| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Reproductive Biology |
Department of Biology, Queen's University, Kingston, Ontario K7L 3N6 Canada
Received for publication June 1, 2004. Accepted for publication January 14, 2005.
ABSTRACT
Biogeographic models predict that geographically peripheral populations should be smaller, more sparsely distributed, and have a lower per-capita reproductive rate than populations near the center of a species' range. Plants in peripheral populations may, therefore, receive less pollinator visitation and outcross pollination, which may select for self-fertilization to provide reproductive assurance. We tested these predictions by comparing population size, plant density, seed production, floral traits, and mating system parameters between 10 populations of Aquilegia canadensis near the northern margin of the range with 10 near the range center. Contrary to predictions, peripheral populations were not smaller, less dense, nor less productive than central populations. Nevertheless, we detected substantial regional differences in key floral traits. Plants in central populations produced larger flowers with 68% greater herkogamy and had 30% more flowers open simultaneously than plants in northern populations. However, there was no regional difference in the mating system. In northern populations, 73% (range = 60–88%) of seeds were self-fertilized compared to 76% (51–100%) in central populations. In both regions, adult inbreeding coefficients were near zero, indicating very strong inbreeding depression despite high selfing. Marked geographic variation in key floral traits does not reflect evolutionary differentiation in the mating system.
Key Words: Aquilegia canadensis • central and peripheral populations • geographic variation • herkogamy • mating system • outcrossing • population size • self-fertilization
In plants, mating patterns may be strongly influenced by the size and physical structure of populations. Regional variation in key population characteristics may, in turn, set the stage for evolutionary differentiation in reproductive and life history traits across the geographic range (Levin, 2000
). In general, species distributions are thought to conform to the "abundant center" model, where population frequency, size, and density are highest at the geographical center of the species' range and decline towards the periphery (Hengeveld and Haeck, 1982
; Brown, 1984
; Lawton, 1993
; Sagarin and Gaines, 2002
). Very few studies have quantified the consequences of geographic variation in population size, density, and isolation for pollination ecology and mating. However, the general effects of variables such as population size, density, and isolation have been investigated outside of a geographical context. For instance, small and/or isolated populations may attract fewer and less diverse pollinators (Sih and Baltus, 1987
; Jennersten, 1988
; Aizen and Feinsinger, 1994
; Groom, 1998
). This, along with the low density of conspecific pollen sources, may reduce outcross-pollination (Byers, 1995
; Ågren, 1996
) and, in self-compatible populations, increase self-fertilization (Barrett and Husband, 1990
; Murawski, 1991
; van Treuren et al., 1993
; Karron et al., 1995
; Routley et al., 1999
). Moreover, pollinators foraging in small, sparse populations may visit more flowers in sequence on individual plants, thereby increasing the likelihood of geitonogamous self-pollination (Pyke, 1979
; Franceschinelli and Bawa, 2000
).
Reduced opportunities for outcrossing in peripheral populations may select for alleles that increase autonomous autogamous self-fertilization to provide reproductive assurance (Cruden and Lyon, 1989
; Lloyd, 1992
). This prediction is supported by many examples of plant groups in which selfing taxa or populations occur in areas that are geographically or ecologically marginal compared to related outcrossers (Baker, 1955
; Stebbins, 1957
; Jain, 1976
; Solbrig, 1976
; Schoen, 1982
; Wyatt, 1988
; Runions and Geber, 2000
). Selection for autonomous autogamy in peripheral populations will be manifested by geographical variation in aspects of floral morphology, development, and physiology that influence the level of self-pollination and/or self-fertilization, and shifts in these key floral traits observed towards range margins are often attributed to selection for reproductive assurance (Barrett et al., 1989
; Klips and Snow, 1997
; Fausto et al., 2001
; Elle and Carney, 2003
). Increased autogamous selfing may evolve via a breakdown of self-incompatibility (Barrett, 1988
) or reductions in traits that separate male and female function within flowers, such as herkogamy and dichogamy (Schoen, 1982
; Wyatt, 1986
). As autogamy evolves, there is likely to be concomitant selection for reduced pollinator attraction and reward (Ornduff, 1969
). However, very few studies have tested for geographic covariation between key floral traits and population characteristics such as size, density, and isolation, and none have determined whether differences in population characteristics and floral traits between geographically central and peripheral populations are associated with the expected differences in mating system parameters.
In this study, we compare population demographic variables, floral characteristics, and mating system parameters between geographically central and peripheral populations of columbine, Aquilegia canadensis L. (Ranunculaceae). This species is a short-lived, spring-flowering perennial found on rocky outcrops and in dry woods throughout eastern North America, from southern Ontario and Quebec to northern Florida and Texas. Large, showy red flowers with long nectar spurs (
30 mm) suggest adaptation outcrossing via hummingbird pollination. However, flowers are self-compatible and will set almost a full complement of seed via autonomous self-pollination when excluded from pollinators (Macior, 1978
; Eckert and Schaefer, 1998
; Routley et al., 1999
). Populations at the northern edge of the geographic range in eastern Ontario are small and patchy, and flowers are only infrequently visited by hummingbirds and bumble bees (C. R. Herlihy and C. G. Eckert, unpublished data, see also Macior, 1966
). Plants in these populations are highly self-fertilizing (average = 73%, range = 17–100%, N = 28 populations, Routley et al., 1999
; Griffin et al., 2000
; Herlihy and Eckert, 2002
; Griffin and Eckert, 2003
).
