E. Nilsson, Department of Plant Ecology, Evolutionary Biology Centre, Uppsala University, Villavägen 14, SE-752 36 Uppsala, Sweden. Tel.: +46 18 471 28 69; fax +46 18 55 3419 e-mail: firstname.lastname@example.org
Theory predicts that the sex ratio of gynodioecious populations (in which hermaphrodites and females coexist) will be affected by the relative female fitness of females and hermaphrodites, and by founder events and genetic drift in small populations. We documented the sex ratio and size of 104 populations of the gynodioecious, perennial herb Plantago maritima in four archipelagos in eastern Sweden and western Finland (from latitude 53 to 64 °N). The sex ratio varied significantly both among and within archipelagos (range 0–70% females, median 6.3% females). The frequency of females was highest in the northernmost archipelago and lowest in the southernmost archipelago. As predicted, females were more frequently missing from small than from large populations, and the variance in sex ratio increased with decreasing population size. The relative fecundity of female plants (mean seed output per female/mean seed output per hermaphrodite) ranged from 0.43 to 2.16 (median 1.01, n = 12 populations). Among the 12 populations sampled for seed production (four in each of three archipelagos), the frequency of females was positively related to relative fecundity of females and negatively related to population size. The results suggest that the local sex ratio is influenced both by the relative fecundity of females and hermaphrodites and by stochastic processes in small populations.
Sexually polymorphic plants offer suitable systems for studies of how the evolution of population differentiation is moulded by selection and stochastic processes. Gynodioecy is a relatively common sexual polymorphism in plants, in which hermaphrodites coexist with female (i.e. male-sterile) individuals (Kaul, 1988). The maintenance of gynodioecy and the evolutionary dynamics of the sex ratio in gynodioecious populations will depend on the relative female fecundity of female and hermaphrodite plants and on the genetic basis of male-sterility. In gynodioecious populations with cytoplasmic inheritance of sex expression, females will increase in frequency as long as their female fecundity is higher than that of hermaphrodites, while in populations with simple nuclear inheritance of sex expression, the female fitness of females needs to be at least twice as high as that of hermaphrodites for females to be maintained (Lewis, 1941, but see Bailey et al., 2003). In the latter case, the frequency of females at equilibrium will be a function of the relative fecundity of females and hermaphrodites and will not exceed 0.5 (Lloyd, 1975). In gynodioecious species, sex expression is often governed by an interaction between cytoplasmic factors causing male-sterility and nuclear genes restoring male function, so-called nuclear-cytoplasmic male-sterility (Kaul, 1988; Frank, 1989; Charlesworth & Laporte, 1998). In these species, founder events and genetic drift in small populations may result in among-population variation in the inheritance of sex expression, and therefore also in the projected dynamics of the sex ratio (Gouyon & Couvet, 1987; Frank, 1989; Couvet et al., 1998). Depending on which cytoplasmic male-sterilising genes and which nuclear restorer genes are present in a given population, sex expression may not vary, may vary and show strict cytoplasmic or nuclear inheritance, or may vary and be determined by an interaction between cytoplasmic and nuclear factors.
Although modelling suggests that founder events coupled with the coevolution of nuclear and cytoplasmic genes can result in rather complex sex ratio dynamics (e.g. Frank, 1989; Gouyon et al., 1991; Couvet et al., 1998; Bailey et al., 2003; Jacobs & Wade, 2003), certain general predictions can be derived. First, stochastic processes in small populations and colonization/extinction dynamics should increase the variance in sex ratio among populations. Moreover, modelling indicates that colonization/extinction dynamics will increase the overall frequency of females at the level of the metapopulation under a broad range of conditions in gynodioecious species with a nuclear-cytoplasmic inheritance of male-sterility (Couvet et al., 1998; Ronce & Olivieri, 2004). On the one hand, stochastic processes and founder events may increase the probability that females are missing altogether from a local population. On the other hand, stochastic processes should also increase the probability that sex expression is inherited cytoplasmically. The latter implies more lax conditions for the maintenance of females, and that females may increase in frequency beyond what is possible under nuclear and nuclear-cytoplasmic inheritance. Based on this argument it has been predicted that frequencies of females should reach particularly high frequencies in young, newly founded populations. Support for this prediction has been obtained in studies of sex ratio variation in Thymus vulgaris (Belhassen et al., 1991; Manicacci et al., 1996). However, the relationship between population size and sex ratio is not well documented in any gynodioecious species.
