Contrasting patterns of genetic variation and structure in plant invasions of mountains

Authors


*Correspondence: Jake M. Alexander, E-mail: jake.alexander@env.ethz.ch

ABSTRACT

Aim  To assess the population genetic consequences of the colonization of two species with contrasting mating systems, Solidago canadensis and Lactuca serriola, along altitudinal gradients in both their native and introduced ranges.

Location  Allegheny Mountains, West Virginia and Wallowa Mountains, Oregon, USA; Valais, southern Switzerland.

Methods  Leaf material was collected from populations along altitudinal gradients and genotyped at seven microsatellite loci for each species. Differences in variability between native and introduced areas and in relation to altitude were analysed using linear models. Differences in the genetic, geographical and altitudinal structure of populations between areas were analysed by AMOVA, cluster analysis and Mantel tests.

Results  Genetic variation within and across populations of S. canadensis was significantly reduced, while populations of L. serriola were significantly more variable, in the introduced area. Genetic diversity decreased significantly with altitude for S. canadensis but not L. serriola. Genetic structure of S. canadensis was similar in both areas, and populations were isolated by geographical but not altitudinal distance. By contrast, population structure of L. serriola was much weaker in the introduced area, and populations were not isolated by distance in either area.

Main conclusions Solidago canadensis has experienced a strong genetic bottleneck on introduction to the Valais, but this has not prevented it from colonizing a wide altitudinal range. Variation in neutral markers is therefore not necessarily a good measure for judging the ecological behaviour of a species. By contrast, the greater variability of L. serriola in the introduced area, where it also occurs over a greater altitudinal range, can be explained by increased outcrossing among admixed populations. This suggests that the ecological amplitude of alien species might be enhanced after population admixture in the new range, especially for species with highly structured native populations. However, even genetically depauperate introduced populations can be expected to colonize the same environmental range that they occupy in the native area.

INTRODUCTION

The importance of evolutionary processes during invasions by alien plant species is becoming increasingly recognized (Ellstrand & Schierenbeck, 2000; Lee, 2002). Introduced populations typically pass through genetic bottlenecks, with allelic diversity being lost through genetic drift (Dlugosch & Parker, 2008), although admixture of previously isolated native lineages through multiple introductions can also generate increased diversity in the new range (Ellstrand & Schierenbeck, 2000; Kolbe et al., 2004). The severity of bottlenecks therefore depends on the distribution of genetic variation within the native range, the breeding system of the species and its introduction history (i.e. the quantity and sources of introduced propagules) (Novak & Mack, 2005; Taylor & Keller, 2007). In out-crossing species, most genetic variation is usually partitioned within populations, so that few introductions are needed to sample a genetically diverse inoculum. By contrast, most variation in highly inbreeding species is partitioned between populations and so these are more likely to experience the effects of a genetic bottleneck (Novak & Mack, 2005).

Because the amount of genetic variation within a population determines its ability to respond to selection, a lack of variation could restrict the spread of alien species along environmental gradients, especially given that variation is expected to be eroded by consecutive founder events in small and isolated peripheral populations at the invasion front (Frankham, 1996; Chauvet et al., 2004). It has been suggested that the failure of some species to extend their ecological range in a new area might be due to insufficient genetic variation (cf. Hoffmann & Blows, 1994; Lee, 2002; Dietz & Edwards, 2006). However, introduced populations of many species appear to span similar ecological amplitudes to populations in the native range, suggesting that when species are introduced into new areas their niches are commonly conserved (Wiens & Graham, 2005). For example, in previous studies we have shown that the altitudinal limits and clinal patterns of trait variation along altitudinal gradients of eight Asteraceae species are remarkably similar between their native and introduced ranges, and that these patterns have emerged within c. 100 years of their introduction (Alexander et al. in press a; in press b). Thus, the link between molecular diversity and additive genetic variation determining the ecological amplitude of invading populations along environmental gradients might not be strong (Reed & Frankham, 2001; Dlugosch & Parker, 2008).

