Intraspecific variation in sex allocation in hermaphroditic Plantago coronopus (L.)


  • Koelewijn,

    1. Netherlands Institute of Ecology, Department of Plant Population Biology, Boterhoeksestraat 22, PO Box 40, 6666 ZG Heteren, The Netherlands
    2. Department of Genetics, University of Groningen, The Netherlands
    Search for more papers by this author
  • Hunscheid

    1. Netherlands Institute of Ecology, Department of Plant Population Biology, Boterhoeksestraat 22, PO Box 40, 6666 ZG Heteren, The Netherlands
    Search for more papers by this author

Hans P. Koelewijn Dr Netherlands Institute of Ecology, PO Box 40, 6666 ZG Heteren, The Netherlands. Tel.: +31 26 4791205; fax: +31 26 23227; e-mail:


Models for sex allocation assume that increased expenditure of resources on male function decreases the resources available for female function. Under some circumstances, a negative genetic correlation between investment in stamens and investment in ovules or seeds is expected. Moreover, if fitness returns for investment in male and female function are different with respect to size, sex allocation theory predicts size-specific gender changes. We studied sex allocation and genetic variation for investment in stamens, ovules and seeds at both the flower and the plant level in a Dutch population of the wind-pollinated and predominantly outcrossing Plantago coronopus. Data on biomass of floral structures, stamens, ovules, seedset and seedweight were used to calculate the average proportion of reproductive allocation invested in male function. Genetic variation and (genetic) correlations were estimated from the greenhouse-grown progeny of maternal families, raised at two nutrient levels. The proportion of reproductive biomass invested in male function was high at flowering (0.86 at both nutrient levels) and much lower at fruiting (0.30 and 0.40 for the high and low nutrient treatment, respectively). Androecium and gynoecium mass exhibited moderately high levels of genetic variance, with broad-sense heritabilities varying from 0.35 to 0.56. For seedweight no genetic variation was detected. Significant among-family variation was also detected for the proportion of resources invested in male function at flowering, but not at fruiting. Phenotypic and broad-sense genetic correlations between androecium and gynoecium mass were positive. Even after adjusting for plant size, as a measure of resource acquisition, maternal families that invested more biomass in the androecium also invested more in the gynoecium. This is consistent with the hypothesis that genetic variation for resource acquisition may in part be responsible for the overall lack of a negative correlation between male and female function. Larger plants had a more female-biased allocation pattern, brought about by an increase in seedset and seedweight, whereas stamen biomass did not differ between small and large plants. These results are discussed in relation to size-dependent sex allocation theory (SDS). Our results indicate that the studied population harboured substantial genetic variation for reproductive characters.


Evolution of various forms of gender expression in higher plants, such as hermaphroditism, monoecy, dioecy and their combinations, has been discussed from the viewpoint of sex allocation theory ( Charnov et al., 1976 ; Charlesworth & Charlesworth, 1981; Charnov, 1982; Lloyd, 1984; Charlesworth & Morgan, 1991). Sex allocation concerns the interrelationships in reproductive success through male function (pollen) and female function (ovules). The basic premise is that there should be a trade-off in resource allocation between male and female function, and that plants should distribute resources to these functions to obtain optimal fitness ( Charlesworth & Charlesworth, 1981; Charnov, 1982; Morgan, 1992). As early as 1877, by comparing reproductive characteristics between male steriles and hermaphrodites in gynodioecious species, Darwin (1877) had recognized the existence of this principle and called it the ‘law of compensation’; reduced investment into one reproductive function may be compensated by the greater resource availability for the other reproductive function. Consequently, a negative (genetic) correlation between investment in male and female function is expected in cosexual species ( Stearns, 1992).

The search for trade-offs between male and female function has resulted in a fast growing and conflicting body of literature. The majority of the studies have detected positive (genetic) correlations (e.g. Goldman & Willson, 1986; Brunet, 1992; Campbell, 1992, 1997; Mazer, 1992; Kudo, 1993; O’Neill & Schmitt, 1993; Agren & Schemske, 1995; Fenster & Carr, 1997), and only few have reported negative correlations ( Atlan et al., 1992 ; Garnier et al., 1993 ; Savolainen et al., 1993 ; Sandmeier & Dajoz, 1997; Mazer et al., 1999 ). Several authors, however, have challenged the expectation of negative (genetic) correlations between fitness components ( van Noordwijk & de Jong, 1986; Pease & Bull, 1988; Houle, 1991; de Jong & van Noordwijk, 1992; de Jong, 1993). They argued that phenotypic variation in fitness components is generated by two independent sets of loci that affect either the acquisition of resources from a common pool used by the two components or the allocation of resources to the two traits. While loci affecting allocation will promote negative covariances between fitness components, loci affecting acquisition will promote positive covariances. Therefore, measurements of the covariance between fitness components should take into account covariance caused by variation in plant size prior to flowering, either determined by resource availability and/or genetic variation among the individuals studied ( Houle, 1991; van Noordwijk & de Jong, 1986; de Jong & van Noordwijk, 1992). Moreover, Mazer & Delesalle (1998) pointed out that the mating system itself should have a strong influence on whether or not a negative (vs. positive) correlation should evolve between pollen and ovule production per flower. They argued that in predominantly selfing species there is no a priori reason to expect a negative correlation between male and female investment.

