Introduction

Asian soybean rust (SBR) is caused by Phakopsora pachyrhizi Syd. & P. Syd. P. pachyrhizi is an obligate, fungal pathogen that was first detected in the continental USA in 2004 (Schneider et al. 2005). This pathogen has the potential to reduce yield by 30 to 75% (Bromfield 1984; Kumudini et al. 2008; Yorinori et al. 2005) and infection can result in reduced numbers of pods, lower oil content, and higher rates of seed abortion (Bromfield 1984). In Brazil, costs of fungicide applications and yield losses from SBR have averaged US $1.98 billion/year from 2004 to 2014 (Godoy et al. 2016). In the USA, SBR is a problem in the southern states of Mississippi, Arkansas, Alabama, and Georgia (SBR.IPMPIPE.ORG), but yield losses have been limited, primarily due to unfavorable environmental conditions in many growing seasons for the spread and reproduction of P. pachyrhizi (Rosa et al. 2015). Despite this, fungicide usage for soybean has greatly increased since the arrival of SBR in North America, and annual fungicide costs for SBR control have averaged US $2.22 million (2005–2013) in Georgia alone (Langston 2009; Martinez-Espinoza 2006, 2007, 2008, 2015; Woodward 2010, 2012, 2013, 2015).

In order to manage soybean rust, host plant resistance to P. pachyrhizi (Rpp genes) is a useful tool. Rpp genes interact with specific pathotypes of P. pachyrhizi and provide either an immune response (IM, no visible sign of infection) or a reddish-brown resistant-type lesion (RB, non-sporulating or reduced sporulation, depending on pathotype virulence and environmental conditions) as compared to the TAN reaction (susceptible with many uredinia and copious amounts of urediniospores) produced by plants with no Rpp genes (Bromfield 1984).

There are six described Rpp loci to date. Rpp1 and Rpp1-b were discovered in PI 200492 and PI 594538A, respectively, on chromosome (Chr) 18 (McLean and Byth 1980; Chakraborty et al. 2009; Hyten et al. 2007). Rpp2 was identified in PI 230970 on Chr 16 (Hartwig and Bromfield 1983; Silva et al. 2008; Yu et al. 2015). A recessive source of resistance, rpp2, that produced a different reaction to a panel of P. pachyrhizi isolates compared to Rpp2 was found in PI 224270 (Garcia et al. 2008; Yamanaka et al. 2015). Rpp3 from PI 462312 was mapped to Chr 6 (Hartwig and Bromfield 1983; Hyten et al. 2009). Rpp4 was discovered in PI 459025B, mapping approximately 26 cM from Rpp1 on Chr 18 (Garcia et al. 2008; Hartwig 1986; Silva et al. 2008). Rpp5 was identified in PI 200526, PI 200487, and PI 471904 and a recessive allele rpp5 in PI 200456 on Chr 3 (Garcia et al. 2008; Pierozzi et al. 2008). Rpp6 was mapped to Chr 18 in PI 567102B, approximately 40 cM from the Rpp4 locus (Li et al. 2012), and a different allele or tightly linked gene, Rpp[PI567068A], was mapped in PI 567068A to the same locus (King et al. 2015).

No single Rpp gene has been shown to provide resistance to all known P. pachyrhizi pathotypes (Bonde et al. 2006; Paul and Hartman 2009; Pham et al. 2009). In addition, P. pachyrhizi populations vary widely by location and can overcome resistance genes over time (Hartman et al. 2005; Paul et al. 2015). In South America, the resistance of Rpp1 and Rpp3 were quickly overcome and the resistance of Rpp2 and Rpp4 is only effective against about one-third of rust pathotypes but Rpp1-b and Rpp5 continue to provide good resistance (Akamatsu et al. 2012; TMG 2016; Yamanaka et al. 2010, 2016). In the USA, Rpp1, Rpp2, Rpp3, Rpp4, and Rpp6 each have provided good resistance to field populations of SBR in most years (Walker et al. 2011, 2014). However, additional Rpp genes or alleles are needed to provide new sources of resistance and be introgressed into elite cultivars (Harris et al. 2015).

