Plasticity in the self-incompatibility system of Solanum carolinense

Authors


Andrew G. Stephenson
Email: as4@psu.edu

Abstract

Solanum carolinense has a gametophytic self-incompatibility (GSI) system that is typical of the Solanaceae in which pistils produce specific S-RNase proteins that disrupt the growth of pollen tubes sharing the same S-allele. However, unlike most self-incompatible plants Solanum carolinense is a weed. Self-incompatibility is uncommon in weeds because disturbed habitats require frequent recolonization (hence populations are repeatedly founded by few individuals bearing a limited number of S-alleles), effective population sizes are small (supporting few S-alleles) and habitats are ephemeral (so there is limited time for the migration of additional S-alleles into populations). We carried out a series of greenhouse experiments using clonal replicates (rhizome cuttings) of plants from two natural populations of S. carolinense to determine if there is variation in the strength of GSI within these populations. We found that the growth rate of self-pollen tubes and self-fertility increases with floral age. That is, flowers become more self-compatible as they age. Moreover, we found that self-fertility increases on plants in which the first 20 flowers receive no cross pollen. That is, when few or no fruits are produced on the first 20 flowers, self-pollination is more likely to result in fruit/seed set. Finally, we found that genotypes differ in their degree of self-fertility indicating that there is broadsense heritability for plasticity in the strength of self-incompatibility. These findings indicate that some genotypes of S. carolinense are capable of producing self-seed when cross pollen is scarce, even though the plants have a functional GSI system.

Introduction

Most flowering plants are hermaphrodites and, consequently, possess the potential to self-fertilize (Yampolsky & Yampolsky 1922). A variety of traits are known to influence the propensity to self-fertilize (the breeding system), such as the temporal and spatial separation of the anthers and stigma within the flowers, the shape of the flower in relation to the morphology of the pollinators, and the number and arrangement of flowers on the plants (e.g. Wyatt 1983; Diggle 1992). Fisher (1941) noted that there is an inherent genetic transmission advantage to selfing because a plant donates two haploid sets of chromosomes to each selfed seed and can still donate pollen to conspecifics. Many others have noted that selfing is also advantageous if it provides reproductive assurance when pollinators are scarce or unreliable and seed set is limited by the availability of cross pollen (e.g. Stebbins 1957; Baker 1965; Lloyd 1992; Lloyd & Schoen 1992; Schoen et al. 1996). The propensity to self-fertilize is believed to be particularly important for weedy species that live in ephemeral habitats, undergo frequent episodes of colonization and often occur in small populations (e.g. Baker 1955). Consequently, genetic variants that promote self-fertilization should increase in frequency unless they are opposed by other evolutionary forces, such as inbreeding depression (the reduction in fitness of inbred progeny relative to outcrossed progeny) and pollen discounting (the adverse affects that a trait that promotes selfing may have on the ability of a plant to donate pollen to conspecifics) (e.g. Nagylaki 1976; Holsinger et al. 1984; Charlesworth & Charlesworth 1987). Consequently, when a genetic variant that promotes selfing arises in a population, its impact on the breeding system is determined by the major forces favoring and opposing self-fertilization which, in turn, depend on environmental conditions within the population (e.g. pollinator availability) and the genotypes (e.g. genetic load) of the individuals that possess the variant.

Nearly half of angiosperm families are reported to include species exhibiting one of several forms of self-incompatibility (heteromorphic, gametophytic or sporophytic) (de Nettancourt 1977). Self-incompatibility (SI) is a genetically based system (typically controlled by a multi-allelic S-locus) whereby self-pollen is recognized by the pistil and rejected prior to fertilization (de Nettancourt 1977). Therefore, SI is a trait with a strong influence on the breeding system (inhibits selfing) that allows a hermaphroditic plant to avoid the adverse affects of inbreeding on progeny performance. Natural populations, however, often exhibit marked phenotypic variation in the strength of SI (e.g. Levin 1996; Tsukamoto et al. 1999; Stephenson et al. 2000). In some species the strength of SI is known to be influenced by environmental conditions such as temperature, by internal stylar conditions such as the age of the flowers, by mutations that directly affect the strength of S-alleles (e.g. weak and strong S-alleles), by mutations that render a specific S-allele functionless, and by unlinked genetic modifiers that can affect the strength of S-alleles in the population (see Levin 1996; Stephenson et al. 2000; Good & Stephenson 2002; Tsukamoto et al. 2003a,b).

