Yellow Rust Epidemics Worldwide Were Caused by Pathogen Races from Divergent Genetic Lineages

We investigated whether the recent worldwide epidemics of wheat yellow rust were driven by races of few clonal lineage(s) or populations of divergent races. Race phenotyping of 887 genetically diverse Puccinia striiformis isolates sampled in 35 countries during 2009–2015 revealed that these epidemics were often driven by races from few but highly divergent genetic lineages. PstS1 was predominant in North America; PstS2 in West Asia and North Africa; and both PstS1 and PstS2 in East Africa. PstS4 was prevalent in Northern Europe on triticale; PstS5 and PstS9 were prevalent in Central Asia; whereas PstS6 was prevalent in epidemics in East Africa. PstS7, PstS8 and PstS10 represented three genetic lineages prevalent in Europe. Races from other lineages were in low frequencies. Virulence to Yr9 and Yr27 was common in epidemics in Africa and Asia, while virulence to Yr17 and Yr32 were prevalent in Europe, corresponding to widely deployed resistance genes. The highest diversity was observed in South Asian populations, where frequent recombination has been reported, and no particular race was predominant in this area. The results are discussed in light of the role of invasions in shaping pathogen population across geographical regions. The results emphasized the lack of predictability of emergence of new races with high epidemic potential, which stresses the need for additional investments in population biology and surveillance activities of pathogens on global food crops, and assessments of disease vulnerability of host varieties prior to their deployment at larger scales.


INTRODUCTION
Crop pathogens with worldwide prevalence and potential for long distance migration and invasions into new areas may pose a serious threat to food security regionally or globally Dean et al., 2012;Beddow et al., 2015). Crops like wheat, which are cultivated worldwide across diverse agro-ecological zones, provide a vast niche for their pathogens at local, regional, and continental scales (von Broembsen, 1989;Brasier and Buck, 2001). Wheat pathogens have been controlled to a large extent via ongoing and large-scale breeding efforts to improve disease resistance, which is economical, environment friendly and sometimes the only available option (Singh et al., 2016). Successful deployment of resistant crop varieties at larger scales and in different regions would, however, require better understanding of pathogen diversity for virulence across regions . Large scale deployment of host varieties with narrow genetic background for disease resistance have been reported to cause the acquisition of virulence at regional and continental scales Singh et al., 2004;Chen, 2005;Kolmer, 2005;Wellings, 2007). This is particularly the case for biotrophic fungal pathogens, which depend on the living host for both onseason and off-season survival, and the host resistance thereby induces strong selection favoring virulence mutants of the pathogen (Hovmøller et al., 1997;McDonald and Linde, 2002;Gladieux et al., 2011).
Yellow rust is a widespread disease across major wheat growing regions with diverse cropping systems, growing seasons and germplasm characteristics (Stubbs, 1985;Manners, 1988;Singh et al., 2004;Wellings, 2011). Resulting losses have been estimated to be at least 5.5 million tons per year at worldwide level (Beddow et al., 2015). Over the last decade a series of regional outbreaks of yellow rust epidemics have been reported worldwide, including Central and West Asia and East and North Africa (Figure 1, Figure S1; www.globalrust.org). A high disease pressure was observed in 2009 and onward in North Africa, particularly in Morocco (Ezzahiri et al., 2009). Since 2010, yellow rust was widely spread in East Africa causing economic losses in low-input farming system (Singh et al., 2016). Widespread epidemics were observed in Tajikistan in 2010 and later on in Uzbekistan and other countries of Central Asia (Rahmatov et al., 2012). In 2010, a high disease prevalence was observed in 2010 in Syria and Lebanon imparting economic losses (El Amil, 2015). These regular epidemics caused not only economic losses and additional need for fungicide sprays, but also threatened seed availability for the next cropping season (Shean, 2010). In Europe, the established P. striiformis population has largely been replaced since 2011 by distinct new lineages, generally known as Warrior, Kranich, and Warrior(−), causing increased epidemics on multiple wheat varieties (Rahmatov, 2016), and another lineage associated with epidemics on triticale in 2009-2010, particularly in Scandinavia (Hovmøller et al., 2011). A better understanding of pathogen virulence structure and the divergence of race(s) associated with major epidemic sites at different continents will be useful to facilitate breeding of resistant or less susceptible crop varieties and the development of appropriate disease management strategies based on host resistance (Johnson, 1992;Hawkesford et al., 2013).
New efforts have been made to investigate the yellow rust population genetic structure at worldwide scale, describing the worldwide population subdivisions, sources of invasions and the existence of center of diversity in the Himalayan and near-Himalayan region (Ali et al., 2014a,b;Thach et al., 2016a;Walter et al., 2016). However, the virulence structure of the pathogen has often only been described at country or regional scales, (e.g., Chen, 2005;Zeng and Luo, 2006;Hovmøller and Justesen, 2007a;Wellings, 2007;Bahri et al., 2009bBahri et al., , 2011Ali et al., 2014c;Hovmøller et al., 2016). Since the work of R.W. Stubbs from 1950s to 1980s (Stubbs, 1988Thach et al., 2015), only a single study about yellow rust virulence at the international scale has been published (Sharma-Poudyal et al., 2013) and attempts to link virulence and race structure with the recent regional yellow rust epidemic outbreaks in different parts of the world are missing.
The present study allowed an assessment of virulence diversity of P. striiformis at the worldwide level and inferring divergence of races prevalent in the recent worldwide yellow rust epidemics. The study was designed to: (i) determine the virulence profile of worldwide collection of P. striiformis isolates representing diverse geographical and evolutionary origin (ii) identify predominant races from recent epidemics worldwide and their divergence and prevalence in different geographical regions, and (iii) describe virulence and race diversity at continental scales using representative samples collected worldwide between 2009 and 2015.