The recent advancements in soybean sequencing and the release of SoySNP50K iSelect BeadChips have provided soybean researchers with a wealth of tools for genomic analysis (Schmutz et al. 2010; Song et al. 2013). The SoySNP50K chips allow for the rapid comparison of polymorphisms between genotypes as well as relatedness comparisons based on the 50 K SNPs across all 20 chromosomes. Haplotype analysis of disease resistance loci enables the prediction of the alleles present in an accession based on its shared ancestry with known sources of resistance (Harris et al. 2015).

P. pachyrhizi populations display considerable variation among locations, possibly due to local selection pressures and the rapid evolution of the species from hyphal anastomosis which leads to heterokaryosis, nuclear fusion, and genetic recombination (Paul et al. 2015; Vittal et al. 2011). In addition, P. pachyrhizi urediniospores are spread by wind currents and can travel hundreds of miles, possibly introducing new pathotypes to existing populations (Rocha et al. 2015; Twizeyimana and Hartman 2012). SBR isolates, which represent a P. pachyrhizi collection at a specific location and year, have been collected internationally and maintained by the United States Department of Agriculture Agricultural Research Service (USDA-ARS) Foreign Disease–Weed Science Research Unit (FDWSRU) in a Biosafety Level 3 (BSL-3) plant pathogen containment facility at Ft. Detrick, MD since 1972 (Melching et al. 1983). Screening soybean accessions with a panel of these isolates provides differential reactions that have been used to discover new resistance alleles (Harris et al. 2015; Kendrick et al. 2011; King et al. 2015).

The Rpp4 gene contributed by PI 459025B has been fine mapped to a 55.3 kb region on Chr 18 (51,511,484–51,566,780 bp, Glyma.Wm82.a2) and a candidate gene, Rpp4C4, has been identified in PI 459025B (Meyer et al. 2009). We report the discovery of a new allele at the Rpp4 locus in PI 423972.

Materials and methods

Population development

The SBR resistance gene in PI 423972 was selected for genetic mapping based on the previous work of Harris et al. (2015), which demonstrated using bulk segregant analysis that the resistance of PI 423972 was within 5 cM of the Rpp4 locus of PI 459025B. PI 423972 had two differential reactions to the P. pachyrhizi isolates IN73-1 (Pantnagar, India, 1973) and TW72-1 (Taipei, Taiwan, 1972) when compared with PI 459025B. Pham et al. (2009) also reported differential reactions between PI 459025B and PI 423972 to P. pachyrhizi isolates AL04-3 (Baldwin Co., Alabama, USA, 2004) and BZ01-1 (Parana, Brazil, 2001). Additionally, PI 423972 shares only 50% similarity to the Rpp4 haplotype (Harris et al. 2015).

A cross was made between Prichard and PI 423972 to create a genetic mapping population. PI 423972 (Takema) is a maturity group (MG) IX soybean landrace that was collected from Kumamoto, Japan in 1976 (ARS-GRIN.GOV) and produces a RB-resistant response to the Georgia 2012 (GA12) P. pachyrhizi bulk isolate. Prichard is a MG VIII cultivar with white flowers and gray pubescence (Boerma et al. 2001) and is susceptible to the GA12 P. pachyrhizi bulk isolate (Fig. 1). The cross was made in the summer of 2009 in Athens, GA at the University of Georgia (UGA) Plant Sciences Farm. The F1 seed was grown in the UGA greenhouse in Athens, GA in the winter (2010–2011) to create the F2 population. The F2 population was advanced at the Plant Sciences Farm in the summer of 2011. Each F2 plant was single plant threshed to generate 140 F2:3 families for genetic mapping of resistance.

Fig. 1
figure 1

Phenotypic reactions of selected lines challenged by the GA12 P. pachyrhizi bulk isolate. a PI 423972, b PI 459025B (Rpp4), c PI 605791A, d PI 566984, e PI 567188, f Prichard, and g G00-3213. PI 459025B (b) produced RB lesions that occasionally sporulated. The G00-3213 and Prichard genotypes produced TAN, highly sporulating lesions (f, g). PI 423972 (a), PI 605791A (c), PI 566984 (d), and PI 567188 (e) all produced similar RB resistant lesions that did not produce urediniospores. The presence of P. pachyrhizi urediniospores are indicated by white arrows. Bar = 1 mm