Here we combine field studies with marker-gene analysis to test three main predictions. (1) According to the abundant center model, northern, geographically peripheral populations of A. canadensis should be smaller, less dense, more isolated, and produce fewer seeds per capita than central populations. (2) As a result of selection for reproductive assurance in small, sparse, isolated populations, plants in northern populations will produce smaller flowers with reduced herkogamy. (3) Northern peripheral populations will exhibit higher levels of self-fertilization and higher parental inbreeding coefficients than central populations.
MATERIALS AND METHODS
Population characteristics
We compared 10 populations near the geographical center of the species' range, in the southern Appalachian region, with 10 populations near the northern margin of the species' range, in eastern Ontario (Appendix 1, see Supplemental Data accompanying the online version of this article). Populations were defined as spatially discrete clusters of plants, separated from other clusters by 
500 m, usually much more. We measured population size as the total number of reproductive individuals. Because A. canadensis exhibits little or no clonal spread, a plant was defined as the basal leaves and inflorescences emerging from a rosette. Plant density was measured at 20 locations within each central population, and 27 locations within each northern population as the number of reproductive individuals within a 1- and 5-m radius of randomly chosen focal plants. Because population means for the two density measures correlated strongly (r = +0.98), and a 5-m radial density also correlated strongly with nearest neighbor distance in a previous study (r = +0.80, N = 9 populations; Herlihy and Eckert, 2004
), we only use 5-m density in the analyses presented below. Data on size and density were not collected for two of the central populations (Appendix 1; see Supplemental Data accompanying the online version of this article).
We previously showed that in northern populations, population size and plant density correlate with other population-level characteristics such as canopy cover, which may also affect the mating system (Herlihy and Eckert, 2004
). Large, dense populations occur on exposed rocky outcrops, whereas smaller sparse populations are typically found in wooded habitat. To confirm that the relation between canopy cover and population size and density is consistent between regions, we measured canopy cover above 20 randomly selected focal plants in both northern and central populations. Mean canopy cover did not differ between regions (mean percent cover ±1 SE: northern = 58.1 ± 7.1%; southern = 67.6 ± 8.0%; t test: t = 0.89, df = 16, P = 0.39). In both regions, canopy cover correlated negatively with population size (northern: r = –0.90, P = 0.0003; central: r = –0.89, P = 0.003) and plant density (northern: r = –0.65, P = 0.04; central: r = –0.90, P = 0.002). These correlations did not differ between regions (Fisher's Z: size, P = 0.9; density, P = 0.2). Accordingly, we report only results from analyses of population size and density, as these characteristics feature explicitly in the abundant center model.
Seed production and floral traits
To quantify geographical variation in seed production, we regularly monitored 60 randomly chosen plants in each population, (total N = 1072) and recorded the number of inflorescences produced by each and total number of flowers produced on one randomly chosen inflorescence per plant. We then estimated the number of flowers that could have potentially donated geitonogamous pollen to a randomly chosen focal flower on that inflorescence. Hereafter, the number of potential geitonogamous pollen donors plus one (i.e., the focal flower plus the number of flowers open simultaneously with it) is referred to as display size. For the focal flower on 45 randomly chosen plants per population (total N = 795), we measured (to 0.1 mm) the length of one nectar spur from the nectar gland to the tip of the lamella and the minimum distance between anthers and stigmas (herkogamy). We use spur length as an index of overall flower size, because it correlates strongly and positively with the size of other floral organs (Herlihy and Eckert, 2004
). Our measurements of spur length and herkogamy were highly repeatable (r = +0.99 for both traits).
In Ontario populations of A. canadensis, there is no evidence of dichogamy, as stigma receptivity begins concurrently with anther dehiscence (Griffin et al., 2000
). In the central populations studied here, stigmatic papillae begin to swell, and peroxide tests indicate stigma receptivity (Kearns and Inouye, 1993
) at the same time as anthers begin to shed pollen (C. R. Herlihy, personal observation). Hence there is no evidence for dichogamy in either central or northern populations.
Each population was visited multiple times throughout the flowering and fruiting period, and floral measurements were taken on at least two different days to account for any variation in flower size across the flowering period. The northern populations used in this study were part of a larger study (Herlihy and Eckert, 2002
, 2004
) and were visited every 3 d throughout flowering and fruiting in 1999, while central populations were visited once weekly throughout flowering and fruiting in 2000. For each of the measured flowers, we collected the fruit if it successfully matured (32 fruits per population, total N = 575), and counted the number of filled seeds and undeveloped ovules within (following Mavraganis and Eckert, 2001
). From these data, we calculated seeds per flower, ovules per flower, seeds per ovule (seed set), and per-capita seed production (= number of inflorescences per plant x flowers per inflorescence x seeds per flower).
Pollinator visitation in both regions was too infrequent to allow reliable quantification of visitation rate and pollinator species composition. However, all floral visitors observed in central populations were observed in our more extensive pollinator observations in northern populations, and the most common visitors in northern populations (ruby-throated hummingbirds, Archilochus colubris, and the bumble bee, Bombus bimaculatus) were also the most commonly observed visitors in central populations (C. R. Herlihy and C. G. Eckert, unpublished data). Hence, there was no indication of any substantial difference in pollinator species composition or visitation rate between northern and central populations.
