Digest: Experimental evolution provides a window into the evolution of generalized pollination *

Do plants with multiple pollinators evolve unique trait combinations or intermediate phenotypes compared to plants with one pollinator? Using experimental evolution, Schiestl et al. (2018) found that plants pollinated by bumblebees and hoverflies evolved trait values not observed in plants pollinated by one taxon, which provides evidence for the existence of a unique generalized pollination phenotype.

While highly specialized plant-pollinator interactions like Darwin's orchids, yucca moths, and fig wasps have captured the imagination of evolutionary biologists for centuries, a number of lines of research suggest that generalization, not specialization, may be more the rule than the exception, particularly in the northern hemisphere (see Johnson and Steiner 2000). For instance, network studies frequently find evidence for generalization (Oleson and Jordano 2002) or asymmetric specialization (Vázquez and Aizen 2004), and large scale analyses of floral phenotypes and pollinator visitation patterns provide mixed support for the idea that pollinators only visit plants with certain floral traits (Ollerton et al. 2009;Rosas-Guerrero et al. 2014).
Despite this evidence for the prevalence of generalization in plant-pollinator interactions, the ways in which pollination by multiple animals or types of animals affects floral trait evolution are relatively unknown. The "most effective pollinator principle" (Stebbins 1970) contends that pollinator-mediated selection is additive and determined by the most frequent or efficient visitor. However, recent studies have demonstrated that selection can be non-additive, such that visitation by multiple pollinators may generate selection for unique trait combinations rather than intermediate phenotypes (Sahli and Conner 2011;Brosi and Briggs 2013;Knauer and Schiestl 2017). * This article corresponds to Schiestl, F. P., A. Balmer, and D. D. Gervasi. 2018. Real-time evolution supports a unique trajectory for generalized pollination. Evolution. https://doi.org/10.1111/evo.13611.
In this issue, Schiestl et al. (2018) utilized an experimental evolution approach to test key questions surrounding the evolution of generalized floral phenotypes. To determine if different pollinators generate different patterns of selection on floral morphology, scent, and rates of spontaneous selfing, they grew fastcycling Brassica rapa (Brassicaceae; Wisconsin Fast Plants R ) in a controlled environment and exposed seven generations of plants to one of three pollination treatments: bumblebees (Bombus terrestris), hoverflies (Episyrphus balteatus), and both insects together. At the end of the experiment, plants in the generalized pollination treatment exhibited lower floral scent production, greater height, and lower nectar production compared to one or both of the single-pollinator species treatment groups. Taken together, these results document the evolution of a unique generalized pollination phenotype that is not merely an average of the phenotypes produced by selection from each pollinator in isolation.
This study is only the third to estimate selection (using the Lande and Arnold (1983) method) via different pollinators on a plant within the same experiment (Figure 1). Across these studies, selection on some traits was consistent between pollinator treatments while patterns of selection on other traits varied. For instance, both bumblebees and cabbage butterflies exerted positive directional selection on floral morphology of B. rapa (Knauer and Schiestl 2017; Figure 1B

DIGEST
Conner 2011; Figure 1A). However, in line with the results of Schiestl et al. (2018), these studies also found unique patterns of selection in their generalized pollination treatments (two or more insect species; differences in color and line type of boxes around one pollinator vs. multiple pollinators within a column in Figure 1). This small body of literature indicates a potentially significant role for species interactions in determining patterns of selection that should be explored in a larger number of systems. In addition to providing valuable insight into the evolution of generalized pollination, the work of Schiestl et al. (2018) highlights how experimental evolution can be used to address outstanding questions in the field of plant-pollinator interactions. While this approach will not be feasible in all systems, future studies can utilize the type of approach presented by Schiestl et al. (2018) to continue to disentangle the effects of different agents of selection and test how species interactions may affect patterns of selection.