Several studies have demonstrated extensive within-population variation in sex allocation ( Horovitz, 1978; Freeman et al., 1980 ; Lloyd & Bawa, 1984; McKone & Tonkyn, 1986; Burd & Allen, 1988; de Jong & Klinkhamer, 1989; Campbell, 1992; Pickering & Ash, 1993; Dajoz & Sandmeier, 1997). This variation could have either a genetic or an environmental basis. A central explanation for this variation has been that, within a population, sex allocation might depend on an individual’s condition or resource status. If female and male components of fitness are differentially affected by changes in size or condition, then individuals are expected to modify their sex allocation according to their size ( Ghiselin, 1969; Charnov, 1982; Iwasa, 1991; Klinkhamer et al., 1997 ). Various sex-differential effects have been proposed as selective mechanisms driving gender adjustments in flowering plants. Hypotheses have incorporated a role for local mate or local resource competition ( Lloyd & Bawa, 1984), pollination syndrome ( Burd & Allen, 1988; McKone, 1987; Bickel & Freeman, 1993; Fox, 1993) and geitonogamous pollination ( Klinkhamer & de Jong, 1997; de Jong et al., 1999 ). Only few studies have considered whether there existed a genetic basis for variation in sex allocation ( McKone, 1989; Mazer, 1992; Agren & Schemske, 1995; Campbell, 1997; Fenster & Carr, 1997; Mazer et al., 1999 ). This lack of information is unfortunate, since the efficacy of selection in bringing about a change in patterns of sex allocation, and therefore the potential for evolutionary change, is determined by the amount of genetic variation available and the genetic correlation structure ( Lande & Arnold, 1983).

This study is concerned with sex allocation and genetic variation in traits related to male and female function at the time of flowering and fruiting (seedset) in the wind-pollinated and gynodioecious species Plantago coronopus. To quantify sex allocation, we determined the biomass of stamens, ovules, floral structures and seeds. We used maternal seed families sampled from one population and raised the offspring at two nutrient levels to separate environmental (resource status) and genetic effects. The specific questions asked in this study are: (1) is there heritable variation for investment in pollen and ovule production upon which natural selection can act?; (2) is there a trade-off between pollen and ovule production as measured as a negative genetic correlation among these characters; (3) is the magnitude of genetic variation influenced by the size of the resource pool; (4) does the resource status of a plant, as indicated by its size or the number of flowers produced, influence gender expression?

Materials and methods


Plantago coronopus is a small, short-lived perennial, in The Netherlands mainly growing in meadows along the coast. Germination takes place either in early spring or late autumn ( Schat, 1982). The species flowers from the beginning of May through September and overwinters as a rosette. It is a wind-pollinated, gynodioecious and predominantly outcrossing species (mean outcrossing rate 0.77; Wolff et al., 1988 ). The outcrossing rate in the studied population was estimated to be 0.80 ( Koelewijn, 1998). The flowers are protogynous, with an overlap in sexual phase. Flowers are formed on long flowering stalks (spikes), which consist of a stem and an ear, and produce a fixed number of five ovules and four stamens. Flowering and subsequent maturation occur from the base of the ear upwards. Each leaf axil has an axillary meristem that can produce either a spike, a lateral rosette or stay dormant. New spikes appear only when new leaves are formed after the first spike has been initiated. Reproductive performance is strongly size-dependent ( Waite & Hutchings, 1982; Koelewijn, unpublished data). Tall plants, with many leafs, form in most instances many large flowering spikes.

The genetics of the sexual polymorphism has been described in Koelewijn & van Damme (1995a, b) and differences in reproductive characteristics between male steriles and hermaphrodites are reported in Koelewijn (1996). Here we are only concerned with reproductive variation in perfect flowering hermaphrodites (cf. Koelewijn & van Damme, 1995a, 1996).


