Some varieties of common bean (Phaseolus vulgaris L.) suffer from pod shattering, which can severely reduce yields, especially in arid conditions. The PvPdh1 locus on chromosome Pv03 has recently been described as a major locus controlling pod shattering in common bean and could be used to mitigate pod shattering in the future. Despite this, the role of a possible second locus on chromosome Pv08 remains unclear and patterns of dominance and epistasis between alleles of these genes have not been resolved. This information will be vital for efficient selection to decrease pod shattering. Further, the genetic diversity around the PvPdh1 gene has not yet been thoroughly explored and there are not yet genetic screens that can be used to evaluate pod shattering in segregating populations. Here, we have developed a recombinant inbred population to determine the roles of genes implicated in pod shattering and evaluate the patterns of dominance among the relevant alleles. Our results suggest that a PvPdh1 allele reduces pod valve twisting, and its dominance varies by phenotyping method. This allele is the only genetic variant that provides environmentally stable and widespread resistance to pod shattering in Middle American common beans grown for grain. Further analyses identified a selective sweep around PvPdh1 with greater nucleotide diversity in individuals with the ancestral, shattering-susceptible allele. Finally, we developed simple, effective CAPS markers to facilitate the introgression of PvPdh1 into new varieties of common bean. These genetic resources will be critical for improving the aridity resilience of a major global staple.

Three populations were evaluated in this study; a biparental population and two diversity panels, which represent each of the two domestication events of common bean. The biparental population was developed to study two shattering-related QTLs, their patterns of dominance, and their interactions. Cultivars ‘Mayflower’ (ecogeographic race Mesoamerica, Kelly et al. 1989) and ‘Bill Z’ (race Durango, Wood et al. 1989) showed total resistance to pod shattering when field-grown in Davis in 2017 (n=27, n=19, respectively). These varieties were among the most distantly related accessions in the MDP, with neither showing any evidence of admixture between ecogeographic races (Moghaddam et al. 2016; Parker et al. 2020). Mayflower is a navy bean type (white, small-seeded), which possesses a SNP allele on Pv08 that is weakly associated with resistance to pod shattering in race Mesoamerica. Bill Z is a pinto bean type and has a SNP variant on Pv03 associated with strong PvPdh1-mediated shattering resistance common in race Durango. The population can therefore be used to determine if a reduction in pod shattering was independently selected in each of these ecogeographic races. An F₃ population of 138 individuals was developed by hybridization between these cultivars. Each F₃ individual was descended from a distinct F₂ plant, and all of the F₂s were the progeny of a single F₁ developed by cross-pollinating Mayflower and Bill Z. This 138-member Mayflower x Bill Z (MxB) population was used to validate the possible alleles on Pv03 and Pv08 and test any patterns of dominance and epistasis between the loci.

The two diversity panels were grown to evaluate the degree of pod shattering across diverse accessions of common bean. In 2016, 98 members of major market classes in the Andean Diversity Panel (ADP, Cichy et al. 2015) were field-grown in Davis, California, to evaluate each variety’s susceptibility to pod shattering. In 2017, 278 varieties of the BeanCAP Middle American Diversity Panel (MDP, Moghaddam et al. 2016) were similarly field-grown in Davis to evaluate pod shattering. At maturity, a sample of pods (mean n = 30) was harvested from each variety.

Mature pods of all phenotyped varieties were harvested and then exposed to seven days of desiccation at 65°C and a further seven days of re-equilibration to room temperature. The desiccation conditions for all varieties were identical and desiccation was conducted using the same drying chamber. The proportion of pods dehiscing in this treatment was recorded, along with the market class of each variety. For evaluation of pod twists in the MxB population, all non-shattering pods were fractured by hand and then all pods were subjected to the desiccation treatment and re-equilibration again. The number of twists was counted for ten pods of each genotype, with “1” indicating a complete 360° rotation of the valve.

DNA was extracted from young trifoliate leaves (approximately 1cm in length) of the greenhouse-grown biparental MxB F₃ generation, using a modified CTAB protocol (adapted from Allen et al. 2006). DNA was quantified with a NanoDrop spectrophotometer and genotyped using the BARCBean6K_3 BeadChip (Song et al. 2015), yielding 5398 initial SNPs. SNPs that were missing or heterozygous in either parent, or identical between the parents, were filtered from further analysis. The remaining SNPs were arranged into a linkage map using the ASMap R package (Taylor et al. 2017). SNPs that did not map to one of the 11 major linkage groups were removed, leaving 1794 SNPs for QTL mapping. QTL mapping was conducted using the expectation maximization method (Lander and Botstein 1989) in R/qtl (Broman et al. 2003). Phenotypes for QTL mapping were generated by harvesting all the pods from each greenhouse-grown F₃ plant (mean n=27 pods/plant), then subjecting them to seven days at 65°C and seven further days of re-equilibration to room temperature. Pods that had fractured to the tip of the beak due to this treatment were counted as shattered, while those with no opening or only fissuring along the sutures were considered non-shattering. The percentage of pods that shattered in this treatment were used for QTL mapping. The maximum LOD score of 1000 randomized analyses of the data was used as a significance threshold. To test dominance, F₃ individuals were subset by genotype at highly significant SNPs, and comparisons were made between groups by t-test.

Next, the 43 SNPs within 100kb of PvPdh1 in the MDP data set were analyzed to identify patterns of selection and diversity around the gene. To simplify and visualize the data, principal component analysis was performed on the SNPs using R. Sequence variation was converted to integer values and the imputePCA() function of the missMDA package was used to impute missing data (Josse and Husson 2016). The genotype data were also sorted to identify unique haplotypes within the populations. The degree of similarity between the PCA and haplotype diversity was then compared. Individuals with missing data for SNPs distinguishing the haplotypes or haplotype clusters were not shown in plots, and not numbered in plots as they could not be unambiguously placed within any haplotype group.

The SNPs tightly linked to PvPdh1 in the MDP data set were then screened for other positions that could be useful for conversion to additional CAPS markers. The SNP closest to PvPdh1 in this data set, at Pv03 position 49,132,438 (accession G19833 genome v2.1), is distinguishable by EcoRI and is highly correlated with pod shattering. Unlike the TaqII-based CAPS marker, the allele cleaved by EcoRI is the shattering-resistant variant, reducing the risk of falsely identifying a susceptible individual as resistant due to technical errors in digestion. The SNP distinguished by EcoRI is separated from the PvPdh1 causal polymorphism by less than 7kb.