Statistical analysis of population characteristics and floral traits
We used nested analysis of variance (ANOVA) to compare plant density, flower number, seeds per fruit, seeds per ovule, per-capita seed production, display size, spur length, and herkogamy between central and northern populations. Region (central vs. northern) was a fixed effect, and population nested within region was a random effect. Because there were unequal numbers of populations sampled in each region (eight central vs. 10 northern) and unequal numbers of plants sampled in each population, F tests involved synthetic denominators calculated using the Satterthwaite method (Sokal and Rohlf, 1995
, pp. 292–300). Population size was compared between regions using a one-way ANOVA. To normalize the data and eliminate associations between residuals and predicted values, population size, plant density, flowers per plant, display size, and seeds produced per plant were log10-transformed, and seeds per ovule was arcsine-transformed. All analyses were performed using JMP (version 5, SAS Institute, 2002
). Means are presented ±1 SE unless otherwise indicated.
Allozyme screen and mating system estimation
We estimated mating system parameters for 10 central populations during the 2000 reproductive season and compared them with previously published estimates for 10 northern populations during the 1999 reproductive season (Herlihy and Eckert, 2002
, 2004
). Previous analyses using northern populations have not revealed any overall year-to-year variation in mating system parameters (C. G. Eckert, B. J. Ozimec, and C. R. Herlihy, unpublished data), hence a comparison of large groups of populations sampled in different years seems justified. First, we screened bulk seed collected from the central populations for allozyme variation at 18 putative loci following a similar screen performed for the northern populations (Routley et al., 1999
). Zymogram phenotypes that were variable and interpretable were found for only two loci, isocitrate dehydrogenase (IDH, EC 1.1.1.42), and peroxidase (PER, EC 1.11.1.7), the same allozyme loci that exhibit variability in northern populations. Electrophoretic procedures followed Routley et al. (1999)
. Allele frequencies (P) in all populations were within the range useful for estimating mating system parameters (i.e., 0.05 < P < 0.95; Appendix 2, see Supplemental Data accompanying the online version of this article).
We collected mature fruits resulting from all measured flowers in all northern and central populations (see above) plus one fruit from each of 50 plants in two central populations (GACWTH and WVPOH1) for which floral and population measurements were not taken, so that mating system estimates could be compared for 10 populations in each region. We screened 10 seeds from each of 36 fruits for each northern population, and approximately six seeds from each of 35 fruits for each central population (Appendix 2, see Supplemental Data accompanying the online version of this article). The parental inbreeding coefficient (F) and the proportion of seeds self-fertilized (the female selfing rate, s) were estimated using the maximum-likelihood program MLTR (Ritland, 1990a
) with Newton-Raphson iteration. For all analyses, iterations starting from any set of initial parameter values always converged on only one maximum-likelihood estimate. Standard errors for each estimate were calculated as the SD of 1000 replicate bootstrap estimates, with the progeny array as the unit of resampling. Because we assayed seeds directly and crossing experiments have revealed little if any early-acting inbreeding depression (Routley et al., 1999
), our estimates represent primary levels of selfing, at fertilization.
Inbreeding depression (
= 1 – the fitness of selfed progeny/fitness of outcrossed progeny) was calculated from the estimates of selfing (s) and the parental inbreeding coefficient (F) using Ritland's (1990b)
equilibrium estimator:
 | (1) |
values for populations in which these were reliably estimated. Accordingly, we do not report estimates of F and in two northern populations and estimates of in two central populations that were associated with very large standard errors and highly irregular bootstrap distributions. In addition, central population NCSMR1 was entirely selfing, so that could not be estimated (Ritland, 1990b
). The assumptions and limitations of this method for estimating , as applied to populations of A. canadensis, are discussed by Routley et al. (1999)
. Although some assumptions, such as inbreeding equilibrium, may be violated by some populations of A. canadensis in some years, this does not strongly bias estimates of and is unlikely to confound our regional comparison.
Differences in s, F, and between central and northern populations were assessed by averaging randomly paired bootstrap estimates across populations within regions and calculating the proportional overlap between averaged bootstrap distributions, which is roughly equivalent to a P value. Mean estimates from central and northern populations were considered significantly different if this overlap (P) was <2.5% for a two-tailed test or <5% for a one-tailed test. Northern populations were predicted to have higher self-fertilization and parental inbreeding coefficients, so comparisons of s and F between regions were one-tailed. Comparisons of between regions were two-tailed because there is no clear theoretical prediction regarding the strength of inbreeding depression in peripheral populations. On the one hand, purging of genetic load in self-fertilizing populations may reduce over the long term (Husband and Schemske, 1996
). However, the more challenging biotic and abiotic environment predicted to occur at range margins might result in stronger expression of inbreeding depression (Roff, 1997
).
RESULTS
Population frequency, size, density, and seed productivity
Contrary to predictions of the abundant center model, northern populations were not less frequent, smaller, or sparser than central populations. Although we did not estimate the frequency of populations in either region, our extensive field observations clearly indicate that populations are much more common in eastern Ontario than in the geographic center of the range. We qualitatively estimate that populations are 10 times more common in the northern than in the central study area. There was no significant difference between regions in either population size or plant density. Most of the variation in population density occurred among populations within regions (Table 1). Population size and plant density correlated positively among populations in both regions (central: r = +0.75, P = 0.03, northern: r = +0.74, P = 0.01), and correlation coefficients did not differ between regions (Fisher's z: P = 0.97).
|
2 = 108.7, df = 1, P < 0.0001). There was also a tendency for plants in northern populations to produce fewer flowers per inflorescence, however the difference between regions was not significant (Table 1). Contrary to predictions, plants in northern populations produced 27% more seeds per flower as a result of flowers containing 9% more ovules (though the difference between regions is not quite significant) and individual ovules having a 17% higher probability of becoming a seed (Table 1). Higher per-flower seed production in northern populations compensated for lower inflorescence and flower production, such that there was no overall regional difference in estimated number of seeds per plant. For all these components of seed fecundity, most of the variation occurred among plants within populations.