Full size image

Phenotyping for rust resistance

The rust phenotyping method of F2:3 populations has been described in detail by Harris et al. (2015). Briefly, for the Prichard × PI 423972 F2:3 population, one family was planted into a black plastic 15-cell tray, of which only the outside 12 cells had pots (Griffin Greenhouse Supplies, Inc., Tewksbury, MA) placed in them and the center three cells were left open for light penetration to reduce crowding. Farfard® 3B potting mix (Sun Gro Horticulture, Agawam, MA, USA) was used to fill each pot and three seeds were planted per pot, with a total of six pots planted per family as well as per each of the parental checks. Plants were later thinned to two plants per pot, but due to variable germination, 8 to 18 plants per family were available for rating. Each parental check was replicated in the experiment for a total of four times.

Seedlings were grown in the greenhouse for 2 weeks and then inoculated with the GA12 bulk P. pachyrhizi isolate. The GA12 bulk isolate that has been described by Harris et al. (2015) was collected from soybean and kudzu leaves with natural SBR infection in 2012 and has been maintained since on susceptible check ‘Cobb’ plants in a greenhouse (Hartwig and Jamison 1975; Harris et al. 2015). Bulk P. pachyrhizi isolates derived from the field have been used to previously map rust genes effectively (Garcia et al. 2008 [ rpp5 and Rpp5]; King et al. 2015 [Rpp(PI567068A)]; Monteros et al. 2007 [Rpp?(Hyuuga)]; Silva et al. 2008 [Rpp2 and Rpp4]). A randomized complete block experiment with three replications was conducted to verify that the 2012 isolate had consistent results with previous isolates collected in Georgia using the PI sources of Rpp1 (PI 200492), Rpp2 (PI 230970), Rpp3 (PI 462312), Rpp4 (PI 459025B), Rpp5 (PI 200526), Rpp?(Hyuuga) (PI 506764), and the susceptible check, G00-3880. Each time the assay was conducted, the results showed that each PI responded with the expected reaction pattern as compared to previous isolates (Harris et al. 2015). The GA12 bulk isolate is tested yearly to ensure that the bulk isolate has not changed.

Inoculations were performed as per Harris et al. (2015). Plants were grown for an additional 2 weeks post GA12 inoculation to develop disease symptoms and all controls and parental checks were evaluated for their SBR reaction phenotypes. Families were scored as susceptible (TAN) or resistant (RB) based on the reaction of the 8 to 18 plants per family. To classify each F2:3 family, a method similar to King et al. (2015) was used. A family was designated as susceptible if at least 80% of the individuals were rated as susceptible (TAN). If 100% of the individuals were resistant (RB) in a family, that family was designated as homozygous resistant. All other families were designated as segregating or heterozygous. The appropriateness of this classification was verified using the expected 1:2:1 segregation ratio for the F2 generation of a single gene (Table 1). In the case of scoring individual plants as RB (resistant) or TAN (susceptible), we observed discrete differences between these reactions illustrated by the PI parents in Fig. 1, whereby little to no sporulation was ever observed on RB resistant progeny. This is similar to the results reported by King et al. (2015) and Harris et al. (2015) using bulk P. pachyrhizi isolates collected in Georgia. Six families were later excluded due to an unexpected phenotype, as they were designated as heterozygous but were expected to be susceptible based on the genotyping scores. The few plants that appeared as resistant are thought to be the escapes of the controlled inoculation of GA12.

Table 1 Chi-square analysis of 134 F2:3 families of the Prichard × PI 423972 population. Plants were phenotyped for resistance after being challenged with the GA12 bulk P. pachyrhizi isolate. The Rpp gene contributed by PI 423972 was expected to segregate as a single dominant gene
Full size table

Evaluation of a panel of lines

The GA12 P. pachyrhizi bulk isolate was used to challenge PI 459025B (Rpp4), PI 423972, susceptible controls Prichard and G00-3213, and resistant checks PI 200492 (Rpp1) and PI 594538A (Rpp1-b). PI 605791A was included in the evaluation as Harris et al. (2015) reported that it has an Rpp gene within 5 cM of the Rpp4 locus. PI 567188 and PI 566984 were also included as they have an unknown resistance locus but share a similar haplotype to PI 423972. The planting, growth conditions, and inoculation were performed as above where each accession was planted into six pots total in half of a tray. The lines were evaluated for SBR reaction phenotypes in May 2015 (Fig. 1; Table 2).

