•Here we examined the roles of sex-differential plasticity (SDP) and gene flow in sex ratio evolution of subdioecious Fragaria virginiana.
•We assessed whether female frequency varied with resource availability in 17 natural populations and then characterized plasticity and mean investment in allocation to female function at flower and plant levels in the sex morphs in the glasshouse. We estimated patterns of population divergence using five microsatellite markers.
•We reveal SDP in fruit production substantial enough to translate into a higher equilibrium female frequency at low resources. Thus SDP can account, in part, for the strong negative relationship between female frequency and resources found in the field. Pollen-bearing morphs varied in plasticity across populations, and the degree of plasticity in fruit number was positively correlated with in situ variation in nitrogen (N) availability, suggesting an adaptive component to sex-allocation plasticity. Low neutral genetic differentiation, indicating high gene flow or recent divergence, may contribute to the absence of population differentiation in fruit-setting ability of pollen-bearing morphs despite considerable sex ratio variation.
•We consider how these processes, in addition to other features of this system, may work in concert to influence sex ratios and to hinder the evolution of dioecy in F. virginiana.
Ecological context can mediate sex ratio and sexual system evolution in plants by influencing the relative fertility of the sex morphs. Previous work in gynodioecious species (i.e. wherein females and hermaphrodites co-occur), in particular, has revealed how ecological factors such as pollen limitation of seed production, context-dependent inbreeding depression, and enemies can shape sex ratios, with consequences for the evolution of dioecy (reviewed in Ashman, 2006). Phenotypic plasticity in sex expression has also been implicated as an influential force in these processes (Delph, 1990, 2003; Delph & Wolf, 2005). For instance, if hermaphrodite genotypes reduce investment in female function in low-quality habitats because of the high cost of seed production, then females can more easily gain the seed-fertility advantage needed to spread in these populations. However if hermaphrodites in high-quality habitats preferentially increase investment in female function, then females would be less able meet the seed-fertility advantage necessary to spread, and female frequency would be lower. An important assumption of this scenario is that females are less plastic than hermaphrodites, and this sex-differential plasticity (SDP) leads to a relationship between the relative fertility of the sexes and resource availability that manifests across populations as a negative correlation between female frequency and resource availability (Delph, 1990; Delph & Wolf, 2005). Consistent with the SDP hypothesis, female frequency tends to be greater in poor-quality environments for several sexually dimorphic species (see Table 1 in Delph, 2003; Table 2 in Ashman, 2006). Plasticity in sex expression has been shown in field studies (reviewed in Delph, 2003; and in Delph & Wolf, 2005) and experimentally in hermaphrodites (e.g. Delph & Lloyd, 1991; McArthur et al., 1992; Sarkissian et al., 2001; Dorken & Barrett, 2004; Ashman, 2006; Bishop et al., 2010), but few studies test the underlying assumption that hermaphrodites are more plastic than females under controlled experimental conditions (but see Delph & Bell, 2008; Dorken & Mitchard, 2008).
An underappreciated premise of the SDP hypothesis is that sex-expression plasticity of hermaphrodites is ubiquitous among populations. However, we might expect population variation in plasticity as a consequence of selection before population invasion by females. First, plasticity may represent an adaptation to heterogeneous environments (Bradshaw & Hardwick, 1989; Alpert & Simms, 2002). With respect to sex allocation of hermaphrodites, in particular, theory and empirical data suggest that increased allocation to female function relative to male function at greater plant size (or resources) can be adaptive because of sex-differential costs and shapes of fitness gain curves (de Jong & Klinkhamer, 1989; Pickering & Ash, 1993; Klinkhamer et al., 1997; Paquin & Aarssen, 2004). Thus, we might expect the degree of sex-expression plasticity to be correlated with within-habitat variability among populations. Second, plasticity in sex expression might vary with mean resource availability across populations (Freeman et al., 1980). For example, selection against plasticity in low-quality habitats, for example, because it is more difficult to meet the costs of maintaining plasticity, would lead to a positive correlation between plasticity and resource availability (Ashman, 2006). Identifying such variation is important for not only understanding the evolution of plasticity itself but also the role of plasticity in sex ratio evolution; in populations with little or no plasticity the SDP hypothesis cannot account for the frequency of females, and other factors that influence the relative seed fertility of the sexes, and thus female frequency (Lloyd, 1976), must be at play.
