Maia F. Bailey, Centre d’Ecologie Fonctionnelle et Evolutive (CNRS), 1919 route de Mende, F-34293 Montpellier cedex 5, France (tel. +33 4 67 61 34 24; fax +33 4 67 41 06 16; e-mail firstname.lastname@example.org).
1Silene vulgaris was introduced from Europe to North America prior to 1800. We evaluated the influence of the resulting population structure on traits related to fitness in a two generation series of glasshouse crosses. Plants were either self-fertilized, outcrossed within local populations, outcrossed between local populations or outcrossed between three geographical regions separated by > 150 km.
2Inbreeding depression following self-fertilization occurred in nearly every maternal lineage studied but its magnitude varied among geographical regions, especially when the fitness component compared was seed germination.
3The consequences of long-distance gene flow varied among geographical regions: crosses between some pairs of regions resulted in fitter offspring than within region outcrosses (i.e. heterosis), whereas other pairs of regions showed outbreeding depression (offspring of more distant crosses were less fit).
4Individuals derived from self-fertilization of F1 individuals were fitter if the first crosses had been made between populations or regions rather than within populations, both in terms of seed germination and seedling survivorship. The benefits of gene flow therefore persist for at least one generation beyond that created by the gene flow event.
5The fitness of offspring of crosses between F1 individuals and their maternal population did not depend on the relatedness of the parents of the F1. Thus, there was no evidence of outbreeding depression following recombination within F1 individuals whose creation involved gene flow.
6Our results imply that alteration of the rate of self-fertilization or the rate of gene flow between populations of S. vulgaris could affect the demographic characters that influence population establishment and persistence in this species.
There is a long-standing interest in the relationship between population structure, inbreeding and those demographic traits likely to contribute to the establishment and persistence of plant populations (recent reviews include Ellstrand & Elam 1993; Dudash & Fenster 2000; Keller & Waller 2002; Tallmon et al. 2004), leading to questions about the demographic consequences of gene flow between populations. If the process of genetic differentiation between populations is largely random, populations will differ in their accumulation of deleterious recessive mutations and their loss of advantageous alleles through genetic drift. In this case we should expect gene flow to be, on average, favourable because it masks deleterious recessive alleles and reintroduces advantageous alleles (Ingvarsson & Whitlock 2000). If, however, the differentiation is driven by adaptation, gene flow could be disadvantageous because it introduces maladaptive alleles and leads to disruption of co-adapted gene complexes (Price & Waser 1979). Because heterosis is maximized in the first generation of interpopulation crosses, while recombination among loci does not occur until the crossing lineages undergo meiosis, heterosis effects are most evident in the first generation following gene flow, while epistatic effects increase in importance in subsequent generations (Lynch 1991).
Many plant population genetic studies have used crosses within and among natural or artificial populations to investigate the effect of inbreeding and gene flow on demographic characters (e.g. Van Treuren et al. 1993; Galloway & Etterson 2005; see a general discussion in Keller & Waller 2002). Many of these studies have focused on the effect of population size and/or the effect of the geographical proximity of the populations exchanging genes. Most studies have demonstrated increased fitness in the F1 generation, particularly when gene flow is into small populations and from increasing distances. Fewer studies have followed the fitness effects of gene flow through to later generations (Tallmon et al. 2004).
The effect of gene flow on demography and population persistence is often couched in terms of genetic rescue (Tallmon et al. 2004), i.e. a gene flow event which results in population fitness increases in excess of those caused by the direct increase in population size resulting from immigration (Ingvarsson 2001). Studies are also generally framed in reference to conservation issues arising when small population size has caused high levels of inbreeding over several generations (e.g. Madsen et al. 1999; Sheridan & Karowe 2000; Vilàet al. 2002). Although the demography of weedy plant species would appear to differ greatly from those of small, remnant populations, genetic rescue may also contribute to their persistence (Thrall et al. 1998), as demonstrated in the weedy plant Silene alba by Richards (2000). Weedy plant species often consist of a large number of ephemeral patches. This patchy spatial distribution can result in a population genetic structure determined by the size and connectivity of extant local populations and the demographic processes associated with frequent turnover. Successive colonization events can deplete genetic variation within populations, depending on the mode of colonization (Wade & McCauley 1988; Whitlock & McCauley 1990; McCauley et al. 1995).