Floral morphology and display
As predicted, plants in northern populations produced smaller floral displays than plants in central populations (Table 1). Although relatively little of the overall variation in display size variation was distributed between geographic regions, display size was 30% higher in central than peripheral populations (Fig. 1). Display size correlated strongly with the total number of flowers per inflorescence among populations in both regions (northern: r = +0.91; central: r = +0.87; both P < 0.005), and this correlation did not differ between regions (P = 0.37).
|
2 = 42.3, df = 1, P < 0.0001). Nectar spur length exhibited less overall variation (CV = 10%) and somewhat less of this variation (16%) was distributed among regions. However, it exhibited a similar regional difference. On average, nectar spur length was 6% greater in central populations and, again, only VAHUF1 fell within the range of means for northern populations. There was little covariation between floral traits among populations within regions. Herkogamy did not correlate with display size in either region (both |r| < 0.15, both P > 0.67). Spur length and herkogamy correlated positively among central (r = +0.74, P = 0.03) but not northern populations (r = –0.10, P = 0.79). However the correlation among central populations was due almost entirely to VAHUF1 (see Fig. 1) and became much weaker and far from significant when this population was removed from the analysis (r = +0.15, P = 0.74). Spur length did not correlate with display size among central populations (r = –0.08, P = 0.85). Among northern populations, spur length correlated negatively, though not significantly, with display size (r = –0.55, P = 0.09), but this correlation was due to population QOR1 (see Fig. 1) and became weakly positive and far from significant when QOR1 was removed from the analysis (both r = +0.15, P = 0.69).
Mating system parameters
Estimates of the proportion of seeds self-fertilized were high for all populations in both regions (Fig. 2) and, contrary to predictions, were not higher for northern than central populations (northern mean s = 0.73 ± 0.03; central mean s = 0.78 ± 0.02; P = 0.96). In populations from both regions, estimated parental inbreeding coefficients (F) were not greater than zero (Fig. 3), with the exception of one central population (NCSMR1) and one northern population (QOR1). Thus estimates of F were not higher for northern than central populations (northern mean F = 0.009 ± 0.051; central mean F = 0.170 ± 0.058; P = 0.95). The combination of high estimates of s with low estimates of F for both regions indicates that although a large proportion of seed was derived from self-fertilization, plants surviving to reproduce were predominantly outbred. Hence, estimates of inbreeding depression () were generally very high and did not differ significantly between regions (Fig. 3; northern mean = 0.983 ± 0.036; central mean = 0.931 ± 0.058; P = 0.43). Estimated was significantly less than 1.0 in only one population.
|
|
Departures from the abundant center model
The abundant center model predicts that population frequency, size, density, and per-capita reproduction are greatest at the center of a species' range and decline towards the periphery (Hengeveld and Haeck, 1982
; Brown, 1984
; Lawton, 1993
). In contrast, our field observations clearly indicate that populations of A. canadensis are much more common in Ontario, near the northern periphery of the species' range, than in the geographic center of the range. Moreover, central populations were not larger or denser nor did they exhibit a higher per-capita seed production than populations near the northern range limit (Table 1).
Although the abundant center model is a widely accepted starting point for thinking about the ecology and evolution of species' geographical distributions, our results join a substantial amount of conflicting empirical evidence. A review of patterns of abundance in species from a wide variety of taxonomic groups found that only 39% of 145 tests directly supported the abundant center hypothesis (Sagarin and Gaines, 2002
). Although plants, by virtue of their sessile nature, would seem to be good subjects for evaluating models of geographic distribution, there has been relatively little effort to determine whether plant species generally conform to the abundant center model (Jump and Woodward, 2003
).
Our simple comparison of geographically central vs. peripheral populations does not provide a strong test of the abundant center model, nor does it reveal why A. canadensis is more common in eastern Ontario than in the geographic center of the range in the Virginias and the Carolinas. The abundant center model derives from the expectation that, on a relatively continuous ecological gradient, a species is most abundant across the range of conditions where survival, reproductive success, and recruitment are highest and becomes less abundant as conditions become more extreme (Hengeveld and Haeck, 1982
). First, this assumes that a species has finished spreading geographically and is at an ecological equilibrium. This would seem likely in a short-lived and widespread species like A. canadensis. Second, the model assumes a relatively smooth ecological gradient. More complex, discontinuous gradients might produce patterns of distribution and abundance that depart markedly from a gradual decline from the center to the edge of the range (Lawton, 1993
). For instance, our extensive field observations suggest that populations of A. canadensis at the geographical center of the range are largely restricted to limestone outcrops in river valleys, whereas populations in Ontario occur in a wide variety of habitats on a variety of parent materials, including the limestone of the Great Lakes basin and granite outcrops of the Canadian Shield. This potentially suggests that the range of edaphic conditions under which the species can persist is actually broader towards the range limit than at the center of the range. This contrasts with general expectations that species should be found in a greater variety of habitat types in central than peripheral areas (Hall et al., 1992
). The narrower edaphic range of the central populations might, for example, result from higher levels of interspecific competition in the more species-rich central region than in the northern region. More extensive geographic sampling of plant abundance and diversity with quantitative data on habitat variables is required to verify and investigate the unexpected relation between edaphic amplitude and abundance observed in A. canadensis.
