| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
0 Botany Department, Duke University, P.O. Box 90338, Durham, North Carolina 27708-0338 USA
Received for publication May 13, 1999. Accepted for publication September 10, 1999.
ABSTRACT
Genotypic trade-offs between male and female reproduction are commonly assumed in theoretical studies of the evolution of gender specialization. Although these trade-offs are supported by higher seed production of females than hermaphrodites in natural populations of gynodioecious species, comparisons between male and female reproductive allocation among hermaphrodite individuals under controlled conditions are rare. We assessed phenotypic and genotypic correlations between stamen and fruit production in fruiting males of the near-dioecious herb Astilbe biternata. In the field, we found a significant negative phenotypic correlation between stamen production and fruit production within individuals that produced both stamens and fruit as well as higher fruit set in females than fruiting males. The negative correlation between fruit and stamen production that was observed in the field was also apparent across clonally propagated genotypes. These results suggest that negative genetic correlations between male and female reproduction may limit the independent evolution of fruit and stamen production in A. biternata.
Key Words: Astilbe • genetic correlation • gynodioecy • phenotypic correlation • reproductive compensation • Saxifragaceae • sex allocation • subdioecy
Negative genetic correlations between male and female reproductive output underlie theoretical models describing the evolution of sexual dimorphic species from hermaphroditic ancestors in plants (for review see Geber, 1999
). In the presence of these genetic trade-offs, selection resulting in increased reproductive output through one sex function also will result in decreased reproductive output in the other sex function. It is impossible to test for these genetic trade-offs in dioecious species because, by definition, all individuals are unisexual and none reproduce by both male and female modes. However, if negative genetic correlations between male and female reproductive output contribute to the evolution of dioecy from hermaphroditism, then species with "intermediate" breeding systems along the evolutionary pathway from hermaphroditism to dioecy should express these trade-offs.
Here we assess the presence of a negative genetic correlation between male and female reproductive output in Astilbe biternata, a subdioecious perennial plant. Subdioecious species comprise individuals that express male, female, or hermaphroditic phenotypes in natural populations and are considered to represent an intermediate step along one common evolutionary pathway between hermaphroditism and dioecy (Webb, 1999
). For clarity we will follow the precedent set by Lloyd (1976)
, Wolfe and Shmida (1997)
, and Barrett, Case, and Peters (1999)
and collectively refer to male and hermaphrodite phenotypes as "males" because the same individual can express male or hermaphroditic phenotypes in different years or environmental circumstances.
Most of the empirical studies that have assessed trade-offs between male and female reproductive output in males of subdioecious species have compared male and hermaphroditic phenotypic classes of individuals (Sakai and Weller, 1991
; Barrett, 1992
; Fleming et al., 1994
; Wolfe and Shmida, 1997
). Three of these studies found evidence apparently contradicting theoretical assumptions: two found that hermaphroditic and male phenotypes produced similar amounts of pollen (Sakai and Weller, 1991
; Fleming et al., 1994
), and one found that fruiting males produced more pollen than nonfruiting males (Barrett, 1992
). One study found evidence that male phenotypes produced more pollen than hermaphroditic phenotypes (Wolfe and Shmida, 1997
), supporting the presence of trade-offs between male and female reproductive output. However, none of these studies explicitly examined genetic correlations between male and female reproduction in hermaphrodites. Genetic correlations between male and female reproduction have been assessed in only two studies of gynodioecious species (Atlan et al., 1992
; Ashman, 1999
). In both cases evidence for negative genetic trade-offs were found.
In natural populations A. biternata expresses a breeding system that is close to dioecy, although we do not know the present evolutionary trajectory of the breeding system (i.e., if it is evolving toward gynodioecy or dioecy). Results from a study of 22 populations showed that females comprise fewer than half the individuals in most populations. Females are invariant in their gender expression from year to year, while males produce fruit in some years but not others. The proportion of males that set fruit usually represents <15% of all males, but occasionally >30% of males produce fruit in some populations (Olson, 1997a
). Although most males do not produce fruit in the field, almost all male genotypes studied produce fruit when moved from the field to a greenhouse environment, whereas females never produce pollen in field or greenhouse environments. In addition, throughout the range of A. biternata, males display variation in the proportion of flowers that produce stamens. Similar patterns of variation in population sex ratio and male sex expression are found in other subdioecious species that are thought to have evolved from gynodioecious ancestors (Delph and Lloyd, 1991
). If trade-offs between male and female reproduction are important for the evolution of increased maleness in hermaphrodites of gynodioecious species, we expect to find support for these trade-offs in A. biternata.
In this study, we investigate both phenotypic and genetic correlations between stamen and fruit production of males in natural and greenhouse populations of A. biternata. Because the flowering period of A. biternata is limited to 3 wk, we developed quick nondestructive methods to estimate whole-plant stamen and fruit production in the field. We take advantage of the ease of clonal propagation to investigate the presence of genetic correlations between whole-plant stamen and fruit production.
METHODS
Study species
False goat's beard, Astilbe biternata (Vent.) Britton, is a herbaceous perennial, that is endemic to the southern Appalachian Mountains and ranges from northern Georgia to southern West Virginia. It typically grows in the shaded understory of rich mesic cove forests and along nearby roadsides. The flowers are small (gynoecia are typically <3 mm in length), and hundreds to thousands of them (estimated average of 2451 flowers per inflorescence at the Coweeta site, see below) are borne in a large determinate inflorescence, which varies between 10 and 80 cm in length. Plants produce copious amounts of both nectar and pollen, and >30 different types of insects have been identified on the inflorescences during flowering. Of these, hymenoptera are thought to be the most efficient pollinators (Mellichamp, 1976
).