Once females are frequent within a population, selection will favor hermaphrodites that invest in male function at the expense of female function (Charlesworth & Charlesworth, 1978; Charlesworth, 1989). Thus, high female frequencies provide the context for invasion by males, that is, female-sterile individuals, and ultimately the loss of hermaphrodites, leading to dioecy (Charlesworth & Charlesworth, 1978). Accordingly, we would predict reduced allocation to female function relative to male function in hermaphrodites where females are more abundant, which, as stated previously, is most often in low-resource habitats. In this scenario, variation across populations in female function of hermaphrodites would not be the result of plasticity but of genetically based differences in sex allocation of hermaphrodites. Although studies have demonstrated in situ variation in fruit or seed set of hermaphrodites in relation to both resources and female frequency (Ashman, 1999; Asikainen & Mutikainen, 2003; Vaughton & Ramsey, 2004), whether such differences among populations are the result of genetic differentiation or plasticity has not been tested experimentally in a common environment. A complication, however, is that the plasticity-based and female-frequency selection-based hypotheses for sex ratio and sexual system evolution are not mutually exclusive, and female frequency may influence plasticity in sex expression of hermaphrodites. For example, even if SDP facilitates the invasion of females and moderates their frequency initially, once female frequency is high, individuals with canalized sex allocation (in particular, males) may be favored over plastic ones if there are costs to plasticity (Ashman, 2006). As a result we would expect to see genetically based differences in hermaphrodites among populations in their sex allocation and plasticity driven by female frequency.
Nonselective forces such as gene flow can also shape among-population patterns of hermaphrodite sex expression, sex ratios, and plasticity and thereby influence sexual system evolution. In fact, metapopulation structure is expected to favor retention of fruit-setting ability in hermaphrodites and plasticity, in general, because of colonization advantages associated with self-fertilization and the ability to readily adjust to new environmental conditions, respectively (Baker, 1955; McCauley & Taylor, 1997; Pannell, 1997; Taylor et al., 1999; Sultan & Spencer, 2002). Yet gene flow, more generally, is rarely measured directly or indirectly in the same systems where the role of ecological context of sexual system has been evaluated. This is a significant omission, because gene flow can dilute the response to within-population selection on sex ratio or sex expression and reduce differentiation among populations. For example, gene flow could facilitate the spread of females (i.e. the male-sterility allele) among populations, homogenizing female frequencies across populations, and thus potentially advancing the evolution of dioecy. Alternatively, if gene flow reintroduces hermaphrodites to populations where they are disfavored and thus rare, it would reduce the effect of sex ratio-mediated selection and retard the evolution of dioecy. Only a combined approach can address these possibilities.
In this study, we take such an approach to evaluate the roles of SDP and gene flow in sex ratio evolution in Fragaria virginiana (Rosaceae), a species with a subdioecious sexual system (i.e. females and two pollen-bearing morphs: males and hermaphrodites). SDP may be important in this system because previous work has shown that female frequency varied with habitat quality in five populations (Ashman, 1999) and that hermaphrodites from at least one population are plastic in sex expression in response to soil resources (Ashman & Hitchens, 2000; Ashman et al., 2001; Bishop et al., 2010). Here, we use a combination of field, glasshouse, and population genetic studies to answer the following questions. Does sex ratio vary with environmental variables across a large set of natural populations, consistent with predictions from the SDP hypothesis? Do sexes and/or populations vary in their plasticity for female function traits? If sexes vary, could this translate into a difference in predicted equilibrium frequency of females across resource environments? If populations vary, can we relate the degree of plasticity in female function expressed by pollen-bearing morphs across populations to environmental variability, mean environmental parameters, or sex ratio? Is there differentiation among populations in mean female function of pollen-bearing morphs and, if so, is it related to female frequency? Do populations show evidence of differentiation at neutral genetic loci, or are they highly connected such that response to within-population selection on female function or plasticity of pollen-bearing morphs would be dampened?