In February 1988, 120 nonflowering, adult plants were collected from a salt-meadow (Westplaat population, see Koelewijn & van Damme, 1995a). Plants were grown in the greenhouse for 6 months in nutrient-rich soil to achieve full flowering and to minimize maternal or age effects. After 6 months 24 maternal plants were randomly selected and self-fertilized. To quantify the amount of among- and within-family variation in reproductive characters we used the selfed seeds from these 24 maternal plants. For each maternal plant, seeds were germinated in Petri dishes in a growth cabinet with the following conditions: 16/8 h day/night, 20/15 °C day/night, photosynthetic photon flux density (PPFD) of 100 μmol m–2 s–1. Thereafter the seedlings were planted in trays filled with washed river sand and supplied with nutrients. The trays were placed in a growthroom for 1 week. Conditions were: 20 °C day/night, 16/8 h day/night, 70% relative humidity and PPFD of ±250 μmol m−2 s−1. Next, seedlings from each maternal parent were transferred to hydroponic growth units containing aerated nutrient solutions. In view of the paramount importance of nitrogen for growth processes ( Lambers et al., 1989 ), we used nitrate as the nutrient to control growth. In the ‘high nutrient’ treatment plants had free access to nitrate. The initial concentration of macro-ions in the nutrient solution were: Ca2+: 1.25 K+: 0.75; Mg2+: 0.5; H2PO4: 0.25; SO42−: 0.5; NO3: 1.5 meq L−1. Microelements were added according to Lewis and Power ( Hewitt, 1966) plus Fe as citrate. In the ‘low nutrient’ treatment the concentration of NO3 was reduced to 0.05 meq L−1. Conditions were otherwise the same. The nutrient solution was replenished every week and the pH was maintained between 5.5 and 6.0 with K2SO4 and adjusted daily. From each maternal parent three seedlings per nutrient treatment were used.


Plants were followed individually during the course of the experiment. When the first spikes appeared, the first one was removed when the anthers dehisced from the flowers at the bottom of the ear and used for determining the carbon content of the androecium, gynoecium and calyx. The second spike was marked with a coloured ring and the plant was harvested at the moment this spike produced ripe seeds. Thus, plants were harvested according to their physiological state and not at a fixed moment in time. At harvest plants were divided into vegetative (leaves and roots) and reproductive biomass (spikes), dried for 48 h at 70 °C and subsequently weighed to the nearest milligram. Sexual state of plants (H – hermaphrodite or MS – male sterile) was determined by eye, based on anther morphology ( Koelewijn & van Damme, 1995a, 1996). Sex expression was independent of nutrient level (G1 = 0.94, P = 0.63).

We quantified allocation to male and female function by determining the carbon content of different floral parts. From each collected spike three flowers, located just above the flowers with already dehisced anthers, were used for determining the carbon content by using a Unicarb Carbon Analyser ( Salonen, 1981). Flowers were dissected and separated in floral structures (corolla and calyx), stamens and gynoecium (ovules and pistil). From each flower two of the four stamens were used and androecium weight per flower was estimated as four times the mean stamen weight. Gynoecium weight is estimated by total ovule and pistil weight. Number of seeds per flower was counted from at least three randomly chosen flowers of the ripe (second) spike and carbon content of individual seeds was determined. If no seedset had occurred in one of the flowers, another flower was used. The empty flower, however, was still used for determining the mean seedset per flower. Pollination was done by hand with a brush when all the flowers on the spike were in the female stage. To prevent possible effects on seedset and seedweight due to pollen limitation or self-fertilization, a random pollen mixture of the other plants was used.

From these measurements, sex allocation was determined both on a per flower and on a per plant basis. At the per flower level, the mean absolute biomasses of the androecium, gynoecium and floral structures were calculated. Biomass of the seeds was obtained by multiplying the mean seeds per flower by the mean seed biomass. The proportion of costs allocated to the male function at the time of flowering, androecium/(androecium + gynoecium), was then calculated, and the final proportion allocated to male function, androecium/(androecium + seed). Allocation was also estimated on a per plant basis by multiplying the per flower biomass for androecium, gynoecium and seeds by the number of flowers.

Because of differences in the composition of different reproductive parts, allocation estimates may vary depending on what currency (e.g. biomass, C-content, nutrients) is used to characterize the allocation patterns ( Goldman & Willson, 1986; Ashman & Baker, 1992). However, Reekie & Bazzaz (1987) concluded that carbon allocation integrated the allocation patterns of other resources, and was therefore an appropriate common currency to assess allocation patterns. Moreover, both McKone (1987) and Ashman & Baker (1992) compared the allocation to reproductive structures based on different currencies (carbon, nitrogen and phosphorous). Although both found a significant effect on allocation resulting from currency, there was also a strong positive correlation between the allocation estimates. We suggest therefore that carbon allocation is likely to be an adequate measure.

Data analyses

Data were analysed by regular ANOVAs with SAS (SAS Institute 1987). Nutrient level is a fixed effect, whereas maternal family, plants within families, and flowers within plants are considered to be random effects. Only 19 of the original 24 families were used, because of segregation of male steriles, causing an unbalanced design and too few observations on hermaphrodites per family.

Within each nutrient level we performed nested ANOVAs to calculate the among- and within-family variance components. The among-family component of variance can be used as an estimate of the genetic variance including both additive and nonadditive effects ( Falconer, 1981). The broad sense heritability of a given character is estimated as the among-family component of variance, divided by the total phenotypic variance of that character, and multiplied by r, a correction factor based on the mating system of the species ( Lawrence, 1984; Becker, 1985). For obligate outcrossing species r is 2, for selfing species r is 1 and when using selfed families from a species with a mixed mating system, like in this study, r is best approximated by 1.5 ( Lawrence, 1984). The standard error of the heritability was calculated according to Lynch & Walsh (1998). The genetic coefficient of variation (CVg) = 100 √Vg/x, where x is the mean phenotypic value of the trait ( Houle, 1992), was also calculated. Genetic correlations between traits were estimated with family mean correlations.

