Genetic diversity of Puccinia kuehnii , the causal agent of orange rust of sugarcane, from Brazil

The use of resistant genotypes is the preferred method to control orange rust of sugarcane ( Saccharum spp) caused by Puccinia kuehnii. This approach has been adopted in Brazil but outbreaks of the disease on previously resistant varieties showed that the efficacy of this method is limited and requires a better understanding of pathogen diversity. Nevertheless, adequate molecular markers for examining


| INTRODUC TI ON
. The first recommendation to cope with this disease was the discard and replacement of susceptible cultivars: SP89-1115, RB72454 and SP84-2025(Barbasso et al., 2010. This strategy was adopted with success by the Australian sugarcane industry to control the first known epidemic of the orange rust, which occurred on cultivar Q124 in early 2000 (Magarey et al., 2001). Therefore, the outbreaks of the orange rust observed on commercial cultivars SP81-3250 and CTC15 in 2012/2013(Raizen, 2013 were unexpected, since they were not included as susceptible before. Nevertheless, it is unknown whether their resistance was broken or they were always susceptible because they were not mentioned by the time of the first outbreak. Later studies showed uncertain results, one indicated both as moderately resistant (Araújo et al., 2013) and the other, SP81-3250 as moderately susceptible (Chapola et al., 2016). The reason why the method of control applied in Australia did not work likewise in Brazil is still unknown.
The success of disease control by plant genotype resistance depends on the knowledge of pathogen diversity (Chen et al., 2006). In pathogens that undergo continuous life cycles of asexual reproduction, such as Puccinia, changes in genetic diversity are more prone to occur (Anderson et al., 2010). Therefore, an effective breeding programme requires a comprehensive information on genetic diversity of the pathogen (Pocovi et al., 2010).
Genetic diversity can be examined by genotypic and phenotypic traits. When P. kuehnii from Queensland and New South Wales in Australia were investigated by sequencing of ITS region, a low variation among 28 field isolates was detected, while greater sequence diversity occurred within rust collection from other countries (Indonesia, Papua New Guinea, China) (Braithwaite et al., 2009). Subsequently, sequencing of the ITS region revealed that the Caribbean population possessed only the 183A allele, whereas Australian and Asian populations showed a further allele (183G), suggesting differences between orange rust populations from both continents . The same ITS region did not discriminate within country the diversity among isolates of P. kuehnii from Brazil (Moreira et al., 2018). Not surprisingly, the ITS region is not adequate to examine diversity at population level, since they are highly conserved intraspecifically, variation occurs among different species (Bruns et al., 1991). Variation at population level or closely related individuals should make use of markers derived from fast changing regions, such as SSR (simple-sequence repeats), due to high mutation rate affecting the number of repeats and making them highly polymorphic (Abdel-Mawgood, 2012). Because such markers are not available for P. kuehnii, information on genotypic diversity at population level has not been clarified.
Phenotypic diversity of P. kuehnii carried out by the time of the first outbreak of orange rust in Australia suggested that a change in the race of the pathogen was the cause of the epidemic of 2000 (Magarey et al., 2001). Similarly, change in pathogenicity was suggested for the orange rust outbreak on CP 88-1762 in Florida, a cultivar that had not shown symptoms previously (Comstock et al., 2010). On the other hand, reason for the orange rust outbreaks in Brazil on SP81-3250 andCTC15 in 2012/13 (Raizen, 2013) is still a matter of controversy, since different fungal races were attributed by some researchers (Urashima et al., 2014) while difference in aggressiveness and not races, by others (Moreira et al., 2018). Fungal races are defined as the existence of significant differential reaction to orange rust between two isolates of P. kuehnii towards a sugarcane cultivar (Shine et al., 2005). A robust research with larger number of rust isolates and accurate inoculation method would reveal existence or not of fungal races, vital for the orange rust control by resistant genotypes. Therefore, to investigate the genetic diversity of P. kuehnii two independent investigations were carried out: first, genotypic diversity of orange rust isolates collected from various geographic locations of São Paulo state, the most important of Brazil, was investigated by SSR markers developed in the present work. Second, phenotypic diversity was examined by single-pustule inoculation technique on P. kuehnii retrieved from three susceptible commercial sugarcane cultivars.

| SSR marker development
Isolates of P. kuehnii were collected from leaves of sugarcane cultivars exhibiting suspected symptoms of orange rust. Subsequently, the causal agent of the orange rust was confirmed in laboratory by light microscopy based on reported morphological traits of urediniospores: orange colour, ellipsoidal format, surrounded by echinulate ornamentation and thickened apical wall (Comstock et al., 2008).

