Author for correspondence:Deborah L. Marr Tel: +1 574 5205564 Fax: +1 574 520 5589 Email: firstname.lastname@example.org
• Sex-dependent infection rates could change the effective sex ratio of a population. Here, I tested whether females and hermaphrodites of Silene acaulis were equally likely to be infected by Microbotryum violaceum, a fungus that sterilizes the host, and whether sex allocation in hermaphrodites differed between low and high disease plots.
• Sex ratios of healthy and diseased plants were estimated in five natural plots. Fitness gained through seed production was estimated by measuring seed quantity and quality for each sex morph in eight plots for 2 yr; four plots had 1–5% disease frequency and four plots had 18–25% disease frequency.
• Sex ratios of healthy and diseased plants did not differ in five plots. The proportion of fitness hermaphrodites gained through ovules varied from 25 to 48%, indicating that this population is near the cosexual end of gynodioecy. Variation in functional gender of hermaphrodites was not explained by sex-dependent infection rates.
• Spatial heterogeneity in resources and microclimate seems to be important in explaining both disease frequency and variation in seed production by females and hermaphrodites.
A number of pathogens directly affect plant reproduction (Jarosz & Davelos, 1995; Clay & Kover, 1996; Lara & Ornelas, 2003), but most plant-pathogen studies have focused on how reproduction is affected in diseased individuals (Burdon, 1987; Jarosz & Davelos, 1995). Alternatively, one could consider population-level consequences such as shifts in allocation to male or female gametes in healthy individuals in populations with low vs high disease frequency. The success of an individual's gametes depends on the gametes produced by other plants in the population (Lloyd, 1976). If a pathogen differentially affects female or male reproduction, then fitness achieved through each gamete type will be altered, and selection will favor the ‘rare’ gamete in that population. This approach is important in understanding how pathogens may contribute to selection on sex allocation in cosexual morphs and on breeding systems.
Anther smut disease, caused by Microbotryum violaceum, is a systemic fungal pathogen that infects species in the Caryophyllaceae, and causes reproductive sterility in infected plants. Host species vary in breeding systems from cosexual to dioecious. Silene acaulis is one host of anther smut and has a gynodioecious breeding system. In the North American Rocky Mountains, S. acaulis has two sexual morphs: female and hermaphrodite (Shykoff, 1988; Delph & Carroll, 2001). Gynodioecy is not a discrete breeding system, but rather represents a continuum from cosexual populations, where each individual, on average, contributes genes equally through pollen and ovules, to near dioecious populations, where an individual contributes genes exclusively through either pollen or ovules (Lloyd, 1976). This continuum reflects the relative performance of an individual as a maternal and paternal parent (i.e. its functional gender). The proportion of fitness transmitted through ovules is referred to as female function (Gi), which ranges from 0 (all fitness gained through pollen) to 1(all fitness gained through ovules) (Lloyd, 1979).
An initial census of 14 sites showed that anther smut disease frequency was positively correlated with female frequency (r = 0.23), suggesting the hypothesis that females and hermaphrodites are affected differently by anther smut disease. There are three ways in which plants infected with a disease that causes reproductive sterility could alter the functional gender of healthy hermaphrodites in a gynodioecious population. First, if females were infected at a greater frequency than hermaphrodites, then the pollen to ovule ratio would increase. This would decrease the relative fitness gained through pollen for hermaphrodites, and there would be no selection for hermaphrodites to decrease their female function. Hence, the breeding system would shift towards the cosexual end of the gynodioecy spectrum. Second, if both sex morphs were infected at equal frequencies (assuming equal effects of infection on pollen and ovules), there would be no change in the pollen to ovule ratio, resulting in no change in functional gender caused by the presence of diseased individuals in the population. Third, if hermaphrodites were infected at a greater frequency, then the pollen to ovule ratio decreases. This should select for increased male function in uninfected hermaphrodites and could drive the population closer to dioecy. I tested which of these outcomes is occurring in populations of Silene acaulis with relatively high frequencies of anther smut disease by addressing the following questions. Are females and hermaphrodites equally likely to be infected by anther smut in natural sites? Does the proportion of fitness gained through pollen and ovules differ for healthy hermaphrodites in areas with many diseased neighbors compared with areas with few diseased neighbors?
