Predicting evolutionary consequences of pollinator declines: the long and short of floral evolution



This article is corrected by:

  1. Errata: Corrigendum Volume 178, Issue 3, 689, Article first published online: 17 April 2008

(*Author for correspondence:
tel +330 972 5122; fax +330 972 8445;

Recent research indicates that pollinator populations across the globe are declining (Allen-Wardell et al., 1998; Biesmeijer et al., 2006; National Research Council Committee on the Status of Pollinators in North America, 2007). This finding has been a prominent feature in recent news stories and headlines in the popular press, for the expectation is that the large fraction of plants that rely on animal pollinators to reproduce may therefore suffer reduced reproduction, and under some circumstances, population reductions (Knight et al., 2005, Biesmeijer et al., 2006; National Research Council Committee on the Status of Pollinators in North America, 2007; Pauw, 2007). Closer to the hearts, and the stomachs, of most newspaper readers are the potential consequences of pollinator declines for human food supplies (National Research Council Committee on the Status of Pollinators in North America, 2007; Klein et al., 2007). Both of these ecological effects are certainly of immediate concern, but little attention has been given to the potential evolutionary consequences of reduced pollinator populations. In this issue of New Phytologist, Fishman & Willis (pp. 802–810) show that the direction and magnitude of selection on floral traits can change dramatically when pollinators are not available. Selection when pollinators are scarce favors smaller flowers that are better able to self-fertilize when unvisited. This raises the possibility that as pollinators decline, flowers may evolve to become smaller and less noticeable, which may accentuate the problem. Evolutionary effects will take longer to arise than will ecological effects, but may be more insidious, fundamentally altering the phenotypic and genetic makeup of plant populations, and the composition of native plant communities.

‘These results imply an antagonism between selection for selfing when pollinators are rare, and selection for a different suite of floral traits when pollinators are more abundant.’

Fishman & Willis worked with Mimulus guttatus, an annual plant that reproduces mostly through outcrossing when pollinators are present. However, when unvisited, the flowers are also able to self-pollinate automatically. This can happen because the female receptive surface (stigma) in the flower is positioned very near the pollen-producing structures (anthers). This separation between anther and stigma is known as herkogamy and varies substantially among plants. The capacity for autonomous seed production provides reproductive assurance when mates or pollinators are scarce (Kalisz et al., 2004), but has the drawback that it often results in fewer, inbred seeds. Inbreeding depression is substantial in this species, reducing offspring performance by 69% (Willis, 1993), so this is not a minor concern. Still, self-pollination is a common reproductive strategy in plants, and many close relatives of M. guttatus (such as Mimulus micranthus) reproduce exclusively through selfing. Those selfing species generally have smaller flowers and very little herkogamy (Fig. 1).

Figure 1.

Mimulus guttatus (yellow monkeyflower). (a) A small portion of the Iron Mountain (Oregon) population of M. guttatus (Phrymaceae). Photo courtesy of Megan Hall, Duke University. (b) Close-up of a single flower of M. guttatus. Fishman & Willis found that when pollinators are present, selection favors larger M. guttatus flowers, but when pollinators are absent, selection favors smaller flowers with anthers and stigma close together. The anthers and stigma are inside the corolla and are not visible in this view. Photo courtesy of Megan Hall, Duke University and NYU.

To understand in greater detail the consequences of pollinator scarcity, Fishman & Willis applied a factorial design of pollen supplementation and pollinator exclusion to over 400 plants in an Oregon population of M. guttatus. They found that pollen supplementation increased seed production and pollinator exclusion strongly reduced it, indicating that seed production was pollen-limited. A particularly novel part of this study was that the authors also quantified selection on floral characters for these plants, allowing more powerful analysis and inference (Ashman & Morgan, 2004; Ashman et al., 2004). When pollen was abundant, selection via seed production favored plants with relatively long and wide flowers. When pollinators were not available, reproduction was much reduced, but selection favored narrow-flowered plants with little herkogamy, a morphology that facilitates autonomous self-pollination.

