Aecial samples were collected from barberry (Berberis diaphana and B. dubia) bushes in Huzhu and Datong counties of Qinghai province from June 11, 2018 to June 25, 2018 (Figure 1). Sampling was performed at an interval space of ~5 m between barberry bushes as a sampling site. A short shoot with leaves bearing aecia were cut off from barberry and then put inside a sampling paper bag. To keep the freshness of aecial samples, the collected samples were brought back to the laboratory in a short time for inoculation. Leaf samples of wheat stripe rust were collected from wheat in a radius of 100 m far from the rusted barberry at 1 week after emergence of aecia on barberry tissues at the mid-to-late June of 2018. Each of the leaf samples was placed in a sampling paper bag, kept at the room temperatures for drying completely, and stored in a desiccator at low temperatures (4–5°C) until use.

To obtain P. striiformis f. sp. tritici isolates from rusted barberry, individual leaf of 10-day old seedlings of wheat cultivar Mingxian 169, susceptible to Chinese races of P. striiformis f. sp. tritici identified so far, were used to inoculate with aeciospores of an individual aecial lesion. Aeciospore suspension was made by adding one drop of deionized water and mixing it well on a clean glass slide. Inoculated wheat plants were transferred to a dew chamber (E-36L2, Percival, IA, USA) and incubated in dark for 36 h at 10°C. After inoculation, wheat plants were moved into a growth chamber in a condition-controlled greenhouse and cultivated in a dual system of 16-h light at 16°C and 8-h night at 13°C.

For each of the wheat stripe rust samples, the routine operation method, described by Liu et al. (2021), was used to establish a pure isolate of single uredinium and to increase urediniospores.

Uredinia produced on leaves of the wheat plants were checked 15–20 days post-inoculation of aeciospores. To obtain a pure isolate, a single pustule (uredinium), before breaking the epidermis of a wheat leaf, was picked using an office pin to be transferred to a leaf of a new wheat plant (cv. Mingxian 169). The same conditions mentioned above were used to incubate single uredinium after transferring. Fresh urediniospores were collected into a glass tube by gently tapping. Inoculation was repeated for multiplying enough urediniospores.

The differential host set of 24 single Yr gene lines (SGLs), AvSYr1NIL(Yr1), including AvSYr5NIL(Yr5), AvSYr6NIL(Yr6), AvSYr7NIL(Yr7), AvSYr8NIL(Yr8), AvSYr9NIL(Yr9), AvSYr10NIL(Yr10), AvSYr15NIL(Yr15), AvSYr17NIL(Yr17), AvSYr27NIL(Yr27), AvSYr32NIL(Yr32), AvS/IDO377s F3-41-1 (Yr43), AvS/Zak 1-1-35-line1 (Yr44), AvSYrSPNIL (YrSP), AvSYrTr1NIL(YrTr1), AvS/Exp1/1-1Li (YrExp2), Tyee (Yr76), Kalyansona (Yr2), Hugenoot (Yr25), AvSYr28NIL (Yr28), AvSYr29NIL (Yr29), Vilmorin 23 (Yr3), AvSYrANIL (YrA), and 92R137 (Yr26), was used to identify avirulence and virulence patterns of isolates. A mixture of urediniospores and talc powder (v:v = 1:20) was used to inoculate seedlings (two-leaf stage) of the differentials according to the method described by Liu et al. (2021). After inoculation, the same conditions mentioned above were used. A 0–9 scale was used to assess avirulence and virulence phenotypes of isolates (Johnson et al., 1972; Line and Qayoum, 1992). Infection types of 0–6 were considered to be avirulent, and those of 7–9 as virulent. Virulence patterns, virulence frequency, and Kosman diversity index (Kosman and Leonard, 2007) of isolates were determined using the software VAT 1.0 (Schachtel et al., 2012).