Geographical differentiation in floral traits but not the mating system
Despite these departures from the abundant center model, the geographic variation in floral morphology, display, and per-ovule seed production we observed is consistent with the hypothesis that there has been greater selection for self-fertilization in peripheral than central populations of A. canadensis. As predicted, plants in peripheral populations produced smaller displays of smaller flowers with greatly reduced herkogamy. This pattern of floral variation has been observed in many groups and is often interpreted as evidence that self-fertilization has evolved to provide reproductive assurance in small, sparse, peripheral populations (Jain, 1976
). Ontario plants also exhibited higher seed set than those in central populations, which is also consistent with a higher level of reproductive assurance in peripheral populations (Larson and Barrett, 2000
). However, neither the estimated level of self-fertilization nor the inbreeding coefficient of reproductively mature plants differed between central and peripheral populations of A. canadensis (Figs. 2 and 3). This result raises two questions: (1) Why were the marked regional differences in floral traits not reflected in differences in the mating system? (2) If the regional differences in floral traits do not affect the mating system, then why have they evolved? Addressing this last question involves determining the extent to which the floral differences observed between northern and central populations have a genetic basis.
It is generally expected that the balance between self-fertilization and outcrossing should be influenced by key floral traits such as display size, flower size, and, especially, herkogamy (Webb and Lloyd, 1986
; Barrett and Eckert, 1990
; Harder and Barrett, 1996
). Yet, we did not detect any regional difference in mating system parameters despite significant differences in floral morphology. One of the challenges in interpreting this result is that central and peripheral populations differed in all three floral traits, so we cannot isolate the effect of each floral trait. Recognizing this, we briefly consider each of the three traits in turn. Display size should have opposing effects on the mating system, by increasing pollinator visitation and outcrossing but also the potential for geitonogamous selfing (Harder and Barrett, 1995
). Display size and self-fertilization correlate positively among plants within populations of several species (Barrett et al., 1994
; Harder and Barrett, 1995
; Vrieling et al., 1999
) or geitonogamy specifically (Eckert, 2000
). However, few studies have examined how selfing and display size covary among populations. Despite substantial regional variation in display size among populations of A. canadensis, most plants have limited potential for geitonogamy because they bear only a single open flower at any given time (62% of plants in northern populations, 44% in central populations). Although substantial geitonogamy has been detected in species with small floral displays (Leclerc-Potvin and Ritland, 1994
), a study of six northern populations of A. canadensis found little evidence of geitonogamy, even in plants with multiple flowers open simultaneously (Herlihy and Eckert, 2004
).
Larger flowers might experience greater pollinator visitation and hence outcrossing. Although comparisons among species or populations that differ strongly in flower size (and probably several other important floral traits) support this prediction (Schoen, 1982
; Holtsford and Ellstrand, 1992
), the effect of more subtle variation in flower size has rarely been investigated. Multivariate flower size did not covary with selfing among plants within a population of Aquilegia caerulea (Brunet and Eckert, 1998
) or among 10 northern populations of A. canadensis (Herlihy and Eckert, 2004
). Likewise, we found no association between regional variation in flower size, as indexed by spur length, and self-fertilization. Perhaps this is because this and other components of flower size did not vary much and relatively little of the existing variation was partitioned among populations or between geographic regions. Direct observation of how display size and flower size influence pollinator visitation and foraging behavior is required to further explore why outcrossing does not seem to covary with these traits. This will be difficult given the infrequent visitation to flowers of A. canadensis.
The most striking difference in floral morphology we observed between central and peripheral populations involved herkogamy, the spatial separation between receptive stigmas and dehiscing anthers within flowers. Herkogamy is widely viewed as a key floral trait that reduces interference between male and female sexual function within flowers, especially self-pollination (Webb and Lloyd, 1986
; Barrett and Eckert, 1990
). Within both northern and central populations of A. canadensis, flowers with more pronounced herkogamy experience less self-fertilization (C. R. Herlihy and C. G. Eckert, unpublished manuscript). The same effect has been demonstrated in A. caerulea (Brunet and Eckert, 1998
) and several other species (Kohn and Barrett, 1992
; Karron et al., 1997
; Motten and Stone, 2000
). Yet, the substantial difference in herkogamy between peripheral and central populations of A. canadensis (Table 1) was not associated with a corresponding difference in self-fertilization. Perhaps the effect of variation in herkogamy at broader scales is obscured by other factors. This is likely because herkogamy varied much more among plants within populations than among populations within regions. Accordingly, herkogamy did not correlate negatively with self-fertilization among either the northern populations (r = –0.13, P = 0.70) or central populations (r = +0.52, P = 0.20) we studied. In a previous study, we found that, at the population level, herkogamy only correlates negatively with the autogamous component of total selfing, the specific component that it is predicted to effect (Herlihy and Eckert, 2004
). Other characteristics such as the size and density of populations vary substantially within regions and probably have a much greater effect on overall self-fertilization in A. canadensis, probably through determining opportunities for outcrossing (Routley et al., 1999
; Herlihy and Eckert, 2004
).
If regional differences in floral traits are not associated with differences in the mating system, then how did they evolve? The first step in addressing this question is determining whether the difference in display size, flower size, and herkogamy between central and northern populations has a genetic basis. In another study (C. R. Herlihy and C. G. Eckert, unpublished manuscript), we grew open-pollinated seed families collected from two of the central populations and two of the northern populations used in this study under a common greenhouse environment and only detected significant differentiation between the central and northern populations for herkogamy. As in the field, herkogamy was substantially higher for central than for northern populations, but there was no tendency for either display size or spur length to be greater for central populations. We conclude herkogamy exhibits strong genetic differentiation between central vs. northern populations of A. canadensis, but regional differences in flower size and display size are likely due to environmental effects.