Relatively few studies have directly compared population genetic structure (i.e. the partitioning of variance among populations) and variability of alien plants between their native and introduced ranges (see Bossdorf et al., 2005 for a recent review), and even fewer relate these to the spread of alien species along environmental gradients (Maron et al., 2004). Altitudinal gradients are particularly useful in this regard, because factors that might limit or delay the invasion, such as climatic conditions and anthropogenic disturbance/propagule pressure, vary over small geographical scales (Arévalo et al., 2005; Parks et al., 2005; Dietz & Edwards, 2006). The establishment of population genetic structure along altitudinal gradients can result from temporal/physical barriers to gene flow such as commonly exist along the gradient. For example, the altitudinal differentiation among populations of Primula farinosa described by Reisch et al. (2005) was attributed to barriers imposed by forests below the tree-line and delayed phenological development at higher altitudes. Such structure might promote, or result from, adaptation to local site conditions along the gradient (Aradhya et al., 1993; Semagn et al., 2001). However, population structure is typically weaker in the introduced range of invading species because at neutral loci it takes time for populations to differentiate through drift, or for new variation to arise after a severe bottleneck (Bossdorf et al., 2005, but see Marrs et al., 2008). Thus, insights into the extent to which changes in genetic variation and structure might affect the progression of an invasion can be gained by comparing populations along altitudinal gradients in the native and introduced ranges (Bossdorf et al., 2005; Hierro et al., 2005).

Here, we compare levels of genetic variation and population structure for microsatellite loci of two Asteraceae species, Solidago canadensis and Lactuca serriola, in mountainous areas in their native and introduced ranges. We chose two mountain areas (canton Valais, southern Swiss Alps and the Wallowa Mountains, northeastern Oregon, USA) with similar continental climates and an additional area (Allegheny Mountains, West Virginia, USA) with some of the broadest altitudinal gradients from the native range of S. canadensis. The two species differ in their reproductive biology, S. canadensis being an obligate out-crosser while L. serriola is mainly self-pollinated. Our aim was to assess the population genetic consequences of the colonization of these two species along an altitudinal gradient in both their native and introduced ranges. We hypothesized that (1) overall genetic variation would be lower in introduced populations due to bottlenecks exacerbated by isolation and restricted gene flow within mountain systems. We expected that (2) population structure would be weaker in the introduced area, also in relation to altitude, because of insufficient time for differentiation to arise. At the same time, we hypothesized that (3) genetic variation of both native and introduced populations would decrease with altitude due to increased effects of genetic drift towards the upper range margin. Additionally, we asked (4) to what extent these patterns might be affected by differences in the breeding system between the two species. Based on these comparisons, we discuss the potential impact of population genetic changes in the introduced range on the progression of plant invasions along altitudinal gradients.

METHODS

Study species

Solidago canadensis L. (Asteraceae; Canada goldenrod) is a perennial, diploid (2n = 18) forb, producing new stems up to c. 120 cm tall each year from underground rhizomes. It is an obligate out-crosser, flowering between the end of August and October, and can produce in the order of 20,000 readily wind-dispersed achenes per ramet (Dong et al., 2006). It also spreads clonally from rhizomes, and forms dense stands that are able to exclude competing species. It is often found abundantly in old-field habitats, and in its introduced range it grows not only in disturbed ruderal areas but also invades semi-natural vegetation (Weber, 1997).

Solidago canadensis is a native of northeastern North America, where it extends from Virginia/Kentucky (USA) into southern Canada, and from the Atlantic to the Great Plains (Semple & Cook, 2006). Its altitudinal distribution is constrained by the fact that few mountain peaks extend above 1200 m a.s.l. (highest point Mt. Washington 1917 m a.s.l.), although it probably occupies much of this altitudinal range (this study). It was introduced on only a few occasions from either the USA (Maryland/Virginia) or Canada to botanic gardens in Paris and London as an ornamental plant before 1750 (Weber, 1994). It was subsequently distributed among botanic gardens, but only began spreading aggressively around 1850, achieving most of its current distribution in Europe by c. 1950 (Weber, 1994). It has been found up to 1530 m a.s.l. in canton Valais in the southern Swiss Alps (Alexander et al. in press b), one of the present study areas, where it was first recorded in 1948 (Hegi, 1979).

Lactuca serriola L. (Asteraceae; prickly lettuce) is an annual, diploid (2n = 18) forb that flowers and sets seed between July and October. It is a species typical of ruderal plant communities, roadsides and sand dunes, and populations tend to be rather transient (Lebeda et al., 2001; Hooftman et al., 2006). Plants grow up to more than 1 m and produce many hundreds of yellow capitula, each bearing on average 16 wind-borne and readily dispersible achenes (Alexander et al. in press a). Pollination is usually autogamous, although occasional insect pollination and out-crossing have been observed (Hooftman et al., 2006).