Materials and Methods
Fragaria virginiana Duch., the Virginian wild strawberry, is a perennial herb native to eastern North America (Staudt, 1989). Plants can reproduce sexually via seed or asexually via creeping stolons or ‘runners’. Sexual phenotype in this subdioecious species (females, hermaphrodites, and males coexist) is controlled by a proto-sex chromosome (Spigler et al., 2008, 2010). ‘Hermaphrodites’ may be either homozygous or heterozygous for the fertility allele at the female-function locus (Spigler et al., 2008) and are highly variable in their fruit set (proportion of flowers that become fruits: 0.05–0.80) (Staudt, 1989; Ashman, 2003, 2006; Spigler et al., 2008), whereas ‘males’ are homozygous for the female sterility allele and set no or very few fruits (0–0.05, as defined in Spigler et al., 2008). Because male and hermaphrodite individuals are indistinguishable at the flowering stage in the field, we refer to these collectively as the ‘pollen-bearing’ morph.
Field sites and surveys
We located 17 wild populations of F. virginiana in northeast Ohio and northwest Pennsylvania, USA (Fig. 1), five of which have been studied previously (Ashman, 1999). Linear distances between populations ranged from c. 2 to 65 km (mean 26.7 ± 12.56 SD). All are located along abandoned railroad right-of-ways or adjacent areas (Fig. 1).
During flowering (May–June 2008), we surveyed populations to estimate sex ratio and measure environmental parameters. At each population we recorded the sex (pollen bearing or female) of a single individual encountered every 2 m along a transect through the population up to 400 m, on average, along the right-of-way for each site. Every 30 m, we alternated the side of the right-of-way sampled. We defined sex ratio as the proportion of females in each population. To estimate soil resources, we collected three soil cores to a depth of c. 10–15 cm evenly spaced through each population. Soil was analyzed for percentage total nitrogen (N) via the combustion method (Agriculture Analytical Services Laboratory, Pennsylvania State University, PA, USA). We estimated soil water availability from four c. 1000 cm3 samples of soil collected per population in July 2008. Soil samples were weighed before and after drying to a constant weight in a drying oven, and we calculated percentage soil moisture from these values. We also recorded the percentage of ground not covered by vegetation (‘open ground’) per m2 at 10 m intervals along five to six transects/population and consider this a phytometric estimate of resource conditions in each population. We used Spearman rank correlation (Proc Corr, SAS software, v. 9.1, SAS Institute Inc., Cary, NC, USA) to determine whether female frequency was negatively related to percentage soil N or moisture and positively correlated with percentage open ground and conducted one-tailed tests of significance. Here and throughout, when correlations were in the opposite direction to that predicted, one-tailed P-values were calculated as 1 − 0.5 × P.
Glasshouse plasticity experiment
At the time of surveys, we randomly collected 10 pollen-bearing and five female individuals spaced evenly throughout each of the 17 wild populations and generated four clones per individual (genotype) in the glasshouse at the University of Pittsburgh, PA, USA. We collected twice the number of pollen-bearing morphs than females to capture the full range of female function in the pollen-bearing morphs. To evaluate plasticity of female and pollen-bearing individuals, we subjected clones to either a high (HR) or low resource (LR) treatment. To create HR treatment, we filled 200 ml pots with a 1 : 2 sand and Fafard™ no. 4 soil mix (Conrad Fafard, Agawam, Massachusetts, USA). To create LR treatments, we filled the bottom half of 200 ml pots with sand and the top half with the 1 : 2 soil mixture. We also instated a fertilization regime over the course of growth that led to HR plants ultimately receiving four times as much Nutricote™ 13 : 13 : 13 N : P : K fertilizer (Sun Gro Horticulture, Bellevue, Washington, USA) as LR plants before flowering (12 vs three beads). Analysis of plant vegetative size (following Case & Ashman, 2007) confirmed the effect of these treatments: plants in the HR treatment were twice as large as those in the LR treatment (mixed model post hoc Tukey–Kramer comparison t837 = 43.30, P <0.0001). After a period winterization (2 months, total dark and 4°C) plants were assembled in the glasshouse in a completely randomized block design.
