Virulence Structure of the Wheat Powdery Mildew Population in Serbia

: Powdery mildew is a common, economically important disease in the wheat growing area of Serbia. A large-scale virulence survey of its causal agent Blumeria graminis f. sp. tritici population was performed in the period 1995–2013. A total of 1013 isolates were recovered from the collected chasmothecial samples. Among them, 862 unique pathotypes were identiﬁed using a differential set of 20 wheat lines with known powdery mildew ( Pm ) resistant genes. The pathogen was highly diverse. Number of virulence genes (virulence complexity) per isolate was large, supporting a constant need to extend the differential set of wheat with newly identiﬁed Pm genes. Virulence frequencies to Pm6 , Pm7 , and Pm5+8 were high throughout the 19-year period, in contrast with that to Pm5+6 , which was consistently at a low level. The most signiﬁcant change in the population was observed for virulence to the Pm2+4b+6 gene combination, with an increasing frequency of virulence to this gene combination over the years. High virulence complexity and genetic diversity of the population are the most inﬂuential factors for the damaging epidemics that this pathogen can cause.


Introduction
Wheat powdery mildew, caused by Blumeria graminis f. sp. tritici, is a common disease in wheat growing areas worldwide. It can cause significant yield and quality losses. The disease occurs every year at a lower or higher intensity in Serbia, reducing yield on average by 30% on susceptible cultivars [1].
B. graminis f. sp. tritici is an obligate, biotrophic fungal pathogen, with a mixed mode of reproduction (sexual and asexual). In Serbia, sexual reproduction occurs in late spring (prior to wheat senescence) by producing chasmothecia. This fruiting body is a means of pathogen survival of unsuitable summer conditions, and also a source of sexual recombination. By maturing, the chasmothecia ascospores may be released and act as a primary source of inoculum for new infections in late spring or early summer. Initial infections are followed by multiple cycles of asexual reproduction. The number of these cycles depends on weather conditions.
Mc Donald and Linde [2] reported B. graminis f. sp. tritici as a high-risk pathogen due to its high adaptability and changeability potential. For these reasons, breeding for resistance to powdery mildew is a difficult and laborious task.
One of the main preconditions for successful wheat production is effective powdery mildew control. This can be achieved by fungicide treatment, and by sowing resistant varieties. Unfortunately, there are limited data about powdery mildew resistance genes (Pm) in the commercial wheat varieties grown in Serbia. Monitoring the virulence structure and diversity of wheat powdery mildew populations is important and necessary in order to identify efficient host resistant genes, as well as to track virulence changes in the pathogen population. These data enable breeders to prioritize Pm genes for utilization in wheat lines [3]. Annual virulence surveys of B. graminis f. sp. tritici populations have been conducted in Serbia since 1961 [4]. In the first period, physiological races were determined based on reactions of a few standard wheat varieties. However, since the mid 1980s, determining the virulence pathotypes on the basis of genetic background of differential wheat lines has been introduced as a more appropriate approach. The most frequent virulence genes in the late 1980s were genes virulent to Pm1, Pm2, Pm3a, Pm5, Pm6, and Pm8 [5]. The most efficient Pm genes in 1991-1992 were Pm4a, Pm4b, Mld, and the combinations Pm5+6 and Pm2+4b+6 [6].
Long-term studies of the powdery mildew population are aimed at revealing shifts in virulence frequencies that lead to changes in the virulence structure of the pathogen population.
The present study includes data from virulence surveys of B. graminis f. sp. tritici in Serbia during the period 1995-2013. It has been hypothesized that changing efficacy of Pm genes would result in shifts in the powdery mildew population. Therefore, the objective of this study was to determine and characterize the virulence structure of the B. graminis f. sp. tritici population in Serbia. The results would give insight into the dynamics of the powdery mildew population and contribute to a better understanding of its pathogenicity potential.