To measure the sex of individual plants, we used Lloyd’s concept of the standardized phenotypic gender ( Lloyd, 1980; Lloyd & Bawa, 1984). Gender is herein defined as the proportion of an individual’s total fitness through male function:

inline image

where fi and mi are female and male fitness of individual i, and F and M are total female and male fitness in the population. Phenotypic gender can vary from zero (all female) to one (all male), and is 0.5 for an entire population. Because we did not measure fitness directly, we used investment in either stamens and ovules or seeds to characterize phenotypic gender.

We used an allometric model to measure relationship between size (x) and reproductive components (y): y = axb, or between the several reproductive components ( Peters, 1983; Prothero, 1986; LaBarbera, 1989). Least-squares regression of the natural log-transformed variables was performed. The regression coefficient estimates b, the exponent for the untransformed relationship and, thus, measures the rate of exponential increase of investment with, for example, size. In addressing the question of whether sex allocation changes with size, the appropriate null hypothesis is that the increases of male (m) and female ( f ) effort with size are isometric, that is when the exponents are equal ( LaBarbera, 1989). The null hypothesis can also be tested when a measure of size is lacking ( Rayner, 1985; LaBarbera, 1989). Isometry implies that the sex allocation ratio is constant over a range of sizes, thus log(m/f ) = constant. Therefore, with isometry the slope of log(m) on log( f ) is +1; a greater (smaller) slope is predicted when there is a male (female) size advantage. We used a t-test to determine whether the regression coefficient b was different from one: ts = (b − 1)/sb, where b is the regression coefficient and sb is the standard error of the regression coefficient ( Sokal & Rohlf, 1994).

The coefficient of variation was used to quantify variation in acquisition and allocation of resources between different traits. Acquisition refers to the total amount of resources available for division between the traits of interest and is measured in absolute units (e.g. g dryweight or μg C), while allocation is a proportional measure referring to the proportion of resources allocated to the traits. Therefore, variances will be different and need to be standardized for fair comparison.


Nutrient effects

The phenotypic variation in growth and reproductive characters at the two nutrient levels is described in Table 1. Nutrient effects were most pronounced at the plant level, with high nutrient plants producing about four times as much vegetative biomass and about eight times as much reproductive biomass. In contrast, at the flower level differences were absent or less pronounced. No difference in resource investment between the two nutrient levels in either androecium (stamens) or gynoecium (ovules and pistil) biomass was observed. Investment in floral structures was slightly higher at the low nutrient level. Consequently, the fraction of resources allocated to male function at flowering did not differ between nutrient levels and was strongly male biased (Table 1, Fig. 1a). Standardized phenotypic gender varied from 0.38 to 0.61 at the high nutrient level and from 0.32 to 0.67 at the low nutrient level ( Fig. 1c), indicating considerable variation in gender among perfectly flowering hermaphrodites.

Table 1.   Phenotypic variation in growth and reproductive characters in Plantago coronopus grown in the greenhouse at two nutrient levels (mean ± SE, n = 53–57 individuals per nutrient treatment). The plants represent 19 selfed seed families collected from one population in The Netherlands. Differences between treatments are tested by two-way factorial ANOVA with nutrient level (nut) and maternal family as the main factors. Significance of the F-ratios are indicated as: ***P < 0.001; **P < 0.01; *P < 0.005; ns – not significant. Thumbnail image of
Figure 1.

 Frequency distribution of maleness (proportion of resources invested in male function) and standardized phenotypic gender in Plantago coronopus grown in the greenhouse at two nutrient levels. (a) Maleness at flowering; (b) gender at flowering (ovule-based estimates); (c) maleness at fruiting; (d) gender at fruiting (seed-based estimates).

Despite abundant pollination, seedset was lower at the low nutrient level, suggesting a higher seed abortion (Table 1). Seedweight was also reduced at the low nutrient level. Structural limitations due to smaller flower size do not seem to be a likely explanation, because floral structures were even somewhat larger at the low nutrient level (Table 1). Thus, plants at the lower nutrient level produced fewer and smaller seeds. Because there was no difference in androecium weight, the fraction of resources allocated to male function at fruiting differed between the two nutrient treatments, with the high nutrient plants having a more female-biased sex allocation pattern (Table 1). Overall sex allocation at the fruiting stage was also female biased ( Fig. 1b). Plants therefore shift from a male-biased to a female-biased sex allocation pattern over time. Standardized gender at the seed stage varied from 0.32 to 0.70 at the high nutrient level and from 0.31 to 0.83 at the low nutrient level ( Fig. 1d).

