθ) n! P(Y = y) = j=0 j=0 (6) y!(n − y)! n−1 ∏(1+ jθ) j=0

88 It follows that the theoretical mean and variance of Y are �� and np(1 – p)(1 + nθ)/(1 + θ), respectively. Thus, the theoretical beta-binomial variance can be seen as the product of the theoretical binomial variance ( ��(1 − �) ) and a heterogeneity component ( (1 + ��) (1 + �) ). Accordingly, when local aggregation increases, the theoretical beta-binomial variance is inflated with increasing θ-value; reversely, as θ approaches 0, the beta-binomial distribution approaches the binomial distribution (Hughes & Madden, 1993; Madden &

Hughes, 1994; 1995; Madden et al., 2007). Therefore, a good fit of the beta- binomial distribution (i.e. the θ-estimate is far from zero) indicates a departure from a completely random pattern, which means local aggregation of isolates carrying the same SNP within a quadrat in our study.

A likelihood-ratio test was performed to determine whether the beta- binomial distribution provided a better fit to the observed frequencies of count data than the binomial distribution. A C(α) test based on the z-statistic was also used to determine if the aggregation in the spatial distribution of each SNP should be described by a beta-binomial distribution (Madden & Hughes, 1994). Finally,

χ2-test statistics of goodness-of-fit were computed for the beta-binomial and binomial distributions.

5.3.5 Sampling curves

Sampling curves were developed from the results obtained with the beta- binomial distribution, so that the mean SNP incidence could be estimated with the

89 “coefficient of variation of the sample mean” (Madden et al., 2007), denoted CV hereafter:

�(1−�) (1+� �−1 )/�� CV = (7) � where � = �/ (� + 1), n is the number of subsamples (isolates) taken in each sampling unit (quadrat), and N is the number of sampling units (number of quadrats). This equation can be rearranged to extract the sample size N:

1−� 1+� �−1 � = (8) ��CV2

The minimum number N of sampling units needed to estimate p with a given level of reliability was calculated from sampling curves developed for CV values of 10, 20 and 30%. Sampling curves were calculated for three values of θ: the maximum, minimum and median of the estimated values of θ for the beta- binomial distributions fitted with our eight data sets (Madden & Hughes, 1999;

Madden et al., 2007).

5.4 Results

5.4.1 SNP incidence

The raw count data collected in this study are mapped in Figure 10; see legend for the sample means and standard errors of the sample means. SNP incidence in B. cinerea populations is not consistently higher for one of the two vineyards, as the mean incidence is higher in the first vineyard (OR) for the

H272R allele, but higher in the second vineyard (TE) for the I86S and H272Y substitution. As for time effects, the mean incidence for the I86S substitution in B.

squamosa populations was much greater (ratio of 2.72) in 2010 than in 2009 (Fig.

10).

5.4.2 Geostatistical analyses

A drift (i.e. large-scale spatial component) was detected with the CRAD method in four of the eight data sets, accounting for 10.4 to 30.0% of the total variation (Table 5). After the removal of estimated drifts, spatial autocorrelation was found in six of the eight omni-directional variograms of residual components.

The estimated range of spatial autocorrelation varied from 21.0 to 31.7m, the mean range being of 24.5m (Table 5). The nugget value tends to increase with increasing mean SNP incidence, whereas the sill of the spherical variogram model remains relatively small in general, implying that a moderate portion of the variance is explained by a spatially autocorrelated structure. Furthermore, directional variograms showed a great variety of spatial dependency, from almost perfect isotropy in spatial autocorrelation (i.e. H272Y for OR in 2011) to various patterns of anisotropy, including the presence of spatial autocorrelation in only one direction (i.e. H272R for OR in 2011) and a large difference in the estimated range with direction (i.e. I86S for VH in 2010) (Table 5).

5.4.3 Spatial statistics

The value of the index of dispersion, D, ranged from 1.71 to 2.97, indicating a departure of medium-to-large magnitude of the frequency distribution of counts from a binomial distribution (Table 6). For all eight data sets, the maximum likelihood estimation procedure converged and the likelihood-ratio test

91 showed that the beta-binomial distribution provided a better fit to count data than the binomial distribution (Table 6). The z-value of the C(α)-test statistic ranged from 5.16 to 14.44 with P-values of 0.0001 (Table 6), meaning there is local aggregation within quadrats and this is well described with the beta-binomial distribution (Fig. 11). The estimates of the beta-binomial distribution parameters varied from 0.16 to 0.68 (average: 0.43) for � and from 0.09 to 0.23 (median:

0.20) for �. The aggregation index values (�) for H272R, H272Y and I86S in B. cinerea populations are all lower in vineyard OR than in vineyard TE, in a twofold ratio (Table 6). By contrast, the difference between θ-estimates for I86S in the B. squamosa populations of 2009 and 2010 is small (Table 6). Nevertheless, for all data sets, the θ-estimate is at least three times greater than its standard error.

5.4.4 Sampling curves

The developed sampling curves are presented in Figure 12, but to ease interpretation, numerical values are also presented in Table 3. For a lower CV (i.e. higher precision), the minimum sample size N required to estimate the mean SNP incidence should be larger. For � = 0.09 (i.e. smallest θ-estimate in our study), about 153, 38 and 17 quadrats should be sampled with 10 isolates from each, to estimate a mean SNP incidence of 0.1 with CV values of 10, 20 and 30%, respectively; for � = 0.20 (median estimated value), the three numbers of quadrats increase to 223, 56 and 25, and for � = 0.23 (greatest estimated value), to 243, 61 and 27 (Table 3). By comparison, for a mean SNP incidence of 0.4 (i.e. fourfold greater than 0.1), the required number of sampling quadrats is much smaller:

about 26, 7 and 3 with � = 0.09 and 37, 8 and 5 with � = 0.20, for CV values of

10, 20 and 30%, respectively (Table 7).

5.5 Discussion

Management strategies exist, either to delay disease onset or to level off disease progress, but plant disease control still depends largely on fungicide applications. Among the limitations regarding fungicide efficacy, the development of resistance by target organisms is becoming a major concern for growers, crop specialists, and manufacturers. Despite the progress made towards understanding fungicide resistance at the individual level, establishing the relationship between theoretical and practical resistance is tedious and our knowledge remains limited. Recent research on resistance mechanisms, discovery of single nucleotide polymorphisms (SNPs) related to fungicide resistance and development of molecular detection tools enables the breakdown of global phenotypic resistance into genotypic resistance. When moving from classical phenotypic to genotypic resistance, several hypotheses regarding practical aspects of resistance can be tested readily. For numerous aspects of fungicide resistance epidemiology, acquiring knowledge about spatial distribution patterns is essential.

