Over the past two decades, elevated public demand for organically produced foods has more than doubled the retail sales of organic food within the United States (USDA ERS, 2024). Consumer interest in health, wellness, and ecological sustainability drives much of this demand (Lee and Yun, 2015). Studies have found that organic crops have higher quantities of antioxidants, lower concentrations of heavy metals, and reduced pesticide residues compared with conventional crops (Barański et al., 2014). Other studies determined that organic agriculture requires fewer non-renewable resources, reduces nonpoint source pollution from agrochemicals, and provides producers with opportunities for higher profit margins (Badgley et al., 2007; Bengtsson et al., 2005; Shreck et al., 2006).

Despite demand, the persistent yield gap between organic and conventional production limits the viability of organic agriculture at scale. Meta-analyses attempting to capture the yield gap found that organic yields average 20-25% lower than conventional (Kniss et al., 2016; Ponisio et al., 2015; Seufert et al., 2012). Though these analyses provide a high-level view of yield gap scale, each study notes high variation across several parameters, including crop type (i.e., cereal, legume, forage, etc.), geographic location, and input intensity of the operations (i.e., high vs. low fertilizer applications). Across all crops with available data, organic fruits and vegetables have the greatest yield gaps, with yields falling 23-62% lower than conventional counterparts (Kniss et al., 2016). This is likely due to the increased impact of insects and pathogens on the visual marketability of these crops, which are frequently sold fresh or direct-to-consumers.

The amount of plant-available nitrogen and the scale of perennial weed pressure are consistently identified as the primary driving forces underlying the yield gap (Knapp and van der Heijden, 2018; Kravchenko et al., 2017). Agronomic solutions to these challenges are under continuous development (Cabrera-Pérez et al., 2023; Reimer et al., 2023; Rowland et al., 2023); yet, incorporating these techniques on-farm may be technically or financially unfeasible for growers. Thus, the breeding and development of new cultivars that are well-suited to organic farm conditions emerge as critical “low-cost, high-impact” solutions to the yield gap.

Unlike conventional crops, organic crops must be able to cope with highly dynamic environments without the buffer of supportive chemical inputs. Plasticity is widely observed in cultivated and wild plant populations, and can be considered a distinct quantitative trait that provides some individuals with the capacity to more rapidly adapt to challenges in their environment. Plasticity may be further characterized as genotype-by-environment (G × E) interactions, wherein the capacity for adaptive plasticity of a given phenotype varies among genotypes (Laitinen and Nikoloski, 2019). While historic plant breeding has attempted to eliminate phenotypic plasticity by controlling the growing environment, the current and future implications of anthropogenic climate change drive demand for cultivars that can exploit adaptive plasticity in unpredictable environmental conditions (Brooker et al., 2022). Similarly, in highly dynamic organic environments, plasticity is a valuable trait that organic breeding strategies should seek to leverage.

The impact of G × E interactions on the performance of a genotype in specific test environments, and the performance of a genotype in the target population environments, underpins the case for developing organic cultivars within organic environments (Allen et al., 1978). When the correlation between performance in a test environment and performance in a target population of environments, referred to as r, is high, the testing environment used is considered typical of the target population of environments (TPEs). When correlation values are low the environments are considered unique in comparison with the TPE. Conventional breeding strategies have historically operated under the understanding that heritability may decrease in stressed environments, thus direct selection should not be carried out under ambient field stress. Instead conventional breeding strategies have relied on a combined method of direct selection for yield under optimal environments and indirect selection of physiological traits under simulated stress conditions (Blum and Jordan, 1985). However, other breeding strategies maintain that heritability is not more variable under stressed conditions, and a breeding approach that optimizes r through direct selection within these environments may also be effective, or preferable in some cases (Allen et al., 1978; Ceccarelli, 1989).

In the context of organic agriculture, several studies have identified improvements in characteristics like marketable weight (i.e., yield), competition against weeds, and disease resistance in cultivars selected under organic management compared to their conventional counterparts. To date, most of this work has materialized within cereal crop breeding programs (Murphy et al., 2007; Schneider et al., 2024; Wolfe et al., 2008), excepting some recent work in potato, carrot, and tomato (Blossei et al., 2024; Colley et al., 2025; Lopez et al., 2024). Other vegetable crops, including common beans, have only been the target of observational studies within organic environments (Ozaktan et al., 2023).