For most traits a significant nutrient-by-family interaction was detected (Table 1), indicating a difference in response to nutrient concentration among maternal families. However, although there was a significant interaction with nutrients for gynoecium, androecium and seed mass, the trait of primary interest, that is proportional allocation to male function, did not display a significant nutrient-by-family interaction (Table 1). This suggests that the male and female components reacted in the same way to nutrient differences, despite the differences in absolute reponse among families.

Among-family variation

The nested ANOVAS indicated significant among-family variation for androecium and gynoecium weight at both nutrient levels (Table 2). Investment in floral structures only showed a significant maternal component at the low nutrient level. No significant among-family effects for seedweight were observed, while for number of seeds per flower a significant maternal component was detected only at the high nutrient level (Table 2). In most cases total phenotypic variation was higher at the low nutrient level (Table 2). Broad-sense heritabilities were estimated as 0.50 and 0.56 for androecium mass and 0.35 and 0.48 for gynoecium mass, for the high and low nutrient level, respectively (Table 3). The genetic coefficient of variation for seedweight and number of seeds per flower was in the same range as for androecium and gynoecium mass (Table 3), suggesting that the lack of significance and low estimate of heritability resulted from relatively high environmental variation. Indeed, the within-plant variation for these traits appeared to be higher at both nutrient levels (Table 2). Genetic variation in costs allocated to male function on a proportional basis could be detected only at the flowering stage (Table 4).

Table 2.   Variance in reproductive characteristics in a population of Plantago coronopus analysed as nested ANOVAs at two nutrient levels. Thumbnail image of
Table 3.   Genetic coefficients of variation (CVg), and within-population broad-sense heritabilities (h2 ± SE) of per-flower biomass traits for Plantago coronopus. Thumbnail image of
Table 4.   Variance in the proportion of resources invested in male function at both the flowering (ovule) and fruiting (seed) stage in a population of Plantago coronopus. Thumbnail image of

Phenotypic and genetic correlations

Contrary to the expectation of a trade-off in sex allocation, maternal families that invested more in male function, as indicated by heavier stamens, did not invest less in female function. In particular, biomass of the androecium showed a positive phenotypic and family mean correlation with gynoecium biomass at both nutrient levels (Table 5, Fig. 2a). The family mean correlation might well represent a genetic correlation, because for both traits genetic variance was detected (Table 2). In contrast, no significant family mean correlation was detected between the seedmass per flower and androecium weight (Table 5, Fig. 2c). Size of a plant, as estimated by vegetative dryweight, did not influence investment in reproductive structures at the flower level, because none of the correlations with size was significantly different from zero (Table 5).

Table 5.   Family mean correlations (above diagonal, n = 19) and phenotypic correlations (below diagonal, n = 52−55) for Plantago coronopus. Values given are Pearson product−moment correlations. Bold typeface indicates that correlations are significant at at least the P < 0.05 level. Thumbnail image of
Figure 2 (a) Androecium mass and gynoecium mass per flower per plant; (b) total gynoecium mass (initial maternal investment) and total androecium mass (paternal investment) per plant; (c) androecium mass and total seed mass per flower per plant; (d) total seed mass (final maternal investment) and paternal investment per plant. Closed symbols – high nutrient treatment (n = 54); open symbols – low nutrient treatment (n = 52.

Figure 2 (a) Androecium mass and gynoecium mass per flower per plant; (b) total gynoecium mass (initial maternal investment) and total androecium mass (paternal investment) per plant; (c) androecium mass and total seed mass per flower per plant; (d) total seed mass (final maternal investment) and paternal investment per plant. Closed symbols – high nutrient treatment (n = 54); open symbols – low nutrient treatment (n = 52.


Genetic correlations may differ depending on whether they are calculated on the basis of the average per-flower investment or whole-plant investment. At the whole-plant level any genetic variation in flower number will contribute to positive correlations of two floral parts. Recalculation of both phenotypic and family mean correlations based on whole-plant totals led to much larger values than those obtained using average per-flower biomasses (Table 5). All correlations between reproductive parts at the time of flowering and fruiting were strongly positive, with estimates ranging from r = 0.76 to 0.96 (Table 5, Fig. 2b,d). Moreover, all correlations between size and reproductive components were positive. The size effects, however, were more pronounced at the lower nutrient level (Table 5).

Size-dependent sex allocation

Although the correlations between investment in male and female function were positive, both at the flower and plant level, linear regressions of log-transformed reproductive components revealed different relationships between male and female function. At the flower level the regression slopes were significantly smaller than one at both nutrient levels (Table 6). Flowers with a higher total resource acquisition allocated proportionally more resources to the gynoecium or to the seeds, resulting in a more female-biased sex allocation pattern with increasing flower size. This pattern was more pronounced at the flowering stage than at the fruiting stage, due to the weak correlation between seedmass per-flower and androecium weight ( Fig. 2a,c). At the plant level, however, three of the four slopes were not significantly different from one, indicating an equal investment in male and female function irrespective of size or resource acquisition (Table 6). This also indicates that size differences among plants are of paramount importance, because even a weak relationship between seedmass and androecium weight at the flower level corresponds with a strong linear relationship between paternal and maternal investment at the plant level ( Fig. 2c,d). At the high nutrient level, the slope of the regression between paternal and maternal investment was less than one (Table 6), suggesting a more female-biased sex allocation pattern with increasing size. The correlation between plant size and proportion of resources allocated to male function was indeed negative (r = −0.38, n = 51, P < 0.05), while at the low nutrient level or at flowering the correlation between maleness and size was not significantly different from zero.