Consequently, the data from two vineyards and two onion fields respectively sampled in 2011 and 2009-2010 in our study were used to characterize the spatial distribution patterns of SNPs related to fungicide resistance in B. cinerea and B. squamosa populations.

93 In both systems, sampling was carried toward the end of the growing season, when disease was considered to be homogenous in each sampled site

(Carisse et al., 2007). However, within an onion field or a vineyard, the results of our geostatistical analyses suggest that quadrat counts on SNPs related to fungicide resistance are spatially autocorrelated, the estimated range of spatial autocorrelation varying between 21.0m and 31.7m in omni-directional analyses.

The spatial autocorrelation measured is largely dependent on quadrat size and the organisms concerned in plant diseases. For example, Larkin (Larkin et al., 1995) found an average range of spatial autocorrelation of about 15m for Phytophthora capsici in bell pepper in soil and water contained in soil. For almond leaf scorch disease, the estimated ranges of spatial autocorrelation averaged at 28m (Groves et al., 2005), and for Venturia inequalis, the range of autocorrelation for potential ascospore dose (PAD) was 40m on average (Carisse et al., 2007). In addition to relatively short ranges of spatial autocorrelation, our results indicated the presence of some anisotropy, which occurred in particular in the onion field sampled in

2010 and was then characterized by spatial autocorrelation restricted to the within-row direction and the direction perpendicular to it and by a much longer range of spatial autocorrelation in the former direction. These results support those of Elmer et al. (1998), who also found spatial autocorrelation in the within- row direction in dicarboximide-resistant isolates from Molinia fructicola populations in stone fruits. In the two vineyards sampled in our study, the vines were planted 0.9m apart on rows separated by 3m, thus forming a compact hedge.

The presence of such hedges could explain the presence of a directional bias, which was moderate here. In the onion field with an anisotropic spatial

94 autocorrelation in quadrat counts on SNPs related to fungicide resistance in 2010, the canopy height was lower and the crop density was much higher. It is unclear why the other onion field sampled the year before did not show a similar anisotropy, and actually did not show any spatial autocorrelation at all in quadrat counts on SNPs related to fungicide resistance. It is true, however, that the high values of the nugget effect relative to the sills of spherical variogram models fitted for the 2010 onion data set indicate that only a small part of the total variance could be explained by a spatially autocorrelated component.

Both the B. cinerea isolates carrying the H272R, H272Y and I86S mutations in the vineyards sampled and the B. squamosa isolates carrying the

I86S homologous mutation in the onion fields sampled were heterogeneously distributed among sampling quadrats. The median estimated value of the parameter θ of the beta-binomial distributions fitted to the eight count data sets

(Table 2) was 0.20, indicating a relatively high level of small-scale spatial (i.e. local) aggregation. In fact, the beta-binomial distribution provides an appropriate description of the frequency distribution of the number of mutations related to fungicide resistance per 10 isolates over sampling quadrats, the resistance status of a Botrytis isolate in a sampling quadrat being dependent on the resistance status of another Botrytis isolate in the same sampling quadrat. Using Lloyd’s patchiness index, Elmer et al. (1998) found that dicarboximide-resistant strains of M. fructicola in stone fruits were aggregated in all blocks and all years. Because few studies have been conducted on the spatial distribution of resistant phenotypes or genotypes, the quantitative results obtained in our study are difficult to compare.

However, the spatial aggregation characteristics of white mold epidemics caused by Sclerotinia in bean (�median = 0.10) and pyrethrum (�median = 0.16), of leaf spot in strawberry (�median = 0.20) and of grape downy mildew (0.02-0.21) were lower or similar (Jones et al., 2011; Madden & Hughes, 1995; Pethybridge et al., 2010;

Turechek & Madden, 1999). For B. squamosa, the heterogeneity in the spatial distribution was well described with Iwao’s patchiness regression technique

(Boivin & Sauriol, 1984). In that study, the authors concluded that the basic components of the Botrytis leaf blight lesion population were aggregates, contagiously distributed in the onion fields (Boivin & Sauriol, 1984). The results reported here are suggestive of a small-scale dispersal pattern of genetic variants

(i.e. the individuals carrying, or not, SNPs related to resistance), by opposition to the dispersal of a well-disbursed pathogen characterized by a completely random spatial distribution pattern (Turechek & Mahaffee, 2004). Small-scale distribution is also an indicator of autoinfection mechanisms, which are known to result in local clustering (Mundt, 2009). Accordingly, the resistant portion of a Botrytis population can be seen as sub-populations with dispersal characteristics different from those of the disease itself.

Using the cluster sampling methodology described by Hughes et al.

(1996), we calculated the minimum sample sizes required to estimate mean SNP incidence with given levels of statistical precision and spatial aggregation. It was found (Fig. 3) that a larger sample size is required when the SNP follows an aggregated spatial distribution, compared to a completely random one. In other

96 studies aimed to survey fungicide resistance using classical phenotypic assays, the sample size was often comprised between 10 and 100 samples per sampling site

(Köller et al., 1997; Köller et al., 1999; Leroch et al., 2011; Pappas, 1997).

Sampling for fungicide resistance may not or should not require the same numbers of samples needed if resistance is tested by classical phenotypic assays or by genotypic assays, because the classical method encompasses all the genotypic variations. In fact, sampling for SNPs related to fungicide resistance requires a high level of precision. It might also be more extensive. For example, sampling for SNPs could be used for different purposes, such as day-to-day monitoring and surveillance of a population shift. Thus, sample size and level of reliability could vary depending on the objectives of the survey.

Because of the clonal reproduction and haploid dispersion of Botrytis species, the expected range of spatial autocorrelation might be longer than the ranges estimated in our study. These were obtained from count data, after removal of a drift when present, for a given size of sampling quadrat. Nevertheless, our results strongly suggest aggregation in the spatial distribution patterns of SNPs related to fungicide resistance, which has important implications for questions of sample size. The large number of samples required when high precision matters represents a possible limitation toward sampling capacities. New detection methods belonging to different approaches allowing the analysis of bulked samples, such as pyrosequencing-based techniques or spore sampling devices for the quantification of mixed populations should then be considered. Finally,

97 thresholds for the proportion of resistant isolates above which fungicide control is inadequate still need to be determined.