Collection and Preparation of Isolates
Senescent wheat leaf samples bearing chasmothecia of the powdery mildew pathogen (B. graminis f. sp. tritici) were collected from wheat fields at 83 locations in Serbia from 1995 to 2013. The number of randomly sampled leaves varied depending on the infection, but was at least 10 leaves per field.
Leaf samples with disease symptoms were placed separately in a clean paper bag and stored at 4 • C until they were used in the experiments. Powdery mildew susceptible wheat variety Barbee was used for the isolation, purification, multiplication, and maintenance of the isolates. The seeds of this variety were sown in 10 × 10 cm clay pots filled with commercial potting mix. Seedlings were grown in a growth chamber at 20 • C and 12 h of light. In order to obtain isolates, 200 chasmothecia from each leaf sample were collected with a sterile needle and placed on moist filter paper inside the lid of a Petri dish. When seedlings were at the second leaf stage, the Petri dish lid was placed on the top of the seedlings. The filter paper was moistened daily. The colonies of powdery mildew developed after 7 to 8 days. A single pustule from each isolate was transferred to the leaves of the susceptible variety Barbee using a dissecting needle. The inoculum multiplication was performed by shaking infected leaves to 7-day-old seedlings (1-2 leaf stage).

Virulence Analysis
Virulence of each single-pustule isolate was assessed by inoculation of a differential set of 20 wheat lines bearing different Pm genes or combinations of Pm genes ( Table 1). The differential set was sown in plastic trays filled with commercial potting mix in four replicates. Seedlings were grown in greenhouse conditions at 20 • C with 12 h of light, covered with plastic cylinders to avoid contamination. When seedlings developed the second leaf (7-8 days from seeding), they were inoculated by gently shaking conidia from 2-week-old isolates.
The reaction type was assessed on each differential line 7 days after inoculation using the modified scale of Mains [7], as follows: 0 = no visible symptoms; 1 = highly resistant, fungal development limited, necrosis, no sporulation; 2 = moderately resistant, moderate mycelium with some sporulation; 3 = moderately susceptible, extensive mycelium, more sporulation; 4 = highly susceptible, abundant sporulation, large colonies.

Data Analysis
Virulence genes of each powdery mildew isolate were determined based on the reaction of each differential line according to Flor's gene for gene hypothesis [8]. If reaction type on the differential line was scored 0-2 or 3-4, such isolate was classified as avirulent or Agronomy 2022, 12, 45 3 of 11 virulent, respectively, to the corresponding Pm gene or combination of Pm genes. Accordingly, a binary virulence profile was postulated for each powdery mildew isolate, where 1 and 0 encoded virulence and avirulence. Individual pathotypes of the powdery mildew isolates were assigned using a hexadecimal code as follows: the set of 20 differentials was divided into five groups of four consecutive differentials each. Then binary patterns of individuals were represented by the ordered sets of five {0, 1}-four-tuples (four-tuple means four ordered symbols) according to the division of differentials into groups. Accordingly, each four-tuple could be encoded by a single letter from a set of the first 20 consonants of the English alphabet [9].
Several descriptive parameters were calculated for each annual population of B. graminis f. sp. tritici: number of pathotypes, singletons, virulence frequencies, and (relative) virulence complexity. Virulence frequencies are presented by the proportion of the virulence gene in each year. Virulence complexity of each isolate equals the number of susceptible differentials to the given isolate and thus ranges from 1 to 20, while the relative virulence complexity values ranges from 0 to 1. The corresponding average value per isolate in a population was considered (relative) virulence complexity of the population.
Diversity within populations was measured using Simpson's index and Kosman KW metric with regard to the simple mismatch dissimilarity codes [10]. Pairwise distance between populations was estimated using Rogers and Nei's standard genetic distance [11]. Coefficient of association ϕ between pairs of differentiating characters was determined by formula 17.5 in [12] related to Fisher's exact test. In addition, the measure of correlation between pairs of differentials was also estimated (formula 1 [13]). All calculations were performed with the VIRULENCE ANALYSIS TOOL (VAT) software [14].
Relationships among powdery mildew populations was analyzed using the unweighted pair group method with arithmetic means (UPGMA) with the matrix of Nei's standard genetic distances. The UPGMA clustering was performed using NTSYS-pc software [16].

Virulence Frequencies and Associations between Virulence Genes
Virulence frequencies of B. graminis f. sp. tritici genes in the period 1995-2013 are given in Tables 4 and 5. The most frequent virulences during the whole period were those to Pm6 and Pm7. Virulences to these genes were constantly high, except in 2001 when virulence to all genes had low or intermediate frequencies.
The greatest shift of virulence was observed for the virulence to the gene combination Pm2+4b+6 (Figure 1).
Pairwise virulence associations were revealed within annual populations of wheat powdery mildew, though many of them were detected only once. Associations between virulences to Pm2 and Pm2+, Pm3a and Pm3b, and Pm3b and Pm4b were observed at least in five annual populations. Pairwise association between virulences to Pm3b and Pm4b was consistent in the last 5 years of the survey. Association of virulences to Mli and Pm2+4b+6 was found in 2004, 2005, and during the period 2010-2013.