Second, spatial variation in the relative fecundity of females and hermaphrodites should influence sex ratio variation. If the relative fitness of females and hermaphrodites varies among habitats, the sex ratio should change along environmental gradients. For some gynodioecious species, it has been suggested that the relative fitness of females is higher in harsh than in more benign environments (Delph, 1990; Wolfe & Shmida, 1997; Ashman, 1999; Asikainen & Mutikainen, 2003). Moreover, a correlation between sex ratio and relative seed output per flower of females and hermaphrodites has been documented in a couple of gynodioecious species (Delph & Carroll, 2001; Asikainen & Mutikainen, 2003). However, no study has quantified the relationship between sex ratio and relative seed fecundity based on estimates of total seed production per plant, and it is therefore not clear to what extent variation in sex ratio can be explained by differences in the relative female reproductive success of females and hermaphrodites. We examined variation in sex ratio, population size and relative female fecundity of females and hermaphrodites in the gynodioecious, perennial herb Plantago maritima (L.) within and among four archipelagos located along a latitudinal gradient in the Baltic Sea and the Gulf of Bothnia. Environmental conditions vary along this geographic gradient (Jerling, 1999). Towards the north, the length of the growing season and sea water salinity decrease, and the rate of land-uplift increases. An increase in the rate of land uplift results in a more rapid succession on the seashores, and should be associated with a higher rate of population turnover. In this study, we addressed the following questions:
1Is the frequency of females higher and more variable in small than in large populations?
2Is the frequency and relative fecundity of females higher in northern than in southern populations?
3Can population size and relative female fitness of hermaphrodites and females together explain variation in sex ratio among populations?
Plantago maritima L. (Plantaginaceae) is a gynodioecious, wind-pollinated, perennial rosette herb. It is common on seashores in Europe (Hultén & Fries, 1986). Individuals are easily distinguished and vegetative propagation is rare (Dinnétz & Jerling, 1997). The flowers are arranged in a dense racemose inflorescence and each plant produces one to many inflorescences. Perfect flowers of P. maritima have four stamens with cylindrical filaments and yellow anthers. Pistillate flowers have shorter filaments and thin brown anthers. As in many gynodioecious species, a variety of morphologically distinct female types, as well as a continuous range of plants with partially reduced male function, can be found (E. Nilsson and J. Ågren, pers. obs.; Dinnétz, 1996). Hermaphrodites are protogynous, but the degree of protogyny varies (Dinnétz, 1997). The fruit is a capsule that contains one to four seeds. Male-sterility shows nuclear-cytoplasmic inheritance in P. maritima (van Damme, 1992), although the genetic basis of male-sterility has not been investigated in detail. Controlled crosses indicate that most plants in the south Swedish Gryt archipelago are self-incompatible, which is consistent with previous studies of European P. maritima populations (Ross, 1970; Dinnétz, 1997), but that hermaphrodite plants in the Korpoo, Öregrund and Skeppsvik archipelagos are self-compatible (E. Nilsson, K. Wolff & J. Ågren, unpublished results).
This study was conducted in four archipelagos in the Baltic Sea and the Gulf of Bothnia: the Gryt (58 °11′N, 16 °50′E), the Öregrund (60 °15′N, 18 °37′E) and the Skeppsvik (63 °47′N, 20 °36′E) archipelago in Sweden and the Korpoo archipelago in Finland (60 °06′N, 21 °40′E, Fig. 1). The northernmost archipelago (Skeppsvik) is situated 880 km north of the southernmost archipelago (Gryt). The land uplift in this region is substantial, ranging from 8 mm per year in the Skeppsvik archipelago to 2 mm per year in the Gryt archipelago (Jerling, 1999). The salinity of the Baltic Sea and the Gulf of Bothnia is lower than 0.8% (Snoeijs, 1999).
The frequency of female plants was scored in the Gryt (34 populations) and Öregrund (nine populations) archipelagos in 1999, and in the Korpoo (30 populations) and Skeppsvik archipelagos (31 populations) in 2000. A population was operationally defined as all plants on an island. To estimate population size, we counted all flower-producing plants during the second half of the flowering period (these counts included also plants that had finished flowering). In the Gryt, Korpoo and Skeppsvik archipelagos, the sex ratio was estimated by recording the sex expression of up to 200 plants in each population. In large populations, plants were scored along a transect across the population. In populations composed of fewer than 200 plants, the sex expression of all plants was determined. Gynomonoecious individuals, i.e. plants with a partially reduced male function, were pooled with the other hermaphrodite plants when the frequency of females was calculated. To determine the sex ratio of populations in the Öregrund archipelago, we collected ripe inflorescences from 200 individuals per population. The inflorescences were brought to the lab, where their sex expression was examined under a stereo microscope.