Variation among populations in the magnitude of inbreeding depression following self-fertilization may similarly be indicative of the structure of genes contributing to fitness. It is well known that the magnitude of inbreeding depression following self-fertilization can differ among families (Koelewijn 1998; Mutikainen & Delph 1998; Pico et al. 2004) and among populations (Dudash et al. 1997; Ouborg et al. 2000), but not how this variation might be partitioned among groups of populations as a consequence of their structure. Since self-fertilization (or close relative mating) must be particularly frequent in the early stages of population establishment, especially when populations are founded by single seeds, the influence of structure on inbreeding depression following self-fertilization could be particularly important in many self-compatible weedy plants whose persistence depends on the continual establishment of a series of local patches.
We examined the effects of population structure and gene flow on characters related to fitness in North American populations of Silene vulgaris (Moench) Garcke. Populations of this weedy species typically consist of patches of 10–100+ individuals found along roadsides or in fallow fields (McCauley et al. 2000). The species is gynodioecious with hermaphrodites being fully self-compatible and known to self at moderate rates (Emery 2001). While inbreeding depression is high within some populations in Giles County, Virginia (Emery & McCauley 2002), little is known about its variation at a larger spatial scale or how population-level inbreeding depression would influence the fitness consequences of gene flow.
Because naturalized populations of S. vulgaris have been observed in North America for less than 300 years (Cutler 1785), this species must have undergone a rapid range expansion following introduction. It is not clear how historical factors associated with introduction and expansion might interact with recent local patch dynamics associated with a weedy habit to influence population structure. For example, multiple introductions from different European sources to different North American localities might create a greater large scale population structure than would a single introduction. Conversely, the gene flow associated with a rapid, recent range expansion might limit the opportunity for local adaptation. Thus, one might want to contrast the demographic effects of crosses among local populations with the effects of longer distance crosses among regions. Further, one might want to replicate the regions involved so as to distinguish between the fitness effects of long distance crosses, per se, with the effects of combining alleles from specific pairs of regions.
We investigate the consequences of population structure for demographic traits related to fitness and population persistence using four crossing treatments (selfing, outcrossing within a population, outcrossing among local populations, and outcrossing among regions) with 12 parental populations of S. vulgaris (four local populations from each of three geographical regions in eastern North America). To test for persistent effects among the long distance outcrosses we produced a F2 generation by selfing some individuals from the F1 generation and crossing others to offspring from the outcross within population treatment for the maternal population (see Fig. 1). We intended the F2 to mimic the fate of genes entering a population by pollen flow. Because seeds tend to fall near the seed parent, the outcrossed individuals were likely to receive pollen from the seed parent's population, and can therefore be thought of as equivalent to a backcross treatment.
All F1 individuals were monitored for seed fitness, germination, early survivorship, late survivorship and flowering. F2 fitness was recorded through early survivorship. In addition to investigating the general effects of scale on the effects of inbreeding and population structure, our design allows us to investigate whether the degree of local adaptation and/or genetic architecture differs among regions.
Silene vulgaris is an insect pollinated herbaceous perennial native to Eurasia, where it occurs in meadows, pastures and along roadsides. Currently, S. vulgaris populations also occur throughout much of temperate North America, growing mainly in disturbed habitats. References to North American naturalized populations of S. vulgaris, under various outmoded names, date back to at least 1785 (Cutler 1785; Torrey & Gray 1840).
S. vulgaris is gynodioecious with most local populations consisting of both female and hermaphrodite individuals. Sex determination is cyto-nuclear with at least three cytoplasmic types and multiple nuclear restorer loci (Charlesworth & Laporte 1998).