Table 2 Phakopsora pachyrhizi isolate reactions of sources of resistance that map near the Rpp4 locus and susceptible checks
Full size table

Evaluation of accessions with a panel of international P. pachyrhizi isolates

In order to determine if PI 423972 harbored a different Rpp allele from PI 459025B (Rpp4), a panel of international isolates was used to test the PIs for their SBR reaction phenotypes. Additionally, PI 200492 (Rpp1) and PI 594538A (Rpp1-b) were tested since the Rpp1 locus is near the Rpp4 locus on Chr 18. Williams 82 (PI 518671) was included as a susceptible control. Harris et al. (2015) previously screened these lines with nine international isolates and found differential reactions between PI 459025B and PI 423972 for the isolates TW72-1 and IN73-1. The other isolates used produced an RB reaction for both PI 459025B and PI 423972. The isolates TW72-1 and IN73-1 were retested in this study to verify the differential reaction that was previously observed. Pham et al. (2009) also found differential reactions between PI 459025B and PI 423972 for the isolates AL04-3 and BZ01-1, but these isolates were not retested in this study.

Isolate reaction tests for TW72-1 and IN73-1 were performed at the USDA-ARS FDWSRU in April, 2014. Experimental details have been described by Harris et al. (2015). In short, each line was tested with TW72-1 and IN73-1 with a total of four biological replicates. Each isolate was tested separately using a randomized complete block design. A biological replicate consisted of two seedlings of a given line in a single pot, and pots were randomized in trays. Seedlings were inoculated with isolates after plants were grown for 3 weeks in the greenhouse. At that time, all plants were transferred to a BSL-3 plant pathogen containment facility for inoculation and inoculated with TW72-1 and IN73-1 as previously described by Harris et al. (2015). Two weeks after inoculation, seedlings were scored as TAN, RB, IM, or INT (intermediate, reddish-brown but relatively smaller with uredinia and urediniospores present) (Table 2).

Bulked segregant analysis

Once phenotyped, a leaflet was collected from each plant of an F2:3 family from the cross of Prichard × PI 423972, for a total of 8 to 18 plants per family. Leaflets were combined into family bulks, lyophilized for 36 h, and subsequently ground into a fine powder using a GenoGrinder (SPEX, NJ, USA).

DNA was extracted from leaf powder as per the CTAB protocol of Keim et al. (1988). DNA samples for genotyping were diluted to 10 to 20 ng μL−1. For bulked segregant analysis (BSA; Michelmore et al. 1991), only families showing no segregation were selected; in this way, 35 homozygous resistant and 22 homozygous susceptible families were selected. For each respective bulk, equal amounts of tissue from each family were pooled and homogenized to create two bulks. DNA was extracted from each bulk separately in the same manner as described earlier, and diluted to 75 ng μL−1. Resistant and susceptible bulks were genotyped in the Soybean Genetics Lab at Michigan State University in East Lansing, MI using the SoySNP50K iSelect SNP BeadChips (Song et al. 2013). GenomeStudio V2011.1 software was used to call the genotypes (Illumina, San Diego, USA). SoySNP50K data of Prichard and PI 423972 were obtained from SOYBASE.ORG (Song et al. 2013). Positive BSA hits were scored when the resistant parent (PI 423972) had the same SNP genotype as the resistant bulk (e.g., both AA) and the susceptible parent (Prichard) genotype was the same as the susceptible bulk (e.g., both GG).

SNP assay design and genotyping

Polymorphic SNPs were identified between Prichard and PI 423972 within the interval identified with BSA. Ten KASP (LGC Genomics, Middlesex, UK) markers were developed using the criteria established by the KASP user guide and manual (LGC Genomics 2013) (Supplementary Table 1). Genotyping was performed using the protocol reported by Pham et al. (2013) for master mix preparation and thermocycling conditions. Endpoint genotyping was completed using a Roche LightCycler 480 II with LightCycler® Software (Roche Diagnostics Corporation, Indianapolis, IN) or a Tecan M1000 Pro Infinite Reader (Tecan Group Ltd., Männedorf, Switzerland) with KlusterCaller software (KBiosciences, Hoddesdon, UK). Allele calls that appeared as ambiguous were called as missing data.