In the spring, new shoots arise from the overwintering corm, and plants flower in late June and set fruit by late July. Along roadsides, individuals can produce several flowering stems from a single corm. However, in understory populations individuals rarely produce more than one inflorescence per year. After the first freeze in the fall, the aboveground stems die back.
The genus Astilbe (Saxifragaceae) contains ~25 species, most of which are found in southeast and central Asia (Mellichamp, 1976
). Astilbe biternata is the only species in the genus that is native to North America. To our knowledge, A. biternata is unique in the genus for having individuals that produce unisexual flowers; all other species produce only hermaphrodite flowers. Inheritance patterns of allozyme alleles (Olson, 1997b
) and cytology (Hamel, 1953
) suggest that A. biternata is allotetraploid, but the progenitor diploid species are unknown.
Field studies
Marked individuals in two populations, Shope Fork and Ball Creek, at Coweeta Hydrological Laboratory near Otto, North Carolina, were censused for gender phenotypes from 1994 through 1997. Each year individuals were marked with flags as they flowered, and both previously marked plants and all additional flowering individuals were censused. By 1997, >900 individuals had been monitored for flowering and sex expression. For the first three years of the census, individuals were scored for percentage fruit set and the presence or absence of stamens. In 1997, percentage stamen production (see below) was also estimated.
Gender classification
Individuals that produced only unisexual pistillate flowers were classified as females. On female flowers, vestigial stamens could often be seen at the base of the ovary. These were very small, often brown and withered, and produced no pollen. In rare cases, females produced a few stray stamens with pollen that may have been viable; in all of these cases only one or two stamens were produced on <10 flowers on the entire inflorescence.
In the field an individual was classified as male when >10 flowers (~0.5%) on the inflorescence each produced ten stamens. All males in the field produced at least 20% of the maximum potential stamen numbers based on all flowers producing ten stamens.
Estimation of relative fruit set and stamen production
Flower production was estimated by measuring the length of the inflorescence (from the axil of the first inflorescence branch to the tip of the main axis of the inflorescence) because this was a good predictor of the log of the total number of flowers produced on the inflorescence (linear effect F1,71 = 29.3, P < 0.0001, quadratic effect F1,70 = 6.34, P < 0.02, R2 = 0.70; y = -0.00022x2 + 0.03957x + 2.10936). The relationship between inflorescence length and flower number was not significantly different for females, males that produced only pollen, and males that produced both pollen and fruit (gender x length: F2,65 = 1.06, P > 0.35; gender x length2: F2,65 = 0.78, P > 0.45; Olson, 1997a
). Moreover, in the field we found no significant differences among genders in the lengths of inflorescences (mixed-model ANOVA treating population and year as random effects: F = 0.74, P > 0.50; Olson, 1997a
). Because females and fruiting males also produced similar numbers of seeds per fruit (mixed-model F = 1.92, P > 0.20; Olson, 1997a
), we used percentage fruit set as a measure of reproduction through seed.
Because A. biternata inflorescences may contain several thousands of flowers, it was not feasible to count the exact numbers of fruit produced on the large numbers of marked individuals. Therefore, percentage fruit production was estimated for each inflorescence and placed into one of the 13 following categories: 0, 2, 10, 20, 30, 40, 50, 60, 70, 80, 90, 98, or 100%. Estimation of percentage fruit production was aided by the tendency for fruit to be spatially clustered on inflorescence branches. Thus, estimation could be done by counting the total number of inflorescence branches of roughly the same size as those that set fruit and then determining the proportion of the total that produced fruit. Inflorescences with fewer than ten fruits on the entire inflorescence were assigned to the 0% category. Inflorescences with between 10 and 90% fruit set were revisited and re-estimated (while blind to the value of the first assessment) within 1 d. Any difference between the first and second assessment were settled by a third, non blind assessment.
A stamen was considered present when it was exserted beyond the sepals. Stamens that were not exserted beyond the sepals were bathed in nectar at the base of the gynoecium and never dehisced. Although pollen from these anthers had viable protoplasts as indicated by Alexander's stain (Alexander, 1980
), this pollen was almost certainly unavailable for pollination. Thus, throughout the remainder of this paper, "stamen production" was measured as the proportion of stamens that were visibly exerted beyond the sepals.
Within an inflorescence, most male flowers had either zero or ten stamens and occasionally had flowers with 0–10 stamens. Most flowers on a particular inflorescence branch had the same number of stamens, either zero, 0–10, or ten. To estimate percentage stamen production, we first counted the number of similar-length inflorescence branches with flowers with zero, 0–10, or 10 stamens. The average stamen number on inflorescence branches with flowers with 0–10 stamens was estimated by counting stamens on ten arbitrarily chosen flowers and computing the mean. Percentage stamen production was calculated as the proportion of inflorescence branches with flowers that produced ten stamens plus the proportion of branches with intermediate stamen numbers multiplied by the average stamen number per flower on these branches (e.g., if 50% of the inflorescence branches produced flowers with four stamens each, these contributed 20% to the estimation of stamen production). Using this estimate, each inflorescence was assigned a percentage functional stamen production value in one of the 13 following categories: 0, 2, 10, 20, 30, 40, 50, 60, 70, 80, 90, 98, or 100%, based on the proportion of stamens exserted on the entire inflorescence. Inflorescences with between 10 and 90% stamen production were revisited to check the accuracy of initial assessments using procedures outlined above for estimation of fruit production.
