Two-thirds of the octoploid strawberry germplasm accessions screened for resistance to Fusarium wilt race 1 in the present study had disease symptom ratings in the resistant range , where ̄y is the estimated marginal mean among replicates and years, y = 1 plants were symptomless, and y = 2 plants were nearly symptomless . The severity of symptoms increased as ̄y increased on our ordinal scale . The race 1 resistance phenotypes observed among resistant and susceptible checks in the present study were consistent with those previously reported. The repeatability of race 1 resistance phenotypes among clonal replicates of resistant and susceptible checks was 0.81. One-fourth of the individuals screened in the present study were symptomless, classified as highly resistant , and appeared to be immune to AMP132 infections . This confirmed our suspicion that resistance to race 1 was widespread in natural and domesticated populations of octoploid strawberry. We suspected this because the only ‘non-California’ individuals screened in our previous study were highly resistant to AMP132 infection, had non-FW1 SNP marker haplotypes, and were presumed to carry novel R-genes . Moreover, the individuals screened in the present study were more diverse than those previously screened from the California population . Slightly more than half of the F. chiloensis and F. virginiana ecotypes screened for resistance to race 1 in the present study were classified as resistant . We did not observe geographic or phylogenetic trends—highly resistant ecotypes were found throughout the natural geographic ranges of both species . Eleven out of 62 F. chiloensis and 12 out of 40 F. virginiana ecotypes were classified as highly resistant . Moreover,plastic square flower bucket highly resistant ecotypes were identified for each of the seven subspecies of F. chiloensis and F. virginiana apart from one ecotype of F. virginiana subsp. platypetala, which was nevertheless classified as resistant .
We did not screen ecotypes of F. chiloensis subsp. sandwichensis, the subspecies found in Hawaii , because none were available when our study was undertaken. Approximately two-thirds of the F. × ananassa individuals screened in the present study were resistant to AMP132 infection . Of these, 20 originated in the California population and were either known or predicted to carry FW1 from SNP marker haplotypes . Other than Wiltguard , the other 141 are heirloom and historically important F. × ananassa individuals originating in North America, Europe, and Japan between 1880 and 1987 . We extracted the pedigree records for Wiltguard and the 141 non-CA individuals from the database described by Pincot et al. to show that the genetic relationships among most of these individuals and their ascendants were complex and intertwined . Of 603 individuals in the pedigree network, 101 were identified to be founders and 180 were phenotyped for resistance to Fusarium wilt race 1 . Twentyone individuals either lacked pedigree records or only had a single-generation of pedigree records with one or both parents known . The orphans are identified in the pedigree database but not shown in Fig. 2, which displays interconnections among the other 120 non-UC individuals and Wiltguard . The pedigree network diagram and database show that the alleles found in these resistant individuals have flowed through several common ancestors . Although the presence of multiple R-genes cannot be ruled out—partly because some of the individuals are orphans or extinct and phenotypes are only known for a subset of the ascendants of resistant individuals—there is a high probability that the number of unique alleles is small and that many of the R-alleles found in cultivars worldwide are identical-by-descent .
Our studies targeted two heirloom cultivars that share three resistant common ancestors , had unique haplotypes for SNPs in LD with FW1, were genetically distant from one another and individuals in the California population, and were predicted to have a high probability of carrying novel R-genes . The pedigrees for these cultivars are shown in Online Resource 7 .Selection of individuals for constructing a host differential panel was informed by insights gained from screening public germplasm collections for resistance to race 1 and insights gained from population structure analyses in strawberry . The host differential panel was assembled to maximize the probability of differentiating races and identifying sources of resistance to different Fof isolates . The phenotypes for resistance to AMP132, MAFF727510, and four other Fof isolates were previously reported for 25 octoploid individuals on the host differential panel: one F. chiloensis ecotype, one F. viriginiana ecotype, and 23 F. × ananassa individuals . MAFF727510 is an Fof race 2 isolate found in Japan . To broaden insights into the frequency and distribution of race 2 resistance sources, the phenotypes for resistance to MAFF727510 are reported here for an additional 116 individuals: 10 F. chiloensis ecotypes, 16 F. viriginiana ecotypes, and 93 F. × ananassa individuals . The latter included cultivars and other individuals selected to broadly sample allelic diversity in California and non-California populations worldwide. Similarly, the ecotypes were selected to sample allelic diversity across the natural ranges of F. chiloensis and F. viriginiana. The race 1 and 2 resistance phenotypes observed in these studies were highly repeatable: estimates of broad-sense heritability were ̂H2 = 0.98 for resistance to AMP132 and ̂H2 = 0.91 for resistance to MAFF727510 . Forty-one individuals on the host differential panel were classified as resistant to race 2 . Thirty-four of these individuals were symptomless and classified as highly resistant , and 34 of the race 2 resistant accessions were resistant to race 1 . Of the 78 F. × ananassa individuals from the California population, only four were resistant to MAFF727510. Conversely, of the 38 F. × ananassa individuals from the non-California population, 21 were resistant to MAFF727510. Slightly more than half of the F. chiloensis and F. virginiana ecotypes were resistant to race 2 , which was comparable to the frequency observed for race 1 resistance .