To ensure full potential fruit production, we pollinated all plants with outcross pollen three times per week until plants ceased flowering. At that time, we counted the total number of flowers and fruits produced per plant. We used fruit number as an estimate of female investment at the plant level. Given that flower number is a major component of male function (e.g. Bell, 1985; Ashman & Penet, 2007), we estimated the relative allocation of female to male function (sex allocation) at the plant level as the ratio of fruits to flowers produced (‘fruit set’) per clone. In addition, to estimate female investment and sex allocation at the flower level, we collected one flower bud per clone during flowering and counted ovule and anther number per flower under a dissecting microscope. We used ovule number as a measure of female investment at the flower level, and the ratio of ovule to anther number (O : A) as an estimate of flower-level sex allocation.
To determine the degree of plasticity in each trait and whether sexes and populations varied in their response to resource treatments, we used mixed-model ANOVA. We included treatment (LR or HR environment), sex (pollen-bearing or female), population, and all possible interactions as fixed variables, and genotype, nested within sex and population, as a random variable. Population was treated as ‘fixed’ because those studied varied in female frequency, and we were interested in making specific predictions about how measured traits would vary accordingly. In addition, as our primary interest was to evaluate plasticity between sexes and among populations, rather than dissect genotype-level variation, we did not evaluate genotype × treatment interactions. We used Proc Mixed (SAS software) to analyze ovule number and O : A. We used Proc Glimmix (SAS software) to analyze fruit number using a Poisson distribution and fruit set as a binomial response variable. For all models Satterthwaite’s approximation was used to estimate denominator degrees of freedom (Littell et al., 1996). A significant treatment effect was interpreted as evidence of plasticity; significant sex × treatment and population × treatment interactions were interpreted as evidence of SDP and population variation in plasticity, respectively. Interactions were removed from the model if they were not significant (P >0.05), unless the three-way interaction was significant. When a three-way or sex × treatment interaction was significant, we proceeded by analyzing data for the different sexes separately. We evaluated whether there was among-population differentiation for female function traits of pollen-bearing morphs by examining whether there was a significant main effect of population. We then evaluated whether mean female function of pollen-bearing morphs was negatively correlated with sex ratio using Spearman rank correlation and one-tailed tests.
To assess whether SDP could translate into a difference in expected sex ratio (p), we used Lloyd’s (1976) formula, p = (1 − 2C )/(2 − 2C ), where C is the relative female fertility of pollen-bearing and female morphs. We used the number of fruits per plant to estimate C for each population in each resource treatment and then estimated p. We evaluated whether C was greater in the HR than in the LR treatment using one-tailed, paired t-tests (Proc Ttest, SAS software) and calculated p based on average C in the LR and HR treatments.
To determine whether among-population variation in plasticity of pollen-bearing morphs was correlated with in situ habitat variables, we first estimated plasticity for the traits with population variation in plasticity (determined earlier). We calculated population mean plasticity from genotypic plasticity estimated as the coefficient of variation (CV). We set plasticity to zero for genotypes that produced no fruits in both environments (i.e. canalized males). We examined whether mean plasticity was correlated with the following variables measured in situ: percentage soil N, percentage soil moisture, percentage open ground, and sex ratio. We estimated within-population environmental variation as the CV for all population resource variables. We used Spearman rank correlation (Proc Corr, SAS software) and one-tailed tests to determine whether plasticity was positively correlated with mean and CV in environmental variables and negatively correlated with sex ratio.
Population genetic differentiation
We characterized neutral genetic differentiation and isolation by distance among the 17 populations using microsatellite markers. In each population, we collected leaf tissue on silica from 16–25 individuals (mean 22 ± 3 SD) separated by ≥ 10 m. DNA was extracted following Ashley et al. (2003). We amplified five locus-specific, polymorphic microsatellite loci (ARSFL_99, FAC-002, Fvi9, Fvi11, F.v.A108; Ashley et al., 2003; Lewers et al., 2005; Spigler et al., 2010) using the ‘poor man’s’ PCR protocol (Schuelke, 2000) and the reaction conditions described in Spigler et al. (2008). Capillary electrophoresis was run on an ABI 3730 DNA Analyzer (Applied Biosystems, Carlsbad, CA, USA), and products were visualized using GeneMapper software (Applied Biosystems). We calculated FST and RST in GenAlEx v. 6.3 (Peakall & Smouse, 2006), suppressing within-individual analysis and using 999 permutations to test for statistical significance. RST is an estimator of genetic differentiation for microsatellite data that uses a stepwise mutation model. We examined isolation by distance using the Isolation by Distance Web Service v.3.16 (IBDWS) (Jensen et al., 2005). Isolation by distance tests whether there is a significant correlation between pairwise genetic distance, calculated here as pairwise linearized FST and linearized RST (Slatkin, 1995; Rousset, 1997), in GenAlEx, and linear geographic distance (both distances log-transformed). Such a pattern would be expected if contemporary gene flow is occurring between populations in equilibrium according to the stepping-stone model (Kimura & Weiss, 1964). Negative linearized RST values were truncated to 0. Significance of the correlations was evaluated using 10 000 randomizations and one-tailed P-values for positive matrix correlations. Because results did not change qualitatively when we used linearized RST, we present results using linearized FST only.
