Feral populations of Brassica oleracea along Atlantic coasts in western Europe

Abstract

There has been growing emphasis on the role that crop wild relatives might play in supporting highly selected agriculturally valuable species in the face of climate change. In species that were domesticated many thousands of years ago, distinguishing wild populations from escaped feral forms can be challenging, but reintroducing variation from either source could supplement current cultivated forms. For economically important cabbages (Brassicaceae: Brassica oleracea), "wild" populations occur throughout Europe but little is known about their genetic variation or potential as resources for breeding more resilient crop varieties. The main aim of this study was to characterize the population structure of geographically isolated wild cabbage populations along the coasts of the UK and Spain, including the Atlantic range edges. Double-digest restriction-site-associated DNA sequencing was used to sample individual cabbage genomes, assess the similarity of plants from 20 populations, and explore environment-genotype associations across varying climatic conditions. Interestingly, there were no indications of isolation by distance; several geographically close populations were genetically more distinct from each other than to distant populations. Furthermore, several distant populations shared genetic ancestry, which could indicate that they were established by escapees of similar source cultivars. However, there were signals of local adaptation to different environments, including a possible relationship between genetic diversity and soil pH. Overall, these results highlight wild cabbages in the Atlantic region as an important genetic resource worthy of further research into their relationship with existing crop varieties.

Keywords: Brassica oleracea; crop wild relatives; domestication; environment–genotype associations; feral populations; isolation by distance.

Figure 1

Population structuring of wild populations of Brassica oleracea. (i) Location of the populations considered here. (ii) Clustering of samples from RAxML (v8.2; GTRCAT model and 1,000 maximum‐likelihood bootstrap replicates), visualized in SplitsTree4. (iii) STRUCTURE plot illustrating shared genetic ancestry for K = 12, ordered by population: 1—Fortrose; 2—Auchmithie; 3—Crail; 4—Kildonan Castle; 5—Tynemouth; 6—Whitby; 7—Tenby; 8—Llantwit Major; 9—St. Aldhelm's Head; 10—West Looe; 11—Fowey; 12—Prussia Cova; 13—Cudillero; 14—Playa de Xágo; 15—Cabo de Peñas; 16—Playas de Viodo; 17—Playa Pedrero; 18—San Juan de Gaztelugatxe; 19—Getaría; 20—San Sebastian. Across the figures the same colors and numbering are used for each population. The dashed lines and letters indicate some clustering: (a) populations in Scotland; (b) populations closest to the Welsh‐English border; and (c) populations in the Basque Country, Spain (excluding San Sebastian)

Figure 2

The distribution of sampled populations in relation to various climate variables: (a) annual mean temperature (°C); (b) mean annual precipitation (mm); (c) mean temperature of wettest quarter (°C); and (d) precipitation seasonality (mm). These are averages between 1970 and 2000 obtained from the WorldClim database (Fick & Hijmans, 2017)

Figure 3

(i) Redundancy analysis (RDA) ordination plot of the association between Brassica oleracea individuals (colored points) and SNPs (dark gray points), with environmental variables. The different colors indicate which population each individual was from. (ii) RDA ordination plot of the SNPs alone, colored for the environmental variable with which they were most strongly associated. For both (i) & (ii), the arrows indicate the environmental predictors and the strength of the association

Figure 4

The relationship between expected heterozygosity and soil pH for 21 individuals from four soil pH classes categorized into: Neutral (6.6–7.3), Slightly acidic (6.1–6.5), Moderately acidic (5.6–6.0), and Strongly acidic (5.0–5.5) based on USDA (1998). A linear model was used to fit a regression line (dashed black line), and the standard error is shown in gray, p‐value > 0.05