- Top of page
Theory predicts that morph ratios in heterostylous populations are governed by negative frequency-dependent selection typically resulting in equal morph ratios at equilibrium. Previous work on the distylous perennial herb Pulmonaria officinalis, however, showed asymmetric mating between floral morphs and a weak self-incompatibility system, with the long-styled morph (L-morph) producing significantly higher seed set following intramorph crosses and even selfing than the short-styled morph (S-morph), two aspects thought to affect female fecundity and morph-ratio variation. Here, we evaluated morph ratios and population size of all known P. officinalis populations in the northern part of Belgium. Morph ratios deviated significantly from 1 : 1 (range 0.09–1 L-morph frequency, mean = 0.58). Relative fecundity of the S-morph (i.e. mean seed set of the S-morph/mean seed set of the L-morph) was on average 0.73, was positively related to the frequency of the L-morph, and reached 1 (similar levels of female fecundity) at an average L-morph frequency of 0.66 in the population. As some small populations had the S-morph in majority, our results suggest that local morph ratios are influenced both by the relative fecundity of L- and S-morph individuals and by stochastic processes in small populations.
- Top of page
Distylous plant populations have two floral morphs (long- and short-styled morph, hereafter referred to as the L- and S-morph) that differ reciprocally in the position of stigmas and anthers. This genetic polymorphism promotes disassortative pollen transfer by reducing interference between pollen removal and deposition within the same morph (Darwin, 1877; Ganders, 1979; Barrett, 1992). In theory, disassortative mating among the floral morphs should result in frequency-dependent selection and equal frequencies in equilibrium populations (Heuch, 1979; Eckert et al., 1996). However, several studies of distylous species have shown morph ratios that deviate significantly from theoretical expectations (e.g. Ornduff, 1980; Philipp & Schou, 1981; Pailler & Thompson, 1997; Endels et al., 2002; Jacquemyn et al., 2002; Kéry et al., 2003; Wang et al., 2005; Brys et al., 2007).
Deviations in morph ratios can be governed by both stochastic and deterministic forces. Random morph loss has been shown to result in skewed morph ratios in populations of several distylous species that have recently experienced strong reductions in population size (i.e. drift) (e.g. Endels et al., 2002; Jacquemyn et al., 2002; Brys et al., 2003; Kéry et al., 2003). In this case, the average morph ratio among distylous populations should not be different from 0.5 and deviations from equal morph ratios should be symmetric around 1 : 1, because the morphs are eliminated randomly. Additionally, founder effects during colonization have been shown to cause similar patterns of morph-ratio variation in populations of tristylous species (e.g. Eckert & Barrett, 1992; Husband & Barrett, 1992).
However, stochastic forces such as drift and founder events cannot explain deviations from equal morph ratios in all cases and some studies have shown that morph-ratio variation can result from maternal fitness differences among morphs (e.g. Ornduff, 1988), especially when they are pollen limited (e.g. Hodgins & Barrett, 2006; Van Rossum et al., 2006). Although most heterostylous plants are characterized by a heteromorphic incompatibility system, which precludes self- and intramorph cross-fertilization and consequently only allows intermorph cross-fertilizations, some species, however, show weak heteromorphic incompatibility (reviewed in Barrett & Cruzan, 1994) either because they are (partly) self-compatible, or because they possess alternative nonheteromorphic incompatibility systems [e.g. Anchusa (Schou & Philipp, 1984) and Narcissus (Barrett & Harder, 2005)]. This can theoretically give rise to self- and intra-morph mating and ultimately to biased morph ratios within populations (Barrett et al., 1989; Eckert & Barrett, 1992; Hodgins & Barrett, 2006, 2008). For example, in tristylous Eichhornia paniculata, simulations revealed that decreased fertility of the S-morph in combination with deviations from disassortative pollen transfer could account for the loss of the S-morph from trimorphic populations (Barrett et al., 1989). Similarly, Hodgins & Barrett (2006) showed theoretically that in tristylous species such as Narcissus, maternal fitness differences in combination with assortative mating can explain biased morph ratios. Although a weak self-incompatibility system combined with low and/or inappropriate pollinator services is one of the most likely mechanisms to cause deviations in morph ratios (Barrett et al., 1989), empirical evidence is still scarce (but see Hodgins & Barrett, 2006, 2008).