Experimental populations
To assess variation and covariation in stamen and fruit production, 12 male genotypes were grown at Duke University. All of the genotypes originated from the Ball Creek population at Coweeta. Two genotypes (100 and 102) were maternal half-sibs grown from seed collected in 1993. The other ten genotypes were transferred as corms to Duke University in the early spring of 1994. All soil was washed from each corm, and each corm was divided into multiple ramets. After division, there was an average of 5.7 ramets/genotype (range 4–8), and ramets averaged 45.7 g (± 3.0 SE). Ramets were placed in separate pots and grown outside under a partial-shade lath house. This was an open-air facility covered by 5 cm wooden slats separated by 5 cm of open space resulting in ~50% direct sunlight when the sun was directly overhead. Flowering in 1995 was not considered so the results would not be confounded by carryover effects of the field environment.
In 1996, at least two ramets from each genotype (a total of 68 ramets) were grown in each of two environments in a completely randomized design with environment and genotype as treatments. The environments were (1) lath house/greenhouse-nutrient treatment consisting of alternating weekly application of 15–0–15 and 20–10–20 Peters Professional fertilizer normally used in the Duke University Greenhouses, and (2) lath house + shade cloth/low-nutrient treatment with all plants covered by 80% shade cloth in addition to the shading provided by the lath house and only fertilized once every 3rd wk. In 1997, 73 ramets (some of which were the same ramets as in the 1996 study) were grown only in the lath house/greenhouse-nutrient environment in an attempt to more accurately assess the effect of genotype on gender expression. In both years the positions of pots was randomized at the beginning of the growing season. It appeared that barriers to pollinator movement in the open-air lath house did not affect fruit production because, generally, females had >95% fruit set (Olson, 1997a
). Percentage stamen and fruit production were measured on all ramets in 1996 and 1997, using the same methods as in field populations.
Effects of the size of individuals on gender expression
To determine whether the size of individuals affected gender expression in the experimental populations, we measured both the mass of corms during the winters of 1996–1997 and 1997–1998 and the length of inflorescences from the axil of the first inflorescence branch to the tip of the main axis of the inflorescence in 1997. Corm masses were measured for each ramet by washing off all soil, drying with towels, and weighing each corm. Corms were immediately repotted with fresh soil. Regression analyses were performed to assess relationships between corm mass and inflorescence length on fruit and stamen production. All analyses were performed using JMP for Windows (SAS, 1995
).
Trade-offs among genotypes
Because there was not sufficient variation in stamen production among clones in 1997, only data from 1996 were used to test for genotypic correlations. The correlation between stamen and fruit production among genotypes was calculated using means of genotypes across clones. The bivariate distribution of percentage stamen and fruit production was not normally distributed, even when transformed (Shapiro Wilk's test P < 0.005; SAS, 1995
). We therefore used nonparametric Spearman's rank correlation to estimate and test significance of correlation coefficients (SAS, 1995
). Furthermore, because these variables were not normally distributed, we could not use standard statistical methods to partition variance into within- and among-genotype components, thus our estimate of genetic covariance is confounded by within-genotype variance (environmental variance). To examine the effects of within-genotype variability on the rank correlation, we performed a bootstrapping procedure whereby we randomly resampled (with replacement) ramets from the nine genotypes, recomputed genotype means, and recalculated the Spearman's rank correlation on 1000 pseudo data sets. The effects of this within-genotype variability was expressed as the 95% confidence interval produced by determining the 2.5 and 97.5% quantiles of the Spearman's rank correlations from the bootstrapped data sets.
To assess the shape of the relationships between stamen and fruit production in the field and experimental populations, we fit the function {f(x) = 1 - x
} (where f(x) = % fruit production and x = % stamen production) to the data. We chose this function because as varies from 0 to +
the equation approximates the concave ( < 1.0), linear ( = 1.0), and convex ( > 1.0) fitness trade-offs (Charnov, 1982
). Moreover, the endpoints are fixed at {0,1}(all allocation to female fitness) and {1,0}(all allocation to male fitness). Fitting was performed by finding the value of that minimized the squares of the distances between the data and the function using the NonlinearFit command in Mathematica (Wolfram, 1991
). The 95% confidence intervals (CI) were estimated for by estimating a pseudo (using sum of squares criterion) for each of 1000 bootstrapped data sets and determining the 2.5 and 97.5% quantiles of the pseudo values. For estimates of phenotypic confidence intervals, bootstrapping was performed by resampling the original data set with replacement. Estimates of genotypic confidence intervals were calculated using pseudo data sets produced by sampling with replacement within genotypes to calculate a new pseudo data set of genotype means.