Sexual differences in reproductive output
We compared the size and fecundity of hermaphrodites and female plants in a total of 12 populations from three archipelagos. In July 2000, at least 30 hermaphrodite and 30 female plants were permanently marked in four populations in each of the Skeppsvik and Korpoo archipelagos. For each plant, we recorded the number of inflorescences, the number of leaves and the length of the longest leaf. At fruit maturation, we collected the inflorescences of marked plants. In four of the populations sampled for sex ratio in the Öregrund archipelago in 1999, we noted for each sampled plant the number of inflorescences produced and the length of the longest leaf. We counted flowers and mature fruits on up to three inflorescences per plant, and quantified fruit set as the ratio between the number of fruits and the number of flowers. For each plant, 10 mature fruits were opened and the number of seeds was counted. To estimate seed mass, we determined the total mass of the seeds from the ten fruits to the nearest 0.1 mg, and divided this value by the number of seeds. Seed output per plant was estimated as (mean number of seeds per fruit) × (mean number of fruits per inflorescence) × (number of inflorescences per plant). For plants in the Skeppsvik and Korpoo archipelagos, we estimated vegetative size by multiplying the number of leaves with the length of the longest leaf (data on leaf number was not available for plants from the Öregrund archipelago).
Variation in the proportion of females, in populations with 20 or more flowering individuals, was analysed with a generalized linear model which included archipelago and population within archipelago as independent factors (procgenmod in sas Version 9, SAS Institute Inc., Cary, NC, USA). We used logistic regression to determine whether the presence of females was related to population size using procfenmod in sas. This analysis included population size, archipelago, and their interaction as independent variables. To examine whether the variance in sex ratio decreased with increasing population size, we determined by how much the frequency of females in individual populations deviated from the mean frequency in a given archipelago, and we regressed the absolute value of this deviation on population size. This analysis was conducted separately by archipelago, and included only populations with 20 or more flowering plants. Analyses of the relationships between sex ratio and population size did not include populations from the Öregrund archipelago, because only three populations smaller than 500 flowering plants were sampled in that area.
Sexual differences in components of reproductive success, and vegetative size were analysed with mixed-model analysis of variance and generalized linear models. We compared the effect of the three main factors sex morph (fixed), archipelago (fixed), and population (random, nested within archipelago). Variation in fruit set was modelled with binomial distribution and logit-link function using the glimmix macro in sas. Variation in other components of reproductive success and in vegetative size were analysed with procmixed in sas; variables deviating from a normal distribution or displaying heterogeneity of variance were log-transformed prior to analysis. To determine the statistical significance of the two random factors population within archipelago and sex morph × population within archipelago we used the likelihood ratio statistic, which is calculated as the difference in the residual log likelihood (DRL) of models with and without the factor of interest included (Littell et al., 1996). We used multiple regression to examine how the sex ratio (arcsine square-root proportion of females) was related to relative female fecundity (mean fecundity of females/mean fecundity of hermaphrodites) and population size [log (number of flowering plants)].
Female frequencies varied significantly both within ( = 499.42, P < 0.001) and among archipelagos ( = 308.58, P < 0.001). Females were 7.7 times more common in the northernmost than in the southernmost archipelago [median (range), Skeppsvik 17.7% (0–70%) females n = 31, vs. Gryt 2.5% (0–25%) females, n = 34].
As expected from stochastic processes, the variance in female frequency increased with decreasing population size. The relationship between population size and presence of females varied among archipelagos as indicated by a significant population size × archipelago interaction in the logistic regression ( = 6.4, P < 0.05). However, the probability that female plants would be present in a population increased significantly with population size in all three archipelagos examined (Gryt, = 10.27, P = 0.001, n = 33; Korpoo, = 5.35, P = 0.02, n = 30; Skeppsvik, = 12.39, P < 0.001, n = 31). The absolute value of the deviation in the frequency of females from the archipelago mean was negatively correlated with population size in the Gryt (b = −0.02, F1,32 = 5.12, P < 0.05) and Skeppsvik archipelagos (b = −0.07, F1,29 = 8.0, P < 0.01), but not in the Korpoo archipelago (F1,18 = 1.0, P > 0.1).
For populations with females present, the frequency of females decreased with population size [log (number of flowering plants)] in the Korpoo (linear regression, F1,22 = 8.7, r2 = 28.3%, P < 0.01) and Skeppsvik archipelagos (F1,25 = 16.2, r2 = 39.4%, P < 0.001), but was not significantly related to population size in the Gryt archipelago (F1,23 < 0.1, P > 0.9) where the overall frequency of females was very low (analyses conducted on populations composed of at least 20 plants; Fig. 2).