Processing of Incoming Samples at the Global Rust Reference Centre
Field samples of yellow rust infected leaves from countries representing five continents were submitted to the Global Rust Reference Centre (GRRC), Aarhus University, Denmark ( Table 1). The samples from outside Europe were mostly sent by collaborators working in wheat breeding, rust pathology or agriculture extension generally within the network of the Borlaug Global Rust Initiative whereas European samples were part of ongoing survey activities in Scandinavia and bilateral agreements with GRRC (Table S1). The field samples contained information on date of sampling, location, crop type (winter/spring wheat,

Isolate Revival and Multiplication
Field samples were processed for recovery and multiplication using standard GRRC procedures: Infected leaves were kept on moist filter papers in petri dishes under humid conditions at 13 • C for 1-2 days to promote formation and release of urediniospores, and then inoculated on susceptible seedlings of wheat lines Cartago, Anja and/or Morocco by rubbing the infected leaf segment on the abaxial side of the leaves. The inoculated seedlings were misted with water and incubated in darkness at 10 • C for 24 h under high relative humidity. After incubation the inoculated plants were transferred to quarantine spore-proof greenhouse cabins with a temperature regime of 17 • C day and 12 • C night and a light regime of 16 h photoperiod of natural light and supplemental sodium light (100 µmol s −1 m −2 ) and 8 h dark. The plants were covered with cellophane bags before sporulation to avoid cross contamination. The spores were collected 15-20 days after inoculation and kept in a desiccator for at least 3 days. The urediniospores were then transferred to liquid Nitrogen storage (−196 • C) for further virulence phenotyping.
Recovery rates varied according to sampling conditions, treatment after sampling and time between sampling and arrival at GRRC. On average, 53% of samples were successfully recovered. Attempts were made to ensure multiplication of single-genotype samples, and on average only 3-5% of samples showed indication of multiple races, which was detected at time of assessment of the virulence phenotype. In case of multiple/contrasting infection types on individual differential lines, the isolate was sub-cultured based on single lesions, multiplied and re-tested for virulence phenotype confirmation (Thach et al., 2015;Hovmøller et al., 2017).
Virulence phenotyping was made through inoculation of differential lines following previously described protocols (Thach et al., 2015;Hovmøller et al., 2016). Both first and second seedling leaf were considered for virulence phenotyping, where infection type 7-9 on a 0-9 scale (McNeal et al., 1971) were generally considered to reflect compatibility (virulence)  2009,2010,2011,2012,2013,2014,2015 Latvia 2  Total 887 2009, 2010, 2011, 2012, 2013, 2014, 2015 and 0-6 incompatibility (avirulence). Conclusions about the phenotype for virulence and avirulence corresponding to the individual resistance genes were in most cases deduced from the infection types on two to three differential lines carrying the considered Yr-gene. The virulence profile/phenotype was inferred based on the overall virulence combination, and each distinct virulence phenotype was considered a distinct race.