Lactuca serriola is native to Eurasia, occurring abundantly across a broad latitudinal range from southern Scandinavia to North Africa, and extending from the Atlantic into Central Asia (Hultén & Fries, 1986). Less is known about the introduction history or sources of L. serriola in North America, although it has probably been introduced many times as a contaminant of seed crops (Mack & Erneberg, 2002), and it is now present in all conterminous US states and even on Hawaii. It has also become more common in Europe in the last 50 years (Lebeda et al., 2001; Hooftman et al., 2006). Settlers probably brought the species to northeast Oregon around the middle of the 19th century, and it spread into the Wallowa Mountains around the turn of the 20th century when mining became important in the region (Pohs, 2000). In Europe it is most abundant below 600 m a.s.l. but has been observed up to 1560 m a.s.l. in the Valais (Swiss Alps; Lebeda et al., 2001 and references therein), though populations and plants above c. 800 m a.s.l. in Europe are usually small and with limited seed set (Lebeda et al., 2001). In the Wallowa Mountains plants have been found up to 1860 m a.s.l. (Alexander et al. in press b).

Study areas

Native populations of S. canadensis were sampled from the Allegheny Mountains (principally in West Virginia, USA, 39°00′ N, 79°00′ W), near one of the putative sources of the European introductions (Fig. 1). These mountains constitute part of the Appalachian range, and experience a humid continental climate (mean monthly temperature (MMT): 3.5–16.5 °C, mean monthly precipitation (MMP): 1035 mm; <http://www.sercc.com>). Source populations for the introduction of L. serriola into North America are unknown, but we sampled its native populations, along with introduced populations of S. canadensis, from the canton Valais (southern Swiss Alps, 46°10′ N, 7°20′ E), near the centre of its native distribution. Introduced populations of L. serriola were sampled from the Wallowa Mountains (eastern Oregon, USA, 45°15′ N, 117°20′ W), which shares a similar continental climate with the Valais (MMT, Valais: –5.7–18.7 °C; <http://www.meteoschweiz.ch>, Wallowa Mountains: –7.2–18.9 °C; <http://www.wrcc.dri.edu>). MMP is somewhat higher in the Valais (1060 mm) than in the Wallowa Mountains (745 mm).

Figure 1.

Topographical maps of (a) the Wallowa Mountains (Oregon, USA), (b) the Allegheny Mountains (West Virginia, USA) and (c) Canton Valais (Switzerland, reproduced by permission of swisstopo (BA071648)), with locations of the sampled populations; Sc, Solidago canadensis; Ls, Lactuca serriola. Note population Sc5-I is located outside the range of the map.

Sampling scheme

We sampled S. canadensis from a total of 20 sites (hereafter referred to as populations) and L. serriola from 12 populations across a broad range of altitude during the summers of 2005 and 2006 (Fig. 1, see Appendix S1 in Supporting Information). Twelve introduced populations of S. canadensis and six native populations of L. serriola were sampled from the Valais. A further eight populations of S. canadensis were collected from its native range in the Allegheny Mountains and six populations of L. serriola were sampled from part of its introduced range in the Wallowa Mountains. From each population, leaf material was collected and dried on silica gel from 6–30 individuals of S. canadensis and 15–24 of L. serriola that were at least 2 m apart. Voucher specimens are lodged in the herbarium of the ETH Zurich.

Microsatellite analysis

DNA of S. canadensis was extracted using a silica gel technique (Elphinstone et al., 2003), and of L. serriola using a modified CTAB procedure. Microsatellite analyses of S. canadensis were performed for seven loci (SS1B, SS4F, SS4G, SS19C, SS19D, SS20E and SS24F) using primers developed for Solidago sempervirens (Wieczorek & Geber, 2002). Forward primers were fluorescently labelled with either NED, FAM or HEX. Polymerase chain reaction (PCR) was performed in a total volume of 10 µL with 2 µL template DNA (5 ng µL−1), 1 × reaction buffer, 1.5 mm MgCl2, 0.2 mm of each dNTP, 0.3 µm of each primer and 0.5 U Amplitaq Gold™ DNA polymerase (Applied Biosystems, Foster City, CA, USA). PCR amplification was performed with an initial amplification of 15 min at 95 °C followed by 30 cycles of annealing temperatures of 53–60 °C, 72 °C and 95 °C for 30 s, with a final annealing step for 1 min and 72 °C for 30 min.