5.6 ACKNOWLEDGEMENTS

The authors are grateful to Mathieu Tremblay for his constant support, and

Annie Lefebvre, Audrey Levasseur and Melanie Gobeil-Richard for their essential technical assistance. This work was financially supported in part by Agriculture and Agri-Food Canada through a DIAP project. The contributions of H. Van der

Heyden and L. Brodeur to this work were partially supported by the Compagnie de Recherche Phytodata Inc.

98 Table 5: Results of CRAD analysis (coregionalization analysis with a drift) and variogram model parameter estimates (range, nugget effect, sill of spherical structure) for SNP distributions within B. cinerea and B. squamosa populations in two vineyards and two onion fields sampled in 2009-2011.

Sampling Range of spatial Year SNP Drift a Direction b Partial sill c Nugget effect site autocorrelation (m)

OR 2011 H272R 0 omni . 4.594 . 0 . 4.734 . 45 0.702 3.940 23.9 90 . 4.200 . 135 . 4.478 . OR 2011 H272Y 0 omni 0.978 2.302 21.3 0 1.308 1.989 21.1 45 0.579 2.743 21.0 90 1.292 2.030 21.5 135 1.161 2.045 20.0 OR 2011 I86S 0.253 omni 1.001 3.032 21.3 0 1.837 2.258 21.0 45 . 4.000 . 90 0.962 3.065 21.5 135 1.417 2.667 21.1 DE 2009 I86S 0 omni . 4.034 . 0 . 4.073 . 45 . 3.914 . 90 . 4.108 . 135 . 3.864 .

99 Table 5 (continued): Results of CRAD analysis (coregionalization analysis with a drift) and variogram model parameter estimates (range, nugget effect, sill of spherical structure) for SNP distributions within B. cinerea and B. squamosa populations in two vineyards and two onion fields sampled in 2009-2011.

Sampling Range of spatial Year SNP Drift a Direction b Partial sill c Nugget effect site autocorrelation (m)

TE 2011 H272R 0.301 omni 1.215 4.227 21 0 . 5.441 . 45 1.093 4.273 20.8 90 3.146 2.469 21.3 135 0.96 4.379 20.2 TE 2011 H272Y 0.23 omni 2.849 1.983 21.3 0 2.517 2.482 33.3 45 3.494 1.368 20.7 90 . 4.854 . 135 2.15 2.444 20.4 TE 2011 I86S 0 omni 0.024 6.392 30.4 0 . 6.42 . 45 0.355 6.152 31.3 90 0.47 5.928 21.5 135 0.426 6.053 21.1 VH 2010 I86S 0.104 omni 0.414 4.415 31.7 0 0.445 4.699 56.6 45 . 4.531 . 90 0.032 4.6 30.2 135 . 4.739 .

100 a Drift estimated by EGLS (estimated generalized least squares) in first phase of CRAD (Pelletier et al., 2009) b Direction used as reference for the computation of directional experimental variograms, expressed in degrees where

0° refers to the within-row direction; “omni” means that the experimental variograms were computed in all directions, i.e., they are omni-directional. c The sill (i.e. variance component) of the spherical structure used in variogram modeling

101 Table 6: Observed values of the index of dispersion and parameter estimates for the beta-binomial and binomial

distributions

Parameter C(α) test b estimates c Index of � � Sampling d 2 e 2 e Year SNP Mean dispersion z P(z) (standard (standard LRT χ BBD P(χ BBD) Site Year site incidence (D) a error) error) 0.6076 0.1107 OR 2011 H272R 0.607 1.87* 6.37 0.0001 (0.02103) (0.03048) 28.72 5.30 0.380 OR 2011 0.1969 0.1185 OR 2011 H272Y 0.197 1.96* 7.01 0.0001 (0.01713) (0.03390) 31.06 2.08 0.555 OR 2011 0.5936 0.0851 OR 2011 I86S 0.593 1.71* 5.16 0.0001 (0.02020) (0.02709) 19.69 1.05 0.975 OR 2011 0.6843 0.2056 TE 2011 H272R 0.684 2.49* 10.94 0.0001 (0.02306) (0.04645) 68.89 8.55 0.200 TE 2011 0.2135 0.2067 TE 2011 H272Y 0.211 2.64* 12.06 0.0001 (0.01976) (0.04766) 69.95 2.06 0.725 TE 2011 0.5494 0.1934 TE 2011 I86S 0.550 2.42* 10.40 0.0001 (0.02434) (0.04145) 65.00 3.39 0.690 TE 2011 0.4238 0.1979 VH 2010 I86S 0.422 2.51* 11.10 0.0001 (0.02451) (0.04369) 70.05 4.54 0.604 VH 2010 0.1592 0.2335 DE 2009 I86S 0.155 2.97* 14.44 0.0001 (0.01760) (0.05449) 79.85 4.41 0.219 DE 2009

a *: Significant at 0.05 level

b Observed values of the z-statistic and corresponding probabilities of significance

102 c Together with n, the number of sampling quadrats here, θ, the local aggregation index, and p, the mean SNP incidence, are the three parameters of the beta-binomial distribution; the binomial distribution has only two parameters, n and p. Standard errors are reported in parentheses. d Values of the likelihood ratio test statistic, obtained by using the binomial distribution as the “null hypothesis” (θ =

0.0) and the beta-binomial distribution as the “alternative hypothesis” (θ > 0.0) e Observed values of the goodness-of-fit χ2-test statistic for the beta-binomial distribution and corresponding probabilities of significance f Observed values of the goodness-of-fit χ2-test statistic for the binomial distribution and corresponding probabilities of significance

103 Table 7: Minimum sample size (N) required to estimate the mean SNP incidence (p) for low, intermediate and high

values of the local aggregation index (θ) and increasing values of the CV (excerpts from Figure 3; see text for details)

Required N (θ = 0.23) Required N (θ = 0.20) Required N (θ = 0.09) p CV=10% CV=20% CV=30% CV=10% CV=20% CV=30% CV=10% CV=20% CV=30%

0.01 2 673.099 668.275 297.011 2 448.298 612.075 272.033 1 688.775 422.194 187.642

0.025 1 053.039 263.26 117.004 964.481 241.12 107.165 665.275 166.319 73.92

0.05 513.019 128.255 57.002 469.875 117.469 52.208 324.108 81.027 36.012

0.1 243.009 60.752 27.001 222.573 55.643 24.73 153.525 38.381 17.058

0.25 81.003 20.251 9 74.191 18.548 8.243 51.175 12.794 5.686

0.5 27.001 6.75 3 24.73 6.183 2.748 17.058 4.265 1.895

0.75 9 2.25 1 8.243 2.061 . 5.686 1.422 .