RESULTS
Field studies
Between 1994 and 1997 at Ball Creek, 487 individuals were monitored. Of these, 48.0% were female, producing only fruit, 41.9% produced only stamens and no fruit, and 10.1% produced both stamens and fruit. At Shope Fork, 416 individuals were monitored. Of these, 45.4% were female, 49.4% produced only stamens, and 4.8% produced both stamens and fruit. Sex ratios (male : female) were not significantly different from 1:1 at Shope Fork (
2 = 3.47, df = 1, P > 0.05) or at Ball Creek (2 = 0.74, df = 1, P > 0.75). There was no statistically significant variation in sex ratio (male: female) over years at Ball Creek (2 = 4.9, df = 3, P > 0.15) or Shope Fork (2 = 3.9, df = 3, P > 0.25).
Reproductive compensation
We assessed reproductive compensation in A. biternata by comparing female fruit set with the fruit set of fruiting males at both Shope Fork and Ball Creek each year from 1995 to 1997. Averaged over 1995, 1996, and 1997 in both the Ball Creek and Shope Fork populations, females produced 2.24 times higher percentage fruit set than fruiting males (means: females = 89.5%, fruiting males = 38.3%, Kruskal Wallis 2 = 63.8, P < 0.0001). Because many females in 1997 produced 100% fruit set, residuals from parametric analyses violated the assumptions of normality even when transformed. Therefore, we constructed six separate nonparametric Kruskal-Wallis tests for each population and year association and used Bonferroni criteria to adjust the significance level for multiple analyses ( = 0.05/6 = 0.008). In 1995, in both populations the fruit set of females was higher than that of fruiting males and approached significance (both Kruskal-Wallis tests, 0.05 > P > 0.008). In 1996 and 1997, females produced significantly higher fruit set than fruiting males in both populations (all Kruskal-Wallis tests, P < 0.002).
Phenotypic correlation
The majority of males that flowered in 1997 produced ten stamens per flower on all of their flowers (e.g., 100% stamen production). No males (out of 104) at Shope Fork had reduced stamen production, while 26 males (out of 123) at Ball Creek had reduced stamen production.
Fruit set for males was inversely correlated with inflorescence-wide percentage stamen production (Fig. 1; Spearman's Rho = -0.75, P < 0.0001), indicating a trade-off in male and female reproductive output. The correlation was still present when only individuals that produced both stamens and fruit were included in the analysis (Spearman's Rho = -0.57, P < 0.005). When we fit the function f(x) = 1 - x to the data, the shape of the relationship between the phenotypic estimates of percentage stamen and percentage fruit production was not significantly different from linear (Fig. 1; the confidence intervals overlapped = 1.0; best fit to data = 1.05, lower CI = 0.57, upper CI = 1.8). At Shope Fork, only two males produced fruit in 1997; each had 100% stamen production and 10% fruit production.
|
2 = 0.01, df = 1, P > 0.90) or fruit production (Kruskal-Wallis 2 = 0.30, df = 1, P > 0.55). However, when all ramets from both environments were pooled by genotype, genotype accounted for a significant amount of variation among ramets in both stamen (Kruskal-Wallis 2 = 22.2, df = 8, P < 0.005) and fruit production (Kruskal-Wallis 2 = 18.5, df = 8, P < 0.02).
In 1997, all ramets were grown in the lath house/greenhouse-nutrient environment. Forty-two ramets flowered (57.5% of the total); all but one produced 100% stamens and 36% produced fruit. Overall, flowering ramets averaged 97% stamen and 15% fruit production. Each of nine different genotypes were represented by at least two flowering ramets; however, only one genotype had ramets that produced <100% stamen production. In 1997, genotype accounted for a significant amount of variation among ramets in both stamen (Kruskal-Wallis 2 = 40.0, df = 8, P < 0.0001) and fruit production (Kruskal-Wallis 2 = 25.9, df = 8, P < 0.002).
Effects of the size of individuals on gender expression
We found no relationships between the mass of the corms and proportions of stamen or fruit production per inflorescence. Unfortunately, we did not know the masses of corms before the 1996 experiment. However, differences in stamen and fruit production per ramet in 1996 were not significantly related to the mass of the corm at the end of the 1996 growing season (regressions: stamens F = 0.49, P > 0.45; fruit F = 2.01, P > 0.15). For 1997–1998, we calculated relative growth rate as (RGR = (corm mass 1998 - corm mass 1997)/(corm mass 1997)). The difference in RGR between ramets that produced both stamens and fruit and ramets that produced only fruit was not statistically different (F1,29 = 0.35, P > 0.55; RGR means: stamens only = +0.027, stamens and fruit = -0.031). In 1997, we also found no effect of preflowering corm mass on the percentage fruit production of ramets (ANCOVA accounting for genotype, F1,36 = 0.002, P > 0.95).
To test whether fruit production was associated with the total number of flowers produced, using ramets flowering in 1997, we regressed the length of the inflorescence onto the percentage fruit production within each genotype. After adjusting for genotype, individuals with longer inflorescences produced lower fruit set (F1,18 = 5.28, P < 0.04).