Linkage mapping

The 134 F2:3 families of the Prichard × PI 423972 population were genotyped using the KASP SNP markers developed in this study (Supplementary Table 1). MAPMAKER software (Lander et al. 1987) was used to analyze linkage of the trait and markers and the results were verified using JoinMap 4.1 software (Van Ooijen 2006) (data not shown). MAPMAKER was used to calculate recombination distances, as it was more robust to inflation of genetic distances, and was executed with error detection turned on and using the commands order and try. Map distances were calculated using Kosombi’s mapping function (Fig. 2a). In addition, MapChart (Voorrips 2002) was used to create a map of the physical position of the genetic markers in this study and those used in previous studies, using the marker GSM0376 as the starting point. The approximate positions of Rpp4/Rpp4-b are based on the genetic distances between flanking markers sc21_3360 and sc21_3420 (Meyer et al. 2009) and GSM0543 and GSM0387 (current study) and all physical positions were taken from SOYBASE.ORG (Song et al. 2013) (Fig. 2b).

Fig. 2
figure 2

a The genetic map of the Prichard × PI 423972 population. The units on the left side of the chromosome map are in centiMorgans (cM) and the KASP marker names are displayed on the right side of the linkage map (Supplementary Table 1). b The physical map of the Rpp4/Rpp4-b locus on chromosome 18. The units on the right side of the chromosome are in kilobases (kb) and the physical position of the genetic markers in this study and those used in previous studies was taken from SOYBASE.ORG (Glyma.Wm82.a2). The approximate positions of Rpp4/Rpp4-b are based on the genetic distances between flanking markers sc21_3360 and sc21_3420 (Meyer et al. 2009) and GSM0543 and GSM0387 (current study)

Full size image

Haplotype analysis at the Rpp4 locus

The SoySNP50K haplotype of PI 423972 at the Rpp4 locus is defined by five SNPs: ss715631686, ss715631689, ss715631702, ss715631707, and ss715631709 (Supplementary Table 1). This haplotype was examined in Williams 82; Prichard and PI 423972 (mapping parents); PI 459025B (Rpp4); PI 605791A (previously mentioned as having BSA hits within 5 cM of the Rpp4 locus); PI 567188 and PI 566984 (with unknown resistance loci); known sources of rust resistance; and the 32 significant North American soybean ancestors (Gizlice et al. 1994; Harris et al. 2015; Monteros et al. 2010) (Table 3).

Table 3 A unique Rpp4-b haplotype that maps to the Rpp4 locus. This haplotype was determined by comparing a panel of susceptible soybean ancestors, the mapping population parents, and PIs with established Rpp genes. Isolate reaction data are from seedling host-plant resistance assays completed in the greenhouse using the Georgia 2012 (GA12) bulk P. pachyrhizi isolate
Full size table

Additionally, FlapJack (Milne et al. 2010) software was used to cluster the lines listed in Table 3 using the SoySNP50K genotypic data across the whole genome (Song et al. 2013). Genotypic data associated with scaffold sequences were removed. The dendrogram was created using hierarchical cluster analysis that takes into account the dissimilarities across the SNP data. If data were missing for a given SNP, it did not count as a mismatch and heterozygous genotype calls were scored as a 50% match to homozygous score (Supplementary Fig. 1).

Results

Resistance reactions

The parents of the mapping population reacted as expected when inoculated with the GA12 bulk P. pachyrhizi isolate. Prichard had TAN, highly sporulating lesions and PI 423972 produced RB lesions that were typically less than 0.5 mm in diameter (Fig. 1a, f). PI 423972 and Prichard were replicated in the experiment four times and showed no segregation in any of the plants observed. The 134 F2:3 families were expected to segregate according to a 1:2:1 ratio (resistant:segregating:susceptible) (Table 1). The chi-square value was not significant, indicating the resistance of PI 423972 when challenged by the GA12 isolate is controlled by a single gene. The gene action of this disease resistance gene is dominant, as indicated by the segregation ratios of 811 F3 plants within heterozygous F2 families that fit the 3:1 (resistant:susceptible) ratio expected for a completely dominant gene (p > 0.05, Table 1).