Common bean (Phaseolus vulgaris L.) was domesticated for use as a crop approximately 8,000 years ago in two separate regions: the Andes and Middle America (Wallace et al., 2018). Snap beans are the form of common beans grown for their stringless, low-fiber vegetable pods; a category that includes green beans, wax beans, and flat-podded Romano beans. These may be harvested for both fresh markets and processing markets, with a distinct crop ideotype for each target market. The processing bean ideotype includes an upright growing habit, suitability for mechanical harvest, palatable appearance after processing, appropriate sieve size at harvest, round pods, concentrated pod set, and white seeds. The fresh market ideotype differs in that oval pods are acceptable, the pods may have higher fiber content for straighter and smoother pods, and pubescence to provide cushioning and prevent damage during handling and transport. Colored seeds may also be acceptable in some markets.

Within the Pacific Northwest, and western Oregon specifically, snap bean production thrives as a product of low disease pressure and irrigation capabilities. Cool, wet spring conditions that hamper crop establishment and early season performance are typical for snap bean production in the region. Factors such as darker seed color and thicker seed coat can improve germination within these challenging environments; however, these traits lie in conflict with processing industry preference for white-seeded cultivars. Within conventional systems, snap bean seeds are coated with fungicidal treatments before sowing, and a pre-emergence herbicide is frequently applied. These approaches eliminate the need for highly vigorous seedlings capable of outgrowing early pressures from pathogens, insects, and weeds. Furthermore, conventional farmers rely on synthetic fertilizers throughout the snap bean growth cycle, a practice which has been found to reduce nutrient-seeking growth in the root system (Duncan et al., 1960; Lammerts van Bueren et al., 2011). While historic breeding practices have met the demands of conventional growers, breeding exclusively within conventional, high-N input systems has resulted in the unintentional loss of essential plant functions, such as the inactivation of legume genes responsible for supporting mutualism with nitrogen-fixing soil bacteria (Toby et al., 2007). A loss of function in this mutualism through the breeding process has minor implications in conventional fields where synthetic nitrogen is available, but can be devastating in organic fields that partially rely on nitrogen fixation. This example emphasizes the need to optimize r when developing new cultivars for a target environment, like organic production systems.

Our work was guided by the hypothesis that selection under organic and conventional management leads to divergent phenotypic expression and changes in the associated genetic architecture. To explore this, we split two populations of snap beans into two target environments, organic and conventional, and allowed natural selection in each system to enact pressure on the sub-populations. We subsequently evaluated sub-populations in both systems to test whether breeding under contrasting environments produced genotypes with distinct performance, levels of plasticity, and adaptive traits relevant to each environment. Additionally, we explored segregation distortion and QTL differences to determine whether differential selection drove measurable shifts at the genetic level.

Organic production environments expose crops to distinct stressors that are typically minimized through chemical interventions in conventional systems. Breeding crops that are resilient to these stressors is a predominant goal within organic agriculture; however, the traits associated with resilience in organic systems are difficult to identify and select for due to the high variation in organic growing systems and the complex, quantitative inheritance of these traits. Our work sought to identify specific traits in snap beans that are beneficial for growth in organic production environments, along with finding regions to develop genetic markers that are associated with these traits. To achieve this, we developed four populations of snap beans that allowed us to make comparisons across parentage and breeding history. We anticipated that the organic populations would outperform the conventional populations within organic production systems, and our findings did point to significant crossover interactions in the organic and conventional populations. Furthermore, the narrow-sense heritability values for many traits indicated similar, and in some cases, higher heritability in the organic environments compared to the conventional environments. These values show that breeding within organic environments is a viable and potentially preferable approach to developing varieties that are well-suited to organic production systems.