Sexual differences in reproductive output and vegetative size
The relative number of seeds produced by females and hermaphrodites varied among populations as indicated by a significant sex morph × population(archipelago) interaction (DRL1 = 7.7, P < 0.01), but did not vary significantly among archipelagos (sexmorph × archipelago interaction, P > 0.45; Table 1, Fig. 3). The relative seed output of females (fecundity of females/fecundity of hermaphrodites) varied from 0.43 to 2.16 (median = 1.01, n = 12 populations; Fig. 3). Total seed output is a function of flower production, fruit set (proportion of flower forming a mature fruit), and number of seeds produced per fruit. A significant sexmorph × population (archipelago) interaction was detected in fruit set (DRL1 = 20.1, P < 0.001), but not in the other components of seed output (flowers per plant; DRL1 = 1.3, NS; seeds per fruit; DRL1 = 1.3, NS). Females produced more flowers than hermaphrodites (least square mean ± SE, females; 86.1 ± 1.2, n = 285, hermaphrodites; 70.5 ± 1.2, n = 315; Fig. 3, Table 1).
Table 1. Effects of sex morph (female vs. hermaphrodite; fixed) and archipelago (fixed) and population nested within archipelago (random) on components of reproductive output and leaf length in Plantago maritima.
Flowers per plant†
Fruits per flower
Seeds per fruit
Seeds per plant†
Mean seed mass
F-ratios and levels of significance are indicated for fixed factors, while the difference in the residual log likelihood (DRL) and levels of significance are indicated for the random factor population (archipelago) and its interaction with sex morph.
*P < 0.05.
**P < 0.01.
***P < 0.001.
Sex morph × archipelago
Sex morph × population (archipelago)
Mean seed mass varied among sexes and archipelagos. Overall, females produced larger seeds compared to hermaphrodites (females; 0.49 ± 0.013 mg, n = 306, hermaphrodites; 0.47 ± 0.013 mg, n = 329; Fig. 3, Table 1). Moreover, plants in the Skeppsvik archipelago produced larger seeds (least-square mean ± SE, 0.56 ± 0.02 mg, n = 229), compared to plants in the more southern Öregrund (0.47 ± 0.02 mg, n = 221), and Korpoo (0.40 ± 0.02 mg, n = 185) archipelagos (Fig. 3).
Leaf length and vegetative size did not vary significantly between females and hermaphrodites [Table 1; main effect of sex morph in mixed-model anova of vegetative size, P > 0.1, sex morph × archipelago, and sexmorph × population(archipelago) interactions, P > 0.3].
Population size, female fecundity, and sex ratio
The relative fecundity of females and population size could explain a considerable proportion of variation in the frequency of females among the 12 populations sampled for seed production (multiple regression, F2,9 = 16.8, r2 = 78.9%, P < 0.001). The proportion of females increased with relative female fecundity (partial regression coefficient, b = 0.19, t = 4.8, P = 0.001), and decreased with population size (b = −0.35, t = 5.7, P < 0.001; Fig. 4).
This study has demonstrated that the sex ratio of P. maritima populations varies considerably both within and among archipelagos. The frequency of females increased with the relative fecundity of females and decreased with population size, suggesting that both spatial variation in selection and stochastic processes contribute to sex ratio variation. To the best of our knowledge, this is the first comparative study to be based on estimates of total seed production per plant, and it adds to previous evidence that variation in relative seed fecundity of females and hermaphrodites may contribute to sex ratio variation in gynodioecious species (Delph & Carroll, 2001; Asikainen & Mutikainen, 2003).
As expected from an increased importance of stochastic processes in small populations, females were more likely to be missing from small than from large populations of P. maritima, and the variance in sex ratio increased with decreasing population size. Moreover, among populations with females, the frequency of females declined with increasing population size, which is consistent with the idea that a mismatch between cytoplasmic male-sterility alleles and nuclear restorer alleles may be particularly likely in small and young populations because of founder events (Couvet et al., 1998; Ronce & Olivieri, 2004). Stochastic processes apparently influence sex ratio variation also in other gynodioecious species. Females have been found to be particularly common in young, small populations of T. vulgaris (Belhassen et al., 1991; Manicacci et al., 1996) and in small and isolated populations of Lobelia spicata (Byers et al., 2005).
