Pollinator and habitat-mediated selection as potential contributors to ecological speciation in two closely related species

Abstract In ecological speciation, incipient species diverge due to natural selection that is ecologically based. In flowering plants, different pollinators could mediate that selection (pollinator-mediated divergent selection) or other features of the environment that differ between habitats of 2 species could do so (environment-mediated divergent selection). Although these mechanisms are well understood, they have received little rigorous testing, as few studies of divergent selection across sites of closely related species include both floral traits that influence pollination and vegetative traits that influence survival. This study employed common gardens in sites of the 2 parental species and a hybrid site, each containing advanced generation hybrids along with the parental species, to test these forms of ecological speciation in plants of the genus Ipomopsis. A total of 3 vegetative traits (specific leaf area, leaf trichomes, and photosynthetic water-use efficiency) and 5 floral traits (corolla length and width, anther insertion, petal color, and nectar production) were analyzed for impacts on fitness components (survival to flowering and seeds per flower, respectively). These traits exhibited strong clines across the elevational gradient in the hybrid zone, with narrower clines in theory reflecting stronger selection or higher genetic variance. Plants with long corollas and inserted anthers had higher seeds per flower at the Ipomopsis tenuituba site, whereas selection favored the reverse condition at the Ipomopsis aggregata site, a signature of divergent selection. In contrast, no divergent selection due to variation in survival was detected on any vegetative trait. Selection within the hybrid zone most closely resembled selection within the I. aggregata site. Across traits, the strength of divergent selection was not significantly correlated with width of the cline, which was better predicted by evolvability (standardized genetic variance). These results support the role of pollinator-mediated divergent selection in ecological speciation and illustrate the importance of genetic variance in determining divergence across hybrid zones.


Site
Type  Campbell et al. (2018).Clines were also measured in 1991 to 1992 (shown in Campbell et al. 2018).For each of 12 populations, Campbell et al. (2018) determined the mean trait value and then standardized it between zero and one by subtracting the minimum across the 12 populations and then dividing by the difference between the maximum and the minimum.Clines were fit to a no-tails model: where Y = standardized trait value, X = distance, a = intercept, c = center of cline, and w = width of cline (Derryberry et al., 2014)

T H A
Supplementary file S4:

Basic demographic results
Overall, 4% of the seeds planted with known genetic background survived to flower.As in previous common gardens, I. tenuituba had low fitness compared to the home species at the I. aggregata site (0.55 times as high for the fitness measure obtained by multiplying survival times seeds per flower; File S1) and, similarly, I. aggregata had low fitness compared to the home species at the I. tenuituba site (0.42 times as high).

Supplementary file S5
Comparison of univariate selection gradients using global scaling versus local scaling of fitness to its mean value.Estimates of selection (β) were obtained from models for each trait with the factor of site and the trait value nested within site.For floral traits, year of blooming was also included as a factor in the model.For vegetative traits, statistical significance of univariate β came from a model of absolute fitness with a binomial distribution, but parameter estimates from a model of scaled fitness with a normal distribution and identity link for direct comparison with selection on floral traits.Concordant difference in selection refers to the difference between β in the two sites in the direction that matches the difference in average observed trait values.Statistical significance of the difference in selection was assessed by the site by trait interaction in a model Y = Site + Year + Trait + Site*Trait, where Y is relative fitness, using a normal distribution and identity link, and the factor Year was only included for floral traits.

Supplementary file S6:
Selection on two traits in the hybrid site for comparison with selection observed at the Ipomopsis aggregata site.A: Relative seeds per flower as a function of standardized corolla length.The fitted curve shows the best-fitting quadratic regression (compare with closed circles and solid line in Fig. 1A).Seeds per flower declined significantly with corolla length in a model that also accounted for year of flowering, shown by fill color (P < 0.01; Table 2).B: Relative survival to flowering as a function of standardized specific leaf area (compare with closed circles and solid line in Fig. 2).Survival to flowering declined significantly with specific leaf area (P < 0.001; Table 2).
Standardized corolla length Clines across the hybrid zone in floral traits measured in 2015 and 2016.Figureis adapted from * P < 0.05.*** P < 0.001.**** P < 0.0001 Supplementary file S3: . T, H, and A indicate the positions of the I. tenuituba, hybrid, and I. aggregata common gardens in the current study.In this figure, the narrowest (steepest) cline is for redness, leading us to predict the strongest divergent selection on that trait.