| Genotyping of P. kuehnii isolates
A total of 96 primer sets were designed and tested by DNA fingerprinting on four rust isolates, two of orange rust (P. kuehnii) and two of brown rust (P. melanocephala). Fingerprinting of these four isolates was performed as reported before (Arias et al., 2011), and amplicons were analysed using GeneMapper v. 4.0 (Applied Biosystems). These SSR markers were employed to examine the genotypic diversity of 29 isolates obtained from Brazilian sugarcane cultivars. Details of these 29 isolates are in Table 1. An isolate (Pkce) from an unknown cultivar of sugarcane in the United States that was initially used to develop the SSR markers was included in the analysis as a positive control of SSR amplification. Nineteen primer sets that showed polymorphism in the initial screening of four isolates were chosen to screen the Brazilian isolates of P. kuehnii. The list of those 19 primers selected and their sequences is shown in Table 2.
The amplicon fragments in base pairs (bp) generated by running capillary electrophoresis for 19 SSR markers on 29 P. kuehnii isolates were converted to a binary matrix to estimate genotypic diversity of the Brazilian P. kuehnii isolates. Cluster analysis by neighbour-joining (NJ) (Nei, 1972) using Jaccard coefficient and 3D principal coordinate analysis (3D PCoA) were performed using software NTSYS 2.0 (F. James Rohlf, Exeter Software, Setauket, NY).

| Host
The following sugarcane cultivars were used in the first screening experiment: RB835054 (72 young plants), RB867515 (308 plants), RB935744 (308 plants), RB966928 (72 plants) and RB975201 (308 plants). They were all classified as resistant to orange rust (Araújo et al., 2013, Porto & Urashima, 2018. In the subsequent experiment, cultivars RB935744 and SP81-3250 were employed, one as resistant and the latter as susceptible. Single-bud cuttings were produced from 9-to 11-month-old sugarcane stalks and planted in 770-ml plastic cups in a mixture (1 : 1 v/v) of soil and substrate (coconut fibre, vermiculite and pine-tree chips) amended with 3 g of 10-10-10 (NPK) fertilizer. Planted cuttings were maintained in a greenhouse for 30 days, at a temperature of 25 ± 3°C, when they were inoculated. Only the soil was watered to keep the aerial part of the plants dry and prevent unwanted fungal infections.

| Inoculum preparation
Leaves of adult plants, around 9-12 months of age, of cultivars RB72454, SP81-3250 and SP89-1115 of sugarcane with suspected symptoms of orange rust were taken to the laboratory for identification of P. kuehnii as previously described. They were classified as highly susceptible and thus employed as orange rust spreader in the sugarcane breeding centre in Araras city, São Paulo state. Fourto six-leaf young plants were inoculated using the single-pustule method specifically developed for orange rust of sugarcane by Porto and Urashima (2018). Briefly, single pustules of type 6 lesion, according to the description of Tai et al. (1981), were selected and detached from the diseased leaves. This tissue piece containing urediniospores was rubbed over an area of 5 cm 2 of the abaxial part of a new sugarcane leaf, previously humidified with autoclaved distilled Milli-Q water and 0.01% of Tween 20; the urediniospore concentration in each inoculation was 1.10 5 spores per ml (Porto & Urashima, 2018). Negative controls were prepared on the distal area of the same leaves, towards the tip of the leaf by repeating the spread motion using a piece of healthy leaf over a 5 cm 2 area. Subsequently,  (Table 4). All inoculation procedures were as described for the initial screening as well as the pathogen concentration (10 5 spores per ml).

| Host reaction
Reaction of sugarcane cultivars to different isolates of P. kuehnii was determined by measuring the size of rust-affected leaf area using Assess 2.0 software (APS Press, American Phytopathological Society, St. Paul, MN, USA). Measurements were taken 21 days postinoculation using a leaf segment of 5 cm 2 , each inoculation was considered one sample. Host/pathogen interactions were considered compatible, that is a virulent fungus on a susceptible host, when lesion areas were 1% or more of the inoculated area (5 cm 2 ). On the other hand, when the diseased area was <1%, the interaction was considered incompatible. This threshold for classification in compatible/incompatible interaction has been adopted by all sugarcane breeding programmes in Brazil to classify breeding lines (Coelho et al., 1990;Chapola et al., 2016;Dinardo Miranda et al., 1998).