To explore population variation for germination-related traits, we enacted high abiotic (i.e., cold and wet soils) and biotic (i.e., pathogenic fungi) pressure on seedlings through extraordinarily early planting dates. As anticipated, the production system significantly influenced germination-related traits. In general, total germination was lower within the organic production environments; yet, an exploration of the role of seed color indicated that this gap between production systems was largely due to the white-seeded cultivars ( Figure 1a ). Our planting dates corresponded to suboptimal germination temperatures (i.e., below 13°C), under which common bean seed imbibes moisture more slowly and thus takes longer to germinate (Walls and Stang, 1985). The reduction in germination seen under cold soil temperatures is exacerbated in white-seeded beans, as white seed is often correlated with a higher rate of imbibitional cracking and injury (Cirak and Myers, 2021). Although our white-seeded families had lower germination, those from an organic breeding history had consistently higher germination than those from a conventional breeding history. In addition to the variation in germination percentage, we identified significantly faster germination among the white-seeded families of organic breeding history compared to their conventional counterparts ( Figure 1b ). Plasticity in germination as a response to environmental conditions may explain this increase in speed along with increases in soil temperature that may have resulted from compost application, as others have reported (Deguchi et al., 2009; Donohue et al., 2005). Furthermore, a growing body of evidence suggests the bean seed microbiome may provide plan-growth-promoting properties to the germinating seeds, allowing for faster germination in the presence of pathogenic microbes (Martins et al., 2018; Mokrani et al., 2019). Given the increased abiotic and biotic pressures that are inherent to the early growing season, this acceleration of germination can be interpreted as an advantageous escape strategy. Accelerated imbibition and germination of white-seeded beans are understood to be flaws, which put the germinating seeds at higher risk of chilling injury or pathogen attack; however, our findings suggest that this characterization of white-seeded bean performance is incomplete and additional genetic factors may have a role in early-season resilience. Vidak et al. (2022) found similar results suggesting that while seed coat color has a significant role in bean germination, it does not explain all of the variation in germination and seedling emergence. Thus, we understand that there may be alternative routes to improving germination in white-seeded beans without altering seed color. This is singularly important for the organic snap bean processing market, where white seeds are essential to the product quality of processed beans but can simultaneously contribute to the organic yield gap through reduced germination in the field.

Similar to results for early-season performance, we found that the production system significantly influenced root morphology and disease. Other factors, including breeding history and cross, also significantly influenced root and disease traits. Many of these root traits were minimally explained by study factors, except for root branching density and disease development. Root disease was significantly lower within organically-bred families, compared to conventional counterparts, and this variation was largely explained by differences among the white-seeded families ( Figure 2a ). Faster germination has been suggested as an effective means through which P. vulgaris seedlings may escape and outgrow soilborne pathogenic pressures, particularly those that cause damping-off (i.e., Pythium, Rhizoctonia, etc.) (Li et al., 2015; Peña et al., 2013). Thus, the changes observed in germination speed among white-seeded families from an organic breeding history may partially explain these specific reductions in root disease. Comparing across parentage, the ORBV families had significantly lower root disease compared to the HYPR families ( Figure 2b ). This difference was not surprising, given the disease resistance package of OR5630 and previous research into the root rot susceptibility of the four parental lines, which determined that Black Valentine has high resistance to the root rot complex prevalent in western Oregon (Huster et al., 2021). These findings deviate from other work with the four RIL populations, which identified a trend towards higher pathogen abundance in the rhizosphere soil of ORBV populations; however, these results were collected from a separate field site and pathogen abundance was determined through a metabarcoding approach that was independent of root disease phenotyping (Park et al., 2023). Branching density was significantly higher among the HYPR families, indicative of a denser root system ( Figure 3 ). Organically-bred families also had significantly higher branching density, compared to the conventionally-bred families. Other work indicates that higher lateral root branching density significantly increases the ability of roots to cope with stressors, particularly in phosphorus-limited soils, suggesting that increased branching density may be an advantageous root system adaptation for nutrient-limited organic environments (Postma and Lynch, 2011).

Positive correlations between seed pigmentation and germination percentage were identified across all years and populations, although these correlations were higher among conventionally-bred populations. Seed pigmentation positively correlated with seed weight, a trend that was stronger among organically-bred populations. Seed weight, however, was positively correlated with days to germination and negatively correlated with germination percentage, suggesting that the larger seeded beans were slower to emerge or failed to emerge altogether. These results correspond to our findings that white-seeded organic families germinate faster, as well as other studies that have identified larger, pigmented seeds to be slower to germinate and thus depleted in field emergence (De Ron et al., 2016; Vidak et al., 2022).