Solanum carolinense L. (Solanaceae) is typical of the self-incompatible members of the Solanaceae in that it has a T2-type RNase-based gametophytic self-incompatibility (GSI) system in which pistils produce S-allele-specific RNase proteins that disrupt the growth of haploid pollen tubes that have an S-allele in common with the pistil (Clarke & Newbigin 1993; Lee et al. 1994; Igic & Kohn 2001). However, unlike most self-incompatible plants, S. carolinense is a weed. Richman et al. (1995) cloned and sequenced S-alleles from plants in two populations (from North Carolina and Tennessee, USA) and estimated that each population contained 12–14 S-alleles. We have cloned and sequenced S-alleles from 15 plants in a population of S. carolinense from Cumberland, MD, USA, and we estimate that this population also contains 12 S-alleles of which nine are also found in the more southern populations examined by Richman et al. (1995).

Not surprisingly, SI is uncommon in weeds (Baker 1955; Byers & Meagher 1992). Consequently, the objective of this study was to determine if there is plasticity in the strength of SI in S. carolinense when outcross pollen is unavailable. Specifically, we addressed the following questions: (i) is there variation in the strength of SI among plants within a population; (ii) is the strength of incompatibility of an individual plant influenced by prior (outcrossed) fruit production; and (iii) does the strength of incompatibility change with flower age?

Materials and methods

Study organism

Solanum carolinense is a weedy, herbaceous perennial that is found in ephemeral habitats and agricultural fields throughout southeastern Canada and central and eastern USA (Britton & Brown 1970). Once established it spreads via rhizomes that may extend a meter or more from the parent stem (Ilnicki et al. 1962). Both growth and reproduction are indeterminate. The flowers are similar in shape to those of its congeners (e.g. potato, Solanum tuberosum) and measure approximately 2 cm in diameter, with five partially fused white to violet petals and five yellow anthers. The flowers are visited by pollen-gathering bees, which must vibrate the flowers to remove pollen from the poricidal anthers (Hardin et al. 1972). Inflorescences consist of 1–20 flowers that mature acropetally. The fruits are round, yellow berries measuring 2–3 cm in diameter and typically contain 60–100 seeds.

Preliminary study

In November 2000, after several hard frosts, the rootstocks of 43 S. carolinense plants were collected from a large population near Cumberland, MD, USA. All of the rootstocks were separated by a minimum of 5 m within the population and subsequent sequencing of the S-alleles and/or hand pollinations revealed that 35 of the plants had a unique combination of S-alleles, indicating that they were unique genotypes (data not shown). The rootstocks were placed into 4 L pots with potting soil and placed into a cold room at 4°C for 60 days. We placed the 35 unique genotypes into a heated greenhouse and self-pollinated each of the first 200 flowers that were produced on each plant to determine if there was variation among the plants in their self-fertility. Pollinations were carried out every 3–4 days on all of the previously unpollinated flowers. Pollen was collected by ‘buzzing’ the anthers with a modified electric toothbrush and catching the pollen in empty gel caps. We then applied the self-pollen onto stigmas by dipping the stigma of each unpollinated flower into the self-pollen. We collected and counted the mature fruits approximately 2 months after pollination.

We found that when the plants received only self-pollinations they varied dramatically in the number of fruits produced following these self-pollinations (Fig. 1). One plant produced 78 fruits whereas other plants produced few or no fruits. Although most of the fruits were small and contained only a few seeds (typically less than 15), this preliminary study indicated that there was variation among plants in this population for self-fertility (i.e. some plants were not 100% self-incompatible) and that more detailed studies were warranted.

Figure 1.

Variation in the number of fruits produced by self-pollinating the first 200 flowers produced on each of 35 plants (genotypes). An * indicates those genotypes that were used in the prior fruit experiment I and a † indicates the genotypes that were used in the prior fruit experiment II.