0.85 4.765 1.191 . 4.364 1.091 . 3.01 . .

0.9 3 . . 2.748 . . 1.895 . .

0.95 1.421 . . 1.302 . . . . .

104 A E

B F

C G

D H

Figure 10: Raw count data maps for B. cinerea in vineyard OR for A) H272R

(observed mean incidence, 0.68, ± 0.023, standard error), B) H272Y (0.21 ±

0.021), and C) I86S (0.55 ± 0.024), and in vineyard TE for E) H272R (0.61 ±

0.021) F) H272Y (0.20 ± 0.018), and G) I86S (0.59 ± 0.020), and for B. squamosa for I86S homologous in D) 2009 (0.16 ± 0.020) and H) 2010 (0.42 ± 0.025).

105 30 A E 30 Obs 25 BBD 25 Bin 20 Mean 20

15 15 Frequency 10 10 Frequency

5 5

0 0 B F 30 30

20 20 Frequency Frequency 10 10

0 0 30 C G 30 25 25

20 20

15 15 Frequency Frequency 10 10

5 5

0 0 30 40 D H 25 30 20

15 20

Frequency 10 Frequency 10 5

0 0 0 2 4 6 8 10 0 2 4 6 8 10 Number of resistant isolates Number of resistant isolates

Figure 11: Frequency distribution of the observed number of Botrytis isolates carrying a given SNP per quadrat (empty bars), together with the expected numbers of Botrytis isolates carrying the SNP per quadrat under the beta-binomial distribution model (filled bars) and the binomial distribution model (dashed bars), for B. cinerea in vineyard OR for A) H272R, B) H272Y and C) I86S, in vineyard

TE for E) H272R, F) H272Y and G) I86S, and for B. squamosa for I86S in D)

2009 and H) 2010.

106 Figure 12: Sampling curves calculated under the assumption of a beta-binomial distribution, using low, intermediate and higher values of the parameter θ that correspond to the maximum (A), median (B) and minimum (C) values of θ - estimates obtained in our study, for three levels of precision given by CV values of 10%, 20% and 30%. The continuous line represents the minimum sample size required to estimate mean SNP incidence with a CV of 10%; the dotted line, with a CV of 20%; and the dashed line, with a CV of 30%.

107 6 General conclusion and consideration for future research

6.1 General conclusion

Recent developments in molecular detection of fungicide resistance have enhanced the surveillance and monitoring capacities of researchers and crop advisors. In our study, resistance was broken down into SNPs, and the resistance levels provided by classical phenotypic assays were generally similar to those obtained with PCR-based assays. The benefit of using molecular tools to test for fungicide resistance is not only to provide the specificity of a polymorphism towards a particular resistance phenotype, but the use of molecular tools could also lead to a more precise characterization in terms of fitness and of practical resistance (resistance factor).

The analysis of the nearest-neighbour distance in the univariate case has been seldom used in plant pathology (Dallot et al., 2003; Gottwald et al., 2002), while it is a common procedure in ecology (Dutilleul, 2011; Perry et al., 2006;

Souris & Bichaud, 2011). In Chapter 3, we performed bivariate spatial point pattern analyses based on inter-SNP distances to characterize spatial relationships between SNPs related to fungicide resistance in the field. The results obtained in this study suggest that a genetic niche occupied by a specific polymorphism might either prevent invasion by other SNPs or co-exist with them in the field. Such results are supportive of a model of co-existence between sensitive and resistant strains (Parnell et al., 2005; Parnell et al., 2006). However, moving from

108

conventional resistance monitoring with a binary outcome (resistant or sensitive) to molecular monitoring implies a more complex level of complexity.

The results reported in Chapter 5 are suggestive of a small-scale dispersal pattern of genetic variants (i.e. the individuals carrying, or not, SNPs related to resistance). Small-scale distribution can also indicate patterns of auto-infection mechanisms, which are known to result in local clustering (Mundt, 2009).

Accordingly, the resistant portion of a Botrytis population can be seen as subpopulations with dispersal characteristics different from those of the disease.

Using the cluster sampling methodology described by Hughes et al.

(1996), we calculated the minimum sample sizes required to estimate mean SNP incidence with given levels of statistical precision and spatial aggregation. It was found that a larger sample size is required when the SNP follows an aggregated spatial distribution, compared to a completely random one. Sampling for SNPs could be used for different purposes, such as day-to-day monitoring and surveillance of a population shift. Thus, sample size and level of reliability could vary depending on the objectives of the survey.

The results obtained in this research strongly suggest aggregation in the spatial distribution patterns of SNPs related to fungicide resistance, which has important implications for questions of sample size. The processes underlying these interactions are not yet fully understood, but the inclusion of spatial

109 relationships, in day-to-day monitoring or in modeling, will facilitate our understanding of fungicide resistance at the population level. Eventually, the monitoring of polymorphisms related to fungicide resistance will lead to a better comprehension of practical resistance.

6.2 Consideration for future research

The results obtained in this project shed light on several areas related to the sampling of mutations related to fungicide resistance, which could require further research. Examples of future research objectives are:

1. The spatial distribution pattern of mutations related to fungicide resistance could be investigated at a larger scale (e.g. at the scale of a growing area or a region).

2. To address the problem of the large number of samples needed to estimate mean incidence when the expected incidence is low, new techniques (e.g. pyrosequencing) aiming at the detection of SNPs of fungicide resistance from bulk samples could be developed. Alternatively, the estimation of mean mutation incidence from air sampling could be considered.

3. Research aiming to relate the proportion of resistant isolates and the level of disease control through fungicide efficacy is presently missing in the scientific community.

110

7 Literature cited

Banno, S., Fukumori, F., Ichiishi, A., Okada, K., Uekusa, H., Kimura, M., &

Fujimura, M. (2008). Genotyping of Benzimidazole-Resistant and

Dicarboximide-Resistant Mutations in Botrytis cinerea Using Real-Time

Polymerase Chain Reaction Assays. Phytopathology, 98(4), 397-404.