Analysis of G × E relationships across traits ( Figure 4 ) confirmed the importance of breeding within a target environment. Days to germination showed significant G × E interactions, with population trends differing by parentage. In HYPR populations, HYPR-C germinated later than HYPR-O in both systems, with a slightly greater difference in the organic system. In ORBV populations, ORBV-C germinated later than ORBV-O in the conventional system, but both populations had similar germination times in the organic system. No interactions were observed for germination percentage, but germination was consistently lower in the organic systems. This trend is likely explained, in part, by the presence or absence of the fungicide seed treatment, which is intended to reduce seedling mortality due to biotic pressures. Notably, we identified a crossover interaction for seedling vigor, with organically-bred populations performing better in organic systems and conventionally-bred populations performing better in conventional systems. This interaction was stable across both parentage groups (ORBV and HYPR), suggesting that breeding within an organic system can enhance seedling vigor of progeny grown under organic conditions. We identified a crossover interaction in ORBV populations for root rot disease severity (higher in ORBV-O in the conventional system but lower in the organic system). This interaction was not observed in the HYPR populations, where HYPR-O had consistently higher disease severity. The interaction in the ORBV populations suggests organically bred families may have developed resistance mechanisms specific to organic environments, whereas the HYPR populations’ inconsistent response raises questions about the stability of this interaction. Previous research into the root rot susceptibility of the Common Bean Coordinated Agricultural Project (BeanCAP) Snap Bean Diversity Panel classified both OR5630 and Black Valentine as highly resistant to the western Oregon root rot complex, which may partially explain the variation observed in G × E interactions (Huster et al., 2021). Other root traits showed less straightforward patterns, suggesting complex interactions between genotype and soil conditions. These findings are not unexpected, as root traits frequently show high plasticity in field environments (Strock et al., 2019; Vieira et al., 2008). This plasticity may itself be advantageous, as suggested by Burridge et al. (2016) and Zhu et al. (2010), but the trends observed in this study do not point to any consistent genotypic variation. Although our sampling protocol sought to characterize true variation in field-grown populations, the use of more controlled assays, such as in vitro germination tests, may result in clearer population trends. The significance of the production system across nearly all traits confirms that environmental factors tied to organic or conventional management can strongly influence phenotypes. These results emphasize the importance of breeding within organic environments to bolster adaptation, particularly for traits showing strong environmental dependencies (i.e., vigor, disease, root architecture). The observed crossover interactions indicate potential trade-offs between conventional and organic selection, underscoring the need for breeders to strategically prioritize traits based on the specific demands of each production system.

Heritability analysis in early-season and root system traits indicated consistent heritability values across production systems. For certain traits, namely germination percentage, seedling vigor, root diameters, and disease rating, the heritability was higher in the organic production system than in the conventional production system. These findings, in conjunction with improved performance of organically-bred populations within the organic environment, support the theory of optimizing the correlation of genotype performance in specific test environments and in the target population of environments within breeding programs.

The genetic linkage maps generated for this work had between 285 and 398 markers, and a coverage ranging from 813.26 to 991.36 cM, which correspond to 78 and 95 percent coverage, respectively, of the Stampede x Red Hawk map (Song et al., 2015). The final marker number was an improvement over initial maps made for the populations, described by King (2019); however, the final maps had fewer markers than those generated for other mapping populations that utilized the BARCBean6K BeadChip or the BARCBean12K BeadChip (Bassett et al., 2021; Heilig et al., 2017). We attribute this primarily to the relatively small population sizes, approximately 90 families within each population. Furthermore, we anticipate that the low marker number in the HYPR maps was a result of the parental background, as both parents originate from the Andean gene pool (Wallace et al., 2018). One of the parents of the ORBV populations, ‘OR5630’, is derived from both Andean and Middle American backgrounds. The other parent, ‘Black Valentine’, originates from the Andean gene pool. The ORBV populations had greater genetic diversity amongst the loci covered, likely due to these differences in the gene pools of the parents (Wallace et al., 2018). Quantitative trait loci (QTL) analysis was conducted to identify chromosomal regions linked to important traits for Phaseolus vulgaris, and many of these loci aligned with regions found in other studies ( Supplementary Table 11 ). QTL with higher values of phenotypic variation (R²) suggest the effect of additive genes in the control of traits. Many QTL were near or overlapping regions noted in the Arriagada et al. (2022) and Izquierdo et al. (2023) meta-QTL analyses for important bean yield traits, including 100-seed weight on LGs 2, 3, 5, and 7; days to flowering on LGs 1, 3, 4, 5, and 11; and seed yield on LGs 1, 6, and 7. Additionally, the QTL we found for 100-seed weight on Pv07 is consistent with several other studies (González et al., 2017; Koinange et al., 1996; Pérez-Vega et al., 2010). A QTL we identified for root disease on LG 3 contained a significant SNP (12,661,037 bp) that Huster et al. (2021) identified for the same trait in a genome-wide association study (GWAS). Hagerty et al. (2015) found a QTL linked to Fusarium solani resistance that was also located on LG 3. We identified several QTL linked to root diameter on LG 7, similar to López-Marín et al. (2009).