Prior fruit experiment I

We conducted an experiment comparing the consequences of self-pollination between plants with and without outcrossed fruits to determine the effect of outcross fruit production on self-fertility. We selected 15 plants (genotypes/genets) from the 35 plants used in the preliminary study (indicated by an * in Fig. 1) that represented a wide range of self-fertility. We made 7–15 clones (ramets) of each genet through the propagation of rhizome cuttings from the original plants. The ramets were grown in a greenhouse (14 h day length), watered daily and fertilized weekly with a phosphorous-rich fertilizer (9-45-15 NPK plus micronutrients) until flowering. After flowering they were watered with a more balanced fertilizer (15-16-17 NPK plus micronutrients). Half of the ramets from each genet were randomly assigned to one of two treatments: the outcross-fruit treatment and the self-pollination-only treatment. In the outcross-fruit treatment, we pollinated the first 20 flowers of each plant with a mixture of outcross pollen from three or more different genets. In the self-pollination-only treatment, we pollinated the first 20 flowers of each plant with self-pollen. We then pollinated the next 40 flowers on all ramets in both treatments with self-pollen. Thus, we could compare the self-seed production of plants that had a substantial fruit load due to outcrossing with plants bearing very few fruits, the self-pollination-only ramets.

Pollinations were conducted three times per week on new flowers that had not been previously pollinated. Pollen was collected and pollinations were carried out using the method outlined in the preliminary study. We recorded the pollination type, node number and inflorescence number of each pollinated flower. We collected the mature fruits on each ramet and counted the seeds 2 months after pollination. Together these data allowed us to calculate fruit set per self and outcross pollination, seed number per outcross and self-pollination and total seed and fruit production per plant.

Floral age experiment

We collected plants from a natural population of S. carolinense growing near State College, PA, USA, in the autumn of 2002 to determine if the age of the flower affects the ability of the SI system to arrest the growth of self-pollen tubes. These plants were potted and given a cold treatment using the method outlined in the preliminary study. In the spring of 2003, the plants were placed into a greenhouse and watered and fertilized using the method outlined in the prior fruit experiment. We found that each flower lasted approximately 7 days if it was not pollinated under our greenhouse conditions and that each flower undergoes a color change as it ages (from white to light blue). We observed that young flowers (1–2 days after anthesis) were white and the petals were slightly reflexed. On middle-aged flowers (3–5 days after anthesis) the petals were flat (perpendicular to the style) and the margins of the petals were just beginning to turn blue. On  old  flowers  (5–7 days  after  anthesis)  the  petals were noticeably blue and beginning to close around the anthers/style. On seven plants that were flowering profusely in the greenhouse we identified five young, five middle aged and five old flowers on each plant. Each of these flowers was self-pollinated. After 48 h the styles from each flower from each plant were harvested, the pollen tubes were stained with aniline blue and pollen-tube growth was examined with a microscope under ultraviolet light (Martin 1959). The progress of the pollen tubes through the style was determined by counting the number of tubes at five equidistant points along the style: 20% (just below the stigma, top of style), 40%, 60%, 80% and 100% of the distance from the stigmatic surface to the base of the style.

Prior fruit experiment II

We repeated the prior fruit experiment to determine the combined effects of outcross fruit production and floral age on self-fertility. The design of this experiment was very similar to that of the original experiment with three exceptions: (i) only 10 of the original 15 genets (indicated by a † in Fig. 1) were used in this experiment because the rootstocks of five of the genets failed to produce sufficient ramets; (ii) we pollinated the flowers on each ramet only once per week and, thus, roughly equal numbers of the flowers were young, middle and old aged; and (iii) we recorded the age category (young, middle and old) of each flower that was self-pollinated. All of the plants used in the second experiment were derived from rootstock of the plants in the self-pollination-only treatment of the original experiment.

Results

In the first prior fruit experiment only 14 of the 15 genets produced enough flowering ramets to be included in the study. Moreover, the ramets did not flower as profusely as did the parent plants in the preliminary study from which the ramets were derived. Consequently, we eliminated all ramets that did not produce at least 25 flowers from the analyses. This left 139 ramets distributed among 14 genotypes and two treatments. The ramets in the outcross-fruit treatment (20 cross-pollinations followed by self-pollinations) produced 40.9 ± 1.7 (mean ± SE) flowers, whereas the ramets in the self-pollination-only treatment produced 45.9 ± 2.0 flowers. Eighty-two percent of the outcross pollinations on the outcross-fruit treatment produced mature fruits containing an average of 81.4 ± 4.4 seeds. As expected in a species that is self-incompatible, fruit and seed production following self-pollination were dramatically lower than outcrossed pollinations on plants in both experimental treatments (Table 1).