Sex ratio variation among populations
Proportion of females varied from 0.16 to 0.48 (mean ± SD, 0.29 ± 0.08) among populations and was negatively related to percentage soil N (rs = −0.70, P1-tailed = 0.001) and percentage soil moisture (rs = −0.52, P1–tailed = 0.015) and positively related to percentage open ground (rs = 0.60, P1-tailed = 0.006) (Fig. 2), demonstrating that females are more common in more stressful habitats. All of these relationships remained significant after Bonferonni correction.
Plasticity and its predicted effect on sex ratio
Both female investment traits (fruit and ovule number) and estimates of sex allocation (fruit set and O : A) were plastic (Table 1; treatment effect). Fruit number and fruit set exhibited SDP (Table 1; sex × treatment effect), and populations varied in plasticity for these traits as well (Table 1; fruit number, population × treatment; fruit set, population × sex × treatment). These results were unchanged when we re-ran the analyses with a reduced data set to balance the sample size between the pollen-bearing morphs and females (data not shown). The sexes differed in all four traits, but only ovule number and O : A varied significantly across populations (Table 1).
Table 1. Results of ANOVA testing for plasticity (treatment) and interactions between treatment, sex, and Fragaria virginiana population
Ovule : anther ratio
F-values are presented for each model term along with asterisks indicating significance. Nonsignificant interaction terms (P >0.05) were removed from the model, indicated by dashes, except when there was a significant three-way interaction. Genotype, nested in sex and population, is included in all models as a random variable. A range is presented for the denominator degrees of freedom (df) as the value depends on the model and sample size for each trait. Significance values: *, 0.05 ≥ P >0.01; **, 0.01 ≥ P >0.001; ***, 0.001 ≥ P >0.0001; ****, P ≤0.0001.
Sex × treatment
Sex × population
Population × treatment
Sex × population × treatment
Pollen-bearing morphs adjusted fruit set in response to the resource treatment, and the degree varied among populations (Table 2, Fig. 3). By contrast, females did not adjust fruit set, and this result was consistent across populations (Table 2, Fig. 3). Both sexes were plastic for fruit number, but there were sex differences in the magnitude of plasticity (Table 2, Fig. 3). Pollen-bearing morphs exhibited a 2.3-fold increase in fruit number in response to increased resources; in comparison, females’ fruit number increased only 1.6-fold. Plasticity in fruit number varied among populations for both sexes (Table 2).
Table 2. Sex-specific ANOVA results for plasticity (treatment) and interaction between treatment and Fragaria virginia population
F-values are presented for each model term along with asterisks indicating significance. Genotype, nested within population, is included in all models as a random variable. A range is presented for the denominator degrees of freedom (df) as the value depends on the model and sample size for each trait. Significance values: *, 0.05 ≥ P >0.01; **, 0.01 ≥ P >0.001; ***, 0.001 ≥ P >0.0001; ****, P ≤0.0001.
Population × treatment
Sex-differential plasticity in fruit number translated into a significant difference in the relative fertility of the sexes across resource treatments. When assessed in the glasshouse and across all populations, females had a 6.4-fold fruit production advantage over pollen-bearing individuals in the LR treatment, but a 4.6-fold fruit production advantage in the HR environment. Relative female fertility (C) was significantly lower in LR (0.16), on average, than in HR (0.22) treatment (t16 = −4.62, P1-tailed = 0.00015).