Table 6.   Allometric analysis of reproductive components in Plantago coronopus (see also Fig. 2). Results are from least-squares regression of log-transformed reproductive components: log(male component) = a + b × log(female component); r2 is the percentage of variance explained by the regression analysis. Also indicated is the result of a test for isometry, i.e. whether the slope of the regression (b) is significantly different from 1 (t-test with 52−54 d.f.; see Material and Methods for explanation). Level of significance as in Table 1. Thumbnail image of


Genetic variation

The results of this study indicate the presence of genetic variation for androecium (stamen) and gynoecium (ovule) biomass in Plantago coronopus. Broad-sense heritabilities varied from 0.35 to 0.56 and were slightly higher for the male trait. The few other studies that have reported genetic variation for these traits suggest lower heritable variation for male traits. In Begonia semiovata ( Agren & Schemske, 1995), Lythrum salicaria ( O’Neill & Schmitt, 1993) and Mimulus guttatus ( Fenster & Carr, 1997) genetic variation in ovule biomass exceeded the variation in stamen biomass. On the other hand, Campbell (1997) reported genetic variation for stamen biomass (h2 = 0.40) and no variation for ovule biomass (h2 = 0) in Ipomopsis aggregata. Moreover, Mazer et al. (1999 ) reported a higher realized heritability for number of anthers (h2 = 0.26) as compared to number of ovules (h2 = 0.12) after two generations of artificial selection for either increased anther or ovule number in Spergularia marina. No genetic variation was detected for seed weight. This last result confirms several other studies (e.g. Mazer, 1987, 1992; Campbell, 1997). Phenotypic variation in seed weight is influenced by maternal effects, position effects among the ear and several developmental sources of variation, resulting in a high within-plant variance component and low to zero heritability (Table 2). Because we estimated genetic variance from data on maternal seed families grown in the greenhouse, our estimated heritabilities represent an upper limit. There are three reasons why estimates derived in this way should be taken only as indications of the available genetic information in the source population. First, the genetic variation for a given character may vary across environments. However, similar results were obtained at both nutrient levels. Second, the among-family component of variance may be inflated by maternal environmental effects. To prevent strong maternal effects, parental plants were kept in the greenhouse for 6 months before being used in crosses. Third, in partially inbred populations, individuals will differ in their degree of inbreeding, resulting in nonrandom genotypic associations. The observed fixation index (F) among adult plants in the studied population was 0.099 ± 0.125 (mean ± SD), indicating a slight excess of homozygotes. The low fixation index suggests, however, that this factor will not have had much influence on our estimates.

The presence of genetic variation in stamens and ovules indicates that selection for these traits is possible and that adjustment of sex allocation at the population level is possible. An important result of this study was the detection of genetic variation in maleness at flowering and the absence of a nutrient-by-family interaction for this trait. Thus, there is a genetic basis for a more male- or female-biased allocation pattern. Abundant phenotypic variation in gender has been shown in several studies (e.g. McKone & Tonkyn, 1986; de Jong & Klinkhamer, 1989; Solomon, 1989; Campbell, 1992; Fox, 1993; Kudo, 1993; Pickering & Ash, 1993; Dajoz & Sandmeier, 1997; Wright & Barrett, 1999). Only few studies, however, have also been concerned with the genetic basis of this variation ( Mazer, 1992; Savolainen et al., 1993 ; Agren & Schemske, 1995; Campbell, 1997; Mazer et al., 1999 ). In contrast to flowering, no genetic variation in maleness was detected at fruiting. There are two possible explanations for this observation. First, limited genetic variation was detected for either seed mass or number of seeds per-flower. Most variation in both traits was within flowers or among individuals, causing a high environmental variance component. The high variance in female investment among plants within families would therefore prevent the detection of differences in male allocation among families. Second, in contrast to investment in ovules, investment into developing seeds is a postanthesis process, and might, due to different developmental processes within a plant or to differences in timing, make use of different resources.