Table 1.  Mean ± SE for three measures of self-fertility obtained from two experiments (prior fruit experiments I and II). Each experiment had two treatments: the outcross fruit treatment in which the first 20 flowers produced on each plant were outcrossed and all later flowers were self pollinated and the self-only pollination treatment in which all flowers on each plant were self-pollinated
TreatmentFruits per self-pollinationSeeds per fruitSeeds per self-pollination
Experiment I
 Outcross fruit0.005 ± 0.00 3.6 ± 0.90.02 ± 0.02
 Self-only pollinations0.027 ± 0.01 8.0 ± 1.70.11 ± 0.03
Experiment II
 Outcross fruit0.027 ± 0.01 2.0 ± 0.80.04 ± 0.02
 Self-only pollinations0.062 ± 0.0213.8 ± 5.80.34 ± 0.13

To assess the influence of the presence of outcross fruits and genet on the propensity for plants to produce fruit and seeds from self-pollinations, we conducted a mixed model anova with treatment and genet (specified as random) as the main effects (proc GLM, SAS Institute 2002). Because the raw data were not normally distributed and included many ramets that produced no seed from self-pollinations, we transformed the raw data to ranks for the anova. This analysis revealed that there were significant effects of both treatment and genet on the number of fruits produced per self-pollination, the number of seeds produced per selfed fruit, and the self-compatibility index (SCI) (Table 2). Ramets in the outcross-fruit treatment produced fewer fruits per self-pollination and the self-fruits contained fewer seeds than ramets in the self-pollination-only treatment (Table 1). The SCI is the number of seeds per self-pollination on each genet divided by the number of seeds produced per outcross pollination on each genet. Therefore, SCI is a measure that ranges from 0 to 1, in which 1 is complete self-compatibility and 0 is complete SI. In this study, SCI ranged from 0 to greater than 0.06 among the 14 genets.

Table 2.  Results of anova on three measures of self-compatibility: fruit set (fruits per self-pollination), seed set (seeds per self-pollinations) and the self-compatibility index (SCI) for the first prior fruit experiment
Dependent variableEffectd.f.FPR2
  1. Genet was included as a random factor testing for differences among plants/genotypes. There were two treatments: outcross fruit treatment and self-pollination-only treatment. Raw data were first converted to ranks and these ranks were used in the analysis. d.f., degrees of freedom; G, genet; P, probability; R2, proportion of the variance explained by the model; T, treatment.

Fruit setModel27, 126 2.89<0.00010.441
Genet13 2.51  0.0051 
Treatment 118.29<0.0001 
G × T13 1.29  0.2303 
Seed setModel27, 126 2.60  0.00030.415
Genet13 2.01  0.0272 
Treatment 117.54<0.0001 
G × T13 1.24  0.2626 
SCIModel27, 126 2.58  0.00040.413
Genet13 1.97  0.0312 
Treatment 117.55<0.0001 
G × T13 1.24  0.2632 

In the floral age experiment, a mixed-effect model anova with genotype (specified as random), floral age and stylar section (distance from stigma to base of the style) and the two–way interactions revealed that each of the main effects and each of the two–way interactions had a highly significant influence (all P < 0.001) on the growth of self-pollen tubes (Table 3). Moreover, the model accounts for over 98% of the variance in pollen-tube growth! As is commonly observed in Solanaceous species with GSI (Stephenson et al. 2003), we found in the floral age experiment that the vast majority of the self-pollen tubes are arrested in the upper half of the style (Table 3). However, some pollen tubes traversed the entire length of the style and entered into the ovary. Furthermore, the number of tubes that traversed the length of the style was greater in some genotypes than others and the number increased with floral age. For example, in genotypes D and E no pollen tubes grew to the base of the style in young flowers, whereas in genotype F more than 30 pollen tubes on average entered the ovary in old flowers (Table 3).