Trade-offs among genotypes
For the nine male genotypes with at least two flowering ramets in 1996 (average ramets/genotype = 2.67), there was a significant negative rank correlation among genotype means for stamen and fruit production (Spearman's Rho = -0.80, P = 0.01; Fig. 2) supporting the presence of a negative genotypic correlation between these traits. Ninety-five percent of the Spearman's rank correlations of genotype means calculated from 1000 bootstrapped pseudo data sets fell between -0.67 and -0.86. The lower of these values is expected to occur less than one time in 100 when the data are uncorrelated (Steel and Torrie, 1980
). This indicates that the within-genotype (environmental) variation was small relative to the among-genotype variation and, in particular, it was not sufficiently large to negate the among-genotype negative correlation. In the experimental populations, the shape of the relationship between the genotypic mean estimates of percentage stamen and percentage fruit production was convex (Fig. 2; the confidence intervals did not overlap with = 1.0; best fit to data = 3.90, lower CI = 2.5, upper CI = 3.9).
|
The phenotypic trade-off between stamen and fruit production in field populations of Astilbe biternata was exhibited both as reproductive compensation in fruit production when we compared females and fruiting males and as an inverse phenotypic correlation between fruit set and stamen production within males. Using experimentally manipulated genotypes, we also found that percentage fruit and stamen production were negatively correlated, suggesting that a genetic correlation underlies the phenotypic correlation between stamen and fruit production found in natural populations. Theoretically, the presence of the negative genetic correlation between stamen and fruit production may contribute to the evolution of dimorphic gender expression in A. biternata. In particular, selection for increased allocation to male reproduction may lead to a corresponding decrease in allocation to female reproduction. Our data therefore lend support to the assumptions underlying theoretical models of the evolution of gender specialization in plants (Charnov, Maynard Smith, and Bull, 1976
; Charlesworth and Charlesworth, 1978, 1981
; Charnov, 1982
).
The confidence intervals determined for the field and experimental populations did not overlap, indicating that the functional relationships between stamen and fruit production differed between the field and experimental populations. This difference may have resulted from variation between the field and experimental environments. The experimental environments probably had higher light availability, warmer growing temperatures, and higher inputs of nutrients than field environments. These differences suggest that our estimates of genetic variation and genetically based correlations between characters would be different than if they had been measured in a natural environment, although it is difficult to predict the magnitude or direction of the differences (Primack and Antonovics, 1981
; Mitchell-Olds and Rutledge, 1986
; Mazer and Schick, 1991
). For example, greenhouse studies may detect larger amounts of genetic variation than studies in natural environments (Mitchell-Olds and Rutledge, 1986
; Mazer and Schick, 1991
) because greenhouse environments are considered more uniform, thus decreasing the contribution of environmental variance to the phenotypic variation. However, when studying demographic traits, greenhouse studies may detect smaller amounts of genetic variation than studies in natural environments because the high nutrient conditions allow all individuals to reach their maximum potential, therefore no variation in the trait is observed (Primack and Antonovics, 1981
).
In the field environment, as males produce more fruit the allocation to stamen production decreased linearly. This pattern suggests that coexistence of hermaphroditic and unisexual genotypes may be evolutionarily stable, assuming that percentage stamen production reflects lifetime male fitness and percentage fruit production reflects lifetime female fitness (Charnov, Maynard Smith, and Bull, 1976
; Charnov, 1982
). In the experimental environment, however, the trade-off between percentage fruit production and percentage stamen production was convex and significantly different from a 1:1 linear trade-off. Here theoretical models predict that genotypes able to allocate to both male and female function would be favored over unisexuals because the combined reproductive output (percentage stamen + percentage fruit production) of fruiting males surpassed the reproductive output of nonfruiting males or females (see Fig 2.; Charnov, Maynard Smith, and Bull, 1976
; Charnov, 1982
)). Taken together, these results suggest that environmental differences affecting resource limits may influence the evolution of the breeding system in A. biternata. Other studies of subdioecious species have also shown effects of resource limits on gender expression (Delph and Lloyd, 1991
; Delph, 1993
; Wolfe and Shmida, 1997
), suggesting that understanding the evolutionary transition from gynodioecy to dioecy may partly rely on a better understanding of the effects of the interplay between genetics and environment on gender determination.
Although our data do indicate negative genetic correlations between fruit and stamen production, the studies were carried out with clonally propagated genotypes, and therefore maternal effects (Roach and Wulff, 1987
) or carryover effects of clonal divisions (Lynch and Walsh, 1998
) cannot be ruled out. Regarding the latter, we attempted to minimize carryover effects of clonal divisions by growing the corms in the greenhouse for one year before beginning estimates of fruit set.
Moreover, by analyzing percentage and not absolute fruit and stamen production per inflorescence, we probably slightly underestimated both (1) reproductive compensation (e.g., differences in absolute seed production between females and males) and (2) differences in stamen and fruit production between fruiting and non-fruiting males. Although neither seed number per fruit nor flower density per inflorescence differed among females and fruiting males in the field (Olson, 1997a
), there were some indications that fruiting males in field populations had slightly shorter inflorescences and thus fewer flowers (and fewer fruit and stamens) than females or nonfruiting males. In an associated study in 1995 of 11 populations (Olson, unpublished data), fruiting males had shorter inflorescences than either females or nonfruiting males (mean for fruiting males: 38.0 cm, nonfruiting males: 43.4 cm; ANOVA F = 22.6, P < 0.0001; mean for females: 46.6 cm; ANOVA females vs. fruiting males F = 9.6, P < 0.006). In the experimental studies presented here, we found similar trends for male ramets growing in pots; although these differences were not statistically significant, on average ramets that had long inflorescences had lower percentage fruit production than ramets with short inflorescences. Unfortunately, inflorescence lengths were not measured for all plants in these studies. However, the correlations would increase in concavity if nonfruiting males produced more flowers (and thus more stamens and pollen) than fruiting males.