Beever, R. E., Laracy, E. P., & Pak, H. A. (1989). Strains of Botrytis cinerea

resistant to dicarboximide and benzimidazole fungicides in New Zealand

vineyards. Plant Pathology, 38(3), 427-437.

Billard, A., Fillinger, S., Leroux, P., Lachaise, H., Beffa, R., & Debieu, D. (2012).

Strong resistance to the fungicide fenhexamid entails a fitness cost in

Botrytis cinerea, as shown by comparisons of isogenic strains. Pest

Management Science, 68(5), 684-691.

Billard, A., Laval, V., Fillinger, S., Leroux, P., Lachaise, H., Beffa, R., & Debieu,

D. (2012). The Allele-Specific Probe and Primer Amplification Assay, a

New Real-Time PCR Method for Fine Quantification of Single-Nucleotide

Polymorphisms in Pooled DNA. Applied and Environmental

Microbiology, 78(4), 1063-1068.

Boivin, G., & Sauriol, P. (1984). Dispersion statistics and sequential sampling

plan for leaf blight caused by Botrytis squamosa in onions.

Phytopathology, 74(11), 1385-1387.

Brent, K. J., & Hollomon, D. (2007). Fungicide resistance in crop pathogens:

how can it be managed? (Vol. Frac Monograph No.1 ): Fungicide

Resistance Action Commitee.

111 Brent, K. J., & Hollomon, D. W. (2007). Fungicide Resistance: The Assessment of

Risk (Second edition ed.): Fungicide Resistance Action Committee.

Carisse, O., Savary, S., & Willocquet, L. (2007). Spatiotemporal Relationships

Between Disease Development and Airborne Inoculum in Unmanaged and

Managed Botrytis Leaf Blight Epidemics. Phytopathology, 98(1), 38-44.

Carisse, O., Tremblay, D.-M., McDonald, M. R., Brodeur, L., & McRoberts, N.

(2011). Management of Botrytis Leaf Blight of Onion: The Québec

Experience of 20 Years of Continual Improvement. Plant Disease, 95(5),

504-514.

Carisse, O., & Tremblay, D. M. (2007). Incidence and significance of iprodione-

insensitive isolates of Botrytis squamosa. Plant Disease, 91(1), 41-46.

Chin, K. M. (1987). A simple model of selection for fungicide resistance in plant

pathogen populations. Phytopathology, 77(5), 666-669.

Clark, P. J., & Evans, F. C. (1954). Distance to Nearest Neighbor as a Measure of

Spatial Relationships in Populations. Ecology, 35(4), 445-453.

Cui, W., Beever, R. E., Parkes, S. L., & Templeton, M. D. (2004). Evolution of an

Osmosensing Histidine Kinase in Field Strains of Botryotinia fuckeliana

(Botrytis cinerea) in Response to Dicarboximide Fungicide Usage.

Phytopathology, 94(10), 1129-1135.

Cui, W., Beever, R. E., Parkes, S. L., Weeds, P. L., & Templeton, M. D. (2002).

An osmosensing histidine kinase mediates dicarboximide fungicide

resistance in Botryotinia fuckeliana (Botrytis cinerea). Fungal Genetics

and Biology, 36(3), 187-198.

112

Dallot, S., Gottwald, T., Labonne, G., & Quiot, J. B. (2003). Spatial Pattern

Analysis of Sharka Disease (Plum pox virus Strain M) in Peach Orchards

of Southern France. Phytopathology, 93(12), 1543-1552.

Dean, R., Van Kan, J. A. L., Pretorius, Z. A., Hammond-Kosack, K. E., Di Pietro,

A., Spanu, P. D., Foster, G. D. (2012). The Top 10 fungal pathogens in

molecular plant pathology. Molecular Plant Pathology, 13(4), 414-430.

Diggle, P. J. (2003). Statistical Analysis of Spatial Point Patterns. London:

Hodder Education.

Dutilleul, P., Haltigin, T. W., & Pollard, W. H. (2009). Analysis of polygonal

terrain landforms on Earth and Mars through spatial point patterns.

Environmetrics, 20(2), 206-220.

Dutilleul, P. R. L. (2011). Spatio-temporal heterogeneity: concept and analyses.

New York: Cambridge University Press.

Elmer, P. A. G., Gaunt, R. E., & Frampton, C. M. (1998). Spatial and temporal

characteristics of dicarboximide-resistant strains of Monilinia fructicola

and brown rot incidence in stone fruit. Plant Pathology, 47(4), 530-536.

Esterio, M., Muñoz, G., Ramos, C., Cofré, G., Estévez, R., Salinas, A., & Auger,

J. (2011). Characterization of Botrytis cinerea isolates present in

Thompson seedless table grapes in the Central Valley of Chile. Plant

Disease, 95(6), 683-690.

Fernández-Ortuño, D., Chen, F., & Schnabel, G. (2012). Resistance to

Pyraclostrobin and Boscalid in Botrytis cinerea Isolates from Strawberry

Fields in the Carolinas. Plant Disease, 96(8), 1198-1203.

113 Ferrandino, F. J. (2004). Measuring Spatial Aggregation in Binary Epidemics:

Correlative Analysis and the Advantage of Fractal-Based Sampling.

Phytopathology, 94(11), 1215-1227. doi: 10.1094/phyto.2004.94.11.1215

Fillinger, S., Leroux, P., Auclair, C., Barreau, C., Al Hajj, C., & Debieu, D.

(2008). Genetic Analysis of Fenhexamid-Resistant Field Isolates of the

Phytopathogenic Fungus Botrytis cinerea. Antimicrob. Agents Chemother.,

52(11), 3933-3940.

Fournier, E., Levis, C., Fortini, D., Leroux, P., Giraud, T., & Brygoo, Y. (2003).

Characterization of Bc-hch, the Botrytis cinerea homolog of the

Neurospora crassahet-c vegetative incompatibility locus, and its use as a

population marker. Mycologia, 95(2), 251-261.

Gatrell, A. C., Bailey, T. C., Diggle, P. J., & Rowlingson, B. S. (1996). Spatial

Point Pattern Analysis and Its Application in Geographical Epidemiology.

Transactions of the Institute of British Geographers, 21(1), 256-274.

Gottwald, T. R., Sun, X., Riley, T., Graham, J. H., Ferrandino, F., & Taylor, E. L.

(2002). Geo-Referenced Spatiotemporal Analysis of the Urban Citrus

Canker Epidemic in Florida. Phytopathology, 92(4), 361-377.