| Data analyses
Test of significance of phenotypic diversity was applied only on the duplicated assay. Data of percentage of disease were transformed to Log 10 (x + 1), if necessary to achieve normality and homoscedasticity, and transformed data were used for analysis of variance (ANOVA). Mean values were compared by the method of Tukey's HSD (Tukey, 1949), Scott-Knott or Kruskal-Wallis one way analysis of variance of ranks (Kruskal and Wallis, 1952). All analyses were performed using the statistical package SigmaPlot v. 12.5 (Systat Software Inc., San Jose, CA).
SSR marker sequence, amplicon sizes and number of alleles detected by each marker are in Table 2. For further analyses, marker NPRL_PK_39a, which presented numerous null alleles, was not included in the cluster analyses. Genetic distances of 29 P. kuehnii isolates showed four main clusters of isolates, as detected by NJ and 3D PCoA (Figures 1 and 2, respectively). The first cluster was Alleles identified by amplification of three SSR markers, NPRL_ PK_108a, NPRL_PK_162_spka and NPRL_PK_221_spka, discriminated the four clusters of isolates. They were named genotypes 1-4, for which the specific alleles of each genotype are listed in Table 3.
Isolates Pk08 and Pk16 were not listed within the four groups since they represented intermediate genotypes.
Because viability of urediniospores of P. kuehnii was above 50%, phenotypic diversity was examined on orange rust pathogens collected from three highly susceptible cultivars. The strategy adopted in this study, that is single-pustule inoculation on resistant cultivars allowed identification of races.  a Compatible interaction (1% or more of lesion area over an inoculated area of 5 cm 2 ).
c Number in parenthesis is the percentage of diseased area caused by each fungal isolate.

F I G U R E 3
Virulence of two different isolates of Puccinia kuehnii on two cultivars of sugarcane with differential reaction carried out in a repeated randomized block design experiment with five replications and one young plant per replication. Isolate PkSp01-01a4 was retrieved from single-pustule inoculation on a resistant cultivar (RB935744), and isolate PkSp01-01 was collected from a susceptible cultivar (SP81-3250) [Colour figure can be viewed at wileyonlinelibrary.com] confirmed existence of distinct fungal races of P. kuehnii. Both isolates had a compatible interaction with the original host SP81-3250, Scpk01-01a4 caused an average of 8.36%, varying between 6.60% to 14.24%, of lesion area whereas ScPk01-01 showed an average of 17.33%, with lesion varying from 9.47% to 27.79%.