Trade-offs in gynodioecious and cosexual species
We reviewed the results from 12 published studies that assessed genotypic correlations between male and female reproductive output (Table 1) to determine whether negative correlations between male and female reproductive output were correlated with breeding system. Nine of the species exhibited cosexual breeding systems (e.g., all individuals could reproduce via both seeds and pollen), and three species were gynodioecious (for this purpose subdioecy is considered a form of gynodioecy). Of the nine cosexual species, three exhibited trade-offs between male and female reproductive output, G. grandiflorus, P. sylvestrus, and Z. mays. All of these studies were conducted in common gardens from commercially bred strains. The other six cosexual species exhibited either positive or no genetic correlation between male and female reproductive output.
|
; Eckhart, 1992
; Wolfe and Shmida, 1997
; Morris and Doak, 1998
; but see also Molina-Freaner and Jain, 1992
; Dinnétz, 1997
). Although these are phenotypic and not genetic measures of reproductive compensation, higher fruit production in females than hermaphrodites may reflect a negative genetic correlation between pollen and fruit production if resources allocated to sexual reproduction are split between male and female modes (Atlan et al., 1992
). The apparent association between gynodioecy and negative correlations may suggest a predisposition for plant species with these trade-offs to evolve gynodioecy or dioecy, while in plant species without such trade-offs hermaphroditism persists. Studies comparing genetic correlations between reproductive traits in closely related hermaphroditic and gynodioecious species may illuminate whether the absence of negative genetic correlations between male and female reproductive output is one factor limiting the evolution of dioecy from hermaphroditism.
On the other hand, the apparent association between breeding system and negative genetic correlations may result somewhat from differences in the traits examined in the studies. The traits examined in the three studies of gynodioecious species are all associated with fruit or seed set per flower, while in three of the studies of cosexual species, only characters expressed prior to seed production such as pistil mass and ovule number were measured. Of particular relevance to this issue is Ashman's (1999)
study of Fragaria virginiana in which the genetic correlation between pollen and ovule production was positive while the correlation between pollen per flower and fruit set per flower was negative. Ashman (1999)
suggests that this may result from a positive correlation between both pollen and ovule production to the size of flowers so that an increase in pollen production will result in an increase in ovule production. Interestingly, although fruit production is quite different among A. biternata male genotypes, preliminary studies suggest that flowers from nonfruiting males produce as many ovules as fruiting males (Olson, unpublished data). In subdioecious species that retain ovules in males allocation of extra resources to pollen production may result in lower seed maturation but not lower ovule production. For this reason, future studies interested in the relationship between breeding system and genetic correlations between characters affecting male and female reproduction should consider measuring both ovule and seed production.
Association between stamen reduction traits and the evolution of dioecy
Although the trade-offs between stamen and fruit production shown in A. biternata are consistent with assumptions from theoretical models for the evolution of dioecy, it is possible that the particular trade-offs that we measured are associated with the evolution of increased outcrossing rates for hermaphrodites or are by-products of genes that partially restore male fertility to individuals harboring cytoplasmic male sterility factors.
When inbreeding depression is severe, traits that lower rates of self-fertilization, such as reduced numbers of stamens, may be favored if fitness gained through increased seed quality exceeds that lost from decreased siring success. Inbreeding in A. biternata results in lower germination rates, higher mortality rates, and increased time from seed to first flowering for selfed compared to outcrossed seeds (Olson, unpublished data). In addition, self-fertilization is common for hermaphrodites and can potentially be quite high (Olson, 1997a
). Therefore, it is possible that individuals with reduced stamen numbers would be selectively favored if reduced stamen number resulted in fewer self-fertilized and more outcrossed seeds.
Alternatively, reduced stamen numbers may result from genes that only partially restore male fertility to individuals harboring cytoplasmic male-sterility genes. Recent theoretical studies have shown that hermaphrodites in species with cyto-nuclear gynodioecy can evolve increased allocation to male reproduction (Maurice et al., 1993, 1994
; Schultz, 1994
) and partial male-fertility traits are not uncommon in species with cyto-nuclear sex determination (for review, see Koelewijn and Van Damme, 1996
). In Plantago coronopus Koelewijn and Van Damme (1995a, b, 1996)
found that reduced stamen production is determined by a combination of cytoplasmic male-sterility and nuclear male-fertility restorer genes (Koelewijn and Van Damme, 1995a
). Expression of reduced stamen production results from incomplete restoration at the nuclear male-fertility loci (Koelewijn and Van Damme, 1995b
). Presently, little is known about the inheritance of sex expression in A. biternata. However, crosses between females and males sometimes result in progeny sex ratios that deviate significantly from 1:1 (Olson, 1997a
), suggesting that the genetics of sex-expression may be quite complex. Moreover, reduced numbers of stamens were associated with particular families of progeny (Olson, 1997a
), supporting the data presented here for a genetic basis for reduced stamen numbers. Thus, it is possible that stamen number in A. biternata is determined by incomplete penetrance of nuclear male-fertility restoration genes. The evolutionary dynamics of partial male-sterility restorers has not been theoretically explored but may suggest new mechanisms for the coexistence of hermaphroditic and unisexual individuals.
FOOTNOTES
1 The authors thank J. Clark, S. Emery, E. Lacey, D. McCauley, B. Morris, C. Richards, J. Stone, two anonymous reviewers, and P. Tiffin for comments on earlier drafts of this manuscript. This research was supported by a Sigma Xi Grant-in-Aid, the Aleane Webb Research Fellowship at Duke University, and a doctoral dissertation improvement grant from the National Science Foundation (DEB-9520754). 