Grabke, A., Fernandez-Ortuno, D., & Schnabel, G. (2012). Fenhexamid resistance

in Botrytis cinerea from strawberry fields in the Carolinas is associated

with four target gene mutations. Plant Disease, 97(2), 271-276.

Groves, R. L., Chen, J., Civerolo, E. L., Freeman, M. W., & Viveros, M. A.

(2005). Spatial Analysis of Almond Leaf Scorch Disease in the San

114

Joaquin Valley of California: Factors Affecting Pathogen Distribution and

Spread. Plant Disease, 89(6), 581-589.

Gubbins, S., & Gilligan, C. A. (1999). Invasion thresholds for fungicide

resistance: deterministic and stochastic analyses. Proceedings of the Royal

Society of London. Series B: Biological Sciences, 266(1437), 2539-2549.

Haltigin, T., Pollard, W., & Dutilleul, P. (2010). Comparison of ground- and

aerial-based approaches for quantifying polygonal terrain network

geometry on Earth and Mars via spatial point pattern analysis. Planetary

and Space Science, 58(12), 1636-1649.

Hughes, G., & Madden, L. V. (1993). Using the beta-binomial distribution to

describe aggregated patterns of disease incidence. Phytopathology, 83(7),

759-763.

Hughes, G., & Madden, L. V. (1995). Some methods allowing for aggregated

patterns of disease incidence in the analysis of data from designed

experiments. Plant Pathology, 44(6), 927-943.

Jarvis, W. R. (1962a). The dispersal of spores of botrytis cinerea fr. in a raspberry

plantation. Trans. Brit. Mycol. Society, 45(4), 549-559.

Jarvis, W. R. (1962b). Splash Dispersal of Spores of Botrytis cinerea Pers.

Nature, 193(4815), 599-599.

Jarvis, W. R. (1977). Botryotinia and Botryis species: taxonomy physiology and

pathogenicity. Canada: Canada departement of agriculture.

Jones, S. J., Gent, D. H., Pethybridge, S. J., & Hay, F. S. (2011). Spatial

characteristics of white mould epidemics and the development of

115 sequential sampling plans in Australian bean fields. Plant Pathology,

60(6), 1169-1182.

Karrandinos, M. G. (1976). Optimum Sample Size and Comments on Some

Published Formulae. Bulletin of the ESA, 22(4), 417-421.

Keller, M., Viret, O., & Cole, F. M. (2003). Botrytis cinerea infection in grape

flowers: Defense reaction, latency, and disease expression.

Phytopathology, 93(3), 316-322.

Köller, W., Wilcox, W. F., Barnard, J., Jones, A. L., & Braun, P. G. (1997).

Detection and quantification of resistance of Venturia inaequalis

populations to sterol demethylation inhibitors. Phytopathology, 87(2),

184-190.

Köller, W., Wilcox, W. F., & Jones, A. L. (1999). Quantification, persistence, and

status of dodine resistance in New York and Michigan orchard populations

of Venturia inaequalis. Plant Disease, 83(1), 66-70.

Krige, D. G. (1953). A Statistical Approach to Some Basic Mine Valuation

Problems on the Witwatersrand. OR, 4(1).

Larkin, R. P., Gumpertz, M. L., & Ristaino, J. B. (1995). Geostatistical analysis of

Phytophthora epidemic development in commercial bell pepper fields.

Phytopathology, 85(2), 191-203.

Latorre, B. A., & Torres, R. (2012). Prevalence of isolates of Botrytis cinerea

resistant to multiple fungicides in Chilean vineyards. Crop Protection, 40,

49-52.

116

Leroch, M., Kretschmer, M., & Hahn, M. (2011). Fungicide resistance phenotypes

of botrytis cinerea isolates from commercial vineyards in South West

Germany. Journal of Phytopathology, 159(1), 63-65.

Leroux, P., Chapeland, F., Desbrosses, D., & Gredt, M. (1999). Patterns of cross-

resistance to fungicides in Botryotinia fuckeliana (Botrytis cinerea)

isolates from French vineyards. Crop Protection, 18(10), 687-697.

Leroux, P., & Clerjeau, M. (1985). Resistance of Botrytis cinerea Pers. and

Plasmopara viticola (Berk. & Curt.) Berl. and de Toni to fungicides in

French vineyards. Crop Protection, 4(2), 137-160.

Leroux, P., Fritz, R., Debieu, D., Albertini, C., Lanen, C., Bach, J., . . . Chapeland,

F. (2002). Mechanisms of resistance to fungicides in field strains of

Botrytis cinerea. Pest Management Science, 58(9), 876-888.

Leroux, P., Gredt, M., Leroch, M., & Walker, A.-S. (2010). Exploring

Mechanisms of Resistance to Respiratory Inhibitors in Field Strains of

Botrytis cinerea, the Causal Agent of Gray Mold. Applied and

Environmental Microbiology, 76(19), 6615-6630.

Ma, Z., & Michailides, T. J. (2005). Advances in understanding molecular

mechanisms of fungicide resistance and molecular detection of resistant

genotypes in phytopathogenic fungi. Crop Protection, 24(10), 853-863.

Madden, L. V., & Hughes, G. (1994). BBD - Computer software for fitting the

beta-binomial distribution to disease incidence data. Plant Disease, 78(5),

536-540.

117

Madden, L. V., & Hughes, G. (1995) Plant disease incidence: Distributions,

heterogeneity, and temporal analysis. Annual Review of Phytopathology

33: 529-564.

Madden, L. V., & Hughes, G. (1999). Sampling for Plant Disease Incidence.

Phytopathology, 89(11), 1088-1103.

Madden, L. V., Hughes, G., & Ellis, M. A. (1995). Spatial heterogeneity of the

incidence of grape downy mildew. Phytopathology, 85(3), 269-275.

Madden, L. V., Hughes, G., & Munkvold, G. P. (1996). Plant disease incidence:

Inverse sampling, sequential sampling, and confidence intervals when

observed mean incidence is zero. Crop Protection, 15(7), 621-632.

Madden, L. V., Hughes, G., and van den Bosch, F. . (2007). The study of plant

disease epidemics.

Madden, L. V., Nault, L. R., Murral, D. J., & Apelt, M. R. (1995). Spatial pattern

analysis of the incidence of aster yellows disease in lettuce. Researches on

Population Ecology, 37(2), 279-289.