Three individuals on the host differential panel were resistant to every Fof isolate tested from California, Japan, Australia, and Spain . According to historical breeding records , Royce S. Bringhurst developed 61S016P006 , an S1 descendant of 43C001P036, by selecting for resistance to Verticillium wilt in Davis, California, nearly a half century before Fusarium wilt was discovered in California . Using the host differential panel as the study population and 50K Axiom SNP array genotypes, we searched the genome for associations between SNP marker loci and race 2 resistance phenotypes. Statistically significant GWAS signals for loci affecting resistance to race 2 were not observed . We repeated this analysis with the host differential panel using race 1 resistance phenotypes and reproduced the strong GWAS signal associated with the segregation of FW1 in the California population . The absence of a significant GWAS signal for race 2 resistance has several possible explanations. First, resistance to race 2 might not be governed by gene-for-gene resistance. Second, resistance to race 2 could be governed by gene-for-gene resistance but undetectable in a highly diverse population where multiple alleles and loci are segregating and the resistant alleles are uncommon. Those alleles, however,plastic plant pots could almost certainly be uncovered and identified by forward genetic analyses of segregating populations developed from crosses between resistant and susceptible parents, as described below for the race 1 resistance loci identified in the present study. Third, the sample size may have been insufficient to detect the presence of gene-for-gene resistance to race 2. This seems unlikely because R-genes have large effects, and we have consistently observed strong GWAS signals for FW1 in small samples of California population individuals, including the host differential panel .The FW1 locus was previously identified and physically mapped using diploid genome-informed GWAS in a closed UCD or ’California’ population of 565 individuals genotyped with a SNP array populated with diploid genome-anchored SNPs . To revisit our original analyses using octoploid genome-informed GWAS, 356 of these individuals were genotyped with either the 50K or 850K Axiom SNP arrays . This substantially increased the density and uniformity of SNPs in the FW1 haploblock and facilitated a search for genes with plant defense annotations because the SNP markers on both arrays were physically anchored in silico to octoploid reference genomes developed since the original study was reported . The physical positions of FW1-associated SNPs in the ’Camarosa’ and ‘Royal Royce’ genomes are shown in Online Resource 8. GWAS and de novo genetic mapping pinpointed the location of the FW1 locus to a near-telomeric haploblock on the upper arm of chromosome 2B spanning approximately 3.3 Mb . The 3.3 Mb haploblock was populated with 1,725 SNP markers from the 50K and 850K Axiom SNP arrays, of which 460 were signifcantly associated with race 1 resistance phenotypes . SNP markers with the strongest GWAS signals on chromosome 2B were AX-184226354 and AX-184176344 . KASP markers were developed for SNPs predicted by GWAS to be in LD with FW1 .
These were genotyped in the Fronteras and Portola S1 populations, shown to be tightly linked to the FW1 locus on chromosome 2B, and estimated to predict race 1 resistance phenotypes with 98.7 to 98.8% accuracy in the Fronteras and Portola S1 populations and 95.2 to 97.6% accuracy in the California population . Even though GWAS signals were observed between resistance phenotypes and SNP markers predicted to reside on homoeologous chromosomes, 84% of the 50K and 80% of the 850K Axiom array SNP markers with statistically significant GWAS signals were predicted in silico to reside on chromosome 2B proximal to the previously genetically mapped FW1 locus . Nevertheless, the strongest GWAS signals were observed for SNP markers that had previously been physically assigned in silico to chromosome 2D: AX-89872358 , AX-184098127 , and AX-184513679 on the 50K SNP array and AX-184055143 on the 850K SNP array . The GWAS signals observed on chromosomes other than 2B were almost certainly caused by incorrect in silico physical assignments of Axiom SNP array probe DNA sequences to positions in octoploid reference genomes . This conclusion was supported by several observations and analyses. First, QTL mapping in the Fronteras and Portola S1 populations only uncovered statistically significant signals for race 1 resistance on chromosome 2B . Second, GWAS was repeated for race 1 resistance phenotypes observed in the California population by fitting AX-184226354 as a fixed effect, then searching the genome for significant GWAS signals—AX-184226354 was the SNP marker on chromosome 2B that was most strongly associated with the FW1 locus . GWAS with AX-184226354 incorporated as a fixed effect eliminated signals on other chromosomes, in addition to eliminating signals for other tightly linked SNPs on chromosome 2B . Hence, fitting and testing multilocus genetic models with SNPs selected from single locus genome-wide searches is a powerful approach for resolving physical address assignment errors, a particular problem in polyploids, and generating more accurate estimates of genetic parameters . A certain percentage of erroneous GWAS signals are expected in octoploid strawberry because a small percentage of the short 71-nt DNA probe sequences for Axiom array SNP markers are incorrectly assigned in silico to physical positions in the reference genome, e.g., Hardigan et al. estimated that approximately 74% of QC-passing 850K Axiom SNP array probes could be assigned to the correct homoeolog in the ‘Camarosa’ genome. That percentage was virtually identical to the percentage of Axiom array SNPs with significant GWAS signals on chromosomes other than 2B in our analyses . Genetic mapping of Axiom array SNPs in the present and previous studies have shown that the percentage of Axiom array SNPs with physical positions that were incorrectly assigned in silico are found on homoeologous chromosomes . This GWAS complication is bound to arise in strawberry and other polyploid and paleopolypoid species with complex repetitive DNA landscapes, especially outbred species with whole genome duplications where homoeologous DNA variation complicates the physical assignment of short DNA sequences to subgenomes . The assignment of highly accurate long-read DNA sequences, by contrast, is straightforward . The octoploid genome-informed GWAS analyses described here were initiated in 2017 immediately after we assembled the ‘Camarosa’ reference genome . We have since expanded our understanding of the complexity of the octoploid genome, built superior haplotype-phased genome assemblies , and unequivocally shown that homologous and homoeologous DNA variation can be differentiated nearly genome-wide in strawberry .