Mutational load and adaptive variation are shaped by climate and species range dynamics in Vitis arizonica

Abstract

Genetic load can reduce fitness and hinder adaptation. While its genetic underpinnings are well established, the influence of environmental variation on genetic load is less well characterized, as is the relationship between genetic load and putatively adaptive genetic variation. This study examines the interplay among climate, species range dynamics, adaptive variation, and mutational load - a genomic measure of genetic load - in Vitis arizonica, a wild grape native to the American Southwest. We estimated mutational load and identified climate-associated adaptive genetic variants in 162 individuals across the species' range. Using a random forest model, we analyzed the relationship between mutational load, climate, and range shifts. Our findings linked mutational load to climatic variation, historical dispersion, and heterozygosity. Populations at the leading edge of range expansion harbored higher load and fewer putatively adaptive alleles associated with climate. Climate projections suggest that V. arizonica will expand its range by the end of the century, accompanied by a slight increase in mutational load at the population level. This study advances understanding of how environmental and geographic factors shape genetic load and adaptation, highlighting the need to integrate deleterious variation into broader models of species response to climate change.

Keywords: adaptation; climate change; crop wild relatives; grapes; mutational load; species distribution models; species' range.

Fig. 1

Site frequency spectra and mutational load in Vitis arizonica. (a) The derived allele frequency for three SNP classes: dSNPs (deleterious), nSNPs (nonsynonymous) and sSNPs (synonymous). (b) The species distribution model in the present (green) based on WorldClim data that summarize bioclimatic averages from 1970 to 2000 and GBIF species occurrence data. The points represent sampling locations for individuals used in genetic analyses and are colored according to the load_M estimate per individual calculated with nonsynonymous variants. The black asterisk indicates the geographic centroid of the predicted range.

Fig. 2

Sample dispersal and feature correlations. (a) Projected species distribution models (SDM) for Vitis arizonica during the Last Glacial Maximum (LGM, c. 22 Kya, brown), for the present (green), and for the overlap between the two SDMs (greenish brown). The dark brown region within the present‐day Gulf of Mexico represents areas where the SDM for the LGM overlaps with land that was exposed due to lower sea levels. Each point on the landscape represents the sampling location for an individual and is colored according to predicted dispersal from the LGM to the present, with more distantly dispersed individuals in warmer colors. The asterisk indicates the geographic centroid in the present, while the X indicates geographic centroid during the LGM. (b) Spearman's correlations between pairs of features used in the random forest (RF) model, as well as between load_M and each individual feature. In addition to the 19 WorldClim bioclimatic variables, the features include distance from the present‐day SDM edge (dist. edge), observed heterozygosity (H _o), the estimated dispersal distance shown in Panel a (dispersal), the distance from the present‐day geographic centroid (dist. geo), and the distance from the present‐day niche centroid (dist. niche). Only correlation tests with P < 0.05 are filled, with the color demonstrating both the magnitude and direction of each correlation (ρ). See also Supporting Information Fig. S8.

Fig. 3

Random forest (RF) regression models to predict load_M. (a) The performance of the RF model compares the predictions (y‐axis) to the observed (x‐axis) values. The red line indicates the linear model fit, and the slope (R ² = 0.61) was highly significant (P = 1.8 × 10⁻⁹). (b) The inferred feature importance used to predict load_M. The features are ranked by their inferred importance. The distance‐related metrics are defined in Table 1; bio8 is the mean temperature in the wettest quarter, and definitions for the remaining bioclimatic variables are provided in Supporting Information Table S2.

Fig. 4

Predicted species distribution models (SDM) and load_M in 2100. (a) Predicted SDMs for Vitis arizonica in 2100 under Shared Socioeconomic Pathway (SSPs) SSP126 (a sustainability‐focused scenario with global warming limited to < 2°C above preindustrial levels) are shown for four Earth Systems Models (ESMs) – IPSL‐CM6A (red), MPI‐ESM1 (blue), MRI‐ESM2 (gold), and UKESM1‐0 (purple) – compared to the present‐day SDM (green). In each map, the predicted future geographic centroid is denoted by a plus (+), while the present geographic centroid is represented by an asterisk (*). Predicted SDMs for additional SSPs are shown in Supporting Information Fig. S10. (b) The predicted area of each SDM in 2100 compared with the present (left), shown for all combinations of ESM and SSP. The height of each bar indicates the total predicted area subdivided by new areas that were not part of the present SDM (top), overlapping area between the predicted future and present SDM (middle), and the area of the present range predicted to be lost (green outline, negative values). (c) The distribution of load_M in the present (left) compared to the predicted distributions in 2100 for each combination of ESM and SSP, as projected by the random forest model. In each beeswarm, the diamond represents the median values per group.

Fig. 5

Adaptive genotypes on the landscape. (a) As in Fig. 1(b), green represents the species distribution model for V. arizonica in the present‐day based on WorldClim (bioclimatic averages from 1970 to 2000) and GBIF species occurrence data, with the black asterisk indicating the geographic centroid of the predicted range. The points represent sampling locations for individuals used in genetic analyses and are colored according to the number of adaptive genotypes (N _G) estimated per individual, with cooler colors reflecting lower N _G. (b) The relationship between observed load_M and N _G across all individuals. (c) The relationship between load_M and genetic offsets. (d) The relationship between N _G and genetic offsets. In (b–d), red lines indicate the fit of linear models; in each case, the slope is highly significant, as reflected by the P‐value and R ².