Taken together, these results suggest strong evolutionary differentiation between northern and central populations in herkogamy without any differentiation in the mating system. At present, we can only offer speculative explanations for this enigmatic pattern. One possibility is that there has been stronger selection for herkogamy to maintain outcrossing in central than in northern populations. Consistent with this, there was much less phenotypic variation in herkogamy among plants within central than within northern populations (mean CV: central = 51%, northern = 74%). Estimates of genetic variation for herkogamy are also much lower for central than for northern populations (C. R. Herlihy and C. G. Eckert, unpublished manuscript), as might be expected if herkogamy were under stronger selection in central populations. Stronger selection could result from stronger inbreeding depression due either to higher genetic load or more stringent ecological conditions (Cheptou et al., 2002
). However, our results do not indicate any regional difference in the strength of inbreeding depression as expressed under field conditions.
Regional differences in selection for herkogamy could also arise from geographical variation in pollinator fauna. For instance, herkogamy may be more effective at limiting self-pollination during visitation by hummingbirds than bees because some bumble bees actively collect pollen when visiting flowers of A. canadensis, whereas hummingbirds simply probe the spurs for nectar (C. R. Herlihy and C. G. Eckert, personal observation). We could not observe enough pollinators to determine whether there might be a regional difference in hummingbird visitation among the populations we studied. However, data from breeding bird surveys (Sauer et al., 2004
) suggest that hummingbirds are more than twice as abundant in the physiographic region that includes our central populations (Blue Ridge Mountains: mean from 1966–2002 = 0.86 birds per survey route) than the region including our northern populations (St. Lawrence River Plain: mean = 0.31 birds per survey). High levels of selfing despite pronounced herkogamy might be further explained if hummingbird abundance had recently declined in central populations. However, the survey data suggest that relative abundances have been stable in the central region (mean = +0.3% change per year, 95% confidence limits = –6.2% to +6.8%) and increasing in the north (mean = +5.8%, 95% CL = +2.3% to +9.3%). Intensive surveys of pollinator abundance and experimental investigation of how variation in floral morphology affects pollination is required to test this hypothesis and to determine why A. canadensis exhibits the classic symptoms of increased selfing in peripheral populations without any geographical differentiation in the mating system.
FOOTNOTES
The authors thank Michael Bhardwaj for help in the field and lab; Jeremy Brown and Celine Griffin for help in the lab; Karen Samis for comments on the manuscript; the Queen's University, Mountain Lake and Highlands biological stations for logistic support in the field; Queen's University for a Dean's Doctoral Field Travel Grant to C. R. H; Highlands and Mountain Lake biological stations for grant-in-aid scholarships to C. R. H, and the Natural Sciences and Engineering Council of Canada for a discovery grant to C. G. E. 
2 Present address: Department of Biology, Indiana University, Bloomington, Indiana 47405 USA
3 Author for correspondence (e-mail: eckertc{at}biology.queensu.ca
)
LITERATURE CITED
Ågren J. 1996 Population size, pollinator limitation, and seed set in the self-incompatible herb Lythrum salicaria. Ecology 77: 1779-1790[CrossRef][ISI]
Aizen M. A. P. Feinsinger 1994 Habitat fragmentation, native insect pollinators, and feral honey bees in Argentine "Chaco Serrano.". Ecological Applications 4: 378-392[CrossRef][ISI]
Baker H. G. 1955 Self-compatibility and establishment after "long-distance" dispersal. Evolution 9: 347-348[CrossRef][ISI]
Barrett S. C. H. 1988 The evolution, maintenance, and loss of self-incompatibility systems. In J. Lovett-Doust and L. Lovett-Doust [eds.], Plant reproductive ecology. Patterns and strategies, 98–124. Oxford University Press, New York, New York, USA
Barrett S. C. H. C. G. Eckert 1990 Variation and evolution of mating systems in seed plants. In S. Kawano [ed.], Biological approaches and evolutionary trends in plants, 229–254. Academic Press, Tokyo, Japan
Barrett S. C. H. L. D. Harder W. W. Cole 1994 Effects of flower number and position on self-fertilization in experimental populations of Eichhornia paniculata (Pontederiaceae). Functional Ecology 8: 526-535[CrossRef][ISI]
Barrett S. C. H. B. C. Husband 1990 Variation in outcrossing rates in Eichhornia paniculata: the role of demographic and reproductive factors. Plant Species Biology 5: 41-55
Barrett S. C. H. M. T. Morgan B. C. Husband 1989 The dissolution of a complex genetic polymorphism: the evolution of self-fertilization in tristylous Eichhornia paniculata. Evolution 43: 1398-1416[CrossRef][ISI]
Brown J. H. 1984 On the relationship between abundance and distribution of species. American Naturalist 124: 255-279[CrossRef][ISI]