In the genus Pulmonaria (Boraginaceae), several species are known that lack strict heteromorphic self-incompatibility [e.g. P. officinalis (Darwin, 1877; Brys et al., 2008) and P. affinis (Richards & Mitchell, 1990)], and that show asymmetric mating (e.g. Olesen, 1979; Richards & Mitchell, 1990; Brys et al., 2008), two conditions that may lead to biased morph ratios within populations. Experimental hand-pollinations in P. officinalis, for example, showed that the L-morph is capable of producing significantly higher seed set following intramorph crosses and even selfing (fertility reduction of respectively 54.4% and 79.6% compared to intermorph crosses) than the S-morph (fertility reduction of respectively 80.4% and 95.8% compared to intermorph crosses) (see Brys et al., 2008). Investigation of pollen deposition rates on visiting insects and stigmas of P. officinalis under natural conditions further showed that pollen transfer rates between morphs were asymmetric (illegitimate pollen flow to the L-morph is on average 2.2 times higher than illegitimate pollen flow to the S-morph; see Brys et al., 2008). Given these conditions, one should thus expect that natural P. officinalis populations should show a general bias towards the L-morph and that morph bias increases with decreasing population size, particularly if small populations are more pollen limited than larger ones. If, on the other hand, stochastic processes determine morph-ratio variation, morph bias should be related to population size, but there should be no bias to the L-morph.
In this study, we examined variation in morph ratio, population size and relative female fecundity of the L- and S-morph in all known P. officinalis populations in northern Belgium (Flanders). To exclude the possibility of an interrelationship between morph bias and population size influencing relative female reproductive success, and to examine the role of pollen limitation in affecting female reproductive success, we additionally created small experimental populations that varied in morph bias, but were constant in size, and investigated pollen deposition rates and relative fecundity. More specifically, we addressed the following questions:
Is morph-ratio variation in natural populations of P. officinalis related to population size, and if so, is this variation symmetric around isoplethy (1 : 1)?
Is the relative fecundity of both morphs related to the observed variation in morph ratio?
Can population size and relative female reproductive success of the L- and S-morph together explain variation in morph ratio among populations?
- Top of page
We scored the morph of a total of 5838 plants in 35 populations of P. officinalis. Based upon the pooled data from all populations, the frequency of the L-morph was 0.58 and was significantly greater than 0.50 (G = 66.15; d.f. = 1; P < 0.0001). At the population level, the frequency of the L-morph ranged from 0.09 to 1.00, and was significantly higher than 0.5 (t34 = 2.06; P = 0.047) (Figs 1 and 2). Of the 35 populations surveyed, the L-morph was the most common morph in 77% of the populations (n = 27). Only six populations contained more S-individuals than L-individuals and were rather small (Figs 1 and 3). Four populations were monomorphic for the L-morph and ranged in size from 39 to 98 flowering individuals (see Figs 1 and 3). Morph bias (defined as the absolute value of the difference in the number of individuals of the L- and S-morph divided by the total number of flowering individuals) was significantly related to population size (r = −0.591; n = 35; P = 0.002), with small populations deviating more strongly from an equal morph ratio than large populations (Fig. 3).
Figure 2. Frequency distribution of L-morph in 35 Pulmonaria officinalis populations. The classes represent proportions of 0, 0.001–0.2, 0.21–0.4, … and 1. Note that the outermost bar (proportion = 1) represents exclusively monomorphic L-morph populations.
Download figure to PowerPoint
Figure 3. Relationship between population size (number of flowering rosettes) and morph bias (|[S-morph individuals − L-morph individuals]/population size|), in 35 natural Pulmonaria officinalis populations. The symbol • represent populations in which the L-morph is in majority whereas the symbol ○ represent populations in which the S-morph is in majority.