2 Author for correspondence, current address: Department of Biology, Vanderbilt University, P.O. Box 1812, Station B, Nashville, Tennessee 37235 USA (e-mail: matt.olson{at}vanderbilit.edu
).
3 Current address: Biology Department, University of Virginia, Charlottesville, Virginia 22903-2477 USA.
LITERATURE CITED
Ågren, J., and D. W. Schemske. 1995 Sex allocation in the monoecious herb Begonia semiovata. Evolution 49: 121–130.
Alexander, M. P. 1980 A versatile stain for pollen, fungi, yeast, and bacteria. Stain Technology 55:107–112.
Ashman, T.-L. 1999 Determinants of sex allocation in a gynodioecious wild strawberry: implications for the evolution of dioecy and sexual dimorphism. Journal of Evolutionary Biology 12:648–661.
Atlan, A., P. H. Gouyon, T. Fournial, D. Pomente, and D. Couvet. 1992 Sex allocation in an hermaphroditic plant: the case of gynodioecy in Thymus vulgaris L. Journal of Evolutionary Biology 5: 189–203.[CrossRef][ISI]
Barrett, S. C. H. 1992 Gender variation and the evolution of dioecy in Wurmbea dioica (Liliaceae). Journal of Evolutionary Biology 5: 423–444.[CrossRef][ISI]
———, A. L. Case, and G. B. Peters. 1999 Gender modification and resource allocation in subdioecious Wurmbea dioica (Colchicaceae). Journal of Ecology 87:123–137.
Campbell, D. R. 1997 Genetic correlation between biomass allocation to male and female functions in a natural population of Ipomopsis aggregata. Heredity 79: 606–614.[CrossRef][ISI]
Charlesworth, B., and D. Charlesworth. 1978 A model for the evolution of dioecy and gynodioecy. American Naturalist 112: 975–997.[CrossRef][ISI]
Charlesworth, D., and B. Charlesworth. 1981 Allocation of resources to male and female functions in hermaphrodites. Biological Journal of the Linnean Society 15: 57–74.
Charnov, E. L. 1982 The theory of sex allocation. Princeton University Press. Princeton, New Jersey.
———, J. Maynard Smith, and J. Bull. 1976 Why be an hermaphrodite? Nature 263: 125–126.[CrossRef]
Delph, L F. 1993 Factors affecting intraplant variation in flowering and fruiting in the gynodioecious species Hebe subalpina. Journal of Ecology 81: 287–296.
———, and D. G. Lloyd. 1991 Environmental and genetic control of gender in the dimorphic shrub Hebe subalpina. Evolution 45: 1957–1964.[CrossRef][ISI]
Dinnetz, P. 1997 Male sterility, protogyny, and pollen-pistil interference in Plantago maritima (Plantaginaceae), a wind-pollinated, self-incompatible perennial. American Journal of Botany 84: 1588–1594.[Abstract]
Eckhart, V. M. 1992 Resource compensation and the evolution of gynodioecy in Phacelia linearis (Hydrophyllaceae) Evolution 46: 1313–1328.[CrossRef][ISI]
Fenster, C. B., and D. E. Carr. 1997 Genetics of sex allocation in Mimulus (Scrophulariaceae). Journal of Evolutionary Biology 10: 641–661.[CrossRef][ISI]
Fleming, T. H., S. Maurice, S. L. Buchmann, and M. Tuttle. 1994 Reproductive biology and relative male and female fitness in a trioecious cactus, Pachycereus pringlei (Cactaceae). American Journal of Botany 81: 858–867.[CrossRef][ISI]
Geber, M. A. 1999 Theories of the evolution of sexual dimorphism. In M. A. Geber, T. E. Dawson, and L. F Delph [eds.], Gender and sexual dimorphism in flowering plants, 97–122. Springer, Berlin, Germany.
Garnier, P., S. Maurice, and I. Olivieri. 1993 Costly pollen in maize. Evolution 47: 946–949.[CrossRef][ISI]
Gouyon, P.-H., and D. Couvet. 1987 A conflict between two sexes, females and hermaphrodites. In S. C. Stearns [ed.], The evolution of sex and its consequences, 245–261. Birkhauser-Verlag, Basel, Switzerland.
Hamel, J. L. 1953 Contribution a l’Étude Taxonomique des Saxifragacées. Thèse Présentées a la Faculté des Sciences de L'Université de Paris. Série A, nombre 2632, nombre d'odre:3504.
Koelewijn, H. P., and J. M. M. Van Damme. 1995a Genetics of male sterility in gynodioecious Plantago coronopus. I. Cytoplasmic variation. Genetics 139: 1749–1758.[Abstract]
———, and ———. 1995b Genetics of male sterility in gynodioecious Plantago coronopus. II. Nuclear genetic variation. Genetics 139: 1759–1775.[Abstract]
———, and ———. 1996 Gender variation, partial male sterility and labile sex expression in gynodioecious Plantago coronopus. New Phytologist 132: 67–76.[CrossRef][ISI]
Lloyd, D. G. 1976 The transmission of genes via pollen and ovules in gynodioecious angiosperms. Theoretical Population Biology 9: 299–316.[CrossRef][ISI][Medline]
Lynch, M., and B. Walsh. 1998 Genetics and analysis of quantitative traits. Sinauer, Sunderland, Massachusetts, USA.