Matheron, G. (1962). Traité de géostatistique appliquée (Vol. 1).

Mavroeidi, V. I., & Shaw, M. W. (2005). Sensitivity distributions and cross-

resistance patterns of Mycosphaerella graminicola to fluquinconazole,

prochloraz and azoxystrobin over a period of 9 years. Crop Protection,

24(3), 259-266.

118

Mavroeidi, V. I., & Shaw, M. W. (2006). Effects of fungicide dose and mixtures

on selection for triazole resistance in Mycosphaerella graminicola under

field conditions. Plant Pathology, 55(6), 715-725.

Milgroom, M. G. (1990). A stochastic model for the initial occurrence and

development of fungicide resistance in plant pathogen populations.

Phytopathology, 80(4), 410-416.

Moyano, C., Alfonso, C., Gallego, J., Raposo, R., & Melgarejo, P. (2003).

Comparison of RAPD and AFLP Marker Analysis as a Means to Study the

Genetic Structure of <i>Botrytis cinerea</i> Populations.

European Journal of Plant Pathology, 109(5), 515-522.

Moyano, C., Gómez, V., & Melgarejo, P. (2004). Resistance to pyrimethanil and

other fungicides in Botrytis cinerea populations collected on vegetable

crops in Spain. Journal of Phytopathology, 152(8-9), 484-490.

Mundt, C. C. (2009). Importance of Autoinfection to the Epidemiology of

Polycyclic Foliar Disease. Phytopathology, 99(10), 1116-1120.

Myresiotis, C. K., Bardas, G. A., & Karaoglanidis, G. S. (2008). Baseline

Sensitivity of Botrytis cinerea to Pyraclostrobin and Boscalid and Control

of Anilinopyrimidine- and Benzimidazole-Resistant Strains by These

Fungicides. Plant Disease, 92(10), 1427-1431.

Myresiotis, C. K., Karaoglanidis, G. S., & Tzavella-Klonari, K. (2007).

Resistance of Botrytis cinerea Isolates from Vegetable Crops to

Anilinopyrimidine, Phenylpyrrole, Hydroxyanilide, Benzimidazole, and

Dicarboximide Fungicides. Plant Disease, 91(4), 407-413.

119

N Ziogas, B., Nikou, D., N Markoglou, A., A Malandrakis, A., & Vontas, J.

(2009). Identification of a novel point mutation in the beta-tubulin gene of

Botrytis cinerea and detection of benzimidazole resistance by a diagnostic

PCR-RFLP assay. European Journal of Plant Pathology, 125(1), 97-107.

Ochiai, N., Fujimura, M., Motoyama, T., Ichiishi, A., Usami, R., Horikoshi, K., &

Yamaguchi, I. (2001). Characterization of mutations in the two-component

histidine kinase gene that confer fludioxonil resistance and osmotic

sensitivity in the os-1 mutants of Neurospora crassa. Pest Management

Science, 57(5), 437-442.

Oshima, M., Banno, S., Okada, K., Takeuchi, T., Kimura, M., Ichiishi, A., . . .

Fujimura, M. (2006). Survey of mutations of a histidine kinase gene

BcOS1; in dicarboximide-resistant field isolates of Botrytis cinerea.

Journal of General Plant Pathology, 72(1), 65-73.

Oshima, M., Fujimura, M., Banno, S., Hashimoto, C., Motoyama, T., Ichiishi, A.,

& Yamaguchi, I. (2002). A Point Mutation in the Two-Component

Histidine Kinase BcOS-1 Gene Confers Dicarboximide Resistance in

Field Isolates of Botrytis cinerea. Phytopathology, 92(1), 75-80.

Pappas, A. C. (1997). Evolution of fungicide resistance in Botrytis cinerea in

protected crops in Greece. Crop Protection, 16(3), 257-263.

Parnell, S., Gilligan, C. A., & van den Bosch, F. (2005). Small-Scale Fungicide

Spray Heterogeneity and the Coexistence of Resistant and Sensitive

Pathogen Strains. Phytopathology, 95(6), 632-639.

120 Parnell, S., van den Bosch, F., & Gilligan, C. A. (2006). Large-Scale Fungicide

Spray Heterogeneity and the Regional Spread of Resistant Pathogen

Strains. Phytopathology, 96(5), 549-555.

Pelletier, B., Dutilleul, P., Larocque, G., & Fyles, J. (2004). Fitting the Linear

Model of Coregionalization by Generalized Least Squares. Mathematical

Geology, 36(3), 323-343.

Pelletier, B., Dutilleul, P., Larocque, G., & Fyles, J. (2009). Coregionalization

analysis with a drift for multi-scale assessment of spatial relationships

between ecological variables 1. Estimation of drift and random

components. Environmental and Ecological Statistics, 16(4), 439-466.

Perry, G. L. W., Miller, B. P., & Enright, N. J. (2006). A comparison of methods

for the statistical analysis of spatial point patterns in plant ecology. Plant

Ecology, 187(1), 59-82.

Pethybridge, S. J., Hay, F. S., & Gent, D. H. (2010). Characterization of the

Spatiotemporal Attributes of Sclerotinia Flower Blight Epidemics in a

Perennial Pyrethrum Pathosystem. Plant Disease, 94(11), 1305-1313.

Saito, Seiya, Suzuki, Shunji, Takayanagi, & Tsutomu. (2009). Nested PCR-RFLP

is a high-speed method to detect fungicide-resistant Botrytis cinerea at an

early growth stage of grapes (Vol. 65). Chichester, ROYAUME-UNI:

Wiley.

Samuel, S., Papayiannis, L. C., Leroch, M., Veloukas, T., Hahn, M., &

Karaoglanidis, G. S. (2011). Evaluation of the incidence of the G143A

mutation and cytb intron presence in the cytochrome bc-1 gene conferring

121 QoI resistance in Botrytis cinerea populations from several hosts. Pest

Management Science, 67(8), 1029-1036.

Souris, M., & Bichaud, L. (2011). Statistical methods for bivariate spatial analysis

in marked points. Examples in spatial epidemiology. Spatial and Spatio-

temporal Epidemiology, 2(4), 227-234.

Staats, M., van Baarlen, P., & van Kan, J. A. L. (2005). Molecular Phylogeny of

the Plant Pathogenic Genus Botrytis and the Evolution of Host Specificity.

Molecular Biology and Evolution, 22(2), 333-346.