Sex allocation theory assumes that genotypes investing more resources in male production have less available for female reproduction. All else being equal, if this assumption is met, the expectation is a negative correlation between investment in the two sexual functions. Here, no evidence was found for such negative genetic correlations. On the contrary, family mean correlations between investment in the androecium and gynoecium were positive at flowering. Other studies attempting to quantify the genetic correlation between allocation to male and female function at flowering also report either zero or positive correlations ( Mazer, 1992; Agren & Schemske, 1995; Campbell, 1997; Fenster & Carr, 1997; Sandmeier & Dajoz, 1997). At fruiting the genetic correlation was not demonstrably different from zero. The same result was obtained by Campbell (1997) in Ipomopsis aggregata, but others ( Atlan et al., 1992 ; Garnier et al., 1993 ; Savolainen et al., 1993 ) have reported negative genetic correlations between pollen production and seed-set. However, because seed-set is a postanthesis process it is not clear whether trade-offs between pollen production and seed-set actually represent allocation trade-offs between male and female function since seed production also influences future female function.

We observed a large difference in the magnitude of the (genetic) correlation at the flower and plant level. The same result was observed in Ipomopsis aggregata ( Campbell, 1997). The results indicate the importance of size differences and modular development in plants, which may lead to uncoupling of processes at different levels (cf. Morgan & Schoen, 1997). Investment in male and female function was therefore analysed at both the flower and the plant level. Positive correlations between the amount of resources allocated to two alternative functions are possible if individuals differ in their ability to acquire resources ( van Noordwijk & de Jong, 1986; Houle, 1991), or if trade-offs involve more than two characters ( Pease & Bull, 1988; de Jong, 1993). Differences in resource acquisition can be caused by environmental differences or by differences among maternal plants. In either case, the basic premise of sex allocation theory that all individuals have the same amount of resources is invalidated (cf. Stanton & Galloway, 1990). Environmental differences were minimized in our design, because all plants received the same amount of nutrients. Therefore, differences in resource acquisition will likely reflect genetic differences. De Jong (1993) modelled resource acquisition in a hierarchial fashion ( Fig. 3) and indicated that: (1) if the variance in resource acquisition is larger than the variance in resource allocation at a node, positive correlations are to be expected, and (2) the sign of the correlation also depends on the variances and mean allocations at higher nodes in the allocation tree. Thus, the covariance between female and male investment (cov(F,M), Fig. 3) is also influenced by the magnitude of c and d, and the covariances at nodes 1 and 2 ( Fig. 3). Moreover, if the variance in acquisition at node 1 is large, it will be difficult to detect negative correlations at the other nodes ( de Jong, 1993). At the plant level acquisition variance exceeded allocation variance at all nodes, while at the flower level acquisition and allocation variance were about equal (Table 7). Also, the covariances at all three nodes were positive (results not shown; calculated according to Table 1 in de Jong, 1993). Our results indicate that differences among families in acquiring resources are large. The detection of trade-offs or compensation will therefore be difficult and seems most likely to be observed at the flower level. Indeed, the only study we are aware off that detected a clear genetic trade-off between male and female function was an artificial selection study for increased ovule or anther number in Spergularia marina ( Mazer et al., 1999 ): an increase in anther number was accompanied by a decrease in ovule number and vice versa. Besides minimizing the environmental variance, genetic variation should also be reduced by using only a few well- characterized lines. One promising approach to detect shifts in allocation is the comparison of male sterile and hermaphrodite siblings ( Poot, 1997), i.e. make use of different individuals within the same genetic background.

Figure 3 Reproductive allocation tree, that is a diagram of repeated consecutive allocation of resources to traits (after de Jong, 1993). Per individual, the amount R of resource acquired is allocated (node 1) to vegetative biomass according to a fraction (1 – c) or the branch representing reproductive biomass and leading to further allocation at the node 2 according to a fraction c. The amount of resource at node 2 is allocated to floral structures according to a fraction (1 – d) or to the branch representing sexual investment and leading to further allocations at the node 3 according to a fraction d. The amount of resource at node 3.

Figure 3 Reproductive allocation tree, that is a diagram of repeated consecutive allocation of resources to traits (after de Jong, 1993). Per individual, the amount R of resource acquired is allocated (node 1) to vegetative biomass according to a fraction (1 – c) or the branch representing reproductive biomass and leading to further allocation at the node 2 according to a fraction c. The amount of resource at node 2 is allocated to floral structures according to a fraction (1 – d) or to the branch representing sexual investment and leading to further allocations at the node 3 according to a fraction d. The amount of resource at node 3.

is allocated to female function according to a fraction (1 – e) and to male function according to a fraction e. At each node the sign of the covariance between the two traits at the node determines whether there is a positive or negative correlation between the traits.