The distribution of chloroplast DNA (cpDNA) and mitochondrial DNA (mtDNA) polymorphism among North American populations indicates that cytoplasmic genes exhibit a very high level of structure among local populations, and an additional level of structure among more widely spaced geographical regions (McCauley 1998; Olson & McCauley 2002; McCauley et al. 2003). This higher level of structure stems from both differences in frequencies of widespread cpDNA haplotypes among regions and from the existence of region-specific haplotypes (McCauley et al. 2003). Allozyme polymorphism displays a moderate amount of structure among local populations in south-west Virginia (McCauley 1998), but little is known about the geographical structure of nuclear genetic polymorphism at a larger spatial scale. Allozyme-based analysis of progeny arrays has estimated the selfing rate of hermaphrodites from Giles County, VA, to be about 40% (Emery 2001).
Four local populations were selected from each of three of the geographical regions studied by McCauley et al. (2003): Delaware County near Stamford, New York (hereafter ‘New York’); Rockingham County near Broadway, Virginia (‘Broadway’); and Giles County, Virginia (‘Giles’), USA. The three regions are separated from one another by 150–500 km and the four local populations within each region are separated by 1–10 km. The Giles region appears to be near the southern limit of the species in eastern North America (J. Antonovics, pers. comm.). Open-pollinated fruits were collected from the 12 experimental populations in late summer 2000. Each fruit was collected from a different individual (chosen at random and not sexed; n = 7 except for South Jefferson, New York, where n = 6) so that its seeds represent an independent maternal half- or full-sib family. Six of these families from each population were used in the crossing design. Seven seeds were germinated and four plants from each maternal half-sib family were used as seed parents, one for each of the pollination treatments, and the remainder as pollen donors as needed. Only hermaphrodite plants were used as seed parents. Germination and pollination protocols have previously been reported for a separate data set using these individuals (Bailey & McCauley 2005).
The four pollination treatments used to produce the F1 generation were (i) self pollen from another flower on the same plant, (ii) outcross pollen from a different maternal family from the same local population, (iii) outcross pollen from a different population in the same region as the maternal plant, and (iv) outcross pollen from a population from a region different from the maternal plant (Fig. 1). Outcross pollen donors were chosen by randomly matching maternal families and populations from the available pool without replacement so that all 72 original maternal families were included as both seed and pollen parents in all treatment crosses.
The F2 generation was produced by randomly selecting two individuals from each outcrossed F1 family (i.e. from F1 treatments 2, 3 and 4). Using the same pollination protocol as for the F1, one sib was self-fertilized and the second was crossed to the maternal population using pollen donors chosen from the within-population outcrosses (i.e. from the F1 treatment 2 using its maternal parent), to give selfing (A) and outcrossing (B) of each treatment 2 family, of each treatment 3 family (C and D) and of each treatment 4 family (E and F) (Fig. 1). Pollen donors were chosen randomly from those outcrosses available that did not share any grandparents. This was done to make the F2 outcrosses more consistent, as B treatment crosses would otherwise be twice as likely to be inbred among their grandparents as the D and F treatment individuals.
For the current analysis a subset of the F2 crosses were grown through to the production of true leaves. The crosses used in this analysis are randomly distributed among the six treatments, three regions and 12 populations. All treatments are represented among each population except treatment B for the Couch population of the Giles region, i.e. outcrossing within the Couch population for a second generation.
Both F1 and F2 seeds were sorted into filled and unfilled seeds. Unfilled seeds are inviable (M. F. Bailey and D. E. McCauley, unpublished data) but were counted for calculating the proportion of filled to unfilled seeds produced by a cross (hereafter referred to as ‘seed fitness’). Up to 50 filled seeds per cross were planted in the glasshouse (randomized spatially by cross treatment, region and population of origin), and F1 seeds were monitored for germination, survivorship and flower production following the protocol in Bailey & McCauley (2005) while a subset of the F2 crosses were grown through to the production of true leaves. Blocks of crosses were germinated as seeds became available such that the F1 was distributed among three and the F2 among two temporal blocks. All offspring from a particular cross were kept together in the glasshouse but trays containing four or more crosses were periodically rotated to reduce spatial effects.