Population differentiation in plasticity of pollen-bearing morphs
Only plasticity in fruit set (CV range: 0.09–0.89) and fruit number (CV range: 0.19–0.93) of pollen-bearing morphs varied significantly among populations (Tables 1, 2). Plasticity in fruit number increased significantly with in situ variation in N availability (rs = 0.50, P1-tailed = 0.02) and tended to increase with in situ variation in soil moisture (rs = 0.40, P1-tailed = 0.06). Plasticity in fruit set similarly tended to increase with in situ variation in N availability and variation in soil moisture (both: rs = 0.36, P1-tailed = 0.08). Plasticity was not related to mean environmental variables or sex ratio (all P1-tailed > 0.10; data not shown). After Bonferroni correction, none remain significant.
Differentiation in female function of pollen-bearing morphs among populations
Although there was high genotypic variation in female function of pollen-bearing morphs at the plant level, with genotypes ranging from completely female sterile – that is, they set no fruit in either resource environment – to almost completely female fertile, with fruit set as high as 0.99 (Fig. 4), populations did not differ significantly in either mean fruit number or fruit set (Tables 1, 2). Furthermore, population mean values were not negatively correlated with sex ratio (both P1-tailed > 0.25; Fig. 4). Populations differed significantly in ovule number and O : A (Table 1), but these were also not significantly correlated with sex ratio (both P1-tailed > 0.40).
Neutral genetic differentiation among populations measured as FST
A high number of alleles per microsatellite locus (mean ± SE, 9 ± 0.20) and high genetic diversity (He, 0.78 ± 0.01) across populations provided for a rigorous test of differentiation. The amount of genetic differentiation among populations was relatively low but statistically significant. Only 3% of the genetic variation was attributable to variation among populations (FST = 0.031, P = 0.001; RST = 0.043, P = 0.01). Pairwise linearized FST values ranged from 0 to 0.069. We did not detect significant isolation by distance (linearized pairwise FST: log-transformed, r = 0.16, P =0.09; untransformed, r =0.08, P =0.27).
Our work reveals roles for SDP and gene flow in sex ratio variation in F. virginiana. Consistent with the SDP hypothesis, we found that pollen-bearing morphs were more plastic for plant-level allocation to female function. This plasticity probably contributes to the strong correlation between female frequency and resource availability in natural populations of F. virginiana. Because plasticity in fruit number among the pollen-bearing morphs tended to increase with variability in in situ N availability, it could indicate selection for retention of female function in pollen-bearing morphs that would maintain hermaphrodites in this species. Indices of population differentiation at neutral loci indicate either recent divergence or extensive gene flow across the study populations. Both mechanisms could contribute to the absence of differentiation in fruit-setting ability of the pollen-bearing morphs despite considerable variation in sex ratio. Below we discuss each of these important findings and their implications for the evolution of the sexual system in F. virginiana.
SDP contributes to sex ratio variation in F. virginiana
Our work builds upon previous studies that have examined SDP (Delph, 1990; reviewed in Delph, 2003; Barr, 2004; reviewed in Delph & Wolf, 2005; Dorken & Mitchard, 2008; Van Etten & Chang, 2009) by taking advantage of the clonality of F. virginiana to experimentally test for SDP under controlled conditions. Moreover, we go beyond previous work on plasticity in F. virginiana (e.g. Ashman, 2006; Bishop et al., 2010) by including females and multiple populations. In doing so, we were able to detect SDP for plant-level female function traits that was pronounced enough to manifest as a significant difference in the relative seed fertility of the sexes (i.e. the C-value) between resource environments. Because C should dictate the frequency of the sexes within a population (Lloyd, 1976), SDP seen in our glasshouse experiment would translate into c. 14% higher expected frequency of females at low resources (P =0.41) compared with high resources (P =0.36). Dorken & Mitchard (2008) similarly examined SDP under controlled conditions in monoecious and dioecious Sagittaria latifolia and found that females gained a significant seed fertility advantage over hermaphrodites in LR but not HR treatments. Taken together, these studies build strong support for the hypothesis that SDP underlies the association between sex ratio and resource availability seen in many species (reviewed in Delph, 2003; and in Ashman, 2006; Caruso & Case, 2007).