Table 3.  Pollen tube counts following self-pollinations on seven plants/genotypes to young, middle-aged and old flowers
GenotypeStylar sectionYoung Mean ± SEMiddle aged Mean ± SEOld Mean ± SE
  1. Stylar section indicates the distance traveled by pollen tubes through the style (1, 20%; 2, 40%; 3, 60%; 4, 80%; 5, 100% of the entire length). Each mean ± SE represents tube counts from five flowers.

A1293.2 ± 37.1159.0 ± 25.7230.8 ± 21.4
2139.8 ± 18.4 69.2 ± 9.9108.4 ± 11.3
3 21.8 ± 6.2 32.0 ± 6.5 28.8 ± 4.6
4 12.6 ± 3.0  9.8 ± 1.1 12.4 ± 2.4
5  8.0 ± 2.4  4.0 ± 0.9  7.4 ± 1.4
B1159.7 ± 48.3167.0 ± 15.1170.4 ± 18.2
2 70.3 ± 23.3 62.0 ± 7.6 72.6 ± 10.6
3 10.0 ± 1.0 10.4 ± 0.9 29.2 ± 3.9
4  3.0 ± 1.5  3.8 ± 0.4 11.4 ± 0.8
5  0.7 ± 0.7  2.2 ± 0.4  7.6 ± 0.5
C1166.8 ± 6.4210.8 ± 33.8268.5 ± 14.5
2 92.8 ± 10.2102.8 ± 19.3142.3 ± 14.1
3 19.2 ± 3.6 27.0 ± 2.3 42.0 ± 10.4
4  9.0 ± 2.1  8.2 ± 1.0 15.3 ± 2.5
5  4.4 ± 1.5  3.4 ± 0.5  8.0 ± 1.7
D1209.7 ± 37.8184.4 ± 13.7203.4 ± 10.1
2117.3 ± 26.6 96.0 ± 6.4142.4 ± 7.8
3 34.0 ± 10.0 38.8 ± 4.6 76.6 ± 8.6
4  0.7 ± 0.7  1.0 ± 0.6 27.6 ± 5.3
5  0.0 ± 0.0  0.0 ± 0.0  4.6 ± 1.0
E1215.2 ± 28.4173.0 ± 39.2168.3 ± 16.0
2 84.4 ± 17.6 85.0 ± 18.4100.0 ± 12.6
3 15.6 ± 3.0 30.3 ± 8.8 35.0 ± 1.1
4  0.0 ± 0.0  7.3 ± 2.2 10.5 ± 2.9
5  0.0 ± 0.0  3.0 ± 1.4  6.3 ± 1.9
F1176.3 ± 29.9117.0 ± 27.4176.5 ± 3.3
2 71.5 ± 18.8 78.5 ± 24.3131.0 ± 10.4
3 14.5 ± 5.6 41.3 ± 18.0 83.3 ± 7.3
4  4.0 ± 4.0 14.0 ± 5.4 52.5 ± 8.7
5  0.5 ± 0.5  4.3 ± 1.1 30.8 ± 3.7
G1198.4 ± 10.4233.0 ± 17.9260.6 ± 16.9
2120.8 ± 10.3160.4 ± 12.5172.0 ± 11.0
3 20.4 ± 2.6 59.4 ± 6.5 65.6 ± 6.7
4  8.4 ± 0.9 15.8 ± 2.0 13.6 ± 1.5
5  2.2 ± 0.7  4.4 ± 0.5  4.4 ± 0.9

As in the first prior fruit experiment, cross-pollinations were far more likely to produce a mature fruit than self-pollinations and the fruits resulting from cross-pollinations contained far more seeds (88.9 ± 10.2) than fruits resulting from self-pollinations in either treatment (Table 1). However, fruit set per self-pollination, the number of seeds per fruit, and the number of seeds per self-pollination were all greater on ramets in the self-pollination-only treatment than on ramets in the outcross-fruit treatment (Table 1). As in the first experiment, an anova using ranked data revealed that both genet and experimental treatment had significant effects on self fruit set, seed number per self fruit and the SCI (Table 4). Ramets in the outcross fruit treatment produced fewer fruits per self-pollination and the self-fruits contained fewer seeds than ramets in the self-pollination-only treatment (Table 1).