Naming System for Races Associated with Epidemics
Prevalent races associated with epidemic outbreaks were assigned a name according to their genetic lineage identified through their molecular genotyping in comparison with the worldwide defined genetic groups (Ali et al., 2014a;Hovmøller et al., 2016;Thach et al., 2016a). A subset of 373 isolates representing epidemics causing races from different geographical regions (Table S2) were microsatellite genotyped , while a subset of 273 isolates were genotyped with SCAR marker (Hovmøller et al., 2008;Walter et al., 2016). Each lineage, consisting of one or more closely related multi-locus genotypes ("strains") of a particular race (virulence phenotype), was named Pst followed by a forthcoming digit. Virulence variants were designated by the additional virulence observed or (−) in case a new variant had fewer virulences than the first defined race within the considered lineage. PstS1 and PstS2 represented two closely related lineages previously defined by AFLP, microsatellite and SCAR markers (Hovmøller et al., 2008;Ali et al., 2014a;Walter et al., 2016).
PstS3 represented a clonal lineage prevalent in southern Europe, North Africa and West Asia (Ali et al., 2014a). Another lineage consisting of races prevalent on triticale in Northern Europe was named PstS4 (Hovmøller et al., 2008(Hovmøller et al., , 2016. PstS5 consisted of two races within a separate lineage with a specific microsatellite profile as compared to the previously described lineages (Ali et al., 2014a;Hovmøller et al., 2016;Thach et al., 2016a,b).
PstS6 represented a lineage prevalent in East Africa with specific microsatellite profile compared to other worldwide populations (Ali et al., 2014a;Hovmøller et al., 2016;Thach et al., 2016a,b

Compilation, Analyses and Interpretation of Data
The virulence phenotype for each isolate was compiled into Excel sheets along with sampling information like country, location, host variety, crop type collection date etc. The compiled data were uploaded to the Wheat Rust Toolbox database (http:// wheatrust.org/wheat-rust-toolbox/) for further data archiving, management and display, which was developed under the framework of the Borlaug Global Rust Initiative (www.globalrust. org). Final outputs of the toolbox are freely accessible via the Global Rust Reference Centre (www.wheatrust.org) and the Global Rust Monitoring System (http://rusttracker.cimmyt.org; Hodson et al., 2012;Hansen and Lassen, 2013).
The virulence corresponding to individual resistance genes and the combined virulence phenotype of the race were considered for analyses and interpretation. Diversity in terms of virulence and race was estimated across geographical locations using Simpson diversity index, 1-D (Simpson, 1949), where each individual virulence or each race was considered a different variant at a given geographical region for calculation of virulence and race diversity, respectively. The distribution of virulences and races were assessed across geographical regions and countries and associated with reported epidemic events based on information obtained through the BGRI rust tracker (http://rusttracker. cimmyt.org/) (Hodson et al., 2012;Hansen and Lassen, 2013).

RESULTS
The virulence phenotyping of 887 P. striiformis isolates from 35 countries representing eight geographical regions resulted in the detection of a total of 79 races (virulence phenotypes) from different genetic lineages. The prevalence of races varied widely across regions and none was detected in all regions.
Divergence in Worldwide Emerging P. striiformis Races P. striiformis remained significantly important around the world during 2009-2015 due to the emergence of races from divergent genetic lineages causing economically important epidemics in various parts of the world (Figure 2). These races represent emerging lineages which have recently been reported as new lineages and are becoming increasingly important epidemic drivers across large geographical area. The previously characterized PstS1 and PstS2 and their variants were prevalent at epidemics sites in North America (only PstS1), North Africa and West Asia (only PstS2) and East Africa (predominantly PstS2 with detection of PstS1), particularly during the epidemics in Morocco in 2009 and Syria in 2009-2010. Three races from distinct lineages, PstS7, PstS8, and PstS10, were prevalent in Europe since 2011, together covering more than 80% of the investigated isolates. Races from the PstS3 were detected in Europe, North Africa, West Asia and South Asia. PstS4 was highly prevalent in epidemics on triticale in 2009-2011 in Europe, particularly in Scandinavia. The Central Asian population was dominated (more than 90%) by PstS5 and PstS9. Races from the related PstS5 and PstS9 lineages were highly prevalent in epidemic areas in Central Asia, particularly in the 2010 epidemics in Tajikistan and in later years in Uzbekistan. In addition to PstS2 variants, the East African population was dominated by a race from the PstS6 (up to 25%), which was associated with huge epidemics in Ethiopia in 2010 ( Table 2). PstS5 and PstS6 were also detected in South Asia, but in very low frequencies. Races from these divergent lineages had characteristic virulence phenotypes along with their microsatellite genotype, which was clearly divergent from one another and from other races ( Table 2, Table S3). The race of PstS7 (a.k.a. Warrior) was the most divergent from other lineages based on the virulence phenotype, while the race of PstS8 (a.k.a. Kranich) was related with races of PstS5 prevalent in Central Asia. The two closely related and widely prevalent lineages PstS1 and PstS2 were the most diversified lineages with multiple variants often differing with a single or few virulences.