Download figure to PowerPoint
Average seed production per flower increased significantly with increasing population size, and was significantly larger in the L-morph than in the S-morph (on average 2.38 ± 0.06 and 1.44 ± 0.07 respectively) (Table 1a; Fig. 4a,b). Mean seed set per flower in the S-morph significantly declined with decreasing population size, whereas seed set per flower did not depend significantly on population size in the L-morph (Table 1b; Fig. 4b). The relative fecundity of the S-morph was positively related to the frequency of L-individuals in the population (r = 0.827; n = 22; P < 0.001; Fig. 5) and equals 1 (equal fecundity rates in both morphs) when the frequency of L-individuals is 0.66.
Table 1. (a) Analysis of covariance (ancova) investigating the effects of style morph and population size on seed set in 25 populations of Pulmonaria officinalis. The homogeneity of regressions assumption is rejected when the interaction term is significant. (b) Parameter estimates (B), significances and 95% confidence intervals for slopes of simple linear regressions between log10 population size and seed set for both the S- and L-morphs.
| Morph||1||2.920||15.91||< 0.001|
| Population size||1||5.585||30.42||< 0.001|
| Morph × population size||1||1.467||7.99||0.006|
| Error||43||0.184|| || |
| ||B||t||Sign.||95% Confidence interval|
|Lower bound||Upper bound|
| S-morph||0.69||5.80||< 0.001||0.45||0.93|
| L-morph||0.22||1.93||< 0.060||−0.10||0.45|
Figure 4. Relationship between population size (number of flowering rosettes) and (a) mean seed set per flower averaged per population (± SE), (b) morph-specific seed set per flower (± SE) averaged per population and (c) relative fecundity of the S-morph in 25 natural Pulmonaria officinalis populations.
Download figure to PowerPoint
Figure 5. Relationship between the relative fecundity of the S-morph and the frequency of the L-morph in 25 natural Pulmonaria officinalis populations.
Download figure to PowerPoint
Although the total number of deposited pollen grains on P. officinalis stigmas did not differ significantly between the morph-ratio treatments in the experimental populations, stigmas of the L-morph received significantly more pollen grains than stigmas of the S-morph (Table 2). The proportion of pollen grains of each morph significantly increased with the number of individuals of that particular morph in the population (Fig. 6a). Overall, stigmas of the L- and S-morph received more L-morph pollen than S-morph pollen, except for the stigmas of L-individuals in S-morph biased populations. Although a significantly larger proportion of L-morph pollen was generally deposited on stigmas of the S-morph than on stigmas of the L-morph (Table 2), the proportion of illegitimate L-morph pollen deposited on stigmas of L-individuals was remarkably high (see Fig. 6a). In each morph, seed set was most successful when the opposite morph occurred in majority, whereas seed production rates were lowest when the opposite morph was in minority (Fig. 6b). Despite the fact that seed production patterns significantly differed between both morphs (Table 2), they generally parallel the proportion of legitimate deposited pollen, especially in the S-morph (Fig. 6a-b). Finally, the relative fecundity of the S-morph was 0.39 in the S-morph-biased populations, reached 0.71 at isoplethy and increased to 1.11 in the L-morph-biased populations (Fig. 6c).
Table 2. Generalized linear mixed model analyses of the impact of treatment ([S-morph < L-morph] = four S-morph: 16 L-morph; [S-morph = L-morph] = 10 S-morph: 10 L-morph; [S-morph >L-morph] = 16 S-morph: four L-morph) and morph (L-morph vs. S-morph) on total stigmatic pollen loads, the percentage S-morph pollen deposited and mean seed set per flower in similar sized (n = 20) experimental Pulmonaria officinalis populations.
