Results

After filtering, a subset of 1225 SNPs across 18 sources and 69 total individuals were utilized for genetic analysis within the ‘Jacob’s Cattle’ populations only. For analyses that included ‘Cal early’ check and University of Minnesota accessions, 9093 SNPs were used.

The number of SNPs within individual ‘Jacob’s Cattle’ sources ranged from 1 to 93. In general, the number of SNPs was substantially less within sources compared to the pooled ‘Jacob’s Cattle’ data (Table 2). Across all five ‘Cal Early’ sources, 624 SNPs remained after filtering, and within-source SNP counts ranged from 1 to 111.  SNP counts within the four single-source heirloom varieties were overall lower than ‘Jacob’s Cattle’ or ‘Cal Early’, with the exception of ‘Peregion’. 72 SNPs were identified within ‘Lina Sisco’s Bird Egg’, 163 within ‘Jacob’s Cattle Gold’, 62 within ‘Tiger’s Eye’ and 2119 within ‘Peregion’. Substantially higher SNP counts within ‘Peregion’ are likely due to its identification as a seed mixture with two distinct seed phenotypes (Swegarden, 2015).

Table 2. Total SNP counts within varieties and seed sources. 18 distinct seed sources of ‘Jacob’s Cattle’, 5 distinct sources of ‘Cal Early’ light red kidney, and single sources of ‘Lina Sisco’s Bird Egg’, ‘Tiger’s Eye’, and ‘Peregion’ were included in the analysis. SNPs were identified via Genotyping-by-Sequencing (GBS). Only individuals with multiple individuals genotyped are included in this table.

A Principal Components Analysis (PCA) of ‘Jacob’s Cattle’ with ‘Cal Early’ and other heirloom beans showed differentiation along Principal Component 1 (PC1), with two identifiable clusters. From the relative positions of known check varieties, genotypes associated with higher mean PC1 values can be assigned to the Middle American gene pool, while Andean genotypes are associated with lower PC1 values. PC1 accounted for 5% of total variance (Figure 1). PC2 primarily differentiated between sub-groups within Andean and Middle American gene pools.  Within the Middle American cluster, ‘Peregion’ clusters more closely with ‘Eclipse’ than ‘Lariat’, indicating that ‘Peregion’ likely belongs to the Mesoamerican race.  ‘Jacob’s Cattle’ shows greater overall differentiation between seed sources when compared to the five ‘Cal Early’ sources.  Three out of four single-source heirloom varieties, ‘Tiger’s Eye’, ‘Lina Sisco’s Bird Egg’, and ‘Jacob’s Cattle Gold’, show little differentiation between individuals, indicating low diversity within single seed sources. The exception is ‘Peregion’, which is a Middle American variety that has been previously characterized as a mixture, and would therefore likely demonstrate higher diversity within source (Swegarden, 2015).  The contextualization of ‘Jacob’s Cattle’ seed sources within a PCA of other traditional bean varieties and market class checks demonstrates the scale of divergence between ‘Jacob’s Cattle’ sources, with evidence of greater intra-varietal differentiation between some ‘Jacob’s Cattle’ sources compared to differentiation between distinct varieties.

Figures 1 a-b. Scatter plot of PC1 and PC2 values for Principal Component Analysis (PCA) of a) ‘Jacob’s Cattle’ (18 sources), ‘Cal Early’ (5 sources), 4 single-source heirloom varieties, and two market class checks, black and pinto, and b) ‘Jacob’s Cattle’ seed sources only. Color of data point indicates seed source origin. PCA of 18 JC seed sources were obtained using GBS data analyzed in TASSEL5 (Glaubitz et al. 2014)

a)

b)

A second PCA of ‘Jacob’s Cattle’ seed sources further demonstrates differentiation within ‘Jacob’s Cattle’ germplasm (Figure 1b).  Fourteen seed sources tightly cluster in the lower left-hand corner of the plot, indicating little genetic differentiation.  Four other sub-populations, however, show significant divergence between sources and a low level of divergence within source.  A single seed source, 712, shows significant divergence across the PC1 axis which comprises 31% of total variance in the analysis.

Discriminant Analysis of Principal Components (DAPC) analysis resulted k=5 genetic clusters. Groups 1, 2, 4 and 5 each consisted of a single seed source comprised of four individual genotypes. Group 3 consisted of 14 seed sources comprised of 53 individual genotypes collectively.  Genetic cluster assignments were subsequently used to explore significant differences between field phenotypes and visualize SNP distributions.

