Biotic interactions promote local adaptation to soil in plants

Although different ecological factors shape adaptative evolution in natural habitats, we know little about how their interactions impact local adaptation. Here we used eight generations of experimental evolution with outcrossing Brassica rapa plants as a model system, in eight treatment groups that varied in soil type, herbivory (with/without aphids), and pollination mode (hand- or bumblebee-pollination), to study how biotic interactions affect local adaptation to soil. First, we show that several plant traits evolved in response to biotic interactions in a soil-specific way. Second, using a reciprocal transplant experiment, we demonstrate that significant local adaptation to soil-type evolved in the “number of open flowers”, a trait used as a fitness proxy, but only in plants that evolved with herbivory and bee pollination. Whole genome re-sequencing of experimental lines revealed that biotic interactions caused a 10-fold increase in the number of SNPs across the genome with significant allele frequency change, and that alleles with opposite allele frequency change in different soil types (antagonistic pleiotropy) were most common in plants with an evolutionary history of herbivory and bee pollination. Our results demonstrate that the interaction with mutualists and antagonists can facilitate local adaptation to soil type through antagonistic pleiotropy.

Supplementary Figure 1: Design of our experimental evolution study.98 full-sib seed families of fast cycling Brassica rapa plants were divided into two replicates (A and B, 49 plants per replicate), shown in the blue-and green shaded squares, and spread among treatment groups, with each treatment containing all seed families.Soil: limestone (blue), tuff (green), herbivory (no aphids, aphids) and pollination (hand, bumblebee) were the factors in the experiment (1).Supplementary Figure 2: Attractiveness of plants measured by first choices of bumble bees.a: in plants that evolved with hand pollination and no-herbivory (N=232), b: in plants that evolved with hand pollination and herbivory (N=224), c: in plants that evolved with beepollination and no-herbivory (N=232), d: that evolved with bee-pollination and herbivory (N=224).The graph shows plants that evolved in limestone (blue bars) and those that evolved in tuff soil (green bars) grown in either limestone-or tuff soil (as indicated on the x-axis label).In the treatment group bee-pollination and herbivory, local tuff-line plants were significantly more attractive than limestone-line plants when grown in tuff ( * F104=12.48,P=4.120 * 10 -04 ).Significance was determined using a two-sided generalized mixed with number of visits as dependent variable, soil and soil lines and their interaction as fixed factors and replicate as random factors.
Supplementary Table 1: Association between plant traits and relative seed set in plants with bee-pollination in generation one and two (combined, (1)), and seven, estimated by two-sided generalized linear model.Bold: P<0.05, italics: 0.05<P<0.10.Because of the zeroinflated distribution of fitness, fitness was treated as binary variable (seeds/no seeds) in A) and assessed only for plants that produced seeds (truncated model; relative seeds seed>0) in B).Plant traits were used as covariates and replicate as random factor in the models (na: traits were not available).
A)  2), the interaction of soil and soil lines, as well as the interaction of soil x soil lines x biotic treatments, in the reciprocal transplant experiment assessed by a general linear model.Significances were determined using two-sided generalized mixed models using soil, soil lines, biotic treatments, and their interactions were included as fixed factors, and replicate as random factor.Bold indicates statistical significance.G x E interaction is indicated by a significant interaction of the factors soil (i.e. the soil the plants were cultivated in at the time of phenotyping) and soil lines (i.e. the soil the plants evolved in).The biotic treatments pollination and herbivory were combined to represent the analysis shown in Figure 2, where the combined effects of pollination and herbivory had the strongest effects on local adaptation.Traits shown in bold are those were the factors (except replicate) and all interactions are significant, thus, patterns of local adaption differ among biotic treatment groups.

Supplementary Table 6: Trait differences (mean ± SD) among plants of generation one and ten, having evolved with herbivory and bee pollination when growing on both tuff and limestone.
Significant differences with generation one (for plants growing on same soil type) are indicated below each treatment group mean.Significance between treatments was determined using two-sided linear mixed models (LMM) with individual traits as dependent variable, soil, soil lines and their interaction as fixed factor, and replicate as random factors.Bold indicate significant factor effects (P<0.05) and traits that showed a significant G x E interaction where G is represented by the factor "Soil lines" (i.e. the soil the plant evolved in; the plant genotype) and E by "Soil" (the soil the plants were cultivated on at the time of phenotyping; environment); this analysis did not include plants of generation one.When Soil and Soil lines interaction was significant multiple-comparison post hoc tests were run to compare local versus foreign and home versus away contrasts using estimated marginal means (EMMs) and their linear contrasts as fixed factor, and replicate as random factors.Bold indicate significant factor effects (P<0.05) and traits that showed a significant G x E interaction where G is represented by the factor "Soil lines" (i.e. the soil the plant evolved in; the plant genotype) and E by "Soil" (the soil the plants were cultivated on at the time of phenotyping; the environment); this analysis did not include plants of generation one.When Soil and Soil lines interaction was significant multiple-comparison post hoc tests were run to 10 compare local versus foreign and home versus away contrasts using estimated marginal means (EMMs) and their linear contrasts (emmeans package; Statistical differences (P<0.05) are indicated in bold).

Table 2 : Association between plant traits and bumblebee first choices in plants of generation ten with an evolutionary history of bee-pollination.
Significances were determined using two-sided generalized linear model, bumblebee first choices were the dependent variable, traits covariates and replicate a random factor.

Table 4 : Impacts of herbivory (aphid herbivory or no herbivory), pollination (bee-pollination or hand-pollination) on the evolution of traits in plants of generation ten in each soil line separately
(limestone and tuff), estimated by two-sided linear mixed models (LMM).Bold indicate significant factor effects (P<0.05).(+) indicates positive effect (increase) of either bees or herbivory on plant traits evolution whereas (-) indicate a negative effect (decrease).

Table 10 :
Candidate genes associated with SNP markers with an evolutionary pattern of antagonistic pleiotropy.HB: plants that evolved with herbivory (H) and bee-pollination (B).NHB: plants that evolved without herbivory (NH) and bee-pollination (B).