There were several notable trends for the quantitative trait loci found for these populations. A number of the identified QTL were stable over two or more years and in multiple growing systems. These regions included the following: days to germination on LG 6_1; germination percentage on LG 6_1, and LG 7; vigor on LG 7; 100-seed weight on LG 2, LG 5, and LG 7. The ORBV and HYPR populations had significant QTL in at least two years on LG 7 for germination percentage, vigor, and 100-seed weight. One-hundred seed weight on LG 7 is consistent with González et al. (2017); Koinange et al. (1996), and Pérez-Vega et al. (2010). Variation in seed related traits is highly influenced by seed color and important seed proteins (Johnson et al., 1996). Both the P locus, which has been linked to seed pigmentation in beans, and phaseolin, one of the most important seed proteins in beans, have been found on LG 7 (Johnson et al., 1996; Koinange et al., 1996; McClean et al., 2018; Moghaddam et al., 2014). One of the two stable regions found for 100 seed weight on LG 7, with a peak in the 32.3 Mb region, was within a meta-QTL identified by Izquierdo et al. (2023) that contained a serine/threonine-protein kinase (STK) gene (Phvul.007G174900). STK genes have been linked to flowering and seed weight in rice, and grain yield in maize (Deng et al., 2017; Hu et al., 2012; Jia et al., 2020). In the same study, this same region on LG 6_1 corresponded with a meta-QTL noted to contain a receptor associated kinase (RAK) (Phvul.006G020700) that was associated with seed yield, seed weight, and days to physiological maturity. These two stable regions on LG 6_1 and LG 7 also overlap with meta-QTL in Izquierdo et al. (2023) that were noted to be correlated with performance in drought conditions. The stable QTL for 100 seed weight found on LG 2 and LG 7 were in the same regions as multiple meta-QTL found by Arriagada et al. (2022). Many of the QTL for important traits were in similar regions for both the organically and conventionally bred populations; however, there were exceptions and QTL for several traits were unique to a single breeding history for a particular cross. For example, we identified a marker related to days to 50% flowering in the ORBV-O population on LG 3, but found QTL for the ORBV-C population on LG 1_2 and LG 8_1. Deviations across breeding history such as these were also found for the days to 50% flowering trait in the HYPR populations and for markers related to seed weight per plant. Some populations had significant QTL in different regions depending on the growing system. For example, the HYPR-C population had a stable QTL for vigor on LG 7 in both organic and conventional systems in 2020 and 2021, but there were unique QTL on LG 4 and LG 6–1 that were only significant in the organic growing system in 2020. Similarly, in 2020, there was a significant QTL on LG 7 in both production systems related to days to germination, and a QTL on LG 8 in both systems in 2021. Yet, in both 2020 and 2021, ORBV-O had a unique QTL in the conventional system – one on LG 4 in 2020 and one on LG 2_1 in 2021. For some traits, the populations only had significant QTL in one growing system. For example, there were only significant QTL in the conventional growing system for ‘100 seed weight’, and in the organic system for ‘Basal Root Diameter 1 and 2.’ Environment-specific QTL and distortion are consistent with studies on other crops, some of which suggest that the same traits selected in different environments may actually be different traits resulting from certain genes, or gene networks, activating or inactivating due to environmental signals (Ceccarelli and Grando, 2007; Lin and Togashi, 2002; Pswarayi et al., 2008).