In order to examine effects on inbreeding among the F1, we used data sets consisting of each of the separate fitness components measured in selfed individuals (treatment 1) and those outcrossed within populations (treatment 2). All proportion variables were arcsine square root transformed prior to Analysis of Variance (Sokal & Rohlf 1995). The Fit Model function of the JMP software package (Sall et al. 2001), set to the least squares option, was used to test the fitness response variables for main effects of block, cross treatment and maternal (dam) region of origin, as well as dam population of origin as a nested effect within dam region. Block was included as random effect. We also included cross effects for block by treatment, block by dam region, and treatment by dam region in the model. Interactions involving populations within regions were not considered owing to limited sample size.
Total fitness was calculated for each cross as the product of proportion filled seeds, proportion of filled seeds that germinated, proportion of germinated seedlings surviving to the end of the experiment and proportion of surviving plants that flowered by the end of the experiment. In order to compare our results with other studies, we also calculated a standardized measure of inbreeding depression for each maternal lineage for which seeds were available from both the selfing and the outcrossing within population treatments, following Ågren & Schemske (1993). If the self treatment had lower total fitness (w) than the outcrossing within population treatment for the same maternal lineage, i.e. ws < wo, inbreeding depression was calculated as:
δ = 1 − ws/wo
If the self treatment had higher fitness than the outcrossing within population treatment, i.e. ws > wo, inbreeding depression was calculated as:
δ = wo/ws − 1
Values for inbreeding depression are therefore bounded by −1 and +1 with negative values indicating higher fitness of selfed individuals than outcrossed individuals. Population-mean inbreeding depression was calculated by averaging over all maternal lineages within that population while regional inbreeding depression was averaged over all populations within that region.
To examine the effects of geographical scale on the consequences of outbreeding among the F1, we used a data set limited to treatments 3 and 4 (outcrossing among populations, within regions and outcrossing among regions). Model effects and software settings were the same as for inbreeding tests above. If fitness is totally additive and normally distributed within populations, fitness of the F1 among-region crosses should equal the mean of the values for the within-region crosses. Heterosis as a consequence of long-distance gene flow is therefore indicated by an increase in the fitness in treatment 4 relative to treatment 3 and outbreeding depression by a decrease in fitness in these comparisons. If general, these phenomena would be reflected statistically by significant treatment effects in the anova and, if limited to certain regions or combinations of regions, by a significant treatment–region interaction.
Fitness components for the F2 individuals are limited to germination and survivorship through to the production of true leaves. Data analysis on the F2 data set included comparisons among selfed treatments (i.e. A vs. C and E) and among outcrosses (i.e. B vs. D and F). Model conditions and effects were the same as those used in analysing the F1 data above. Here the residual effects of heterosis due to gene flow would be reflected by a significantly higher fitness in F2 offspring of F1 parents that were the result of among-population gene flow within either self or outcross treatments (i.e. C or E > A, D or F > B). Similarly, outbreeding depression would be indicated if gene flow results in lower fitness (i.e. C or E < A, D or F < B).
We also compared mean fitness values for F2 crosses resulting from among-region gene flow (i.e. treatments E or F) with values predicted from means of F1 parental fitness. If fitness is a totally additive trait and normally distributed within populations, fitness of F2 individuals that are the result of selfing (treatment E) should equal that of their (treatment 4, outcrossed) F1 parents. Inbreeding depression is indicated if treatment E offspring values are lower. The fitness of outcrossed F2 individuals (treatment F) is predicted by the mean of the appropriate F1 among-region (treatment 4) and within-region (treatment 3) crosses. Heterosis is indicated when observed values exceed the predicted mid-parental value and outbreeding depression when observed values fall below predicted values.