Brunet J. C. G. Eckert 1998 Effect of floral morphology and display on outcrossing in blue columbine, Aquilegia caerulea (Ranunculaceae). Functional Ecology 12: 596-606[CrossRef][ISI]
Byers D. L. 1995 Pollen quantity and quality as explanations for low seed set in small populations exemplified by Eupatorium (Asteraceae). American Journal of Botany 82: 1000-1006[CrossRef][ISI]
Cheptou P. O. J. Lepart J. Escarre 2002 Mating system variation along a successional gradient in the allogamous and colonizing plant Crepis sancta (Asteraceae). Journal of Evolutionary Biology 15: 753-762[CrossRef][ISI]
Cruden R. W. D. L. Lyon 1989 Facultative xenogamy: examination of a mixed mating system. In J. H. Bock and Y. B. Linhart [eds.], The evolutionary ecology of plants, 171–208. Westview Press, Boulder, Colorado, USA
Eckert C. G. 2000 Contributions of autogamy and geitonogamy to self-fertilization in a mass-flowering, clonal plant. Ecology 81: 532-542[CrossRef][ISI]
Eckert C. G. A. Schaefer 1998 Does self-pollination provide reproductive assurance in wild columbine, Aquilegia canadensis (Ranunculaceae)?. American Journal of Botany 85: 919-924[Abstract]
Elle E. R. Carney 2003 Reproductive assurance varies with flower size in Collinsia parviflora (Scrophulariaceae). American Journal of Botany 90: 888-896
Fausto J. A. Jr. V. M. Eckhart M. A. Geber 2001 Reproductive assurance and the evolutionary ecology of self-pollination in Clarkia xantiana (Onagraceae). American Journal of Botany 88: 1794-1800
Franceschinelli E. V. K. S. Bawa 2000 The effect of ecological factors on the mating system of a South American shrub species (Helicteres brevispira). Heredity 84: 116-123
Griffin C. A. M. C. G. Eckert 2003 Experimental analysis of biparental inbreeding in a self-fertilizing plant. Evolution 57: 1513-1519[ISI][Medline]
Griffin S. R. K. Mavraganis C. G. Eckert 2000 Experimental analysis of protogyny in Aquilegia canadensis (Ranunculaceae). American Journal of Botany 87: 1246-1256
Groom M. J. 1998 Allee effects limit population viability of an annual plant. American Naturalist 151: 487-496[CrossRef][ISI]
Hall C. A. S. J. A. Stanford F. R. Hauer 1992 The distribution and abundance of organisms as a consequence of energy balances along multiple environmental gradients. Oikos 65: 377-390[CrossRef][ISI]
Harder L. D. S. C. H. Barrett 1995 Mating cost of large floral displays in hermaphrodite plants. Nature 373: 512-515[CrossRef]
Harder L. D. S. C. H. Barrett 1996 Pollen dispersal and mating patterns in animal-pollinated plants. In D. G. Lloyd and S. C. H. Barrett [eds.], Floral biology: studies on floral evolution in animal-pollinated plants, 140–190. Chapman & Hall, New York, New York, USA
Hengeveld R. J. Haeck 1982 The distribution of abundance. I. Measurements. Journal of Biogeography 9: 303-316[CrossRef][ISI]
Herlihy C. R. C. G. Eckert 2002 Genetic cost of reproductive assurance in a self-fertilizing plant. Nature 416: 320-323[CrossRef][Medline]
Herlihy C. R. C. G. Eckert 2004 Experimental dissection of inbreeding and its adaptive significance in a flowering plant, Aquilegia canadensis (Ranunculaceae). Evolution 58: 2693-2703[ISI][Medline]
Holtsford T. P. N. C. Ellstrand 1992 Genetic and environmental variation in floral traits affecting outcrossing rate in Clarkia tembloriensis (Onagraceae). Evolution 46: 216-225[CrossRef][ISI]
Husband B. C. D. W. Schemske 1996 Evolution of the magnitude and timing of inbreeding depression in plants. Evolution 50: 54-70
Jain S. K. 1976 The evolution of inbreeding in plants. Annual Review of Ecology and Systematics 7: 69-95
Jennersten O. 1988 Pollination in Dianthus deltoides (Caryophyllaceae): effects of habitat fragmentation on visitation and seed set. Conservation Biology 2: 359-366
Jump A. S. F. I. Woodward 2003 Seed production and population density decline approaching the range-edge of Cirsium species. New Phytologist 160: 349-358[CrossRef][ISI]
Karron J. D. R. T. Jackson N. N. Thumser S. L. Schlicht 1997 Outcrossing rates of individual Mimulus ringens genets are correlated with anther-stigma separation. Heredity 79: 365-370[CrossRef][ISI]
Karron J. D. N. N. Thumser R. Tucker A. J. Hessenauer 1995 The influence of population density on outcrossing rates in Mimulus ringens. Heredity 75: 175-180[ISI]
Kearns C. A. D. W. Inouye 1993 Techniques for pollination biologists. University Press of Colorado, Niwot, Colorado, USA
Klips R. A. A. A. Snow 1997 Delayed autonomous self-pollination in Hibiscus laevis (Malvaceae). American Journal of Botany 84: 48-53[Abstract]
Kohn J. R. S. C. H. Barrett 1992 Experimental studies on the functional significance of heterostyly. Evolution 46: 43-55[CrossRef][ISI]
Larson B. M. H. S. C. H. Barrett 2000 A comparative analysis of pollen limitation in flowering plants. Biological Journal of the Linnean Society 69: 503-520[CrossRef]
Lawton J. H. 1993 Range, population abundance and conservation. Trends in Ecology and Evolution 8: 409-413
Leclerc-Potvin C. K. Ritland 1994 Modes of self-fertilization in Mimulus guttatus (Scrophulariaceae): a field experiment. American Journal of Botany 81: 199-205[CrossRef][ISI]