|Total stigmatic pollen load (n = 120)|
| Treatment||4.958||10.024||2, 120||0.77|
| Treatment × morph||7.011||−15.774||2, 120||3.15|
| Morph||5.500||−4.647||1, 120||16.26***|
|Percentage SS-pollen of the total stigmatic pollen load (n = 120)|
| Treatment||2.585||7.382||2, 120||42.83***|
| Treatment × morph||3.656||−5.166||2, 120||2.10|
| Morph||2.868||1.649||1, 120||3.19*|
|Mean seed set per flower (n = 158)|
| Treatment||0.196||−0.713||2, 152||1.32|
| Treatment × morph||0.244||1.953||2, 152||32.91***|
| Morph||0.158||−1.753||1, 152||69.07***|
Figure 6. Differences in (a) percentage of S- and L-morph pollen grains (respectively indicated by the symbols and ○) deposited on stigmas of S- and L-morph individuals (± SE), (b) morph-specific seed set flower (± SE) and (c) relative fecundity of the S-morph (± SE) (indicated by the symbol ♦) in similar sized (n = 20) experimental Pulmonaria officinalis populations characterized by a different morph ratio. The three morph ratios applied were: 16 S-individuals: four L-individuals when [S-morph > L-morph]; 10 S-individuals: 10 L-individuals when [S-morph = L-morph]; and four S-individuals: 16 L-individuals when [S-morph < L-morph].
Download figure to PowerPoint
- Top of page
This study has demonstrated that morph ratios of P. officinalis populations varied considerably within the study area. Because the species shows a weak heteromorphic self-incompatibility system, variation in female reproductive success among morphs may consequently play a important role in governing morph-ratio evolution, in particular under pollen-limited conditions (Charlesworth, 1979; Washitani et al., 1994; McCauley & Taylor, 1997; Hodgins & Barrett, 2006, 2008). Data from both the natural and experimental populations indeed clearly indicated morph-specific differences in female reproductive success between the L- and S-morph and suggest that the conditions that could cause morph-ratio variation are met in P. officinalis. However, because we also found some small populations showing an S-biased morph ratio, stochastic processes associated with small population sizes cannot be ruled out completely as a potential moderator of morph bias variation in P. officinalis.
Female fecundity of the L-morph was on average 1.36 times greater than that of the S-morph. However, the difference in reproductive output between both morphs declined with increasing population size, suggesting that a certain level of pollen limitation has to be reached before differences in female reproductive success may affect morph-ratio variation in P. officinalis. Indeed, seed set significantly declined with decreasing population size, suggesting increased pollen limitation in small populations. Further evidence for morph-dependent pollen deposition and for pollen limitation affecting female reproductive success was obtained from the pollination experiment, where we created similar sized but small populations that strongly varied in morph ratio. First, the study of stigmatic pollen loads obtained in these experimental populations revealed two important aspects about female fecundity in P. officinalis: (1) pollen limitation is prevalent in small populations and (2) pollen deposition rates varied substantially among morphs. Second, female fecundity of the L-morph always exceeded that of the S-morph, except in the case when SS-individuals were in majority and female fecundity did not differ between both morphs.
Pollen limitation can result from both low pollen quality and low pollen quantity (Wilcock & Neiland, 2002; Ashman et al., 2004; Aizen & Harder, 2007). Comparison of the stigmatic pollen loads obtained from two large (n > 5000) and balanced P. officinalis populations (see Brys et al., 2008) with those obtained from the experimental populations (n = 20) indicated that total stigmatic pollen loads per stigma were on average three times lower in the small experimental populations, indicating reduced pollen quantities and thus pollen limitation in small populations. Seed set in these large populations was also on average twice as large as observed in the experimental populations, indicating that low pollen deposition translates into reduced seed set and that pollen limitation thus significantly affects female reproductive success in P. officinalis. Although L-morph stigmas received significantly larger pollen loads than S-morph stigmas, the study of stigmatic pollen loads in the experimental populations further revealed asymmetric pollen flow patterns. Moreover, L-morph pollen were generally deposited in significantly higher proportions on both S- and L-morph stigmas (except on stigmas of S-plants in the S-morph-biased populations) and that L-morph stigmas received significantly higher rates of illegitimate pollination than stigmas of the S-morph. These findings strongly correspond with earlier observations in two large and natural P. officinalis populations in the same area (Brys et al., 2008), and can be attributed to the fact that L-morph individuals produce much more pollen than S-morph individuals and that the principal pollinators of P. officinalis in the study area are mainly short-tongued Bombus species (Brys et al., 2008). The limited proboscis length of the latter together with the occurrence of floral hairs at the corolla entrance of L-morph flowers further explains the large amount of illegitimate pollination in the L-morph (Brys et al., 2008). These observations thus suggest that the impact of asymmetric mating and weak self-incompatibility on reproductive differences between both morphs is magnified by an increase in pollen-limitation, especially in the S-morph.