Materials and Methods
Natural history and study sites
Silene acaulis L. ssp. subacaulescens (F.N. Williams) C. L. Hitchc. & Maguire, is a long-lived perennial that occurs throughout the Northern hemisphere in arctic and alpine tundra ecosystems (Gleason & Cronquist, 1991). Both sex morphs (females and hermaphrodites) in S. acaulis can be systemically infected by M. violaceum (Pers.) Deml. & Oberw. (= Ustilago violacea Pers.), and both are reproductively sterilized as the fungus disrupts ovary development and inhibits pollen production (Fisher & Holton, 1957). Stamens develop to full size (in both sexes), but the anthers are filled with diploid spores rather than pollen. Spores are transmitted from plant to plant by pollinators. Queen bumblebees, Bombus sylvicola, are the most common pollinator of S. acaulis in Colorado, USA, (Shykoff, 1992). Spores germinate on floral styles, undergo meiosis, and then mitotically bud until they fuse with a compatible mating type (Fisher & Holton, 1957). Floral infection is the major mode of transmission for this disease (Alexander & Antonovics, 1988; Jennersten, 1988) but, unlike other hosts, vegetative infection has not been observed in S. acaulis (Marr & Delph, 2005). In S. acaulis, switches between diseased and healthy status were rare; fewer than 1% of > 2500 plants studied gained or lost infection over an 8-yr period (Marr & Delph, 2005).
Field work was carried out on Pennsylvania Mountain in central Colorado (Park County, 39°15′ N, 106°15′ W). Four sites were chosen to represent the altitudinal gradient where anther smut occurs on Pennsylvania Mountain. Two plots, 6 × 10 m, were established at each site (for a total of eight plots) such that one plot had few diseased plants (1–5%) and the other plot had a relatively high number of diseased plants (18–26%) (Table 1). Density of S. acaulis ranged from 2 to 27 plants m−2 per plot (Table 1). Within each site, plots were separated by 15–40 m and were selected to represent typical S. acaulis patches for that local area. Studies of genetic structure of S. acaulis have shown that these subpopulations are structured on the scale of meters (Gehring & Delph, 1999).
Table 1. Sex ratios of healthy and diseased Silene acaulis for eight plots on Pennsylvania Mountain, Park County, CO, USA
Density (plants m−2)
Sex ratio of healthy plants
Sex ratio of diseased plants
The sex ratio of disease plants was based on the number of plants for which sex could be unambiguously determined. The number sex known per total number of diseased plants in the plot is listed in parentheses underneath the sex ratio of diseased plants. h, Hermaphrodites; f, females.
A χ2 goodness-of-fit test was used to test the null hypothesis that the sex ratios of healthy and diseased plants are equivalent.
70% h, 30% f
73% h, 27% f (15/19)
57% h, 43% f
67% h, 33% f (21/24)
66% h, 34% f
57% h, 43% f (7/12)
60% h, 40% f
59% h, 41% f (46/116)
77% h, 23% f
72% h, 28% f
76% h, 24% f (17/37)
69% h, 31% f
67% h, 33% f
68% h, 32% f (75/299)
Sex ratio was measured as the number of healthy flowering female and hermaphroditic plants (only healthy plants can contribute to the pollen and ovule pool for the population), and was recorded for each of the eight plots for 4 yr. The sexual identity of diseased S. acaulis individuals cannot be determined directly since both sexes respond to infection by producing full-size stamens and small ovaries. To determine the sex of diseased individuals, shoots were collected and raised in the glasshouse at Indiana University, Bloomington, IN, USA. Plants were subjected to temperatures of 35–40°C, which caused the plant to lose infection at least temporarily (Fisher & Holton, 1957; Alexander & Maltby, 1990), and in many cases plants lost infection for the duration of the study (D. Marr, pers. obs.). The sex of the ‘cured’ plants was based on observation of at least six flowers, because the first flowers are often pistillate even though subsequent flowers may be perfect. Approximately 25% of the shoots collected from the field survived long enough to monitor healthy flower production. Shoots were collected from diseased plants for three summers to increase sample sizes. In addition, partial disease expression in the field, where some flowers did not exhibit disease symptoms, enabled the sex to be determined for some diseased plants. For plots with adequate sample sizes, χ2 analysis was used to compare the sex ratio of healthy plants with that of diseased plants.