The GA12 bulk isolate was used to challenge PI 459025B (Rpp4), PI 423972, PI 566984, PI 567188, and PI 605791A; the SBR susceptible controls Prichard and G00-3213; and the resistant checks PI 200492 (Rpp1) and PI 594538A (Rpp1-b). PI 423972, PI 566984, PI 567188, and PI 605791A all produced similar RB-resistant lesions that did not produce uredinia or urediniospores and were generally <1 mm in diameter. PI 459025B (Rpp4) also produced RB lesions; however, the RB lesions were often >1 mm in diameter, and would occasionally coalesce. Additionally, uredinia were observed on every plant of PI 459025B, and often associated with relatively low amounts of urediniospores compared to the susceptible control, although some lesions of PI 459025B were RB with no sporulation. G00-3213 and Prichard had TAN lesions with uredinia and high levels of sporulation (Fig. 1).

Bulked segregant analysis and linkage mapping

A total of 17 positive BSA hits were observed from 50,325,784 to 52,979,027 bp (Wm82.a2 genome sequence). Based on the BSA interval, ten KASP markers were developed from polymorphic SoySNP50K SNPs (Supplementary Table 1). The KASP markers GSM0387 and GSM0390 behaved as dominant SNP markers but all other SNP markers behaved co-dominantly as expected (Supplementary Table 1).

For the Prichard × PI 423972 population, none of the SNP markers showed significant segregation distortion from what was expected (data not shown, p > 0.05). Linkage mapping using MAPMAKER created a 18.7 cM map distance on Chr 18 that included 10 SNP markers and spanned from 50,325,784 bp (GSM0376) to 52,821,191 bp (GSM0394) (Glyma.Wm82.a2) (Fig. 2; Supplementary Table 1). The PI 423972 Rpp resistance gene was mapped to a 187,553-bp region between the markers GSM0543 (51,397,064) and GSM0387 (51,584,617). This interval overlaps with the fine-mapped Rpp4 locus of PI 459025B (Rpp4) defined by sc21_3360 (51,511,484) and sc21_3420 (51,566,780) (Meyer et al. 2009) (Fig. 2).

Haplotype analysis at the Rpp4 locus using the SoySNP50K Infinium Chip data

The Rpp4-b haplotype of PI 423972 within the mapped resistance region is defined by five SoySNP50K SNPs: ss715631686, ss715631689, ss715631702, ss715631707, and ss715631709 (Supplementary Table 1). Of these SNPs, ss715631702 and ss715631707 had been used previously by Harris et al. (2015) to define the Rpp4 haplotype. However, the Rpp4-b haplotype proved to be unique to the lines PI 423972, PI 605791A, PI 566984, and PI 567188. All other genotypes tested including lines with known Rpp genes and the 32 susceptible soybean ancestors, Prichard, and Williams 82 did not have this haplotype.

The SoySNP50K data were used to create a dendrogram of relatedness across the whole genome to see how closely related the genotypes were that harbor an Rpp gene within 5 cM of the Rpp4 locus. PI 566984, PI 567188, and PI 605791A all may have an Rpp gene at the Rpp4 locus and clustered together (Supplementary Fig. 1). PI 566984 was collected from Indonesia and PI 567188 and PI 605791A were collected from Vietnam, respectively (Table 3). PI 423972 and PI 471904 (Rpp5) clustered tightly together (Supplementary Fig. 1); however, they were from different countries, Japan and Indonesia, respectively (Table 3).

Discussion

The resistance of PI 423972 was mapped using 134 F2:3 families. The Rpp gene from PI 423972 was flanked by GSM0543 (51,397,064) and GSM0387 (51,584,617) and a 187,553-bp interval overlaps with the Rpp4 interval of PI 459025B (51,511,484–51,566,780) (Fig. 2; Supplementary Table 1). PI 423972 and PI 459025B have different haplotypes at the Rpp4 locus and are phenotypically different when tested with a unique panel of P. pachyrhizi isolates, as well as the GA12 bulk isolate (Tables 2 and 3). When challenged with IN73-1, PI 423972 produced TAN lesions and PI 459025B produced RB lesions, and when challenged with TW72-1, PI 423972 produced mostly TAN lesions and a few RB lesions and PI 459025B produced only RB lesions. In this study, we confirmed the same differential reaction of PI 423972 and PI 459025B when challenged with IN73-1 observed by Harris et al. (2015) and Pham et al. (2009). The mostly TAN reaction of PI 423972 to TW72-1 in this study was also similar to the mixed reaction observed by Harris et al. (2015) and the TAN reaction observed by Pham et al. (2009). PI 459025B reacted with RB lesions that occasionally sporulated when challenged with GA12, and PI 423972 produced relatively smaller RB lesions that did not sporulate (Table 2).