Table 4.  Results of anova on three measures of self-compatibility: fruit set (fruits per self-pollination), seed set (seeds per self-pollinations) and the self-compatibility index (SCI) for the second prior fruit experiment
Dependent variableEffectd.f.FPR2
  1. Genet was included as a random factor testing for differences among plants/genotypes. There were two treatments: outcross fruit treatment and self-pollination-only treatment. Raw data were first converted to ranks and these ranks were used in the analysis. d.f., degrees of freedom; G, genet; P, probability; R2, proportion of the variance explained by the model; T, treatment.

Fruit setModel19, 62 4.92<0.00010.685
Genet 9 8.26<0.0001 
Treatment 111.23 0.0017 
G × T 9 0.80 0.6187 
Seed setModel19, 62 2.35 0.01040.509
Genet 9 3.12 0.0057 
Treatment 1 6.79 0.0125 
G × T 9 0.81 0.6117 
SCIModel19, 62 2.25 0.01390.499
Genet 9 3.10 0.0058 
Treatment 1 6.30 0.0159 
G × T 9 0.74 0.6710 

We expected that the three floral age categories (based on petal orientation and color) would be equally represented in our self-pollinations in the second prior fruit experiment. However, the three flower age categories were not equally represented in the flowers that were self-pollinated: over twice as many pollinations were made on flowers that were categorized as young (968) and middle aged (992) than on flowers categorized as old aged (365). These findings imply that we consistently placed some older flowers into younger aged categories and that the flowers categorized as old were within 1–2 days of abscission. We found that the proportion of self-pollinated flowers that subsequently produced fruit differed among the three floral age categories (χ2 = 9.03, d.f. = 2, P < 0.005). Approximately 4% of the old flowers produced fruit following self-pollination compared to 5.6% and 8.1% for young and middle-aged flowers, respectively. We also found that prior fruit treatment significantly influenced the proportion of self-pollinations yielding fruits with seeds (χ2 = 20.4, d.f. = 1, P < 0.001), but that the impact of the two treatments differed among the floral age categories (Fig. 2). Fruit set following self-pollination increased from young to middle-aged flowers for both treatments, but the proportion of fruit set in the self-pollination-only treatment was more than twice that of the self-fruit set in the outcross-fruit treatment. In contrast, there was little difference between treatments in self-fruit set for old flowers. The number of seeds per fruit ranged from 1 to 121 and averaged approximately four seeds per fruit across all flower types. There were no differences among the floral age categories in the number of seeds per fruit (Kruskal–Wallis test: χ2 = 2.50, d.f. = 2, P > 0.05).

Figure 2.

The proportion of self-pollinations on young, middle-aged and old flowers that produced mature fruit on ramets of 10 genets. Half of the ramets from each genet received only self-pollinations (control, ▪) whereas the first 20 flowers on the other half of the ramets were cross pollinated prior to the start of the self-pollinations (fruit treatment, ░).

Discussion

Solanum carolinense has a functional T-2 type RNase-based GSI system that is characteristic of the SI systems in the Solanaceae, Rosaceae and Scrophulariaceae (McCormick 1998; Igic & Kohn 2001). As is typical for species with this type of GSI (Anderson et al. 1986), the pistils of S. carolinense produce an abundant, highly polymorphic RNase that is developmentally regulated, 28–34 kDa in size, and has a highly basic isoelectric point (see Richman et al. 1995; Stephenson et al. 2003). Furthermore, these proteins co-segregate with the S-phenotype (Lee et al. 1994). The data presented in this study reveals that: (i) there is variation in self-fertility among individuals in a natural population; (ii) that there is environmentally induced plasticity in the strength of SI; and (iii) that this plasticity varies among individuals. Specifically, we have shown that self-fertility increases when there are no outcrossed fruits developing on a plant, that the number of self-pollen tubes that traverse the length of the style increases in older flowers, and that fruit and seed set following self-pollinations increases from young to middle-aged flowers.