Cetyltrimethylammonium bromide (CTAB) method with modification was used to extract genomic DNA of urediniospores of isolates (Aljanabi and Martinez, 1997). Concentration of DNA solution was diluted to 50 ng/ml for polymerase chain reaction (PCR) amplification with simple sequence repeats (SSR) markers. Totally, 12 pairs of SSR primers, CPS11, CPS32, and CPS34 (Chen et al., 2009), RJ3N and RJ11N (Bahri et al., 2009), RJ27 (Enjalbert et al., 2002), PstP031(Cheng et al., 2012), SUNIPst15-30, SUNIPst11-10, SUNIPst11-44, and SUNIPst10-48 (Tian et al., 2016), and WU6 (Ali et al., 2011), were used for the genotyping of isolates. Primers were synthesized by Sangon Biotech (Shanghai) Co., Ltd. (Shanghai, China), and the 5′-end of forward primers were fluorescently labeled. PCR products were analyzed on a DNA Analyzer (3730XL, Applied Biosystems, Waltham, MA, USA). SSR amplicons were evaluated using the software GeneMarker HID (Holland and Parson, 2011). Multilocus genotypes (MLGs) were detected based on microsatellite loci using GenClone 2.0 program (Arnaud-haond and Belkhir, 2007). The expected heterozygosity, Shannon's diversity index, and allele analysis in populations were conducted using GenAlEx 6.5 (Peakall and Smouse, 2012), and analysis of molecular variance (AMOVA) was used to evaluate microsatellite variation. Using NTSYS 2.10e program, a similarity matrix was generated based on the simple matching (SM) coefficient in the Qualitative module. The sequential, agglomerative, hierarchical, and nested (SAHN) clustering method was chosen for constructing a dendrogram for isolates by the unweighted pair group method with the arithmetic means (UPGMA) (Rohl, 2000).

Patterns of population structure and admixture were analyzed using the Bayesian model method, and implemented in software STRUCTURE 2.2 (Pritchard et al., 2000). The Monte Carlo Markov Chain scheme was run, as recommended by 100,000 and K value ranging from 1 to 10 with at least 100 repetitions to check the convergence of likelihood value for each K value. The optimal number of population subdivisions was determined by plotting the graph of estimated values of logarithm likelihood for each K value, and the maximum logarithm likelihood was used. For each K value, the Large K Greedy algorithm implemented in the program CLUMPP version 1.1 was used to search for the best alignment of multiple replicate cluster analyses (Jakobsson and Rosenberg, 2007). The population structure was visualized using DISTRUCT version 1.1 (Rosenberg, 2004). To supplement the output from STRUCTURE, discriminant analysis of principal components (DAPC) was performed in the adegenet package in R environment (Jombart et al., 2010).

Pycnia was initially observed on barberry tissues in the early June, and aecia was produced from middle June to early August in Qinghai (Figure 2A). Totally, 375 single aecium, collected from 52 rust-infected barberry bushes at 11 sampling sites located in Huzhu and Datong counties of Qinghai, were inoculated on wheat cv. Mingxian 169. Seven (1.87%) of all the 375 collected aecial samples produced typical P. striiformis f. sp. tritici uredinial symptoms on wheat leaves with different infection types (Figure 2B). As a result, a population of 29 single-uredinium isolates, derived from the seven aecial samples, were established for subsequent analysis. Meanwhile, 54 pure isolates were obtained from sample collections of wheat plants adjacent to barberry bushes.

Phenotype testing indicated that frequency of all isolates from barberry virulent to Yr. genes, except for Yr44, were <70%. Conversely, isolates from barberry showed higher virulence frequency than those from wheat at several resistance loci, including Yr8, Yr10, Yr26, Yr32, Yr2, and YrExp2. Particularly, frequency of isolates from barberry virulent to Yr8 was remarkably higher than that from wheat plants close to barberry. Comparatively, most of the isolates (>70%) from wheat were virulent to Yr7, YrSP, Yr6, Yr17, Yr76 (YrTye), Yr28, Yr9, Yr25, YrA, Yr1, Yr29, Yr1, and Yr44. Regardless of isolates from barberry or from wheat, all of them were avirulent to Yr5 and Yr15 (Figure 3).