These results suggest that M. guttatus plants experiencing different pollinator abundances will be subject to strongly divergent types of selection. Importantly, the pattern of selection when pollinators are absent is in the direction of the patterns found in selfing relatives. If pollinator abundance varies strongly in time or space, as seems to be common in wild populations of many plants (e.g. Moeller, 2005), the selective surface (averaged over years and sites) for M. guttatus might be visualized as having two fitness peaks (Fig. 2). One of the peaks applies when pollinators are available but plants must compete for visitation. In this situation, selection favors large flowers with substantial herkogamy. The other peak occurs when pollinators are rare. There, selection favors small flowers with little herkogamy. The results of Fishman & Willis suggest that over time (and assuming that fitness through seed production mirrors total fitness) the morphology of M. guttatus populations will wander between these two fitness peaks, with the direction of travel related to the history of pollinator abundance and the genetic correlations among these traits. The position and relative height of the peaks might change if seed fertility does not correlate well with overall fitness. This depends on the extent to which male function (pollen donation) and survival to reproductive age covary with these morphological characters. Other complicating factors to consider include the quality of pollen arriving on stigmas (Aizen & Harder, 2007) and the level of inbreeding depression. Future research that more fully documented selection via lifetime total fitness would be valuable, as would work that assessed a range of pollinator abundances, and thus would allow exploration of and characterization of the entire selective surface.

Figure 2.

Hypothetical fitness surface for plants that experience variation in selection as a result of variation in the pollinator environment. The green area (peak on the left) applies when pollinators are rare and most flowers will only produce seeds through autonomous selfing. The yellow area (peak on the right) applies when pollinators are sufficiently abundant that most flowers produce some outcrossed seeds, but not so abundant that all flowers are fully pollinated. The gray dot indicates the location of a population on the fitness surface, and the two arrows show expected trajectory under two conditions: no pollinators (white arrow on the left); or pollinators present (red arrow on the right). The direction of evolutionary change depends on the pollinator environment at that time.

What is especially intriguing is that these results also imply an antagonism between selection for selfing when pollinators are rare, and selection for a different suite of floral traits when pollinators are more abundant. The balance between these deserves both empirical and theoretical investigation, which could provide answers to important questions about, for example, the conditions and amount of time required to generate a selfing species such as M. micranthus from a primarily outcrossed species like M. guttatus, or the evolutionary conflicts faced by plants with mixed-mating systems (for example, is a jack-of-all-trades phenotype possible, and, if so, what features would it have?).

The factorial manipulation of pollen addition and pollinator exclusion, as used in this study, allows another surprising insight. In the open-pollinated treatment, selection favored larger flowers. Most studies of selection on floral traits include only such unmanipulated plants. That part of the analysis for this population of M. guttatus indicates that the cause of selection is pollinator response to flower size. The role of pollinators could, in principle, be confirmed or denied with pollinator observations. However, large-flowered plants were also favored when pollen was added to plants that were never visited by pollinators, a treatment that equalizes pollen receipt of large-flowered and small-flowered plants, and eliminates the opportunity for pollinator preference to operate. Fishman & Willis suggest that this association between large flowers and high seed production, regardless of pollinator activity, may be mediated by correlations of floral traits with other factors such as individual environment or inbreeding history. More generally, this result joins a growing literature suggesting that floral traits often are developmentally correlated with other traits that have important effects on fitness (Herrera, 1995; Frey, 2004), and raises important questions about the extent to which floral phenotypes serve only to affect pollination (Galen, 1999).

One might be tempted to conclude that the results of Fishman & Willis suggest that in a world of declining pollinator populations the flowers themselves may begin to evolve to be less attractive and less reliant on pollinators, which might then reinforce pollinator declines. But in truth, their results highlight how little we know mechanistically about the link among pollinator abundance, natural selection and likely evolutionary trajectories.