Estimating functional gender
The average female function of hermaphrodites was estimated for each of the eight plots for two years (1993 and 1994). In populations where outcrossed and selfed seeds differ in quality, not all ovules and pollen have equal chances of contributing to the next generation. Therefore, the number of seeds must be weighted by their fitness from estimates of inbreeding depression and the number of seeds produced by selfing and outcrossing (Lloyd, 1980). Lloyd (1980) developed a model for quantifying the proportion of genes transmitted through ovules, G, in populations with inbreeding depression. I modified Lloyd's equivalence factor, Ex, to include both females and hermaphrodites. A brief explanation of how data were collected to estimate each variable is provided here (a more detailed explanation of the equation and calculations is provided in Appendix 1).
Flower, fruit and seed production (mix, diH, diF)
Flower production by hermaphrodites (mix) was used as an estimate of pollen production. The total number of flowers and fruits produced by 30 females and 30 hermaphrodites was counted in each plot in 1993 and 1994. Seed production per plant (di) was used as an estimate of the fitness gained through ovules for both hermaphrodites (diH) and females (diF). Total seed production was determined for each of these plants by counting seed from 80 to 100% of the fruits produced by the plant. In plants where a subset of seed was counted, total seed production was estimated from the total number of fruits and the mean number of seeds per fruit. The sex ratio of healthy plants was used to scale this data to the relative proportion of seeds (diH and diF) and polleniferous flowers (mix) produced by each sex morph in each plot.
The effect of site, sex, and disease level (fixed factors) on flower, fruit, and seed production was analysed using a general linear model for each dependent variable (flower, fruit and seed) (GLM procedure in SPSS 11.0 for Macintosh; SPSS Inc., Chicago, IL, USA). Disease level was nested within site, and interaction terms for site × sex and sex × disease level were included in the model; type III sum of squares were used. Number of flowers and number of seeds per plant were normally distributed, but did not meet the assumption of equal variances. Neither log nor square-root transformations improved homogeneity of variances. Analyses were done separately for each year to reduce heterogeneity of variances.
An analysis of variance with year, sex, and disease frequency was also calculated to determine differences in reproduction between years. To control for the total number of flowers made per plant, fruit production was analysed as a proportion, and was arcsine square-root transformed (Zar, 1996). To examine significant interactions between site and sex, a composite variable of site and sex was made and one-way analyses of variance were used to test for differences between the sexes at each plot. To examine significant interactions between disease frequency and sex, a composite variable of sex and plot was made and one-way analyses of variance were used to test for differences between low disease and high disease plots within sites. Kruskal–Wallis anova using ranked data gave similar results as a Welch anova. The results from a Welch anova (assuming unequal variances) are presented here. A sequential Bonferroni adjustment for alpha levels was used to determine the significance levels for all sets of anovas (sequential alphas were 0.006, 0.007 and 0.01). All statistical analyses were performed either by hand or with SPSS 11.0 for Macintosh computers (SPSS Inc. 1999).
Quality of seed produced by hermaphrodites and females (qix, Ii and qx)
Both the number of seed produced by hand-pollinations and the percentage of seeds surviving to 5 wk were included in the estimate of seed quality for each plot. The quality of hermaphroditic outcrossed seed (qix) and selfed seed (Ii) was calculated relative to the quality of female outcrossed seed (see Appendix 1). Estimates of population seed fitness for hermaphrodites (qx) were based on relative quality of open-pollinated seeds produced by hermaphrodites compared with females.
Outcrossed and selfed pollinations were done in June and July of 1994 in the following manner. At each site, 20 females and 20 hermaphroditic plants were chosen near the established plots. At the Summit site, 30 females and 26 hermaphrodites were chosen to increase the likelihood of obtaining seeds from each cross type. At all sites, pollinators were excluded by covering flower buds about to open with an inverted 1.5 ml centrifuge tube, and flowers were pollinated with freshly dehisced pollen collected from two hermaphroditic plants at least 10 m away. On hermaphrodites, six flowers were emasculated before the anthers dehisced; three of these flowers received selfed pollen and the other three received pollen collected from two hermaphroditic plants at least 10 m away. Females were only pollinated with outcross pollen. Pollinator-exclusion caps were removed when styles were no longer receptive. Hand-pollinated fruit were collected when ripe and seeds were counted. In addition, open-pollinated fruit were collected from each plant and seeds were counted.