All four F1 crossing treatments were completed for 57 maternal lines. Partial treatment arrays were produced for 12 additional maternal lines. We were unable to successfully produce offspring from only three maternal lines [one each from the Couch (Giles), Saddle (New York) and Windy Ridge (New York) populations]. Overall, 11 482 seeds from 278 crosses were planted. Data on F1 germination, survivorship and flowering were recorded until January 2004, by which date 8103 plants, or 70.6% of planted seeds, had produced at least three flowers.
Self-fertilization reduced total fitness of F1 offspring compared to our outcrossing treatments (Fig. 2). When comparing fitness traits of offspring under selfing (treatment 1) to offspring of outcrosses within populations (treatment 2), crossing treatment had a significant effect on seed fitness (% filled) and survivorship to flowering (Table 1). Germination was significantly affected by the dam's source population, block by region and treatment by region interactions (Table 1). Treatment by region effects on germination seem to be due to offspring from New York dams having disproportionately higher germination with outcrossing compared to other regions (Fig. 3). Block by dam region effects were due to higher germination rates for offspring of New York dams in block 3, perhaps due to different ambient light conditions during this block, which was started in the glasshouse in July 2003, as opposed to blocks 1 and 2 which were initiated in November 2002 and January 2003, respectively.
Table 1. Model effects on offspring fitness components for treatment crosses of self-fertilization and outcrosses within population. % filled was tested using a different model because this fitness component has no block effect. Results are given as the F-statistic in italics and associated probability. Statistically significant results are in bold
Prob > F
Prob > F
Prob > F
Prob > F
Prob > F
Dam pop (Dam region)
Block × Treatment
Block × Dam region
Treatment × Dam region
We also calculated inbreeding depression for each maternal lineage for which we have both treatment 1 and 2 data (Fig. 4). A few lineages have higher offspring fitness when selfed; however, the generally consistently lower fitness of selfed offspring indicates significant genetic load within populations. Although not statistically significant (F2 = 1.95, P = 0.151) at the regional level, New York had the highest inbreeding depression (δ = 0.404 ± 0.096), Giles the lowest (δ = 0.145 ± 0.091), and Broadway intermediate values (δ = 0.291 ± 0.085).
Among outcross treatments, there is no consistent effect of cross distance on offspring fitness (Fig. 2 and a significant treatment by region interaction term in the model comparing among-population outcross treatments in Table 2). For crosses using dams from the Giles region, pollen donors from greater distances produce offspring with increasingly greater average fitness whereas within-region crosses produced the highest offspring fitness for New York dams but the lowest fitness of all outcrosses for Broadway dams. When we examine the components of total fitness, both germination and survival to flowering are significantly affected by an interaction of treatment and dam region (Table 2, Fig. 3).
Table 2. Model effects on offspring fitness components for treatment crosses of outcrosses among local populations and outcrosses among widely spaced populations. % filled was tested using a different model because this fitness component has no block effect. Statistically significant results are in bold
Prob > F
Prob > F
Prob > F
Prob > F
Prob > F
Dam pop (Dam region)
Block × Treatment
Block × Dam region
Treatment × Dam region
Mean fitness of treatment 3 crosses was used as a surrogate for the population parental mean and compared with treatment 4 means to evaluate heterosis, outbreeding depression and reciprocal cross effects (see Methods). In crosses between Broadway and Giles individuals, F1 offspring show clear heterosis with average fitness exceeding the mid-parental mean (Fig. 5a, the open symbols for the F1 are above the solid line between the two parents). In contrast, the fitness of New York crosses depends on which region is used as the dam and which as the sire. When Giles and New York are crossed, only offspring from New York dams do better than predicted (Fig. 5b); however, in crosses among Broadway and New York individuals, offspring from New York dams appear to have outbreeding depression while those of Broadway dams are as predicted by parental means (Fig. 5c).