Both P. striiformis f. sp. triatic populations from barberry and wheat showed a high level of diversity in virulence. In total, 28 out of 29 (one isolate invalid) isolates from barberry produced 17 virulence patterns (Table 1), and 8 isolates of them were avirulent to all of the 24 SGLs (Supplementary Table 1). In contrast, 44 virulence patterns were detected in a population of 48 out of 54, of which 6 were invalid isolates from wheat, significantly higher than those from barberry (Table 1; Supplementary Table 1). The population from barberry, with Nei's diversity index (Hs = 0.402) and Kosman index (K = 0.649), showed a higher virulence diversity than that from wheat with a Hs value of 0.265 and a K value of 0.384 (Table 1). Genotyping analysis showed that 24 MLGs were identified among 29 isolates from barberry, and that 38 were detectable among 54 isolates from wheat (Table 1). Both populations showed high diverse MLGs, and MLG ratio of the former (82.8%) was higher than those of the latter (70.4%). However, no same MLGs were detected in two populations from barberry and wheat. Based on genotypic diversity analysis, the mean Shannon's diversity index (I) and expected heterozygosity (He) of the population from barberry was 0.691 and 0.362, and those from wheat were 0.635 and 0.343, respectively (Table 1). The number of the virulence patterns and MLGs, and virulence and genetic diversity index of two populations from barberry and wheat were obviously high, indicating that the two populations were not strictly clonal.

Detection of the allelic variation uncovered that thirty same alleles were observed among 12 loci between isolates from barberry and those from wheat. The allele frequencies of isolates from barberry and wheat are presented in Figure 4. Although isolates from barberry and wheat shared most of the alleles, unique alleles at 7 SSR loci were detected among the isolates from barberry. These unique alleles included 111 and 113 at locus CPS34; 285 at locus PstP031; 256, 259, and 265 at locus CPS32; 344 and 346 at locus RJ3N; 194 and 224 at locus SUNIPst15-30; 230 at locus RJ27; 287, 299, and 327 at locus SUNIPst10-48. Similarly, some alleles, including 258 at locus PstP031, 210 at locus WU6, 337 at locus SUNIPst11-10, and 178 at locus RJ11N, were unique in the population from wheat, but not in that from barberry. For each locus, the number of alleles of the population from barberry was higher than that from wheat. All results mentioned above indicated that alleles of population from barberry were more abundant than that from wheat, and overlapping alleles were detectable between the population from wheat and those from barberry.

To examine the variation between the populations from barberry and wheat (Table 2), the AMOVA was used, and it showed that the variation between the populations from different hosts was 11.78%, that among populations from a common host was 4.17%, and that within populations was 84.05%, respectively. These results indicated that there was not a significant differentiation between the populations from barberry host and wheat host, and a large proportion of the variation presented within the population.

Phenotyping analysis, based on virulence in the NTSYS 2.10e program, indicated that isolate B1-15 from barberry and isolate W2-8 from wheat had the same virulence pattern, and that isolate B1-9 from barberry and isolate W2-9 from wheat were clustered in a clade with the simple matching coefficient value of 0.95 (Figure 5). In terms of the calculated virulence differences between each of the isolates, principal coordinates analysis (PCoA) with GenAlEx 6.5 software indicated that isolates from barberry were clustered together with those from wheat, and that 43.22 and 17.44% of the total virulence differences between the individuals of all isolates are represented in the first axis and the second axis, respectively (Figure 6). All results indicated that a close relationship existed between the populations from barberry and wheat in virulence.

Population structure analysis with 12 pairs of polymorphic SSR primers revealed a close relationship between wheat population and barberry population. Using the structure program, the number of optimal inferred cluster (K) was 3 (Figure 7A). When K = 3, isolates from barberry showed that ancestry primarily derived from two inferred clusters (blue and green bars), either single or combined, and that isolates from wheat showed ancestry mainly originated from two clusters (green and yellow bars). Some isolates from wheat shared identical ancestry with those from barberry (blue and yellow bars) (Figure 7B). Non-parametric DAPC for supplementary analysis using structure program showed that genotypes of P. striiformis f. sp. tritici from different hosts were obviously clustered together. Based on the Bayesian information criteria in the DAPC analyses, the whole population generated 3 clusters (Figure 8A). Isolates from barberry and those from wheat were grouped into one subcluster in group 1 or group 2 (Figure 8B). This indicated that both P. striiformis f. sp. tritici populations from barberry and wheat in the vicinity of barberry bushes had a close genetic relationship.

Principal coordinates analysis showed that isolates from barberry and wheat were clustered in a large group, and that there was no obvious separate cluster for isolates from both hosts (Figure 9). The first axis of the PCoA explained 27.63% of the differences between the individual isolates, and the second explained 23.92% of the differences between the individual isolates.