Up to 10 filled seeds from each cross per individual (i.e. outcross, self, and open) were planted in vegetable plug mix, bottom-watered and placed in a glasshouse room kept at 26°C in October 1996. Seeds were monitored for germination and survival once a week for 6 wk. At the Summit site, 26 hermaphroditic plants received the hand-pollination treatments, but many of the flowers did not produce any seed, which reduced the sample size to 18 plants with open-pollinated seed, 14 with outcrossed seed, and eight with selfed seed. At all other sites, sample sizes ranged from 19 to 25 female and 18–24 hermaphroditic maternal plants. For hermaphrodites, a total of 700 open-pollinated seeds, 663 outcrossed seeds and 593 selfed seeds were planted, and for females 710 open-pollinated seeds and 695 outcrossed seeds were planted.
The effect of pollination treatment (outcrossed, selfed and open-pollinated) on seed number in hermaphrodites was analysed using a two-way anova with site and pollination treatment as fixed factors. A two-way anova analysing effect of site and pollen treatment on seedling survival in hermaphrodites was also performed. Seed number was square-root transformed and percentage of seedlings surviving was arcsine square-root transformed; both transformed variables met assumptions of normality and equal variances. Differences between females and hermaphrodites in seedling survival were analysed for outcrossed and open-pollinated seed using a two-way anova with site and sex as fixed factors.
Outcrossing rates (ti and si)
Outcrossing rates were estimated for six plots by assaying four polymorphic isozyme loci from approx. 20 hermaphroditic plants and 10 of their offspring using cellulose acetate gels. For the maternal plants, shoots were collected and used for isozyme analysis. For offspring, all of the seed produced by the maternal plant was collected, combined and thoroughly mixed, then 20 filled seeds from this mix were planted. Seedlings were grown for 1–2 wk before being assayed. Four polymorphic loci were used that showed consistent banding intensity for both adults and seedlings: 6-phosphogluconate dehydrogenase (6-Pgd-1), glutamate oxaloacetate transamine (Got-1), phosphoglucose isomerase (Pgi-2) and shikimic dehydrogenase (SkD). Loci were numbered according to their migration rates, with the most anodal (fastest) being assigned number 1. The electrophoresis methods used in this study are described in Gehring & Delph (1999).
In this population of S. acaulis, isozyme variation was low and most of the polymorphic loci had one very common allele (> 90%). This may cause outcrossing events to be underestimated. Inbreeding depression is unlikely to have strongly biased outcrossing rates because seedlings were harvested early and seed germination was high (75–85%); except for seeds from the Summit site where families varied in germination rate from 0 to 80%. Previous studies have also shown no difference in germination between outcrossed and selfed seed (Shykoff, 1988) or between hermaphrodites and females growing in the glasshouse (Delph & Mutikainen, 2003). The Newton–Raphson method in Ritland's multilocus mating system program (MLTR) was used to estimate outcrossing rates (Ritland, 1986; Ritland, 1990). Standard errors of outcrossing values were estimated using 200 bootstraps where family was the unit of resampling.
Estimating proportion of fitness gained through seeds in hermaphrodites (Gi)
Average functional femaleness was calculated for hermaphrodites in each low disease and high disease plot using the mean values for each variable in 1993 and 1994. Standard errors were estimated by jackknifing the data (Dixon, 1993). Hermaphrodite seed production (diH) was chosen as the unit of resampling because it represents the best indicator of fitness gained through ovules. Standard error estimates were based on resampling the data for 29–50 plants using Microsoft Excel version 5.0 for Macintosh; sample sizes are provided in Fig. 1. Because outcrossing and seed quality estimates were only made in 1994, 1993 G coefficients were calculated using the lowest and highest mean outcrossing values (0.55 and 0.91) and the lowest and highest mean seed quality (outcrossed = 0.77 and 1; selfed = 0.60 and 0.97) from the 1994 data. Thus, 1993 G estimates have larger standard errors than 1994 G estimates.