Maurice, S., E. Belhassen, D. Couvet, and P.-H. Gouyon. 1994 Evolution of dioecy: can nuclear-cytoplasmic interactions select for maleness? Heredity 73: 346–354.
———, D. Charlesworth, C. Desfeux, D. Couvet, and P-H Gouyon. 1993 The evolution of gender in hermaphrodites of gynodioecious populations with nucleo-cytoplasmic male-sterility. Proceedings of the Royal Society of London, Series B 251: 253–261.[CrossRef]
Mazer, S. J., and V. A. Delasalle. 1996 Covariation among floral traits in Spergularia marina (Caryophyllaceae): geographic and temporal variation in phenotypic and among-family correlations. Journal of Evolutionary Biology 9: 993–1015.[CrossRef][ISI]
———, and C. T. Schick. 1991 Constancy of population parameters for life-history and floral traits in Raphanus sativus L. II. Effects of planting density on phenotypes and heritability estimates. Evolution 45:1888–1907.
Mellichamp, T. L. 1976 A comparative study of Aruncus (Rosaceae) and Astilbe (Saxifragaceae), and the problem of their relationships. Ph.D. dissertation, University of Michigan, Ann Arbor, Michigan, USA.
Mitchell-Olds, T., and J. J. Rutledge. 1986 Quantitative genetics in natural plant populations: a review of the theory. American Naturalist 127: 379–402.[CrossRef][ISI]
Molina-Freaner, F., and S. K. Jain. 1992 Female frequencies and fitness components between sex phenotypes among gynodioecious populations of the colonizing species Trifolium hirtum All. in California. Oecologia 92: 279–286.[CrossRef][ISI]
Morris, W. F., and D. F. Doak. 1998 Life history of the long-lived gynodioecious cushion plant, Silene acaulis (Caryophyllaceae), inferred from size-based population projection matrices. American Journal of Botany 85: 784–793.[Abstract]
Olson, M. S. 1997a The genetics and evolution of subdioecy in Astilbe biternata. Ph.D. dissertation, Duke University, Durham, North Carolina, USA.
———. 1997b Bayesian procedures for discriminating among hypotheses with discrete distributions: Inheritance in the tetraploid Astilbe biternata. Genetics 147: 1933–1942.[Abstract]
O'Neil, P., and J. Schmitt. 1993 Genetic constraints on the independent evolution of male and female reproductive characters in the tristylous plant Lythrum salicaria. Evolution 47: 1457–1471.[CrossRef][ISI]
Primack, R. B., and J. Antonovics. 1981 Experimental ecological genetics in Plantago. V. Components of seed yield in the ribwort plantain Plantago lanceolata L. Evolution 35: 1069–1079.[CrossRef][ISI]
Rameau, C., and P.-H. Gouyon. 1991 Resource allocation to growth, reproduction and survival in Gladiolus: the cost of male function. Journal of Evolutionary Biology 4:291–307.
Roach, D. A., and R. D. Wulff. 1987 Maternal effects in plants. Annual Review of Ecology and Systematics 18: 209–235.
Sakai, A. K., and S. G. Weller. 1991 Ecological aspects of sex expression in subdioecious Schiedea globosa (Caryophyllaceae). American Journal of Botany 78: 1280–1288.[CrossRef][ISI]
SAS. 1995 JMP statistics and graphics guide. Version 3.1. SAS Institute Cary, North Carolina, USA.
Savolainen, O., K. Kärkkäinen, A. Harju, T. Nikkanen, and M. Rusanen. 1993 Fertility variation in Pinus sylvestris: a test of sexual allocation theory. American Journal of Botany 80: 1016–1020.[CrossRef][ISI]
Schultz, S. T. 1994 Nucleo-cytoplasmic male sterility and alternative routes to dioecy. Evolution 48: 1933–1945.[CrossRef][ISI]
Spongberg, S. A. 1972 Genera of Saxifragaceae in the southeastern United States. Journal of the Arnold Arboretum 53: 409–498.[ISI]
Steel, R. G. D., and J. H. Torrie. 1980 Principles and procedures of statistics. McGraw-Hill, New York, New York, USA.
Webb, C. J. 1999 Empirical studies: Evolution and maintenance of dimorphic breeding systems. In M. A. Geber, T. E. Dawson, and L. F Delph [eds.], Gender and sexual dimorphism in flowering plants, 61–96. Springer, Berlin, Germany.
Wolfe, L. M., and A. Shmida. 1997 The ecology of sex expression in a gynodioecious Israeli desert shrub (Ochradenus baccatus). Ecology 78: 101–110.[CrossRef][ISI]
Wolfram, S. 1991 Mathematica: a system for doing mathematics by computer, 2nd ed. Addison-Wesley, Redwood City, California, USA.
This article has been cited by other articles:
|
|
 |
|
  N. N. Gomez and R. G. Shaw Inbreeding effect on male and female fertility and inheritance of male sterility in Nemophila menziesii (Hydrophyllaceae) Am. J. Botany, May 1, 2006; 93(5): 739 - 746. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. L. Parachnowitsch and E. Elle Variation in sex allocation and male-female trade-offs in six populations of Collinsia parviflora (Scrophulariaceae s.l.) Am. J. Botany, August 1, 2004; 91(8): 1200 - 1207. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
