This study demonstrated that (1) numerous soybean PIs previously reported to be susceptible to P. pachyrhizi isolates from other continents were actually resistant to field populations of P. pachyrhizi in the southern United States; (2) the majority of the resistant accessions originated from either Japan, Vietnam, or Indonesia; and (3) GWAS analysis of BLUP values calculated from disease reactions in multiple years and locations could be used to identify eight genomic regions putatively associated with resistance to SBR. Six of those regions have not previously been reported as far as we know.
Rpp genes and genomic regions of the soybean genome associated with resistance to SBR have been identified in previous studies using biparental molecular mapping, bulked segregant analysis, or GWAS with seedling reaction data from greenhouse assays (Chang et al. 2016; Garcia et al. 2008; Harris et al. 2015; Hyten et al. 2008, 2009). In most of those assays, plants were inoculated with an isolate collected in a single location and growing season. To our knowledge, this is the first GWAS of SBR resistance that used field data from multiple years and locations, and it is only the second GWAS study for SBR resistance. Unlike greenhouse assays, in which seedling reactions are typically classified into three infection types two weeks after inoculation, our plant reaction data provided assessments of the resistance of adult plants to different field populations of the rust fungus based on sporulation intensity and disease severity. Calculation of BLUP values from the semi-quantitative disease ratings allowed us to obtain estimates of the relative resistance of the germplasm accessions from an unbalanced data set collected in multiple locations over several years. GWAS analysis of the BLUP values made it possible to identify regions of the soybean genome associated with resistance to P. pachyrhizi pathotypes prevalent in the southern United States, including regions containing the Rpp3 and Rpp6 resistance genes.
The disease ratings demonstrated SBR resistance in at least 132 germplasm accessions that Miles et al. (2006) had previously reported to be susceptible or only partly resistant to four foreign P. pachyrhizi isolates. This was less surprising after the extent of pathotype diversity among and within P. pachyrhizi populations became evident (Pham et al. 2009; Twizeyimana et al. 2009; Akamatsu et al. 2013). The BLUP values calculated from the disease rating data indicated the relative levels of resistance or susceptibility of the accessions, and several of the PIs with the lowest BLUP values also had high levels of resistance in previous field and greenhouse tests conducted in the southern United States (Walker et al. 2011, 2014a, 2014b). The reactions of some of the PIs also showed that pathotypes of P. pachyrhizi populations from fields in the southern United States differ from some South American pathotypes (Garcia et al. 2008; Miles et al. 2008). This illustrates the importance of verifying that soybean germplasm has resistance to local and regional populations of the rust fungus.
Phakopsora pachyrhizi populations have high levels of pathogenic diversity within and among populations (Yamaoka et al. 2002; Akamatsu et al. 2013; Twizeyimana and Hartman, 2012). It is therefore important to identify germplasm with Rpp genes that condition resistance to local pathotypes across years and locations. Data from Stone et al. (2022) reaffirmed that the pathotypes of the Fort Detrick isolates used by Miles et al. (2006) are very different from pathotypes in U.S. P. pachyrhizi populations. After finding that only about 50 of the 805 PIs that Miles et al. (2006) selected were resistant to P. pachyrhizi pathotypes in the southern United States (Walker et al. 2011), we hypothesized that some accessions with susceptible or mixed reactions to the four foreign isolates would have resistance to at least some U.S. pathotypes. This study proved that hypothesis to be correct; 84 accessions that Miles et al. (2006) reported to be susceptible to SBR had BLUP values ≤ -0.10 and can thus be considered resistant.
Soybean cultivars with pyramids of two or more Rpp genes are likely to have broader and more durable resistance than cultivars with single Rpp genes (Yamanaka et al. 2012, Yamanaka and Hossain 2019). Since many of the known resistant soybean accessions have Rpp genes at the same loci, however, it is important to continue searching for additional Rpp loci and novel alleles. We focused primarily on evaluating PIs originating from Japan, Vietnam and Indonesia after results from early tests revealed that most of the resistant accessions with resistance in the United States were from those regions (Walker et al. 2011, 2014b). In addition, most of the named Rpp genes were discovered in PIs from southern Japan (Rpp1, Rpp2 and at least three alleles at the Rpp5 locus), central Indonesia (Rpp6), or Vietnam (Rpp7). Although PI 462312 (‘Ankur’), the source of the Rpp3 gene, was selected in Uttar Pradesh in north-central India, it was selected from a cross between unknown parents made in the United States, so the original source of the resistance gene is not known (Germplasm Resources Information Network). In this study, 84% of the accessions from Japan, 69% of those from Vietnam, and 49% of those from Indonesia had negative BLUP values, compared to only 11% of the 27 Chinese PIs tested. Nevertheless, some PIs from Japan, Vietnam, or central Indonesia were susceptible in our tests; the Japanese accessions PI 200456, PI 200526 (Shiranui), and PI 224270 (Hougyoku) had positive BLUP values, even though Yamanaka et al. (2010) had reported that they were highly resistant to a bulk fungal isolate from central Japan. In contrast, PI 200487 was resistant in central Japan and also in our field tests.