The sex ratio of diseased plants was not significantly different from the sex ratio of the surrounding healthy plants (Table 1). The sex ratio of healthy plants varied among plots from 23 to 43% female. Within plots, sex ratios varied less than 2% during the 5 yr of census (1992–96), and disease level remained constant among years (Marr & Delph, 2005). High-disease plots tended to have more females than the low-disease plots, although these differences were small at False Summit and Summit sites.
Reproduction in eight natural plots over 2 yr
Differences in flower and fruit production in sex morphs and low- and high-disease plots Sites differed significantly in flower and fruit production in both years (Fig. 1). There was no significant effect of disease frequency on flower production in 1993, but there was a significant effect in 1994 when flower production was at least threefold higher in most plots. Flower production did not differ consistently between the sexes between years (Fig. 1a). Hermaphrodites produced more flowers than females in 1993, but there were no significant differences between females and hermaphrodites in total flower production in 1994 in any of the eight plots (GLM model 1993 flower production: site F3,520 = 3.1, P = 0.03, Sex1,520 = 13.5, P < 0.0001, sex × disease1,520 = 4.6, P = 0.03, disease levels nested within site and site × sex were not significant; 1994 flower production: site F3,480 = 13.0, P < 0.0001, disease level nested within site F3,480 = 12.7, P < 0.0001, sex × disease F1,480 = 3.0, P = 0.08, sex and site × sex were not significant).
Females produced a higher percentage of fruits than hermaphrodites in all plots in 1994 (Fig. 1b). In 1993, females produced significantly more fruit than hermaphrodites in only one of eight plots. Fruit set differed significantly between low disease and high disease plots, although the direction of these differences differed among sites and across years (Fig. 1b, GLM model 1993 proportion fruit set: site F3,520 = 18.6, P < 0.0001, disease level nested within site F3,520 = 37.3, P < 0.0001, site × sex F3,520 = 6.9, P < 0.0001, sex × disease F1,520 = 17.7, P < 0.0001, sex not significant with P = 0.07; 1994 proportion fruit set: site F3,480 = 118.7, P < 0.0001, sex F1,480 = 143.8, P < 0.0001, disease level nested within site F3,480 = 7.8, P < 0.0001, site × sex and sex × disease were not significant).
Seed production in low disease and high disease plots and across years
Seed production was highly variable across years (Fig. 2). Plant size did not differ between 1993 and 1994 within plots (D. Marr, unpublished). Thus, differences in flower and fruit production between years cannot be explained by differences in plant size. More flowers were produced in 1994 than in 1993 (Fig. 1a, effect of year on flower number anova F1,994 = 87.2, P = 0.0001). Fruit-set and seed set were also higher in 1994 compared with 1993 (Figs 1b and 2; effect of year on fruit set anova F1,994 = 698.6, P < 0.0001, seed set anova F1,993 = 51.4, P < 0.0001). However, the total number of seeds per plant was not consistently different between females and hermaphrodites as indicated by the significant interactions between sex × site and sex × disease in both years (Fig. 2, Table 2).
Table 2. Differences in number of seeds per plant among sites, between sex morphs, and between low vs high disease frequency plots
Site, sex, and disease level was nested within site, and were considered fixed factors in the general linear model. Seed number was normally distributed, but did not meet assumption of homogeneity of variances. Years were analysed separately to reduce heterogeneity of variances. Significant values are in bold type.
Sex (female or hermaphrodite)
Disease level nested within site
Site × sex
Sex × disease
R2 for model = 0.153
Sex (female or hermaphrodite)
Disease level nested within site
Site × sex
Sex × disease
R2 for model = 0.230
The ratio of female to hermaphrodite seed production was calculated for each plot by weighting average seed production per sex morph by its frequency in the population (Fig. 3). This figure shows that in both years females produced more seeds in high disease plots than in low disease plots (Wilcoxon paired t-test Z = −2.52, df = 7, two-tailed P = 0.012). At the Summit high disease plot, the ratio was above one for both years, indicating that females produced more seeds than hermaphrodites even though there were fewer females than hermaphrodites (sex ratio was 33% females to 67% hermaphrodites).