Diversity within Populations
The Simpson's index of diversity within population (Table 2)  Average relative virulence complexity was high in total ( Table 2). The lowest relative complexity was recorded in 2001 (0.21 on average), and the highest in 2000 (0.81 on average).

Relationships among Populations
The number of common pathotypes between populations in two successive years varied from 9 (1995 and 1996, and 1998 and 1999) to 0 in most consecutive years from 2000 to 2010 (Table 6). Twenty-four common pathotypes were found in two groups, the populations of 1995-2000 and 2001-2013.

Diversity within Populations
The Simpson's index of diversity within population (Table 2) Average relative virulence complexity was high in total ( Table 2). The lowest relative complexity was recorded in 2001 (0.21 on average), and the highest in 2000 (0.81 on average).

Relationships among Populations
The number of common pathotypes between populations in two successive years varied from 9 (1995 and 1996, and 1998 and 1999) to 0 in most consecutive years from 2000 to 2010 (Table 6). Twenty-four common pathotypes were found in two groups, the populations of 1995-2000 and 2001-2013.
According to the estimates of the Rogers distance, high distances were observed for most successive years (Table 6). In contrast, Nei's gene distance revealed the greatest distance only between 2000 and 2001 populations (0.749). The lowest distance was observed between the populations in 2011 and 2012 (0.013) and 1998 and 1999 (0.017).   According to the estimates of the Rogers distance, high distances were observed for most successive years (Table 6

Discussion
Virulence structure of the powdery mildew population in Serbia was surveyed for the period 1995-2013. Overall, our study revealed highly diverse annual populations, with considerable differences between consecutive years. The high variability of the Serbian B. graminis f. sp. tritici population was expected due to the fact that the population structure is shaped to a large extent by sexual reproduction of the pathogen.
We found that pathotypes of this pathogen are ephemeral, changing from year to year via sexual recombination. Therefore, year-to-year comparisons of virulence gene frequencies seems to be more suitable than comparisons of pathotypes. However, some consistency when it comes to the most frequent pathotypes in the population was noticed. The most frequent was the group of TTT** pathotypes, characterized by high relative virulence complexity. These pathotypes were observed mainly in the population of 1995-2000. By contrast, in the last 4 years of the survey (2010-2013), the most frequent pathotypes had lower virulence complexity. Overall, virulence complexity in total was high, and since it represents the ability of isolates to overcome multiple Pm genes, these results