Diversity in Virulence and Races across Geographical Regions Worldwide
Out of the 19 Yr genes investigated, virulence were observed to 17 when the overall worldwide population was considered ( Table 3,  Table S4). Virulence to Yr5 and Yr15 were absent, while virulence to Yr10 and Yr24 was rare (up to 12%) in most areas except for samples from Triticale, e.g., PstS4. None of the virulences was fixed in the overall worldwide population, however, virulence to Yr2, Yr6, Yr7, Yr9, and Yr25 were generally in high frequencies and across many regions (>70%). Virulence frequencies varied substantially across regions. In addition to Yr5 and Yr15 in all regions, virulence was not detected to Yr27 in Europe; to Yr3, Yr4, Yr17, Yr32, YrSp, and YrAmb in West Asia; Yr10 and Yr24 in North Africa and South Asia; to Yr10, Yr24, and YrSp in Central Asia; and to Yr10, Yr32, and YrSp in East Africa.
Geographical regions varied with respect to diversity, both in terms of virulences and races detected ( Table 3, Table S4). The maximum number of virulences, regardless of combination in single races, was detected in Europe (16), while the minimum was observed in West Asia and South America (11). The highest number of races was detected in the pathogen center of diversity in South Asia (31) and the minimum in South America (3). Similarly, race diversity was the highest in South Asia (0.941) and lowest in North Africa (0.489). Europe, with the highest sampling intensity, had a high race diversity (0.799) due to the presence of both pre-2011 NW European races and post-2011 races like PstS7, PstS8, and PstS10.

Races Prevalent across Worldwide Geographical Regions
The 79 races detected in the worldwide population represented a wide range of races from the simplest (avirulent on all the tested Yr genes) to complex races like the one in PstS7 and PstS8 lineages (Tables 4A-D). None of the races was present on all continents representing eight geographical regions. While races from the emerging lineages like PstS2, PstS4, PstS5, PstS6, PstS7, PstS8, PstS9, and PstS10, were prevalent in distinct epidemic situations, most of the other races (34 out of 79) were detected only once ( Table 4, Table S5).
A total of 17 races were detected in the 462 isolates representing the European population during 2009-2015. PstS8, PstS7, PstS10, and PstS4 were the most frequent representing more than 80 percent of the population, while the pre-2011 races remained in low prevalence during the period (Figure 2, Table 4A). Among these 17 races, seven were detected only once in Europe during the period. Four, including PstS7 and PstS8, were re-sampled in other geographical regions.
In South America, three races were detected among the 28 tested isolates, none of which was detected in any other geographical region (Table 4B). Three PstS1 races were present in the isolates from North America, along with another race, none of which were re-sampled in any other geographical region (Table 4B).
A total of 33 races were detected in the larger area comprising Central Asia, West Asia, North Africa, and East Africa; seven of these were shared by most of these regions (Table 4C). North Africa and West Asia was dominated by races from the PstS2 lineage. In East Africa, PstS2 and PstS6 were the most prevalent during the studied period. Six races were typical to Central Asia and 18 were typical to East Africa, with limited re-sampling in other geographical regions. These regions had diverse races ranging from virulence to a single Yr gene (e.g., v27) to races of diverse virulence phenotypes (e.g., PstS9,v17).