Quality of hermaphroditic seed relative to female seed
The relative fitness of hermaphroditic seed was generally lower than female seed whether flowers received outcrossed pollen, self pollen or were open-pollinated, indicating that inbreeding depression is present at these sites (Table 3). There were significant differences among sites and among pollen treatments in both the number of seeds produced and in percentage seedling survival (Table 4). The significant pollen source effect resulted from low seed numbers and seedling survival in the self-pollinated treatment compared with outcross and open-pollinated treatments (Scheffe's contrast: number of seeds P < 0.0001, seedling survival P = 0.017). The effect of site for both hermaphrodites and females (Tables 4 and 5) was primarily caused by low seed production and survival at the Summit. For example, in females 25% of the Summit seedlings survived to 5 wk compared with 62–70% at the other sites (Fig. 2 shows open-pollinated seed production per plot).
Table 3. The quality of hermaphroditic selfed, outcrossed and open-pollinated seed relative to female outcrossed seed
Quality of outcrossed seed (qix) Mean (low–high)
Quality of selfed seed (Ii) Mean (low–high)
Quality of open-pollinated seed (qx) Mean (low–high)
Seed quality estimates (measure of inbreeding depression) were based on number of seeds per fruit and percentage seedling survival to 5 wk. A low and high seed quality estimate was calculated using values based on ± 1 SE of the mean. Seed quality ranges from 0 (hermaphrodites gained zero fitness from seeds) to 1 (fitness of hermaphroditic seed equal to female seed). Sample sizes were 20 hermaphrodites and 20 females per site, except for the summit site in which 30 females and 26 hermaphrodites were pollinated (see the Materials and Methods section). Refer to Appendix 1 for details on calculations.
Table 4. Effect of pollen treatment (self, outcross, or open-pollinated) and site on seed number per fruit and seedling survival in hermaphrodites
Pollinations were carried out at four sites; Slope, Flat, False Summit and Summit sites, and germinations were done in the glasshouse. Significant values are in bold type.
Seed number per fruit
Proportion of seedlings surviving to 5 wk
Table 5. Comparison of seedling survival to 5 wk in females and hermaphrodites receiving either outcross pollen or open-pollinated
Site and maternal sex (female or hermaphrodite) were considered fixed factors in the two-way anova. Significant values are in bold type.
Fewer open-pollinated seed from hermaphrodites survived compared to open-pollinated seed from females, but there was no difference in survival between the sex morphs when hand-pollinated with outcross pollen (Table 5). Open-pollinated hermaphroditic flowers may have received both self and outcross pollen. The seed quality measure (qx) is based on the open-pollinated seeds, thereby taking into account this difference between females and hermaphrodites.
Outcrossing rates and functional femaleness
Outcrossing rates varied from 0.55 to 0.91 (Table 6). The variability in outcrossing rates was striking and differences within sites were as great as among sites. However, there was no correlation between outcrossing rate and disease frequency. Differences in outcrossing rates can partly be explained by pollinator activity and differences in timing of flowering between plots (D. Marr, pers. obs.).
Table 6. Multilocus and single-locus estimates of outcrossing rates in Silene acaulis for 1994
Multilocus t estimate
Single-locus t estimate
Mean (± 1 SE) based on 200 bootstraps.
The average proportion of fitness hermaphrodites gained through seed production varied among plots from 25 to 48% in 1993, and from 27 to 42% in 1994 (Fig. 4). Hermaphrodites in high disease plots tended to have lower G coefficients (more fitness gained through pollen) than hermaphrodites in low-disease plots, although G estimates for low disease and high disease plots overlapped at some sites. G coefficients were similar between years at the Slope and Flat sites, but differed across years at False Summit and Summit sites. Sex ratio and disease occurrence remained constant across years, so changes in G are caused by seed production differences in both females and hermaphrodites.
In this study both sex morphs appear to be infected with anther smut disease in proportion to their frequency in the population. This result does not support the idea that presence of individuals with anther smut disease directly affects the functional gender of hermaphrodites. However, the functional gender of hermaphrodites was consistently lower in high disease plots compared with low disease plots in both years of this study. In general, hermaphrodites had relatively high G-values (25–48% fitness from seeds), and are therefore closer to the cosexual end of the gynodioecy spectrum in this population of S. acaulis. Another study of a S. acaulis population located c. 180 km miles from this study population found that S. acaulis ssp. subacaulescens was also near the cosexual end of gynodioecy (Delph & Carroll, 2001). By contrast, studies of S. acaulis ssp. exscapa and cenisia indicate that populations in the French Alps are much closer to the dioecy end of the spectrum, and have three sexual morphs (Maurice et al., 1998). Of all the variables measured in this study, sex ratio and seed production were the most important factors in explaining the differences in functional femaleness observed across the eight plots. These factors are explored in more detail to better understand spatial variability in function gender and the apparent pattern of decreased female function in hermaphrodites in high-disease plots.