Discussion
Virulence structure of the powdery mildew population in Serbia was surveyed for the period 1995-2013. Overall, our study revealed highly diverse annual populations, with considerable differences between consecutive years. The high variability of the Serbian B. graminis f. sp. tritici population was expected due to the fact that the population structure is shaped to a large extent by sexual reproduction of the pathogen.
We found that pathotypes of this pathogen are ephemeral, changing from year to year via sexual recombination. Therefore, year-to-year comparisons of virulence gene frequencies seems to be more suitable than comparisons of pathotypes. However, some consistency when it comes to the most frequent pathotypes in the population was noticed. The most frequent was the group of TTT** pathotypes, characterized by high relative virulence complexity. These pathotypes were observed mainly in the population of 1995-2000. By contrast, in the last 4 years of the survey (2010-2013), the most frequent pathotypes had lower virulence complexity. Overall, virulence complexity in total was high, and since it represents the ability of isolates to overcome multiple Pm genes, these results point to a large potential of the B. graminis f. sp. tritici population in Serbia to break down the resistance of local wheat varieties.
The analysis of virulence associations in terms of pathotype frequencies is especially important in the study of structural alterations in populations of plant pathogens [17].  (Tables 2 and 3). Association of virulence between Mli and Pm2+4b+6 was frequently found (51.0% of isolates) during the last 4 years of the study but it was not detected in the most frequent pathotypes that have been observed in this period. Additionally, association between Pm3b and Pm4b virulences was found only in the period 2009-2013, yet it was not detected in the most frequent pathotypes from this period. The high frequency of these associations accumulated from a large number of rare pathotypes carrying the corresponding virulence genes. These associations were not found in the period 1995-2003. We can assumed that significant changes that were observed in the 2010-2013 period could be explained from this strongly associated virulence. These findings are in accordance with the results of the leaf rust population survey from Israel [17], where it was concluded that some strong virulence associations could be missed if only the predominant pathotypes are considered. High frequency of virulence to most tested Pm genes was observed during the entire period of the survey. This pointed to a constant need for extending differential sets of wheat with newly identified Pm genes. The high frequency of B. graminis f. sp. tritici virulent genes was also confirmed in previous studies in other countries [18][19][20][21], and recently in China [22], Ukraine [23], Australia [24], Egypt [25], Israel [26], and the USA [27].
On the other hand, combination of wheat resistance genes Pm5+6 remained effective in Serbia during the whole survey period. Virulence frequency to Pm5+6 was consistently low, leading to a conclusion that the Coker 983 line must have some additional Pm gene(s), in addition to Pm5+6, that is (are) more effective than this combination of genes. This conclusion was derived since the virulence frequency to Pm5 and Pm6 was intermediate or high during the whole period (with the exception of 2001). However, some increase in virulence frequency to Pm5+6 during the last 5 years was observed. The efficacy of combination of the resistance genes Pm5+6 was also detected in 1988, 1989, and 1991 in the studies of the powdery mildew population in Serbia [5,6]. However, the efficacy of the genes Pm4a, Pm4b, and Mld detected in these years has been lost since 1995, and also the efficacy of Pm2+4b+6 since 2004. These results indicated the constant threat of losing resistance, presumably due to the selective pressure of varieties.
The most significant change in the B. graminis f. sp. tritici population was observed for the virulence to Pm2+4b+6. This Pm combination was efficient for the period 1995-2002, when the corresponding virulence was at a low level (9.1% on average). In the period 2003-2013, the average virulence frequency increased to 58.9%. Unfortunately, there is no information about Pm genes present in the commercial wheat varieties grown in Serbia. However, the Pm2+4b+6 gene combination is present in the wheat line C39 [32], which was widely used in the wheat breeding program in Serbia. Based on these results, it can be presumed that the selective pressure of commercial wheat varieties led to the change in virulence to Pm2+4b+6.
The Simpson's index showed a high phenotypic diversity within annual powdery mildew populations during 19 years. However, according to the Kosman index, high diversity was observed in populations of 2005 and during the last 4 years of the survey. The same issue of inconsistency was noticed when distance between populations was measured by different metrics. According to the Rogers index, distances between most successive years were very large (around the theoretical maximum 1), while the corresponding values of Nei's distances were relatively small, with the greatest distance between 2000 and 2001 populations. Weak correlation between estimates obtained with different diversity/distance indices was already observed and demonstrated [10]. This inconsistency can be attributed to the nature of these indices: the Simpson's index and Rogers distance are based only on pathotype frequencies in the population. On the other hand, Nei's gene distance is based on virulence frequencies, whereas the Kosman index uses dissimilarity between isolates and maximizes the overall dissimilarity within a population. Similar discrepancy between estimates of different types was also shown in the study of powdery mildew population originated from barley [33].
The Rogers index showed a high dissimilarity between populations in the two periods: 1995-2000 and 2001-2013, with only 24 common pathotypes out of 1013 tested isolates. However, clustering of years using Nei's standard gene distance dissimilarity between B. graminis f. sp. tritici annual populations was detected between 1995-2008 (except 2004) and 2009-2013. These findings also indicate that there was a shift in the virulence structure of the powdery mildew population over the years.
Our results revealed that in the last period of the survey (2010-2013), the B. graminis f. sp. tritici population changed remarkably (the most frequent pathotypes, pairwise virulence associations, and also virulence frequency and complexity). This can be attributed to changes in wheat varieties grown in Serbia during the period 2000-2010. Although there is no information on the genetic background of the newly introduced varieties, one can assume that the variety shifts might have an influence on the pathogen as a result of the selection pressure.

Conclusions
Powdery mildew surveys that took place over a long period of time provided data of great importance for keeping track of the change in virulence occurring in powdery mildew populations and planning wheat breeding program strategies.
The obtained results showed that the virulence structure of the powdery mildew population in Serbia is complex and highly diverse. Furthermore, these results suggested that disease resistance can easily be overcome by the pathogen, resulting in severe epidemics in years with favorable agro-climatic conditions.
The gene combination Pm5+6 has remained effective during the whole period of the survey. Therefore, the Coker 983 wheat line possessing these resistance genes should be used in wheat breeding programs for resistance to powdery mildew. Nevertheless, introduction of new wheat lines with Pm resistance genes is necessary for further research on powdery mildew populations in Serbia and also for possible incorporation of these genes into wheat breeding material.