DISCUSSION
We report on the race and virulence structure of P. striiformis isolates from five continents including several yellow rust epidemic areas during 2009 to 2015. Our results showed that races from relatively few divergent lineages were associated with huge yellow rust epidemics in different parts of the world, resulting in economic losses in the respective regions. High frequencies of virulence to widely deployed resistance genes in the regions, and the absence of virulences to other host resistance genes, are discussed in the context of sustainable use of host resistance in crop varieties. Finally, the impact of invasions on shaping the pathogen population across geographical regions is discussed. The results were based on a set of samples from important wheat varieties and breeding lines in different geographical regions and years. Comparable sample sizes were considered for most of the regions, except North America where isolates consisted of important references from previous studies *Based on Simpson diversity index "1-D" (Simpson, 1949).   -,-,6,-,-,9,-,-,17,-,25,-,32,-,AvS,--1 *v24/Avr24 could not be assessed due to avirulence to Avocet S. **High sampling activity in Scandinavia resulted in higher than expected frequency of PstS8 (Kranich) in Europe. Figures and symbols designate virulence and avirulence (-) corresponding to yellow rust resistance genes: Yr1, Yr2, Yr3, Yr4, Yr5, Yr6, Yr7, Yr8, Yr9, Yr10, Yr15, Yr17, Yr24, Yr25, Yr27, Yr32, and the resistance specificity of Spalding Prolific (Sp), Avocet S (AvS), and Ambition (Amb), respectively. (Milus et al., 2015a,b). The study did not include the Australian population, which is known to be dominated by races emerging from an incursion of a particular race from NW-Europe in 1978 (Wellings and McIntosh, 1990) and PstS1 related races (Wellings, 2007;Hovmøller et al., 2008). Northern Europe was over-represented, due to ongoing, intensive surveillance activities .

Worldwide Race and Virulence Diversity
Seventy-nine races were detected in the worldwide P. striiformis population, here represented by 887 isolates from 35 countries and eight geographical regions. Although none of the races were detected in all geographical regions, PstS7 and PstS2 variants (Hovmøller et al., 2016;Walter et al., 2016) were found in several, distant geographical regions, reflecting the long-distance dispersal capacity of rust pathogens (Zadoks, 1961;Hermansen, 1968;Brown and Hovmøller, 2002;Hovmøller et al., 2008;Ali et al., 2014a). Interestingly, 35 races were detected only once in the worldwide population, and the overall population was dominated by the above described genetic lineages associated with regional epidemics, which resulted in economic losses (Wellings, 2011;Beddow et al., 2015;Singh et al., 2016). In many of the geographical regions, relatively low race diversity was observed along with predominance of regionally important lineages. In East Africa, despite high race diversity, the overall population was dominated by races from the PstS2 and the PstS6 lineage, which contained virulence to Yr17 and Yr27, two widely deployed resistance genes in the region El Amil, 2015). In the South Asian recombinant population (Ali et al., 2014a;Thach et al., 2016a), a high race diversity was observed with no clear prevalence of any particular race. Recombination in a highly diverse population with temporal maintenance through a sexual cycle may generate new variants, including those carrying virulence to the deployed resistance genes. This will maintain the high virulence and race diversity in a recombinant population as observed in China and South Asia (Mboup et al., 2009;Duan et al., 2010;Ali et al., 2014c), even if the related host resistance is not deployed.

Worldwide Virulence Structure in the Context of Host Resistance Deployment
The observed worldwide virulence structure could be explained to a large extent by the regional deployment of host resistance (Table S6). Virulence to most of the considered resistance genes was observed in Europe, reflecting the large-scale deployment of these genes in Europe in the past (Bayles et al., 2000;Hovmøller, 2007;de Vallavieille-Pope et al., 2012;Hovmøller et al., 2016). The West and South Asian population showed fixation or a very high frequency of virulence to resistance genes widely deployed in the region like Yr2, Yr6, Yr7, Yr8, Yr9, Yr25, and Yr27 Bahri et al., 2011;El Amil, 2015). Interestingly Yr5 and Yr15 were the only resistance genes to which virulence was not observed in this study. These two genes have so far very rarely been reported deployed on large scale Chen, 2005;Hovmøller, 2007;Bahri et al., 2011;de Vallavieille-Pope et al., 2012;El Amil, 2015). Nonetheless, virulence toward Yr5 and Yr15 do exist in the center of diversity as well as in spontaneous virulence mutants elsewhere (Wellings and McIntosh, 1990;Hovmøller, 2007;Ali et al., 2014c), but as yet neither has been subject to strong selection by host resistance. The role of selection is further reflected by the predominance of races from major lineages carrying virulences to deployed resistance genes, even in populations with high race diversity, e.g., races from PstS6 carrying v17 and v27 dominating the East African population. These results emphasize the role of host selection on the virulence structure of pathogen populations.