Our data on the impact of floral age on self-fertility are, however, somewhat ambiguous. Because the number of self-pollen tubes that reach the base of the style and enter into the ovary is greatest in old flowers, we expected that self-fertility would be greatest in the old flower category in the second prior fruit experiment. However, this expectation was only partially met because the middle-aged flowers produced more mature fruits from self-pollinations than did the youngest flowers. Unexpectedly, the oldest flowers actually produced fewer fruits per self-pollination than the younger flowers. We suspect that this decrease in self-fertility in our old flower category may result from ovaries in the oldest flowers being earmarked for abscission regardless of the number of ovules that are fertilized just prior to abscission. The greatest window of opportunity for self-fertility would fall just prior to the physiological decision to abscise.

This study joins a growing list of studies that reveal genetically and environmentally induced variation in the strength of SI in natural populations (see Levin 1996; Stephenson et al. 2000, 2003). For example, Tsukamoto et al. (1999) examined natural populations of Petunia axillaris in Uruguay and found that self-fertility varied dramatically among populations. In one of the more self-fertile populations, self-fertility was caused by a modifier locus that suppressed the expression of an S-RNase allele (Tsukamoto et al. 2003a). In a second population, self-fertility resulted from mutations that rendered two of the S-alleles functionless (Tsukamoto et al. 2003b). In Campanula rapunculoides the growth of self-pollen tubes and self-fertility varied among individuals in two populations and also increased with the age of the flowers (Vogler et al. 1998; Stephenson et al. 2000). A quantitative genetic study revealed that these variations were heritable and resulted from the action of 3–4 unlinked modifier loci (Good-Avila & Stephenson 2002).

This study did not explore the genetic or physiological mechanisms underlying the observed variations in the strength of SI among individual plants. The data, however, indicate that the variations are not likely to result from mutations that render specific S-alleles functionless because such mutations are unlikely to change with floral age or prior fruit production (see Good-Avila & Stephenson 2002) and because seed in selfed flowers with functionless S-alleles is likely to be greater than that recorded here, particularly when the stigmatic surface is saturated with self-pollen as occurred in this study. The leakiness in the SI system that we observed is more likely to be the result of factors that directly influence the expression of SI (e.g. the transcription or translation of the S-RNase, its turnover or its RNase activity) or factors that indirectly influence the ability of self-pollen tubes to achieve fertilization (e.g. the longevity of the flowers or the nutritional characteristics of the stylar-transmitting tissue). Genetically, these factors could result from differences among S-haplotypes (e.g. weak and strong S-alleles) or unlinked modifiers. Studies are currently underway that explore these possibilities.

Variations in the strength of SI within and between populations are well known and are well documented for a variety of species (see de Nettancourt 1977; Levin 1996; Stephenson et al. 2003 for reviews), but only recently have these variations been considered within the context of dynamic evolving breeding systems (Lloyd & Schoen 1992; Levin 1996; Stephenson et al. 2000). As a weed, it is reasonable to predict that S. carolinense would benefit from plasticity in the strength of SI. Our data show that when cross pollen is scarce (i.e. when old flowers have not been cross pollinated and/or when many flowers have failed to set fruit because of a lack of cross-pollination) the plants become more self-fertile. However, if cross pollen is available the same plants would produce few or no self-seed and thereby avoid the adverse affects of inbreeding. If there is heritable genetic variation for plasticity in the strength of SI (as found in other species) then natural selection might favor the more plastic genotypes, particularly during the colonization process (when few S-alleles are present and effective population sizes are small). However, at this point in time, we do not know whether the plasticity in SI that we observed is actually an important component of the breeding system of S. carolinense or whether it is just neutral genetic/environmental variation. The importance of this plasticity in SI in the breeding system awaits the outcome of further studies that examine inbreeding depression and its variability among genotypes, the genetic basis of the plasticity, and the reproductive behavior of plants in natural populations in which cross pollen limits reproduction.

Acknowledgments

We thank Tony Omeis and his staff at the Buckhout Greenhouse for assistance with the cultivation of the plants and Ashley Kelley, Tia Kinney, Julia Thaller and Nicole Myers for greenhouse and laboratory assistance. This research was supported by NSF grant DEB99-82086 to A. G. Stephenson.

Ancillary