Levin D. A. 2000 The origin, expansion, and demise of plant species. Oxford University Press, Oxford, UK
Lloyd D. G. 1992 Self- and cross-fertilization in plants. II. The selection of self-fertilization. International Journal of Plant Sciences 153: 370-380[CrossRef]
Macior L. W. 1966 Foraging behaviour of Bombus (Hymenoptera: Apidae) in relation to Aquilegia pollination. American Journal of Botany 53: 302-309[CrossRef][ISI]
Macior L. W. 1978 The pollination of vernal angiosperms. Oikos 30: 452-460[CrossRef][ISI]
Mavraganis K. C. G. Eckert 2001 Effect of population size and isolation on reproductive output in Aquilegia canadensis (Ranunculaceae). Oikos 95: 300-310[CrossRef][ISI]
Motten A. F. J. L. Stone 2000 Heritability of stigma position and the effect of stigma-anther separation on outcrossing in a predominantly self-fertilizing weed, Datura stramonium (Solanaceae). American Journal of Botany 87: 339-347
Murawski D. A. 1991 The effect of density of flowering individuals on the mating system of nine tropical tree species. Heredity 67: 167-174[ISI]
Ornduff R. 1969 Reproductive biology in relation to systematics. Taxon 18: 121[CrossRef]
Pyke G. H. 1979 Optimal foraging in bumblebees: rules of movement between flowers within inflorescences. Animal Behaviour 27: 1167-1181[CrossRef][ISI]
Ritland K. 1990a A series of FORTRAN computer programs for estimating plant mating systems. Journal of Heredity 81: 235-237[ISI]
Ritland K. 1990b Inferences about inbreeding depression based on changes of the inbreeding coefficient. Evolution 44: 1230-1241[CrossRef][ISI]
Roff D. A. 1997 Evolutionary quantitative genetics. Chapman & Hall, New York, New York, USA
Routley M. B. K. Mavraganis C. G. Eckert 1999 Effect of population size on the mating system in a self-compatible, autogamous plant, Aquilegia canadensis (Ranunculaceae). Heredity 82: 518-528
Runions C. J. M. A. Geber 2000 Evolution of the self-pollinating flower in Clarkia xantiana (Onagraceae). I. Size and development of floral organs. American Journal of Botany 87: 1439-1451
Sagarin R. D. S. D. Gaines 2002 The ‘abundant centre’ distribution: to what extent is it a biological rule?. Ecology Letters 5: 137-147
SAS Institute. 2002 JMP user's guide, version 5.0. SAS Institute, Cary, North Carolina, USA
Sauer J. R. J. E. Hines J. Fallon 2004 The North American breeding bird survey, results and analysis 1966–2003, version 2004.1. USGS Patuxent Wildlife Research Center, Laurel, Maryland, USA
Schoen D. J. 1982 The breeding system of Gilia achilleifolia: variation in floral characteristics and outcrossing rate. Evolution 36: 352-360[CrossRef][ISI]
Sih A. M. Baltus 1987 Patch size, pollinator behavior and pollinator limitation in catnip. Ecology 68: 1679-1690[CrossRef][ISI]
Sokal R. R. F. J. Rohlf 1995 Biometry, 3rd ed. W. H. Freeman, New York, New York, USA
Solbrig O. T. 1976 On the relative advantages of cross- and self-fertilization. Annals of the Missouri Botanical Garden 63: 262-276[CrossRef][ISI]
Stebbins G. L. 1957 Self-fertilization and population variability in the higher plants. American Naturalist 91: 337-354[CrossRef][ISI]
van Treuren R. R. Bijlsma N. J. Ouborg W. van Delden 1993 The effects of population size and plant density on outcrossing rates in locally endangered Salvia pratensis. Evolution 47: 1094-1104[CrossRef][ISI]
Vrieling K. L. Saumitou P. J. Cuguen H. van Dijk T. J. de Jong P. G. L. Klinkhamer 1999 Direct and indirect estimates of the selfing rate in small and large individuals of the bumblebee pollinated Cynoglossum officinale L. (Boraginaceae). Ecology Letters 2: 331-337[CrossRef][ISI]
Webb C. J. D. G. Lloyd 1986 The avoidance of interference between the presentation of pollen and stigmas in angiosperms. II. Herkogamy. New Zealand Journal of Botany 24: 163-178[ISI]
Wyatt R. 1986 Ecology and evolution of self-pollination in Arenaria uniflora (Caryophyllaceae). Journal of Ecology 74: 403-418[CrossRef][ISI]
Wyatt R. 1988 Phylogenetic aspects of the evolution of self-pollination. In L. D. Gottlieb and S. K. Jain [eds.], Plant evolutionary biology, 109– 131. Chapman and Hall, London, UK
This article has been cited by other articles:
|
|
 |
|
  S. Kubota, Y. Kameyama, A. S. Hirao, and M. Ohara Adaptive significance of self-fertilization in a hermaphroditic perennial, Trillium camschatcense (Melanthiaceae) Am. J. Botany, April 1, 2008; 95(4): 482 - 489. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Mimura and S. N. Aitken Increased selfing and decreased effective pollen donor number in peripheral relative to central populations in Picea sitchensis (Pinaceae) Am. J. Botany, June 1, 2007; 94(6): 991 - 998. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. L. Stone, M. A. Sasuclark, and C. P. Blomberg Variation in the self-incompatibility response within and among populations of the tropical shrub Witheringia solanacea (Solanaceae) Am. J. Botany, April 1, 2006; 93(4): 592 - 598. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. W. Busch The evolution of self-compatibility in geographically peripheral populations of Leavenworthia alabamica (Brassicaceae) Am. J. Botany, September 1, 2005; 92(9): 1503 - 1512. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||