Noticeably, our glasshouse-based estimates of equilibrium female frequencies are relatively high, near the upper end of the range of that observed in natural populations. This suggests that either both of our experimental conditions were close to low-resource levels in the field, leading us to underestimate C in the HR treatment and so overestimate the expected frequency of females, or there are biological factors other than SDP that limit female frequencies at high resources in the field. For example, our experiment did not incorporate effects of pollen limitation. If females are more pollen-limited in high than in low resources, then we would expect to see fewer females in HR environments than predicted from SDP alone. Pollinators are known to discriminate against females in favor of pollen-bearing morphs in F. virginiana (Ashman et al., 2000; Case & Ashman, 2009), with similar results in other systems (e.g. Bell, 1985; Ashman & Stanton, 1991; Eckhart, 1992). If sexual dimorphism in attractive traits becomes greater at high resources, for example, because hermaphrodites increase petal size more than females, then females may suffer greater discrimination at high resources. This hypothesis remains to be tested. Another possibility is that pollen-bearing morphs have greater survival or clonal growth rates than females at high, but not low, resources. Studies have shown differences between the sexes in physiological capabilities (e.g. photosynthetic rates, nutrient uptake rates, and/or water-use efficiency) that can lead to differential survival and clonal growth, although with varied results (Eckhart & Chapin, 1997; Dawson & Geber, 1999; Delph, 1999; Obeso, 2002; Case & Ashman, 2005). In our experiment, however, females and pollen-bearing individuals did not differ in runner production in the HR treatment, although females showed an advantage in the LR treatment (planned contrasts: F1,324.7 = 0.22, P =0.64; F1,1096 = 5.59, P =0.018, respectively, data not shown). This mechanism may contribute to the general relationship between female frequency and resources seen in the field. Regardless of what other factors are at play, the relative difference in C shown here between our two resource environments clearly indicates a role for SDP in sex ratio variation among F. virginiana populations.
Population variation in plasticity for female function
While a relationship has been demonstrated between environmental heterogeneity and plasticity for photosynthetic and functional plant traits (e.g. Balaguer et al., 2001; Nicotra et al., 2007; Molina-Montenegro et al., 2010), ours is the first study to show a relationship between sex allocation plasticity and environmental variation. Average plasticity for fruit number of pollen-bearing morphs tended to increase with increasing variability in soil N availability measured in situ, and although the statistical significance was reduced by correction for multiple tests, it is highly suggestive of an adaptive response to heterogeneity in this clonal, weedy species. Size-based sex-allocation theory provides a rationale for why pollen-bearing morphs that retain fruit-setting ability and plasticity may have higher fitness than canalized hermaphrodites or males: whereas diminishing fitness returns are associated with increased investment in male function at high resources, increased investment in female function should lead to increasing fitness returns (e.g. Klinkhamer et al., 1997; Ashman, 2006). The fitness gains from a plastic strategy should be maximal in patchy environments (e.g. Freeman et al., 1980) and/or when migration is high (Sultan & Spencer, 2002). Future work should empirically test whether resource-based sex allocation plasticity is adaptive and especially whether such plasticity confers an advantage over canalized males.
Does gene flow maintain hermaphrodites in F. virginiana?
While our estimate of FST was statistically significant, it is reflective of quite low amounts of neutral genetic differentiation among our study populations and suggests high amounts of gene flow. Such low differentiation is perhaps not surprising among F. virginiana populations given the species’ clonal, perennial habit and ingested mode of seed dispersal, which lend to greater gene movement (Hamrick & Godt, 1996), and relatively small spatial scale of our study. The absence of significant isolation by distance suggests that genes are exchanged across populations according to an island model (Wright, 1931), wherein gene flow is equal and reciprocal among populations. However, a model that can account for distance between populations along the railroad right-of-ways rather than Euclidean distance, since seeds may be spread by animals that use these corridors, might better explain patterns of genetic distance among populations, similar to findings for hydrochorous Hibiscus moscheutos in relation to tidal streams (Kudoh & Whigham, 1997). Low FST and absence of isolation by distance could also indicate recent divergence among populations. Strawberry is a relatively weedy species, and populations may have spread quickly along the railroad right-of-ways after abandonment (Staudt, 1999), in which case they may not have yet had time to diverge substantially from one another at neutral loci (Fig. 1).