Figure 2. Results from Discriminant Analysis of Principal Components (DAPC) indicate k=5 clusters across 18 seed sources using Bayesian Information Criterion (BIC) to assess goodness of fit. Results are displayed in a bar chart of posterior membership probability for each seed source across k=5 genetic clusters. 4 individual genotypes for each seed source were included, except for source 706, for which only one individual remained after filtering.

Distribution of minor alleles of individual seed sources, grouped by genetic cluster and plotted by physical location, show the presence of large minor allele haplotype blocks present in some genetic clusters, but not others (Figure 3a). In particular, clusters 4 and 5, which consist of seed sources 712 and 703 respectively, show particular haplotype divergence from the majority of seed sources represented by cluster 3. Clusters 1 and 2 also demonstrate some areas of divergent haplotypes, though blocks are less frequent and smaller in size. In the plot comparing minor allele frequency of five ‘Jacob’s Cattle’ clusters to heirloom and market class checks, minor allele haplotypes in clusters 4 and 5, which consist of sources 712 and 703 respectively, show several regions of similarity to ‘Anasazi’, black and pinto genotypes, which are of Middle American origin (Figure 3b). Minor allele distributions suggest that two ‘Jacob’s Cattle’ seed sources, cluster 4 and 5, likely outcrossed with a Middle American genotype. Cluster 4 displayed several large haplotype blocks shared with Middle American genotypes, particularly on chromosomes 2, 3 and 8. Cluster 5 displayed a large haplotype block on chromosome 6, with smaller regions homologous to Middle American checks on chromosomes 1, 2, 7, 10 and 11. These regions of shared haplotypes likely explain divergence of sources 712 and 703 seen in the PCA of ‘Jacob’s Cattle’ seed sources (Figure 1b). 

Figures 3a-b. Distribution of minor allele genotype at SNPs identified across chromosomes. Homozygous calls are displayed in blue and heterozygous calls are displayed in red, for: a) all Jacob’s Cattle seed sources grouped by genetic cluster along y-axis; b) representative genotypes from each of five Jacob’s Cattle clusters, heirloom and market class checks across all chromosomes. ‘Lina Sisco’s BE’ represents ‘Lina Sisco’s Bird Egg’, ‘Cal Early LRK’ represents ‘Cal Early’ light red kidney. Exact chromosome lengths vary, for individual chromosome lengths see Phaseolus vulgaris genome v2.1 (DOE-JGI and USDA-NIFA).

a)

b)

Field phenotypes demonstrated significant differences between k-means clustered group assignments for three out of five measured traits as determined using Tukey’s HSD (α=0.05).  In particular, clusters 4 and 5 had significantly smaller mean 100-seed weight than the three other clusters. Cluster 4 also had significantly longer mean stem length and higher mean pod number per plant (Figure 4). No significant differences were observed for mean plant yield. Diversity in seed shape, color and pattern were also noted, though these were not measured quantitatively (Figure 5). 

Figure 4.  Boxplots of phenotypic trait values of 18 ‘Jacob’s Cattle’ seed sources for 100-seed weight in grams (top left), plant yield in grams (top right), pods per plant (bottom left) and stem length in centimeters (bottom right), grouped by five assigned genetic clusters. Clusters denoted with the same letter are not significantly different as determined using Tukey’s HSD (α=0.05).

Figure 5. Seed phenotypes of each of the 18 seed sources of ‘Jacob’s Cattle’ included in the experiment, labelled by source number in bottom right corner of each panel. Differences in seed phenotypes can be observed.

A significant relationship between mean trait value for 100-seed weight and PC1 values for all genotypes was identified through regression, resulting in an R² value of .60 and a p-value of 1.73 x 10^-13.  This further supports the hypothesis that genetic divergence is a driver in observed diversity of measured phenotypic traits across ‘Jacob’s Cattle’ seed sources, and suggests that genetic divergence occurred due to historical introduction of Middle American genetic material into one or more ‘Jacob’s Cattle’ seed sources. This admixture could be the causal factor for phenotypic indicators of Middle American origin, including smaller seed size, higher pod number and greater plant height for cluster 4, which is represented by data points with highest PC1 values (Figure 6) (Singh et al., 1991).

Figure 6. PC1 values for all 18 ‘Jacob’s Cattle’ seed sources plotted against measurements for 100-seed weight phenotypic trait.