As of December 2004, 228 F2 crosses had been grown in the glasshouse through to the production of true leaves. Of these, 42 treatment A, 28 treatment B, 42 treatment C, 39 treatment D, 37 treatment E and 40 treatment F crosses are represented in the analysis.
When F1 individuals resulting from outcrossing were selfed, the origin of their sire (i.e. whether from the same, local or distant populations), affected the fitness of the F2 offspring (Table 3). Individuals resulting from among population gene flow (treatments 3 and 4) produced offspring with higher germination and survival rates when selfed (treatments C and E) compared to selfed individuals derived from within population crosses (treatment A) (Fig. 6). Seed fitness among selfed offspring also depended on the maternal lineage's region, resulting in a significant treatment by region interaction term in the model (Table 3).
Table 3. Model effects on offspring fitness components for F2 self treatments (A, C and E). Statistically significant results are in bold
Prob > F
Prob > F
Prob > F
Prob > F
Dam pop (Dam region)
Block × Treatment
Block × Dam region
Treatment × Dam region
When F1 individuals from outcrosses were further outcrossed with their maternal population, the presence or absence of gene flow in their pedigree does not seem to affect the fitness of their offspring (Table 4). For germination, the three F2 outcross treatments produce remarkably similar results (Fig. 6). Raw seedling survival rates for outcrosses using individuals derived from among region crosses (treatment F) are lower than survival rates for the other outcross treatments (B and D), although this difference is not statistically significant.
Table 4. Model effects on offspring fitness components for F2 outcross treatments (B, D and F)
Prob > F
Prob > F
Prob > F
Prob > F
Dam pop (Dam region)
Block × Treatment
Block × Dam region
Treatment × Dam region
Comparison of the fitness of F2 offspring resulting from among-region gene flow (i.e. treatments E and F) to values predicted from their F1 parents (see Methods) suggests that inbreeding depression, heterosis and outbreeding depression depend on population; however, large standard errors for this data due to small sample sizes prevents us from drawing many conclusions. Fitness of selfed offspring with a pedigree of long-distance gene flow fall below the F1 parental value (closed F2 symbols are below open F1 symbols in Fig. 5) although in most cases error bars overlap. For treatment F crosses among New York and Broadway regions, Broadway outcross individuals appear to marginally exceed the predicted value (closed symbol – and its error bar – is above dotted line in Fig. 5c) indicating some residual heterosis in these crosses.
By crossing within and among S. vulgaris populations we found examples of inbreeding depression, heterosis and outbreeding depression. In particular, the magnitude of inbreeding depression following self-fertilization, and the fitness consequences of long-distance gene flow, depend on which specific geographical region, or combination of geographical regions, was considered. It was particularly interesting that lineages derived from between-population gene flow are protected from inbreeding depression, should they subsequently self-fertilize. Taken together, our results suggest that S. vulgaris has developed a population structure during its range expansion across North America that could have important consequences for some of the demographic characteristics that determine weediness.
Specifically, in the F1 generation, we found high inbreeding depression following self-fertilization across all regions and populations. Very few maternal lineages had higher or equivalent fitness when they were self-fertilized compared to outcrosses within the same population. In the Broadway region, offspring from self-fertilization have nearly equivalent germination rates when compared to offspring from outcrosses within populations, while both New York and Giles region populations expressed high inbreeding depression for this measure of fitness. This suggests that the amount and/or expression of genetic load varies from region to region. Although variation among populations in the magnitude and expression of inbreeding depression has been shown in studies of other species (e.g. Dudash et al. 1997; Ouborg et al. 2000; Richards 2000), studies that demonstrate geographical variation in the magnitude of inbreeding depression at a more regional scale are not common.