Analyses of L. serriola were performed for seven loci (A001, A004, B101, B104, D106, D108 and E011) using primers designed for L. sativa by van de Wiel et al. (1999). PCR conditions were as above, except that each reaction contained 0.4 µm of each labelled primer, 1.5–3.0 mm MgCl2 and 5 µL of template DNA (1 ng µL−1). PCR amplifications consisted of an initial denaturation at 95 °C for 9 min followed by 30–35 cycles of 95 °C, annealing temperatures of 55–60 °C depending on the locus and 72 °C for 30 s each, with a final elongation at 72 °C for 5 min.

PCR products were denatured at 92 °C for 3 min and separated on an ABI PRISM 3130xl Genetic Analyser (Applied Biosystems) with GeneScan ROX or LIZ as the internal size standard. Individuals were genotyped using the GeneMapper version 3.7 software (Applied Biosystems). Cases where loci failed to amplify after three attempts were scored as null alleles. In total, amplification failed at at least one locus for 20% of S. canadensis and 15% of L. serriola individuals. Null alleles were detected at all 14 loci.

Statistical analysis

Genetic variation within populations, measured as Nei's (1987) gene diversity (HT) and allelic richness (A), as well as the inbreeding coefficient (FIS), was calculated using fstat version 2.9.3.2 (Goudet, 2002). The proportion of unique multilocus genotypes (PGT) was also calculated as a measure of genetic variability within populations. The calculation of allelic richness controls for differences in sample sizes by using rarefaction (El Mousadik & Petit, 1996). Thus allelic richness was adjusted to a minimum population size of four individuals for S. canadensis and 15 individuals for L. serriola. The classic method for estimating the selfing rate (s) from heterozygote deficiencies (FIS) is sensitive to technical artefacts such as null alleles. We therefore calculated a maximum likelihood estimation of s based on the multilocus structure of the data using rmes, with statistical differences between areas tested by likelihood ratio (David et al., 2007). Observed heterozygosity (HO), private alleles within populations and allele frequencies were determined using GenAlEx version 6 (Peakall & Smouse, 2006). Multiple regression models were fitted in R (R Development Core Team, 2006) to investigate how patterns of genetic diversity (A, HT, PGT) and mating system (HO, FIS) of populations changed with altitude in each area.

The hierarchical partitioning of genetic variation within and between populations and areas was assessed using analysis of molecular variance (AMOVA) (Excoffier et al., 1992) using arlequin version 3.01 (Excoffier et al., 2005), with significance tested by 1023 permutations of individuals within populations/areas, and populations within areas.

The significance of deviations from Hardy–Weinberg equilibrium (HWE) was tested for each locus by permuting alleles among individuals within populations in fstat based on 1000 permutations and applying Bonferroni corrections for multiple comparisons (Goudet, 2002). fstat was also used to test for significant genotypic disequilibrium between pairs of loci based on 2100 randomizations of alleles within loci after Bonferroni corrections.

Correlations between Nei's (1972) genetic distance (D) and (1) (log) geographical and (2) (log) altitudinal distances distance between pairs of populations were assessed with ibdws version 3.12 and tested using a Mantel test with 1000 permutations (Rousset, 1997; Jensen et al., 2005). To additionally investigate the structuring of populations in each area, cluster analysis was performed based on Nei's (1972) genetic distance between pairs of populations using the phylip version 3.66 package (Felsenstein, 1989). A total of 1000 bootstrap replicates of the allele frequency data were generated using seqboot and distance matrices for each calculated using gendist. UPGMA unrooted tree topologies were created for all replicates using neighbour, and the consensus tree drawn using consense.

RESULTS

Allelic diversity at the species level

In total we detected 83 alleles at seven loci in S. canadensis (mean 11.9 per locus), 41 of which were present in both the native and the introduced areas. Thirty-six alleles were found only in the native area (Allegheny Mountains), while six were present only in the introduced area (Valais). The shared alleles included the most common in both areas (mean frequencies 0.14 and 0.17 in the native/introduced areas respectively; see Appendix S2 in Supporting Information), and there was a strong correlation between their frequencies in the two areas (r = 0.73, d.f. = 39, P < 0.001). There was a large variation in allele frequencies (range 0.001–0.591, mean, standard deviation (SD); 0.084, 0.136). Ninety-six per cent of the 432 S. canadensis individuals that we analysed were genetically distinct using these markers. Three multilocus genotypes occurred twice, and one three times, in population Sc14-N, one occurred twice in population Sc2-I and one occurred twice in Sc10-I. No multilocus genotype occurred in more than one population.