Statistique Canada, Division de l’Agriculture, Section des cultures. Productions

de fruits et légumes. June 2008. Statistique Canada, Ottawa, Canada.

Catalogue No 22-003-X, vol 77, no 1.

Thébaud, G., Peyrard, N., Dallot, S., Calonnec, A., & Labonne, G. (2005).

Investigating disease spread between two assessment dates with

permutation tests on a lattice. Phytopathology, 95(12), 1453-1461.

Tremblay, D. M., Talbot, B. G., & Carisse, O. (2003). Sensitivity of Botrytis

squamosa to different classes of fungicides. Plant Disease, 87(5), 573-578.

Turechek, W. W., & Madden, L. V. (1999). Spatial pattern analysis and sequential

sampling for the incidence of leaf spot on strawberry in Ohio. Plant

Disease, 83(11), 992-1000.

Turechek, W. W., & Madden, L. V. (2000). Analysis of the Association Between

the Incidence of Two Spatially Aggregated Foliar Diseases of Strawberry.

Phytopathology, 90(2), 157-170.

122

Turechek, W.W., & Mahaffee, W.F. (2004). Spatial pattern analysis of hop

powdery mildew in the Pacific Northwest: Implications for sampling.

Phytopathology, 94(10), 1116-1128.

Mahaffee, W.F., Turechek, W.W., Ocamb, C.M. (2003). Effect of variable

temperature on infection severity of Podosphaera macularis on hops

Phytopathology, 93, pp. 1587-1592;

Nyrop, J.P., Binns, M.R., Van Der Werf, W. (1999). Sampling for IPM decision

making: Where should we invest our time and resources? Phytopathology,

89, pp. 1104-1111.

Ocamb, C., Klein, R., Barbour, J., Griesbach, J., Mahaffee, W. (1999). First report

of hop powdery mildew in the Pacific Northwest Plant Dis., 83, p. 1072.

Pethybridge, S.J., Turechek, W.W. (2003). Analysis of the association among

three viruses infecting hop in Australia Plant Pathol., 52, pp. 158-167;

Rao, J.N.K., Scott, A.J., A simple method for the analysis of clustered binary data

(1992) Biometrics, 48, pp. 577-585.

Royle, D.J., (1978). Powdery mildew of the hop. The Powdery Mildews, pp. 381-

409. , D. M. Spencer, ed. Academic Press, London.

Scott, J.B., Hay, F.S., Wilson, C.R., Cotterill, P.J., Fist, A.J. (2003).

Spatiotemporal analysis of epiphytotics of downy mildew of oilseed poppy

in Tasmania, Australia Phytopathology, 93, pp. 752-757.

Tomlan, M.A., (1992) Tinged with Gold: Hop Culture in the United States, ,

University of Georgia Press, Athens.

123

Turechek, W.W., Ellis, M.A., Madden, L.V. (2000). Sequential sampling for

incidence of Phomopsis leaf blight of strawberry Phytopathology, 91, pp.

336-347.

Turechek, W.W., Madden, L.V. (1999). Spatial pattern analysis of strawberry

leaf blight in perennial production systems Phytopathology, 89, pp. 421-

433.

Turechek, W.W., Madden, L.V. (1999). Spatial pattern analysis and sequential

sampling for the incidence of leaf spot on strawberry in Ohio Plant Dis.,

83, pp. 992-1000.

Turechek, W.W., Madden, L.V. (2003). A generalized linear modeling approach

for characterizing disease incidence in a spatial hierarchy Phytopathology,

93, pp. 458-466.

Turechek, W.W., Mahaffee, W.F., Ocamb, C.M., (2001) Development of

Management Strategies for Hop Powdery Mildew in the Pacific

Northwest, Online. Plant Health Progress

Xu, X.-M., Madden, L.V. (2002). Incidence and density relationships of powdery

mildew on apple Phytopathology, 92, pp. 1005-1014.

Young, L.J., Young, J.H., (1998) Statistical Ecology, , Kluwer Academic

Publishers, Boston

Upton, J. G. G., & Fingleton, B. (1985). Spatial Data Analysis by Example: Point

Pattern and Quantitative Data: John Wiley & Sons Inc. van den Bosch, F., & Gilligan, C. A. (2008). Models of Fungicide Resistance

Dynamics. Annual Review of Phytopathology, 46(1), 123-147.

124 Van der Heyden, H., Carisse, O., & Brodeur, L. (2012). Comparison of

monitoring based indicators for initiating fungicide spray programs to

control Botrytis leaf blight of onion. Crop Protection, 33(0), 21-28.

Veloukas, T., Leroch, M., Hahn, M., & Karaoglanidis, G. S. (2011). Detection

and Molecular Characterization of Boscalid-Resistant Botrytis cinerea

Isolates from Strawberry. Plant Disease, 95(10), 1302-1307.

Walker, A.-S., Micoud, A., Rémuson, F., Grosman, J., Gredt, M., & Leroux, P.

(2013). French vineyards provide information which opens ways for

effective resistance management of Botrytis cinerea (grey mould). Pest

Management Science, n/a-n/a.

Weber, R. W. S. (2011). Resistance of Botrytis cinerea to Multiple Fungicides in

Northern German Small-Fruit Production. Plant Disease, 95(10), 1263-

1269.

Williamson, B., Tudzynski, B., Tudzynski, P., & Van Kan, J. A. L. (2007).

Botrytis cinerea: the cause of grey mould disease. Molecular Plant

Pathology, 8(5), 561-580.

Xiao, C. L., Hao, J. J., & Subbarao, K. V. (1997). Spatial patterns of

microsclerotia of Verticillium dahliae in soil and verticillium wilt of

cauliflower. Phytopathology, 87(3), 325-331.

Yarden, O., & Katan, T. (1993). Mutations leading to substitutions at amino acids

198 and 200 of beta-tubulin that correlate with benomyl-resistance

phenotypes of field strains of Botrytis cinerea. Phytopathology, 83(12),

1478-1483.

125

Yin, Y. N., Kim, Y. K., & Xiao, C. L. (2011). Molecular Characterization of

Boscalid Resistance in Field Isolates of Botrytis cinerea from Apple.

Phytopathology, 101(8), 986-995.

Zhonghua, M. A., Michailides, & Themis, J. (2005). Genetic structure of Botrytis

cinerea populations from different host plants in California (Vol. 89). St.

Paul, MN, ETATS-UNIS: American Phytopathological Society.

126