Overall, results indicate that observed genetic diversity is correlated to phenotypic diversity. Most notable were differences in 100-seed weight, pod number and stem length, and a high correlation between mean 100-seed weight and PC1 values. Along with the presence of large haplotype blocks distributed across the eleven chromosomes that appears to account for the majority of SNPs present, these findings indicate that observed genetic and phenotypic diversity within some seed sources is likely to be the result of outcrossing. Furthermore, it appears that some instances of outcrossing occurred with genotypes of Middle American races highly unrelated to ‘Jacob’s Cattle’, which is of Andean origin.

Discussion

Polymorphism within ‘Jacob’s Cattle’ was largely observed between seed sources rather than within seed sources, indicating that genetic bottlenecks within networks of seed distribution (i.e. the exchange of very small amounts of seed), are likely commonplace.  The 14 seed sources assigned to cluster 3 appeared to be highly related, which may suggest a relatively recent common seed source, or simply an absence of genetic divergence after seed sources were distributed. This cluster may offer limited potential for selection or adaptation when faced with contrasting environments. However, seed sources 712 (‘Jacob’s Cattle Amish’), 714 (‘Mammoth Trout’), 703 (‘Wink’s Jacob’s Cattle’) and 718 (‘Jacob’s Cattle’) demonstrated significant differentiation from each other and all other seed sources, forming distinct genetic clusters in the DAPC model.  Sources 703 and 712 in particular show evidence of one or more historical outcross events with a Middle American genotype. While these sources still largely cluster with Andean genotypes in a PCA (Figure 1), several minor allele haplotype blocks that are shared between these sources and Middle American genotypes (Figure 3b) indicates that such a cross likely occurred. This outcrossing event may have been followed by seed saver selection back towards a ‘Jacob’s Cattle’-like phenotype that could explain the relatively low frequency of Middle American haplotypes in the current genotypes, as segregating material with Middle American traits were removed from the gene pool.

It should be noted that three out of four of the divergent seed sources were named distinctly, indicating an understanding that they represent a unique strain of the variety. However, other uniquely named strains ‘Coach Dog’, ‘Dutch Cattle’, ‘Deep Red Trout’, and ‘Jacob’s Cattle Gasless’ clustered with the majority of other seed sources.  This indicates that naming of a variety or strain may or not capture meaningful strain diversity at the genetic level.  Two of the four divergent sources were obtained from the Seed Saver’s Exchange network, the third from a regional farm-based seed company, and the fourth from a commercial farm in Maine.  The two accessions procured from the NPGS-GRIN collection clustered with the majority of other sources in cluster 3.  This finding that ex-situ germplasm conservation resources do not capture the extent of intra-varietal diversity maintained within In-situ seed saver networks is supported by similar comparisons in European seed systems (Negri and Tiranti, 2010; Enjalbert et al., 2011).  ‘Jacob’s Cattle’ sources also demonstrated significantly higher diversity between sources than did the commercial kidney variety, indicating that more centralized seed production systems typical for elite cultivars may not generate as much diversity with varieties, perhaps due to stricter controls on varietal purity and identity preservation (AOSCA, 2021).

More broadly, this case study indicates that, for traditional varieties being cultivated within decentralized seed saving networks, meaningful intra-varietal diversity is likely to be found at both genetic and phenotypic levels, even in autogamous crops such as common bean. However, local adaptation via allele and haplotype frequency change is likely contingent upon either an initial population with sufficient genetic variance, or the occurrence of an outcrossing event to generate such variance. Interestingly, past findings indicate that outcrossing in autogamous crops such as common bean increases when presented with a stressful environment, perhaps one evolutionary mechanism to initiate short-term crop adaptation (Klaedtke et al., 2017). Though beyond the scope of this study, it is important to note that phenotypic divergence within crop varieties in response to environmental conditions can also occur at the heritable epigenetic level, changes that in some cases contribute to local adaptation (Galloway, 2005).

Within the context of institutional plant breeding efforts in the Global North, germplasm maintained within decentralized stewardship networks comprised of home gardeners, farmers and freelance plant breeders is rarely recognized as a significant source of genetic diversity or adaptive traits, as for instance, compared to formal germplasm repositories (Almekinders 2000; Galluzzi et al. 2010). In part, strict commodity market class standards such as canning quality may play a role in disincentivizing use of novel germplasm in formal breeding programs (Kelly, 2010). In contrast, freelance plant breeders operate primarily in organic or low-input farming systems and emphasize distinct breeding goals such as adaptation to local environment and novel flavor and culinary quality (Deppe, 2021). Freelance plant breeders have made extensive use of seed saver networks such as Seed Saver’s Exchange as well as regional farm-based seed companies in sourcing germplasm (Deppe, 2021).