| D ISCUSS I ON
Data of the present work demonstrated that P. kuehnii infecting Brazilian sugarcane is genetically diverse, both at genotypic level and at phenotypic level. This finding was obtained by two independent studies since isolates employed at the genotypic assay were not the same from the phenotypic experiment. Therefore, any link between genotypic diversity and pathogen race was not possible in the present work. The molecular diversity was identified by SSR markers described in this work. They were developed from genome sequence of P. kuehnii collected from an American sugarcane cultivar (Techen et al., 2010) and were useful for Brazilian rust pathogens from various geographic locations and cultivars, and, thus, have potential application for P. kuehnii of other sugarcane producing countries where orange rust disease is present. The usefulness of these markers was demonstrated since fragments generated by these primers in PCR amplification were clearly visible and of the expected size.
The diversity of the rust pathogens showed four distinct genotypic populations among the 29 of P. kuehnii isolates submitted to SSR analyses suggesting that even larger variability can be found in the most important sugarcane producing state should the number of rust isolates increases. This is best exemplified by SP81-3250, a source of P. kuehnii isolates collected from seven locations of six different municipalities of São Paulo state, which isolates grouped in three out of four genotypic groups (Figures 1 and 2). The diversity shown by isolates from this cultivar suggests that a clearer picture of genotypic diversity of P. kuehnii can be obtained by a broader sampling of orange rust on SP81-3250, a cultivar that still covers large areas in the central southern Brazil (Braga et al., 2018). The other two (RB72454 and SP81-3250) were already discarded by the Brazilian sugarcane industry. Therefore, there is no possibility of examining further diversity of P. kuehnii sampling other cultivars, but SP81-3250.
The great genotypic diversity detected in our work revealed two clusters (2 and 3) not present in a sugarcane breeding site, the location where new clones (or breeding lines) are challenged against biotic/ abiotic stresses and then selected. Although conclusions may have been a little too hasty due to the limited number of isolates, the absence of sub-populations of P. kuehnii where selection of sugarcane genotypes is carried out leaves the possibility for a cryptic error in the breeding process (Correa-Victoria & Zeigler, 1993). This type of error may result in selection of germplasms that were not challenged against all variants of the pathogen population. This is a pivotal aspect in the selection process since a pathogen with great diversity has higher probability of rapid adaptation to the host and/or environment (Prasad et al., 2017). In this sense, one important step of breeding programmes to avoid cryptic error is increase the number of susceptible cultivars employed as spreader of P. kuehnii during selection of breeding lines, usually only one is employed (Chapola et al., 2016). to the one produced by the existing pathogen races (Draz, 2019).
Pathotypic diversity was also detected in our work by the existence of different races of P. kuehnii, where a significant interaction of two rust isolates on cultivar RB935744 was observed (Figure 3).
The existence of races was first suspected on P. kuehnii on SP81-3250 because it was not mentioned as susceptible by the time of the first outbreak in 2009 (Barbasso et al., 2010), but subsequently showed severe epidemics in 2012-2013 (Raizen, 2013 Barrera et al., 2013;Millus et al., 2009). Our data showed that the change in cultivar reaction from resistant to susceptible was due to the existence of races in P. kuehnii.
Races were identified in our work by the inoculation method employed. The single-pustule inoculation technique has previously proved a powerful tool in the case of UG99 of P. graminis f.sp. tritici. UG99 was a new race that emerged in Uganda, Africa, in 1999 with virulence to resistance gene Sr31, widely employed source of resistance to stem rust of wheat, reason why was considered a serious threat to wheat production worldwide (Pretorius et al., 2000). The race UG99 was then controlled by the resistant gene Sr24 (Jin & Singh, 2006). Later however, the single-pustule inoculation revealed that UG99 actually harboured another race, virulent to the gene Sr24 (Jin et al., 2008). This is an illustrating example of the strength of the single-pustule inoculation to examine the existing diversity of a pathogen, paramount for the success of disease control based on genotype resistance. Similarly, the single-pustule inoculation in the present work revealed that P. kuehnii infecting susceptible cultivars harboured a tiny sub-population capable of inflicting considerable symptoms on resistant cultivars.
The existence of fungal races is the main reason for breakdown of resistant cultivars in many host-pathogen interactions (Correa-Victoria & Zeigler, 1993;Prasad et al., 2017). Our phenotypic data proved that P. kuehni also features races, which is common in many Puccinia species: P. polysora (Ryland & Storey, 1955), P. striiformis (Johnson et al., 1972) and P. recondita (Park et al., 2000), to name just a few. Moreover, our data showed that the virulent sub-population of P. kuehnii still corresponds to a tiny fraction of P. kuehnii present in susceptible cultivars and that a correct orange rust management control should be adopted to maintain this population in such level and thus minimize chances of future outbreaks of orange rust in Brazil. Conditions that favour orange rust symptoms include warm and humid conditions (Magarey et al., 2004), predominance of a single cultivar, covering large areas (Magarey, 2010), higher number of ratoon crops and inoculum pressure (Chapola et al., 2016), extended period of leaf wetness  were already pinpointed. Nevertheless, one of the most important measures towards this strategy is the discard of susceptible cultivars, as recommended by the time of the first orange rust outbreak in Brazil (Barbasso et al., 2010). In sugarcane mills where this recommendation is followed, the urediniospore concentration of the virulent sub-population is likely to remain low and resistant cultivars, symptomless. Nevertheless, situations favouring increase of this sub-population can result in increase of urediniospores of the virulent sub-population resulting in disease symptoms on currently resistant cultivars. The first symptoms of orange rust were noticed when pathogen concentration reached 10 3 urediniospores per ml in a controlled inoculation assay . One factor that favours augment of this sub-population is the widespread cultivation of susceptible cultivars. In this sense, a recent survey on the central south region, responsible for more than 90% of the Brazilian sugarcane production, revealed that five out of 10 most cultivated cultivars in the central southern Brazil during 2017-2018 crop season were susceptible to orange rust, namely RB855156, RB92579, SP83-2847, CTC15 and SP81-3250, representing 26.9% of 6.3 million hectares (Braga et al., 2018;Chapola et al., 2016). Of the most concern is SP81-3250 due to its popularity, been the fourth most cultivated in the 2017-2018 growing season with 5.8% (Braga et al., 2018), and a record of previous orange rust epidemics in 2012-2013 (Raizen, 2013). Therefore, it is pivotal for the industry to discard this cultivar despite its popularity; possibility exists of severe symptoms of orange rust on resistant cultivars such as RB867515, RB935744 and RB966928, due to the new fungal race present in small proportion.
Brazil is the largest sugarcane producer in the world and control of orange rust is an essential part of the industry. The outbreaks of this disease on popular cultivars revealed knowledge of pathogen diversity is crucial for the orange rust control based on resistant cultivars. The set of SSR markers developed and validated in this work revealed genotypic diversity even in a limited number of pathogen isolates. Existence of different fungal races was also disclosed here. Our data showed the need to implement measures to improve control of orange rust by resistant cultivars in Brazil: to breeding programmes, the importance of challenging breeding lines to the whole diversity of P. kuehnii, and to sugar mills, the urgent need to discard susceptible cultivars.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/jph.12937.