Table 7.   Acquisition and allocation of resources in Plantago coronopus at fruiting. Data were analysed according to the allocation tree in Fig. 3. Because acquisition was meaured in absolute units (g dry weight or μgC) and allocation is a fraction between 0 and 1, coefficients of variation have been used to compare the variability in acquisition (CVac) and allocation (cval) (n = 52–55). Acquisition was determined at both the flower the plant level. The fraction of resources allocated to reproduction, sexual structures and male function corespond to fractions c, d, and e in the allocation tree of Fig. 3. Thumbnail image of

Sex allocation

The mean proportion of resources invested in male function was 0.86 at flowering and varied from 0.30 (high nutrient level) to 0.40 (low nutrient level) at fruiting. Thus, the sex allocation pattern shifts from male-biased to female-biased. Hermaphroditic plants can change their sex allocation either by producing different amounts of pollen and ovules per flower, by producing sterile stamens and pistils or through abortion of seeds or entire fruits ( Stephenson, 1981). The difference at fruiting came about by reducing the number and size of seeds. Stamen mass and ovule mass were not different between the two nutrient levels and mass of floral structures differed only slightly. Individual plants therefore seem to have a fixed investment in flowers and use their excess of resources by producing more, and not larger, flowers. Field estimates of the Westplaat population (Koelewijn, unpublished data) did not differ from the greenhouse ((mean ± SE) androecium: 177.2 ± 5.02 μg C; gynoecium: 27.8 ± 1.27 μg C; seedmass per-flower 315 ± 15.0 μg C; floral structures 123.4 ± 3.73 μg C; n = 20). Male investment therefore seems to get priority in Plantago coronopus, because it did not differ among the field and nutrient treatments. Our estimates of the proportion of resources invested in male function at fruiting are high compared with animal pollinated species (range 0.05–0.12; Lloyd, 1984; Campbell, 1997), and somewhat lower than the few estimates available for wind-pollinated outcrossing species (range 0.42–0.48; McKone, 1987; Poot, 1997). However, our estimates are in close agreement with the model of Ross & Gregorius (1983), who predict that the ESS for investment in male function in a self-compatible species will be ½(1 − s), where s is the selfing rate of the population. Assuming a selfing rate of 0.20 for the Westplaat population ( Koelewijn, 1998), the predicted allocation to male function would be 0.40.

Sex allocation theory and ESS models predict the optimal allocation of resources to male and female function ( Charnov, 1982; Lloyd, 1984; de Jong et al., 1999 ). The most well known predictions of sex allocation theory are that the evolution of self-fertilization should lead to reduced expenditure on male function and that plants adjust their gender according to size. These predictions, however, depend on the shape of the curve relating male and female contributions to fitness ( Charnov, 1982; Morgan, 1992; Klinkhamer et al., 1997 ). If fitness returns are equal for both male and female contributions, the optimal sex allocation will be 0.5. However, if fitness returns differ between the two sex functions, different optimal allocation patterns are expected. The majority of studies in animal-polllinated species indicate that femaleness increases with size. The male fitness curve levels off more quickly with plant size than the female fitness curve, because of local mate competition (LMC) or geitonogamy ( de Jong & Klinkhamer, 1989; Snow et al., 1995 ; Klinkhamer & de Jong, 1997). For wind-pollinated species there is less agreement. Either more male-biased ( Burd & Allen, 1988; Solomon, 1989; Bickel & Freeman, 1993; Fox, 1993; Klinkhamer et al., 1997 ) or female-biased ( de Jong et al., 1999 ) patterns are predicted with increased size. In wind-pollinated species the male fitness curve is expected to be linear (no LMC), and the female fitness curve is expected to decelerate, as in animal-pollinated species, because seedlings from plants that produce many seeds suffer more from local resource competition (LRC) and are therefore less successful. Maleness would therefore increase with plant size in wind-pollinated plants ( Burd & Allen, 1988; Klinkhamer et al., 1997 ). However, de Jong et al. (1999 ) argued that this conclusion only holds for self-incompatible species. The reversed situation could be true in self-compatible wind-pollinated species that suffer from inbreeding depression when selfing increases with size. Our results indicate an increase of femaleness with size in Plantago coronopus. First, female allocation was higher at the high nutrient level. Secondly, within the high nutrient level larger plants invested proportionally more in seeds. There are two alternative explanations for this pattern. Both explanations rely on LMC, but focus on different aspects of size. Petersen & Fischer (1996), working with sea basses, argued that smaller individuals contribute a smaller proportion of the sperm released in spawns with multiple spawners and thus are under more intense sperm competition than larger indivduals, which should select for increases in male allocation in smaller individuals. Plant size and number of flowering spikes are closely related in Plantago coronopus. Small individuals only have a few spikes, while large individuals continuously produce new spikes. Thus, small individuals have only few opportunities to release their pollen. This mimics the situation in sea basses. On the other hand, de Jong et al. (1999 ) focus on large individuals to explain increasing femaleness. Having many spikes flowering at the same moment will increase the possibility of selfing, causing diminishing returns for male investment and selecting for increased femaleness.

Finally, it is important to note that the existence of genotypes with a more male- or female-biased allocation pattern within a population will give rise to sexual asymmetry and continuous frequency-dependent selection ( Ross, 1989), thus generating possibilities for the maintenance of genetic variation in reproductive characters.


Thanks to Olga Clevering, Jos van Damme and two referees for reading and commenting on the manuscript. H.P.K. was supported by a BION-NWO (Netherlands Organization for Scientific Research) postgraduate scholarship. This is publication number 2600 from the Netherlands Institute of Ecology.