Are Pollinator Preferences Strong Enough To Cause Assortative Mating Between Plants With Similar Flowers?
Are Pollinator Preferences Strong Enough To Cause Assortative Mating Between Plants With Similar Flowers?
Yes, and further, hawkmoths' preferences for longer corolla tubes (panels d-f) are manifest in the next generation by an overabundance in the F3 populations descended from natural pollination (F3nat) versus control F3 populations descended from hand pollination among randomly chosen F2 plants (F3ran).
The net effects of any generalist pollinators or other selective agents, observed or not, are included in the F3nat populations. F3ran populations controlled for all unmeasured evolutionary forces and constraints, and for genetic segregation that would confound F2-F3 comparisons. Environmental effects were randomized by the planting designs of the F3 populations. Therefore, differences in floral trait distributions between F3nat and F3ran are genetically-based and can only be due to the different mating structures applied to the F2 (natural vs. random) since the genetic and ecological histories of the F3 populations are otherwise identical.
Hummingbirds were only observed in one of five F2 plots, and their preferences were for shorter-tubed red flowers (panels a-c). The smallest flowers were more common in F3nat than expected (panels h,i), although the small number of expected flowers in the 0-10% bin renders this conclusion tentative. At a minimum we suggest that hummers' preferences may have kept the tail at the small end of the distribution from disappearing, despite hawkmoths preferring longer tubes.
Floral tube length, limb width and color were all correlated with pollinator visitation rate (r = 0.26, p<0.0001; 0.23 p<0.001; -0.10, p=0.05, respectively). The latter two traits are correlated with tube length (r=0.49, p<0.0001, r= -0.17, p=0.001) so we focus here on floral tube length. Studying only one trait is a conservative approach that avoids potentially inflating pollinators' inferred roles in causing floral evolution of a suite of traits in a hypothetical ancestral species. Gametic-phase disequilibrium (D) among floral trait genes would certainly be greater in our F2 populations than in the ancestral species since there has only been one generation of random mating (F1 to F2) to disrupt statistical associations between floral trait alleles. Strong D in the F2 might lead to a larger multivariate evolutionary response than would be possible if the floral trait genes were in equilibrium, as would be expected in a highly outcrossing ancestral species (both extant species are self-incompatible).
We now also know that pollinators' preferences lead to a genetic
correlation between these traits in the F3nat, whereas these traits
are uncorrelated in F3ran. Those data are here.
Figure 1. Pollinator preferences for corolla tube length in the F2 and genetic response in the F3. a, Hummingbird preferences in situ in F2-3, the only population where hummingbirds were regularly observed: the relative frequencies of corolla lengths of all plants (black outline, N=137 plants) and of hummingbird-visited plants (grey, N=78 observed visits). b, Hummingbird preference presented as the difference in the absolute frequencies of corolla lengths of hummingbird-visited plants minus corolla lengths of all F2 plants. c, Hummingbird preference presented as the ratio of the relative frequencies of corolla lengths of visited plants vs. corolla lengths of all plants. d, The relative frequencies of corolla lengths of all plants in 5 F2 populations and all pollinator-visited plants (N=672 plants [black outline], N=2431 visits [grey bars], 96.2% of which were by hawkmoths). Triangles indicate the phenotypic means of N. forgetiana, F1 hybrids and N. alata, from left to right. e, f are plotted as in b, c but for all pollinator visits. g, Corolla lengths of F3 plants descended from naturally pollinated F2 plants (F3nat, N=681, grey bars) and a control population of F3 plants descended from hand pollination among randomly-chosen F2 plants (F3ran, N=673, black outline). h, The genetic response expressed as the difference of the standardized relative frequencies (F3nat - F3ran). i, The genetic response expressed as the ratio of the standardized relative frequencies (F3nat / F3ran). The dashed lines represent null hypotheses that the two distributions in the top row of panels are identical (grey = observed; black outline = expectation). The bins on the y-axes contain 10% of the range of F2 phenotypes. Asterisks indicate that a bin contributed significantly (** = p<0.01, *** = p<0.001) to the overall c2. Parentheses indicate that expected counts in that bin were below five.