PI 635999 (‘DT 2000’) and PI 423972 were also resistant in field studies conducted in Hanoi, Vietnam, in 2006 and 2007 (Pham et al. 2010). Some other accessions were susceptible in our tests but showed resistance to field populations of P. pachyrhizi in northern Vietnam. Among those were the Rpp differentials PI 230970 (Rpp2), which had an RB reaction in Hanoi, and PI 459025B (Rpp4), which had an intermediate reaction there. In contrast, PI 200492 (Rpp1), PI 462312 (Rpp3) and PI 417089A (allele at Rpp3 locus) were resistant in our tests but developed TAN infection types in Hanoi (Pham et al. 2010). The reactions of some Rpp gene differentials in 2006 field tests in eastern Paraguay also differed from their reactions in our tests (Miles et al. 2008). PI 200492 and PI 615437, which has an Rpp3 resistance allele, were both susceptible in Paraguay, whereas PI 230970 and some PI 567099A plants (with the recessive rpp3 allele) were resistant there. These results attest to the pathogenic diversity among P. pachyrhizi populations.
Accessions from other countries that were resistant in our tests included PIs 368039, 379621 and 518295 from Taiwan, PI 423972 from Nepal, and PIs 203398, 417503 and 628932 from Brazil. PI 368039 was one of the few resistant PIs from this study that was also highly resistant to an unpurified rust isolate from central Japan (Yamanaka et al. 2010). The resistant accession PI 476897 is reported to be from China but was obtained from a germplasm collection in Hanoi, Vietnam. Harris et al. (2015) reported that PI 518295 has a resistance allele at the Rpp1 locus, PI 417503 has one at the Rpp3 locus, and PI 476905A has a resistance gene at the Rpp6 locus. It was somewhat surprising that very few of the accessions from southern China were resistant in our tests, especially since at least seven of them have Rpp genes that are effective against foreign isolates and populations of the fungus (Supplemental Table S1). Six of 24 Chinese PIs screened are known or thought to have a resistance gene at the Rpp1 locus, but like the Rpp1-b gene of PI 594538A and the Rpp1 allele from PI 561356, none of their alleles provided resistance in the southern United States (Supplemental Table S1; Walker et al. 2011, 2014b). A few other Chinese accessions in our study had negative BLUP values that were only slightly less than zero, so they unlikely to be of value as sources of SBR resistance genes.
A high percentage of soybean accessions with resistance in the United States have a resistance allele at the Rpp3 locus (Harris et al. 2015), and many of them originated from southern Japan (Supplemental Table S1). At least 14 (28%) of the accessions with the 50 lowest BLUP values in the present study have a resistance allele at the Rpp3 locus, so it is not surprising that a marker near this locus had the highest significance level in the GWAS (Fig. 3). Alleles at the Rpp3 locus provided resistance to accessions from Indonesia (e.g., PI 567046A and PI 567034), Vietnam (PI 635999), and several from Japan, such as PI 416826A and PI 200488. Hyuuga and PI 462312 had similar BLUP means (-0.64 and − 0.66), suggesting that the Rpp5 allele in Hyuuga may not have enhanced resistance against U.S. rust populations. Resistance genes at the Rpp6 locus on Chr 18 appear to be less common, but the allele from PI 567102B and PI 567104B conditioned very high levels of resistance, resulting in the lowest and fourth lowest BLUP means. PI 567090, which had the third lowest BLUP value, has a resistance allele at the Rpp3 locus and another on Chr. 18, likely at the Rpp6 locus (Harris et al. 2015).
The reactions of most of the Rpp gene differentials to field populations in this study differed considerably from the reactions of these accessions to a diverse collection of international and pre-2005 U.S. isolates in a recent study by Stone et al. (2022). In that study, all 16 isolates defeated the Rpp1 gene in PI 200492 to some extent, even though PI 200492 had one of the lowest BLUP values in the present study. In contrast, the Rpp2 and Rpp4 genes, which were ineffective in this study, conditioned resistance to all or most of the 14 isolates in the Stone et al. (2022) assays, as did the Rpp1-b gene from PI 594538A and the Rpp1 alleles in three other accessions from China. PI 567102B (Rpp6) was the only highly resistant accession from the present study that also had resistance to most of the 16 isolates. These results and those of Pham et al. (2009) indicate a possible pathotype shift between the founder populations of P. pachyrhizi discovered in late 2004 and the predominant pathotypes of fungal populations along the Gulf Coast of the United States a few years later.