As in most distylous species the L-morph represents the recessive (ss) and the S-morph the heterozygous (Ss) genotype, such as in P. officinalis (Vuilleumier, 1967), weak incompatibility in the L-morph may theoretically result in L-morph-biased populations. Given that the dominant S-allele, governing the expression of the S-morph, is only carried by S-individuals, and that the higher intramorph-compatibility in the L-morph may increase homozygous L-individuals in the populations, both increase the chance that populations become L-morph monomorphic populations, especially when they are small (Arroyo, 2002). Again this corresponds with our findings that all monomorphic populations in the study area are indeed composed of the L-morph. On the other hand, our data also showed that once a certain L-morph bias is reached or exceeded, female reproductive success of the S-morph equals or exceeds that of its L-morph congeners. Based on Fig. 5, the frequency of L-morph individuals where female reproductive success of both morphs is the same is 65.6%. This value approximates the average frequency observed over all populations (57.7%) and therefore it might be considered as the equilibrium morph ratio. It may also explain why the S-morph is not entirely lost from most of the surveyed populations, as negative frequency-dependent mating causes a reproductive advantage to the S-morph once this equilibrium morph ratio is exceeded.
More evidence that interactions between mating patterns and female reproductive success affect morph-ratio variation in P. officinalis was found when we enter the observed relative fecundity rates of S-morph individuals obtained from the experimental populations into the equation of the regression line representing the relationship between relative fecundity of S-morph individuals and the frequency of the L-morph in the natural populations (see Fig. 5). The latter results in frequencies of the L-morph of 28.9%, 48.3% and 72.6% respectively, values that are remarkably equivalent to the actual L-morph frequencies applied in the experimental populations (25%, 50% and 75% respectively). Again this suggests a clear association between morph-ratio variation on the one hand and female reproductive success on the other.
These findings cannot explain why in some small populations S-morph individuals occurred in a majority. Probably, stochastic processes may have resulted in the observed morph-ratio variation in these populations. Several other studies have already shown that stochastic processes influenced morph-ratio variation in small populations of distylous species (e.g. Endels et al., 2002; Jacquemyn et al., 2002; Kéry et al., 2003; Brys et al., 2007). In the case of the studied P. officinalis populations, recent population bottlenecks due to inappropriate forest management may have caused the observed deviations in morph ratios. Although we have no historical data of the surveyed populations, field observations suggest that a lack of appropriate woodland management (in particular a lack of coppicing) can lead to a rapid decline of the species. Together with the high degree of fragmentation of the studied forest patches this may have resulted in a substantial reduction in population size of the remaining populations during the last decades.
In conclusion, using both natural and experimental populations, we have clearly shown that the female fertility disadvantage of the S-morph, brought about by distorted mating patterns and weak self-incompatibility, can play an important role in governing morph-ratio variation in distylous species, particularly in small populations. The observed bias in morph ratios in P. officinalis thus not only arises from stochastic events, but also and more importantly from variation in female reproductive success between both morphs. These findings largely confirm earlier suggestions of Barrett et al. (1989) that relaxation of the self-incompatibility system under low and/or inappropriate pollinator services is one of the most likely mechanisms causing deviations in morph ratios and may lead in the most extreme scenario to the establishment of monomorphic populations. It might, however, be clear that more research that intends to study morph-ratio variation and pollination and fertility patterns over a larger geographical range of the species distribution area is needed to elucidate whether the observed patterns either represent a general pattern or a local evolutionary response mediated by a change in the pollinator assemblage at the border of the species distribution area.