When considering the effect of geographical scale on the results of crossing between populations (Treatment 3 vs. Treatment 4), germination and survival were best predicted by an interaction between treatment and region. New York individuals produced more fit offspring when crossed within the region than when pollen donors from other regions were used. When among-region crosses are compared to within-region crosses for total fitness, the poor performance of New York dams in among-region crosses can be traced to F1 outbreeding depression when pollen from Broadway populations is used. Conversely, Giles region populations produce offspring with higher germination and survival when pollinated by other regions, i.e. heterosis is higher when there is among-region gene flow, while Broadway region populations show nearly equivalent germination and survival in among-population and among-region crosses. Thus, the simple dichotomy of comparing the effects of local vs. long-distance gene flow events on fitness is complicated by these region-specific interactions that were detectable by our level of replication at the region level.
It is also interesting that the decomposition of among-region crosses into specific dam and sire regions, shows maternal effects on crosses among New York and Giles region individuals. Thus when the dam is from New York and the sire is from the Giles region, offspring have high fitness, similar to among-population crosses within the New York region. When the dam is from Giles and the sire is from New York, offspring fitness is near the average for among-population, within-region treatments of the two regions. This difference between reciprocal treatments implies an effect such as maternal provisioning or nuclear–cytoplasmic interactions. Because S. vulgaris has nuclear–cytoplasmic sex determination, much is known about the distribution of cytoplasmic genetic variation. New York and Giles regions differ both in the frequencies of common chloroplast and mitochondrial DNA haplotypes and in the presence of region-specific haplotypes (McCauley et al. 2003; D. E. McCauley, unpublished). Bailey & McCauley (2005) found evidence for between-region nuclear–cytoplasmic effects that influenced sex determination. Therefore, nuclear–cytoplasmic interactions, either directly or indirectly through effects on sex determination, may cause this fitness difference. Both cross-by-region effects (Thompson et al. 2004) and maternal effects (Galloway & Etterson 2005) among interpopulation crosses have also been found in other studies of non-threatened, native plant species.
In the F2 self-fertilization treatments, prior among-population (treatment C) and among-region (treatment E) gene flow improved both germination and seedling survival rates as compared to those self-fertilizations in which each pair of grandparents were derived from the same population (treatment A), even though some descendants of crosses between specific regions apparently suffered from outbreeding depression. This effect of gene flow would be expected if F1 heterozygosity is considerably higher in between- than within-population outcrosses. While self-fertilization would always be expected to decrease heterozygosity by 50%, the absolute amount of heterozygosity remaining depends on the amount in each parent. Thus, the fitness advantage derived from gene flow could persist for at least two generations if a reasonable proportion of the hermaphrodites self-fertilize. Emery (2001) found that about 40% of hermaphrodite S. vulgaris self-fertilize in Giles County, Virginia. This advantage would be particularly important if some F1 seeds dispersed and founded new populations. Because self-fertilization is more likely in young populations consisting of only a few individuals (Taylor et al. 1999), any factor that decreases the associated inbreeding depression could lead to higher colonization success. The general question of how the demographic consequences of gene flow are propagated beyond the F1 generation is not well studied, either experimentally or theoretically, and depends, in part, on the nature of epistatic interactions among loci (Tallmon et al. 2004).
Crossing individuals derived from separate populations to representatives of their maternal population did not affect the fitness of F2 individuals as compared to within-population outcrosses. All such treatments produced similar values for all three fitness components and fitness means were close to the mid-values of their F1 and regional parents. Thus, there is no evidence of the epistatic outbreeding depression that might be revealed by such ‘backcrossing’.
In summary, we found evidence of inbreeding depression, due to genetic load within maternal lines, populations and regions and of additive but not epistatic outbreeding depression. These observations are consistent with expectations for an introduced plant a few hundred generations into a range expansion. Repeated sampling of the ancestral genetic pool through colonization processes should cause a random loss of genetic diversity at the level of the local population and, eventually, at the regional level, leading to spatial structuring of those alleles contributing to genetic load. Outbreeding depression in the F1 generation is consistent with other studies of native, non-threatened plant species in which inter- and intrapopulation crosses have been compared (Fenster & Galloway 2000; Montalvo & Ellstrand 2001; Thompson et al. 2004). In contrast, studies of threatened or declining species tend to find evidence for more consistent heterosis effects among first generation interpopulation crosses (Luijten et al. 2002; Tallmon et al. 2004; Vergeer et al. 2004).