Detection of the genomic regions containing the Rpp3 and Rpp6 loci demonstrated that the GWAS approach used was effective. Of the eight genomic regions that were significant at P < 1.445 × 10− 5; -log10(P) = 4.8), only ss715594707 on Chr 6 and ss715632525 on Chr 18 were locations with known Rpp loci. The former is near the Rpp3 locus, a common location for genes that condition resistance to SBR in the United States (Supplemental Table S1; Harris et al. 2015), and the latter is near the Rpp6 locus. Since at least four of the 20 PIs with the lowest BLUP values (ranging from − 1.457 to -0.754) are known or thought to have an SBR resistance gene at the Rpp3 locus (Harris et al. 2015; Vuong et al. 2016), it was not surprising that a marker near the locus was significantly associated with disease. In contrast, resistance alleles at the Rpp6 locus have only been reported in a few PIs, but those alleles have conditioned much higher levels of resistance than any known Rpp3 alleles. This was demonstrated by the low BLUP values of the Indonesian accessions PI 567102B and PI 567104b and the significance of the ss715632525 marker near the Rpp6 locus on Chr 18. Since PI 567090 also has the same base residue at that marker, it might also have a resistance gene at the Rpp6 locus. The Rpp6 gene is one of the few named genes that has been effective in both North and South America (Miles et al. 2008; Walker et al. 2014b). PIs 476905A, 567068A, 567076 and 567129 may also likely have a resistance gene at the Rpp6 locus (Harris et al. 2015), but their BLUP values ranged from − 0.35 to -0.56, suggesting that they carry a different allele from the one in PI 567102B and PI 567104B.
When Chang et al. (2016) performed a GWAS for SBR resistance using infection type data from the Miles et al. (2006) greenhouse assays, they found one significant marker near the Rpp1 locus on Chr 18 and another on Chr 15. The Rpp1 and Rpp6 loci are both on Chr 18 but are located on different telomeres (Kim et al. 2012; Li et al. 2012; Yu et al. 2015). Although PI 200492 was one of the most resistant accessions in our study, we did not detect any significant markers close to the Rpp1 locus. At least 11 of the PIs in the panel have a resistance allele at the Rpp1 locus, but the allele(s) in seven Chinese accessions were ineffective against the U.S. field pathotypes. As a result, the frequency of Rpp1 alleles with significant phenotypic effects in this study was low, which probably explains the reason that the locus was not detected using GWAS. Other than PI 200492, PIs 417120 and 423958 from Japan, and PI 518295 from Taiwan were among the few accessions screened that had an effective resistance allele at the Rpp1 locus (Supplemental Table S1).
The failure of the GWAS analysis to detect significant markers close to Rpp2 on Chr. 16; (Yu et al. 2015), Rpp4 at the opposite end of Chr. 18 from Rpp1 and Rpp6 (Silva et al. 2008), Rpp5 on Chr. 3 (Garcia et al. 2008), or Rpp7 on Chr. 19 (Childs et al. 2018b) was likely due to a low frequency of resistance genes and/or weak phenotypic effects. Although PI 605823 (Rpp7) and PI 200487 (allele at Rpp5) were among the 20 accessions with the lowest BLUP means for disease in this study, Rpp7 has not been reported in any other accessions, and PI 471904 appears to be the only other PI with an effective allele at the Rpp5 locus. A failure to detect genes that are present at a low frequency in a population is one limitation of GWAS analyses (Bandillo et al. 2015).
The detection of significant markers in regions of six chromosomes that have not been reported to have Rpp loci was unexpected, and the fact that the marker significance levels were higher than that of the marker close to the Rpp6 locus is encouraging. Some of the resistant accessions that we tested may have novel resistance genes in those regions that could be used in Rpp gene pyramids to improve resistance. Because a majority of the PIs in the GWAS panel had been reported by Miles et al. (2006) to be susceptible or to have had mixed reactions, few of them have been used to develop biparental mapping populations. It is possible that the SBR disease rating data that we used for the GWAS may have allowed contributions of Rpp loci in the six regions to be detected for the first time. Much of the phenotypic data used the BLUP values used for GWAS in this study reflected both disease severity and intensity of urediniospore production. Those semi-quantitative data provided a more accurate assessment of reactions that were intermediate between heavily sporulating TAN infection types and Type 0 or RB infection types with few uredinia and low sporulation. Yamanaka et al. (2010) also recognized the value of using semi-quantitative criteria (i.e., rating scales) for more accurate assessments of host reactions to SBR. Although Chang et al. (2016) also reported a novel putative resistance QTL between positions 10,659,000 and 10,859,000 bp on Chr 15, it did not correspond to the genomic region that we detected near position 15,694,109 on that chromosome.
Findings from this study should be useful for the development of cultivars with broader and more durable SBR resistance. Some accessions previously reported be susceptible to SBR should be re-evaluated for resistance in some other countries where rust is an economically important disease. Confirmation of an unknown Rpp locus in any of the six previously unreported genomic regions that were detected by GWAS would offer soybean breeders additional options for developing cultivars with novel combinations of resistance genes. Data collected early in this study led to the discovery of the Rpp7 locus in PI 605823 (Childs et al. 2018b). After the resistant accessions responsible for the significant GWAS markers are identified, their resistance can be characterized using a panel of pathogenically diverse P. pachyrhizi isolates and other biparental mapping populations can be created to fine-map the locations of the novel loci.