Preliminary work with our data looking at broad-scale genomic differences suggested that populations from the same cross are quite similar at a broad level. Following this, we investigated segregation distortion to determine whether certain chromosomal regions showed differential selection depending on breeding history, and whether there were regions distorted in both crosses. Segregation distortion is common in many plants, including Phaseolus vulgaris, and may be linked to incompatibility genes or genes related to sporophytic selection (Blair et al., 2003; Córdoba et al., 2010; Wang et al., 2005). Incompatibility between gene pools is common in beans and may explain some of these selection pressures, although distortion was higher in the HYPR populations. Notably, the HYPR population parents are from more similar gene pools than the parents of the ORBV populations.

In both crosses we utilized, there were significant differences in the SNPs and regions that were distorted in each system ( Figures 5A and 5B; Supplementary Tables 11 and Supplementary Table 12 ). Among all sampled bi-allelic SNPs, 9% were skewed in ORBV-O, 8% in ORBV-C, 11% in HYPR-O, and 15% in HYPR-C ( Supplementary Table 12 ). Differences between populations from the same cross were compared by linkage group in Supplementary Table 12 and noted in results. These differences were likely caused by selection in the respective systems, as they came from the same F₁ populations. There were a few notable regions that had high levels of segregation distortion in all populations. There was a region of high segregation distortion on LG 2 (47–49 Mbp) in all populations, with the exception of ORBV-C, which had some distorted SNPs near this region. In the HYPR populations, there was a strong skew towards the maternal allele in both populations – with a stronger skew in the organic population. In contrast, the paternal allele was favored in this region in the ORBV populations. There were no significant QTL found in this region on LG 2; however, there was a QTL found for root disease in ORBV-C in the organic growing system in 2021 that was skewed towards the maternal allele. All populations had many distorted SNPs that skewed strongly towards the paternal allele on LG 7, along with many distorted SNPs skewed towards the maternal allele on LG 10.

To our knowledge, this work comprises the first combined look at the role of parentage and breeding history on bean performance in organic production systems. Our findings suggest that breeding within organic environments is an effective, and in some cases preferable, approach for harnessing qualitative fitness. We found significant crossover G × E interactions in seedling vigor and root rot disease, indicating breeding under organic management can significantly improve progeny performance in those environments. We also identified important shifts in germination speed and rate within the white-seeded families of organically-bred populations, along with overlapping QTL on linkage groups 4 and 7 for these traits. These QTL may serve as a basis for marker-assisted selection for early-season performance in future organic bean breeding efforts. Although our work identified significant variation within the white-seeded populations, a specific explanation for the higher germination rate in the organically-bred families remains elusive. Further explorations of the seed coat morphology in the white-seeded families of these populations may provide insight into these changes, and how they may best be leveraged in subsequent breeding work. While we were able to improve the linkage maps for our study populations compared to preliminary maps generated by King (2019), the final marker number in the maps remained low compared to similar studies (Bassett et al., 2021; Heilig et al., 2017). This may be attributed, in part, to the relatively small population size of approximately 90 families. Furthermore, we anticipate the low marker number in the HYPR populations is influenced by the genetic relatedness of the parents, which both originate primarily from the Andean background. Vargas et al. (2021) encountered similar issues when SNP genotyping populations generated from parents of the Andean gene pool, although our final marker numbers were higher than their final maps. Given our small population sizes, our work may be improved by using a genotyping-by-sequencing (GBS) approach to generate maps with higher marker numbers.

Broadly, our results suggest that enhanced performance in organic production systems may be captured through breeding and selection within organic production environments. This work led to the identification of specific traits relevant to establishing vigorous and adaptable plants for organic systems. Furthermore, we identified genetic markers associated with these traits, comprising a toolbox of loci that are useful for organic bean breeding. If breeders are able to improve and release varieties that are better suited to organic production, it is reasonable to expect declines in the organic-conventional yield gap. This work has critical implications for the organic industry, as seed companies seek to improve the cultivars that are available for organic production and farmers seek suitable varieties that are competitive in organic environments.