Female frequency varied nearly two-fold across sites (23–43% females). Spatial heterogeneity in sex ratio has also been observed in other populations of S. acaulis ssp. subacaulescens (Delph & Carroll, 2001) and in ssp. exscapa and ssp. cenisia (Hermanutz & Innes, 1994; Maurice et al., 1998). To determine how much sex ratio differences among plots contributed to functional gender estimates, calculations were done holding all variables constant except for sex ratio. A 13% difference in female frequency (32% vs 45% females) resulted in a 5–7% change in G (36% vs 41%G-value). At the Slope and Flat sites, the plots differed the most in female frequency (13% and 6%, respectively), thus the separation in their G-values is, in part, due to differences in sex ratio. The underlying causes of variability in female frequency are poorly understood, and are likely to be a combination of factors, including genetic factors determining sex (Frank, 1989; Dudle et al., 2001), localized seed dispersal and environmental factors that may favor one sex morph over the other (Lovett Doust et al., 1987; Delph, 1993; Delph & Carroll, 2001). Nevertheless, differences in sex ratio did not completely explain the pattern of G-values. Sex ratio and disease level remained constant from year to year, but the G coefficients changed.
The variation in functional gender across years and across some of the sites is largely caused by differences in seed production. Female seed production differed more than 10-fold between years in some plots. Hermaphrodite seed production did not differ as dramatically between years (greatest difference was 3.8 times more seed). Generally, the higher production of seed by females in 1994 lowered the G coefficients for hermaphrodites. This effect was most pronounced at both False Summit plots and the Summit low-disease plot (Fig. 2), where females produced 9–13 times more seed in 1994 than in 1993. Plant size and flower production did not differ between sex morphs, and plant size did not differ between years: thus, differences in seed production are primarily caused by environmental differences between years. A snowstorm that occurred July 1993 is partly responsible for the difference in seed set between years. Some of the variation in seed production across years could reflect sampling error. I was most interested in average G-values, so I did not sample the exact same group of individuals from one year to the next. However, individuals that did happen to be measured in both years followed the pattern of overall greater flower, fruit, and seed production in 1994 (e.g. some individuals produced five times more flowers in 1994 than in 1993).
Given the lack of difference between healthy and disease sex ratios and lack of correlation between outcrossing rates and disease frequency, it was somewhat unexpected that the ratio of female to hermaphrodite seed production showed a consistent pattern between low disease and high disease plots (Fig. 3). The relatively higher seed production by females in the high-disease plots resulted in a lower functional gender for hermaphrodites in these plots (Fig. 4). The higher seed production by females in the high disease areas was also consistent with the pattern of slightly higher female frequencies in high-disease areas compared with the low disease areas (Table 1). Based on 1994 pollen and spore deposition data, the difference in seed production between low disease and high disease plots does not result from pollen or spore deposition differences between plots or sex morphs (Marr, 1997).
Seed production differences between low disease and high-disease plots appear to be driven by local resources or microclimates that differentially affect seed production in females and hermaphrodites. If the resources important to plant reproductive success are also important to fungal success, then this may explain the appearance of a pattern in functional femaleness between the low disease and high disease plots. High disease sites flowered later than the low-disease sites because of differences in snow accumulation (Marr, 1997). The opposite pattern for female frequency and anther smut disease frequency was observed for S. acaulis on Baffin Island, Canada (Hermanutz & Innes, 1994). Low disease areas (< 2%) had high female frequencies (72–80%) and occurred in dry, well-drained sites, whereas the one high disease area (22%) was a moist site and had a relatively lower frequency of females (50%) (Hermanutz & Innes, 1994). Together, the Baffin Island study and the pattern of snow accumulation on Pennsylvania Mountain suggests that differences in soil moisture or cooler soil temperatures may be important in understanding why anther smut disease is more frequent in some areas than others.
Is anther smut disease an important selective agent acting on gynodioecious plant breeding systems?