The relative fecundity of females and hermaphrodites varied considerably among populations, and females were not overall more fecund than hermaphrodites. Females produced between 0.4 and 2.2 times as many seeds as hermaphrodites did (median 1.01, n = 12 populations). This is in contrast to the findings of a female fecundity advantage in most gynodioecious species (reviewed by Gouyon & Couvet, 1987; Budar et al., 2003; Shykoff et al., 2003; but see Asikainen & Mutikainen, 2003). The present study suggests that a fecundity advantage of females contributes to the maintenance of females in some, but not all P. maritima populations. It is conceivable that females outperform hermaphrodites in other components of female reproductive success in P. maritima. In particular, females may have an outcrossing advantage in P. maritima populations in the Korpoo, Öregrund and Skeppsvik archipelagos, where hermaphrodites are self-compatible (E. Nilsson, K. Wolff & J. Ågren, unpublished data). Differences in outcrossing rate may result in differential survival and performance of the offspring of female and hermaphrodite plants (Charlesworth & Charlesworth, 1978; Delph, 2004; Medrano et al., 2005). In the present study, females overall produced somewhat larger seeds compared with hermaphrodites in the Korpoo, Öregrund and Skeppsvik archipelagos (Table 1; Fig. 3). Seed mass is positively correlated with germination probability in P. maritima (Dinnétz & Jerling, 1997), suggesting that this difference may influence probability of seedling establishment. Interestingly, females were found to produce smaller rather than larger seeds than hermaphrodites in a self-incompatible P. maritima population about 80 km NE of the Gryt archipelago (58°57′N, 17°36′E; Dinnétz & Jerling (1997). In a perennial herb such as P. maritima, sexual differences in other life-history traits contributing to lifetime female reproductive success could potentially give females an advantage over hermaphrodites. Females have been reported to have a higher survival than hermaphrodites in the congener P. lanceolata (van Damme & van Delden, 1984). However, no sexual difference in survival was detected in a previous study of one self-incompatible P. maritima population (Dinnétz, 1996).
The frequency of females tended to increase from the southernmost to the northernmost archipelago, but this variation was not associated with differences among archipelagos in population size or relative fecundity of females and hermaphrodites. There was no clear relationship between the median population size and the median frequency of females among archipelagos (Fig. 2). Moreover, there was no evidence that females performed relatively better in the northern Skeppsvik archipelago than in the other archipelagos (Table 1). In ongoing studies, we examine whether the outcrossing rate of hermaphrodites varies among archipelagos, and whether inbreeding depression influences the relative performance of offspring produced by hermaphrodites and females.
Extinction/colonization dynamics should increase the overall frequency of females in metapopulations of gynodioecious species with a nuclear-cytoplasmic inheritance of male-sterility (Couvet et al., 1998; Ronce & Olivieri, 2004), and latitudinal variation in the rate of population turnover may contribute to the variation in sex ratio among archipelagos. The rate of land uplift increases from the southernmost to the northernmost archipelago, and this should contribute to a higher population turnover in the north. This possibility could be assessed by long-term demographic studies. Alternatively, the structuring of genetic variation at neutral loci within and among archipelagos could be examined to determine whether signatures of recent population bottlenecks (cf. Nei et al., 1975; Cornuet & Luikart, 1996) are more frequent among the northern populations than among the southern populations.
Among the 12 populations sampled for seed production, population size and relative female fecundity explained together almost 80% of the total variation in sex ratio. The frequency of females decreased with population size, as was true for the larger samples of populations for which only morph frequencies were available in the Korpoo and Skeppsvik archipelagos. Moreover, the proportion of females increased with the relative seed output of females indicating that some of the among-population variation in sex ratio can be attributed to variation in the relative performance of females and hermaphrodites. Sexual differences in seed fecundity may contribute to sex ratio variation also in other gynodioecious species. The proportion of females is negatively related to seed output per flower in hermaphrodites relative to females in Silene acaulis (Delph & Carroll, 2001), and Geranium sylvaticum (Asikainen & Mutikainen, 2003), and to fruit set in hermaphrodites in Hebe strictissima (Delph, 1990), and Fragaria virginiana (Ashman, 2003). To substantiate the relationship between sex ratio and sexual differences in female fecundity, future studies should collect data that are as good estimates as possible of total life-time fecundity of females and hermaphrodites.
Taken together, the results of the present study suggest that the sex ratio of P. maritima is governed by an interaction between selection and stochastic processes. Further studies are needed to determine whether the considerable geographic variation in sex ratio in this species reflects differences in the genetic basis of sex expression and the importance of extinction/colonization dynamics, and to what extent it is a function of variation in the mating system of hermaphrodites.
We thank Jos van Damme, Patrik Dinnétz, Lenn Jerling, and Hans Koelewijn for valuable discussions, and Larsgunnar and Gunilla Nilsson, Magnus Andersson, Nora Rustum and Stefan Sjögren for assistance in the field. This research was financially supported by Grants from B. Lundmans fond and H. Ax:son Johnssons fond (to EN) and the Swedish Research Council (to JÅ).