Regardless of the cause of the low neutral genetic differentiation, the high connectivity among populations has implications for the evolution of the sexual system. High connectivity may explain why we did not find population differentiation in plant-level female function (fruit set or number) of pollen-bearing morphs or a negative correlation between it and sex ratio (Table 2, Fig. 4). While SDP may account, in part, for this homogeneity (e.g. because hermaphrodites can appear phenotypically male at low resources), not all F. virginiana pollen-bearing morphs are plastic. In fact, we see canalized males in all populations (Fig. 4), and there is a genetic component to fruit-setting ability (e.g. Ashman, 1999; Spigler et al., 2008). Thus we expected female presence to exert selection to reduce female function of pollen-bearing morphs (Charlesworth & Charlesworth, 1978; Charlesworth, 1989), even with plasticity, such that the frequency of males and/or hermaphrodites with genetically-based low female function would increase with female frequency, and this would manifest as a negative relationship between female function of pollen-bearing individuals and female frequency. However, this pattern may be obscured if female-mediated selection is weak relative to gene flow. Acknowledging the assumptions needed to do so (reviewed in Whitlock & McCauley, 1999), we estimate between six and eight migrants per generation among populations (based on RST and FST, respectively; Wright, 1951), a number more than sufficient to reduce the effects of within-population selection. Thus, while SDP may influence C and thereby the frequency of females, subsequent gene flow may serve to maintain hermaphrodites and males in all populations and, ultimately, subdioecy.
It is worth noting that in addition to gene flow, another feature of our system that could reintroduce hermaphrodites within populations and obscure patterns of within-population selection is recombination between the sex-determining loci on the proto-sex chromosome in F. virginiana (Spigler et al., 2008). Such recombination, while rare, can create hermaphrodite morphs with exceptionally high fruit-setting abilities de novo within populations and thus could account for their presence across all populations (Fig. 4). Moreover, recombination would also create male- and female-sterile plants (i.e. neuters), which have been identified in a F. virginiana mapping population (Spigler et al., 2008) and in the field (T.-L. Ashman, unpublished; Valleau, 1923). When considering this possibility as well, we would then predict that one would not see a strong signature of selection (differentiation of fruit production of pollen-bearing morphs) at the population level until recombination is suppressed in the sex-determining region.
We demonstrate that SDP can contribute to the relationship between female frequency and resource availability in populations of F. virginiana, as expected according to the SDP hypothesis (reviewed in Delph, 2003; Delph & Wolf, 2005). And, while SDP may account, in part, for why we do not see the predicted population differentiation for female function of pollen-bearing individuals at the plant level despite wide variation in female frequencies, population variation in plasticity and the presence of males combined with high connectivity of populations suggest that a full evaluation of role of SDP in sex ratio and sexual system evolution requires consideration of selection on plasticity in general as well as the degree of gene flow. Our work underscores the importance of integrating genetic studies that can inform on population history and isolation when evaluating the role of local ecological context in the evolution of sexual systems. This importance has been recognized in and modeled for systems with demonstrated metapopulation dynamics (Pannell, 1997; Olson et al., 2005; Obbard et al., 2006), but ought to also be considered more generally where gene flow is possible (see, e.g., Fleming et al., 1998). A recent study, for example, modeled the combined effects of selection, drift, and migration on sexual system stability, and demonstrated different implications of pollen vs seed flow for maintaining a polymorphic sexual system (Dufay & Pannell, 2010). Male sterility in F. virginiana is dominant (Ahmadi & Bringhurst, 1991) and can only migrate via seed flow, so if low differentiation among our populations is mainly the result of pollen movement, we predict that both gene flow and SDP will promote the stability of subdioecy in this system by reintroducing hermaphrodites that plastically adjust sex allocation at low resources.
The authors wish to thank D. Bacher-Hicks, E. Bishop, A. Houck, D. Jackson, M. Jenkins, A. Johnson, E. McAuley, B. McTeague, and E. York for field, laboratory, and glasshouse assistance and the staff at Pymatuning Laboratory of Ecology for logistical assistance. We also thank three anonymous reviewers for helpful comments on an earlier version of the manuscript. This work was supported by the National Science Foundation (DEB 0449488 and 1020523 and REU supplements, to T.-L.A.). This is contribution number 282 to the Pymatuning Laboratory of Ecology.