It is perhaps surprising that effects of population structure on fitness could be detected in a weedy species, given that large population size and high dispersal rates are often associated with weediness. However, results somewhat similar to ours were found by Richards (2000) and Keller et al. (2000) in studies of Silene alba (=latifolia); a relative of S. vulgaris with a similar weedy life history that was also introduced to North America several hundred years ago. Richards (2000) examined genetic rescue in North American populations of S. alba and found strong evidence of heterosis among F1 generation interpopulation crosses. Keller et al. (2000) made crosses among four European and one North American populations of S. alba and found heterosis among European crosses and outbreeding depression in an intercontinental cross.
If changes in the value of the fitness components measured influence population size or growth rate, our study could aid understanding of how population structure and gene flow influence the demography of weedy populations. For example, gene flow from the Broadway or New York regions into the Giles region enhances survivorship under our experimental conditions. If increased survivorship translates into larger populations or genetic rescue, then conditions that promote long-distance gene flow into the Giles region would increase the invasive potential of S. vulgaris in that region. However, the leap from documenting an increased rate of survivorship following gene flow in the glasshouse to predicting demographic changes in natural populations is a long one.
Because of the size and logistical constraints of this study, we elected to conduct a glasshouse common garden experiment, recognizing the trade-off between statistical control and realism. Therefore, we have measured only intrinsic selection and adaptation to glasshouse conditions, and not external selection due to field conditions. It has been shown that the expression of inbreeding depression can be environment-dependent (e.g. Dudash 1990; Pray et al. 1994), and is often magnified under stressful conditions (Keller & Waller 2002, and references therein). Stressful factors not encountered in the glasshouse include extremes in water availability and temperature, as well as biotic factors such as competitors, herbivores and pathogens. The question is whether all Dam Region × Cross Treatment combinations would react to field conditions equally. Further, in the glasshouse the fitness components that could be measured reliably were related to early survivorship. Seed productivity was not measured. Finally, because each individual was raised in its own container, fitness was not measured in a population context. The fitness of inbred and outbred individuals could be frequency-dependent or density-dependent, as has been shown in some plants (e.g. Schmitt & Ehrhardt 1990; Cheptou & Schoen 2003). In that case relating changes in the survivorship of individuals raised in isolation to proportional changes in the size of populations subject to frequency or density dependent processes is problematic.
Still, it is quite clear that the genetic variation that influences early survivorship is geographically structured in North America, both at a local and a regional scale. Local structure is most likely due to recent stochastic effects associated with the recurrent extinction and recolonization events that are associated with patch dynamics (McCauley et al. 2003). The regional structure could, in some way, reflect historical factors associated with the introduction and spread of the species. The role of range expansion in generating structure could be partly understood by comparing the population structure of S. vulgaris in North America with that in its native range in Europe. Although there have been previous marker-based studies of population structure of S. vulgaris in both Europe (Runyeon & Prentice 1997) and North America (McCauley 1998; Olson & McCauley 2002; McCauley et al. 2003) differences among genetic markers and the geographical scale employed in the different studies have made comparisons difficult. However, it is noteworthy that a recent study by Storchova & Olson (2004) compares genetic diversity and population structure for cytoplasmic genes in North American and European S. vulgaris and finds that North American populations have a lower diversity of mitochondrial alleles within populations. It is impossible to know whether this difference in population structure extends to the nuclear genome and whether it would change the effect of gene flow among populations and regions but it would be interesting to compare the results reported here for North American S. vulgaris to similar crosses for European populations.
The authors would like to thank Jonathan Ertelt for his expertise and assistance in the glasshouse. Financial support was provided by National Science Foundation award DEB-0078531 and United Stated Department of Agriculture award 2002–02213.