Anther smut disease has been studied in two other hosts with a gynodioecious breeding system, Dianthus sylvestris and Gypsophila repens. In D. sylvestris, evidence for sex-specific differences in M. violaceum spore deposition differed between studies. One study found higher spore loads on females and pistillate flowers and higher seed predation in hermaphrodites (Collin et al., 2002), but another study did not find differences in spore load among the sex morphs (Shykoff et al., 1997). It is not known whether higher fungal spore loads result in sex-specific differences in infection frequency for this species. If high spore loads increase female susceptibility to anther smut disease, and hermaphrodites are more susceptible to seed predation (Collin et al., 2002), then natural enemies reduce the fitness of both sex morphs. In this case, the relative strength of seed predation and anther smut disease frequency would affect which sex morph had the fitness advantage.
Gypsophila repens can also be infected with M. violaceum, but unlike S. acaulis, individuals are frequently partly infected and infected individuals can produce seed. No difference in pollen limitation between sex morphs has been observed (Lopez-Villavicencio et al., 2003), but spore deposition affected seed production in hermaphrodites more severely than in females (Lopez-Villavicencio et al., 2005). If these seed production differences remain significant in field populations then, even if spore deposition patterns and risk of infection were the same for the sex morphs, one would predict that females would have a reproductive advantage in populations with a high frequency of anther smut disease. This is not the case in S. acaulis because seed production is depressed in both healthy hermaphrodites and females that receive both pollen and spores (Marr, 1998).
In summary, anther smut disease does not appear to affect the balance of fitness gained through pollen and ovules in S. acaulis. However, the average fitness achieved by hermaphrodites through male and female gametes varied significantly both spatially and temporally. Interestingly, spatial heterogeneity at the scale of a few meters seems important in explaining both the patterns of disease incidence and reproductive success of sex morphs observed in S. acaulis. Based on the results presented here, experiments that tease apart the effects of resources and pathogens on progeny fitness would be critical to understanding sex allocation in hermaphrodites.
I thank K. Clay, S. Davis, L. Delph, D. Dudle, G. Gastony, C. Lively and three anonymous reviewers for critically reading and providing comments on earlier versions of the manuscript. Special thanks to L. Delph for many discussions on functional gender that helped sharpen my thinking. I gratefully acknowledge D. Dudle for assistance in the field. The University of Colorado at Colorado Springs provided access to their Alpine Research Station on Pennsylvania Mountain. This work was supported by a grant from NSF to D. Marr (DEB-9411951) and by an American Fellowship from the American Association of University Women.
Model for functional femaleness (Lloyd, 1980) and explanation for how each variable was calculated in the equation.
Total number of seeds produced by hermaphrodites adjusted by sex ratio. Assuming 50 individuals and 70% hermaphrodites, then there would 35 hermaphrodites. If 7888 seeds were produced by 34 hermaphrodites, then diH = 8120 seeds.
Total number of seeds produced by females adjusted by sex ratio. Assuming 50 individuals in population and 30% females, would give 15 females. If 4321 seeds produced by 27 females, then diF = 2401 seeds.
Total number of pollen-producing flowers (e.g. hermaphrodites produced a total of 3049 flowers). Adjusting for 70%H: 30%F sex ratio, and assuming 50 individual in the population, then expect 2824.8 polleniferous flowers
Outcrossing rate = 0.67 ± 0.16
Selfing rate = 0.33 ± 0.16
Quality of hermaphroditic outcrossed seeds relative to female outcrossed seed, e.g. qi = number of outcrossed
seed × proportion of seedlings survived to 5 wk:
hermaphrodite qi = 10 × 0.7 = 6.3 seed
female qi = 10 × 0.8 = 7.2 seed
hermaphrodite qix = 6.3/7.2 = 0.88 fitness
female qix = 1.0 fitness
Quality of hermaphroditic selfed seeds relative to female outcrossed seed, e.g. selfed seed quality = number of selfed seed × proportion of seedlings survived to 5 wk, weighted by female seed quality:
10 × 0.6 = 5.4
Ii = 5.4/7.2 = 0.75 fitness
Population seed quality – similar to qix, except based on open-pollinated seed produced by hermaphrodites and females
Equates seed production with pollen production for the population being considered