Unisexual flowers as a resolution to intralocus sexual conflict in hermaphrodites

Abstract

In dioecious populations, males and females may evolve different trait values to increase fitness through their respective sexual functions. Because hermaphrodites express both sexual functions, resolving sexual conflict is potentially more difficult for them. Here, we show that hermaphrodite plants can partially resolve sexual conflict by expressing different trait values in different male and female modules (e.g. different flowers, inflorescences, branches etc.). We analysed the flowering phenology, sex allocation and selection gradients on floral traits of flowers of the andromonoecious plant Pulsatilla alpina, which produces both bisexual and male flowers. Our results indicate that strong protogyny prevents early bisexual flowers from profiting from high siring opportunities early in the reproductive season at a time when male flowers could achieve high siring success. The production of unisexual male flowers thus resolves this sexual conflict because it allows the flowers to express their male function without waiting until after the female function has been performed. Our study illustrates the resolution of sexual conflict arising from phenological constraints via modular divergence in sex allocation. We discuss the extent to which modular variation in sex allocation in the context of other sexual systems may be similarly explained.

Keywords: androdioecy; dichogamy; monoecy; sex allocation; sexual antagonism; sexual interference.

Figure 1.

Theoretical conception of trait evolution in (a) dioecious and (b) hermaphroditic populations, as elaborated by Morgan [6]. Circles represent lines of equal total fitness for given combinations of the male and female values of trait z, with the smallest circle indicating the point of maximum fitness. Populations are envisaged to evolve along the arrowed lines. In (a), the separation of the two sexes into male and female individuals in a dioecious population allows populations to evolve towards the trait optimum for both sexual functions. In (b), the fact that the same individuals must express both a male and female function in a hermaphroditic population means that the fitness optimum is unreachable for traits measured at the individual level. In the text, we argue that the modularity of hermaphroditic organisms such as plants allows them to produce different male and female modules, each expressing a different (potentially optimum) value of trait z. Figure panels are redrawn from Morgan [6].

Figure 2.

(a) Number of pistils in bisexual and female flowers (red line) and functional stamens from bisexual flowers (light blue line) and male flowers (dark blue line) over the course of the flowering season in the study population. (b) Pistil availability per stamen calculated for the study population and accounting for all bisexual and male flowers observed. (c) Pistil availability per stamen calculated for the study population in which only bisexual flowers are accounted for. In b and c, pistil availability was calculated by dividing the total prospective number of available pistils by an estimate of the total number of functional stamens produced by both bisexual and male flowers (b) or by only bisexual flowers (c) within each time window.

Figure 3.

The delay in the onset of the male function of bisexual flowers as a function of the number of their pistils. The delayed onset of male function was quantified by the difference between the day a flower first opened and the beginning of its male function. Shown is the least-squares regression (line) with the 95% confidence interval (shaded).

Figure 4.

Selection gradients for contributions to fitness made through female function (red) and male function (blue) (a–c), and through both male and female functions together (grey) (d–f) as a function of pistil number (a,d), stamen number (b,e) and flowering date (c,f) in bisexual flowers. Two points are displayed for each bisexual flower in a, b and c (N = 57), one for female and one for male function. Table 1 for statistical details. Dashed regression lines indicate that neither linear nor quadratic gradients were significant. In c, the linear and quadratic gradients via the female function were marginally non-significant and non-significant, respectively. The shaded ribbons indicate 1 s.e. of the corresponding regression line. In a–c, optimal trait values of female and male reproductive success are indicated by red and blue vertical lines, respectively (solid lines: inference for directional selection; dashed lines: inference for stabilizing selection). For d–f, see electronic supplementary material, table S4 for the values of the linear and quadratic selection gradients.

Figure 5.

Resolution of sexual conflict via the production of male-only flowers. Histograms showing the distribution of (a) the number of pistils, (b) the number of stamens and (c) the flowering date for bisexual (N = 129, light orange bars) and male flowers (N = 46, light green bars) in the population before experimental manipulation of stamen number. The overlaps of the two bars are shown in dark green. Inferred optimal trait values for female and male contributions to fitness by bisexual flowers under directional and stabilizing selection are indicated, as inferred from figure 4. Note that no significant selection was detected on stamen number (b) for female function. Traits are standardized to the scales used in the selection gradient analysis.