Field resistance to SBR in Paraguay showed some difference between PI 423972 and PI 459025B (RB reaction in three of the four replications for PI 423972 with 5.2% severity, and TAN reaction for PI 459025B with 3.0% severity) (Miles et al. 2008). In the southeastern USA, PI 423972 had a lower average rust severity compared to PI 459025B (1.2 points on a 1 to 5 scale) in three locations in 2008 and a lower rust index (RI; 0.68 points on a 1 to 5 scale) between 2009 to 2012 in all locations tested (except Blackville, SC, 2009, which had a 0.35 point higher RI) (Walker et al. 2011, 2014). In addition, PI 423972 had 33% fewer uredinia per lesion compared to PI 459025B when inoculated with the isolates AL04-1 (Mobile Co., Alabama, USA, 2004), LA04-1 (Ben Hur, Louisiana, USA, 2004), TH01-1 (Chaingmai, Thailand, 2001), and TW72-1 (Taipei, Taiwan, 1972) (Pham et al. 2009). PI 423972 was collected from Japan, whereby PI 459025B was collected from China, representing geographically distanced regions (Table 3). These data indicate that the Rpp gene of PI 423972 is allelic to Rpp4, and the Soybean Genetics Committee has approved the designation Rpp4-b for the resistance of PI 423972. Rpp4-b appears to provide resistance to a narrower range of pathotypes, but shows a greater level of resistance than that provided by Rpp4.

The Rpp4-b (PI 423972) haplotype at the Rpp4 locus was used to examine PI 423972, a panel of diverse genotypes including the 32 North American soybean ancestors, known sources of Rpp resistance, Prichard, Williams 82, PI 605791A (that had BSA hits within 5 cM of the Rpp4 locus), and two other unmapped sources of resistance (PI 566984 and PI 567188). PI 423972, PI 605791A, PI 566984, and PI 567188 all possess a unique 5-SNP haplotype that no other PI in the panel possessed, including PI 459025B (Rpp4) (Table 3). This suggests that they may all possess Rpp4-b.

Interestingly, PI 423972 did not cluster with the other genotypes that had the unique haplotype, including PI 605791A (which had BSA hits within 5 cM of the Rpp4 locus) (Supplementary Fig. 1; Table 3). PI 605791A had a unique isolate pattern when compared to PI 423972 (Rpp4-b) and PI 459025B (Rpp4) and therefore may harbor another allele or tightly-linked gene at the Rpp4 locus and needs to be investigated further. PI 566984 and PI 567188 have not been tested with a panel of isolates (Table 3).

The 187.5 Kb interval (51,397,064–51,584,617), to which Rpp4-b has been mapped, contains 10 Glyma.Wm82.a2.v1 annotated genes in the Williams 82 reference genome (SOYBASE.ORG) (Supplementary Table 2). Of these, Glyma.18g226300 and Glyma.18g226500 are possible candidate genes, as they belong to the NBS-LRR gene family that has been associated with Rpp genes (Meyer et al. 2009; Yu et al. 2015). Similar to Rpp4C4, that was identified in the PI 459025B source of Rpp4 resistance, Rpp4-b is likely a sequence or copy number variant of one of these Williams 82 candidate genes in the PI 423972 genotype (Meyer et al. 2009). Further studies should be done to identify the sequences encoding resistance in Rpp4-b.

This study has mapped the resistance of PI 423972 (Rpp4-b) and demonstrated that Rpp4-b could be a valuable resistance allele for cultivar development in the southeastern USA. The KASP SNP assays developed here, including the GSM0543 and GSM0387 SNP markers, offer a tool to introgress Rpp4-b into elite germplasm.