Worldwide Virulence Structure in Relation to Genetic Structure
Our results added into our knowledge on the global landscape of P. striiformis. The Himalayan and near Himalayan populations have been shown to be recombinant and the center of diversity (Ali et al., 2014a,b;Thach et al., 2016a), which is endorsed by the high race diversity observed in the region. The Central Asian population has been shown to be closely related in 2000-2005 to the West Asian population based on samples from Kazakhstan, Kyrgyzstan, and Uzbekistan (Hovmøller et al., 2008;Ali et al., 2014a), which however was replaced by races from the new PstS5 causing widespread epidemics in the region since 2010. Races of PstS5 and PstS9 are now widely prevalent in the Central Asian region. West Asia has been reported to be invaded by PstS2 (Hovmøller et al., 2008;Walter et al., 2016), which has now successfully been established in the region, reducing the overall diversity as observed in the population before 1990 (Thach et al., 2016a). The East African population, which was closely related to the Middle Eastern population (Ali et al., 2014a;Thach et al., 2016a), was in this study dominated by PstS6. The Mediterranean population has been reported to represent a sink for different races with overall selective advantages (Enjalbert et al., 2005;Thach et al., 2016a). Mediterranean races prevalent in the region have been reported to be replaced by PstS2 in post-2000 epidemics (Hovmøller et al., 2008;Bahri et al., 2009b;Ali et al., 2014a). Since the appearance of the PstS7 (a.k.a. Warrior) in Europe, it has spread to North Africa and has become widely established in the region (Hovmøller et al., 2016). Indeed the European population has been shown to be replaced by the PstS7 and PstS8 lineages since 2011, which was confirmed in our study (Hovmøller et al., 2016). Regional studies over several years, based on both molecular and virulence data along with consideration of locally deployed host resistance, would further improve our understanding on the changes in global landscape of P. striiformis.

CONCLUSION
We report virulence and race diversity of worldwide P. striiformis populations, with the emphasis on races from regionally prevalent lineages causing epidemic outbreaks resulting in widespread economic losses for wheat production. Although resistance gene deployment in consideration of pathogen population variability played a significant role in protecting European wheat against the pathogen (Mboup et al., 2012;Sørensen et al., 2014;Walter et al., 2016), the emergence and prevalence of races from few divergent lineages highlights the lack of predictability of invasive races in terms of their origin and adaptability. This underlines the need for collaborative efforts from all stakeholders to understand the biology of crop pathogens, drivers of epidemics, surveillance of the pathogen population and vulnerability of host varieties. Concordance of virulence structure and establishment of certain races with regionally deployed host resistance emphasized the role of host selection on pathogen virulence structure and emphasized the need for greater regional and local diversification of host resistance. Efficient sharing of knowledge, germplasm, rust diagnostic facilities and information at national, regional and continental scales will be crucial to meet future challenges by the yellow rust pathogen.

AUTHOR CONTRIBUTIONS
SA, KN, DH, and MH contributed to collection of isolates; JR, TT, CS, and MH assisted in recovery and multiplication of isolates and performed the race phenotyping; JR and MH interpreted and quality controlled the race phenotype data; SA, JR, JH, PL, DH, and MH compiled and uploaded data into the database; SA, JH, and MH analyzed the data and designed the research; SA, TT, AJ, and MH identified and defined the lineages by molecular genotyping; SA and MH wrote the manuscript. All authors read and contributed to the revision of the manuscript.

ACKNOWLEDGMENTS
We are grateful to our worldwide colleagues for their contribution to the extensive sampling of isolates (mentioned in