L. serriola showed a very different pattern. We recorded a total of 87 alleles at seven loci (mean 12.4 per locus), of which 46 were detected in both areas, 18 only in the native area (Valais) and 23 only in the introduced area (Wallowa Mountains). The shared alleles were among the most common overall (mean frequencies 0.13 and 0.11 in the native/introduced areas respectively; see Appendix S2), although there was no significant correlation between their frequencies in the two areas (r = 0.03, d.f. = 44, P > 0.8). Furthermore, some alleles that were very common in one area were absent from the other. Only 47.1% of the 278 individuals analysed were genetically distinct using these markers. Of the 26 multilocus genotypes that occurred more than once, some occurred up to 16 times (mean 5.7 times). Only two multilocus genotypes (one in the Valais, one in the Wallowa Mountains) were found to occur in two populations. More than twice as many multilocus genotypes were detected in the introduced (111) than in the native area (46).

The frequency of null alleles at each locus for S. canadensis was estimated using cervus 3.0 (Kalinowski et al., 2007), and these estimates did not differ between areas (t = 0.36, d.f. = 6, P > 0.7). cervus could not be used for the inbreeding L. serriola. However, for this species the mean frequency across loci of failed amplifications (i.e. homozygous null alleles) was higher in the introduced than the native area (0.056 vs 0.003).

Genetic diversity within populations

Genetic diversity (HT and A) and observed heterozygosity (HO) of S. canadensis were significantly greater within populations in the native area (Table 1). The selfing rate (s) in both the native and the introduced areas was not statistically different from zero (χ2 < 2.89, d.f. = 1, P > 0.09). In multiple regression models, allelic richness and gene diversity both decreased significantly with altitude (r = –0.45, F1,16 = 9.07, P = 0.008 and r = –0.45, F1,16 = 6.68, P = 0.020, respectively) and were significantly greater in the native area, although the relationship between diversity and altitude did not differ between areas (non-significant interactions; Fig. 2, see Appendix S3 in Supporting Information). Deviations from HWE (observed heterozygosity, FIS) did not vary significantly with altitude (see Appendix S3).

Table 1.  Comparison of genetic diversity in native and introduced populations of Solidago canadensis L. and Lactuca serriola L. A, allelic richness; PA, number of private alleles; HO, observed heterozygosity; HT, gene diversity; FIS, inbreeding coefficient; PGT, proportion of unique multilocus genotypes. Significant differences (P < 0.05) are indicated by bold-face type. SD, standard deviation.
 Native rangeIntroduced ranget-test
MeanSDMeanSDP
S. canadensis
 A3.2470.2922.6610.268 < 0.001
 PA0000NA
 HO0.4680.0500.3940.0490.004
 HT0.6190.0440.5370.0580.003
 FIS0.2190.0920.2000.1190.712
 PGT0.9720.0800.9890.0310.581
L. serriola
 A3.0200.8594.9440.6340.001
 PA1.8332.2291.5001.3780.762
 HO0.0200.0160.0580.0330.030
 HT0.4200.1220.6410.1260.011
 FIS0.9500.0370.9120.0390.114
 PGT0.3390.1370.7830.144< 0.001
Figure 2.

Relationship between allelic richness and altitude in populations of (a) Solidago canadensis L. and (b) Lactuca serriola L. in the native (filled circles, solid line) and introduced (open circles, broken line) areas. In both (a) and (b), R2 = 0.66 for the multiple regression models (see Appendix S3).

Populations of L. serriola were significantly more diverse (A, HT) in the introduced area and contained a greater diversity of multilocus genotypes (PGT) and greater HO (Table 1). The selfing rate (s) was significantly greater in the native (estimate (95% CI); 0.952 (0.885, 0.979)) than in the introduced area (0.856 (0.768, 0.904); χ2 = 4.68, d.f. = 1, P < 0.04). Genetic diversity (A, HT, PGT) also tended to decrease with altitude in both areas, although these trends were not statistically significant (Fig. 2, see Appendix S3).

Population genetic structure

Across both areas, most genetic variation of S. canadensis (84%) resided within populations, and relatively little (6%) between mountain areas (Table 2). Variation among populations was low but 1.5% greater in the introduced area (Table 2). A similarly low portion of genetic variation of L. serriola was attributed to differences between areas (6%). However, populations of L. serriola were genetically much more structured than those of S. canadensis (36% of the variation among populations within areas), and this structure was substantially greater in the native than in the introduced area (Table 2).

Table 2.  Results of AMOVA (analysis of molecular variance) showing the distribution of genetic variation (% var.) among areas, populations and individuals. SS, sums of squares; VC, variance components.
SourceS. canadensisL. serriola
d.f.SSVC% var.d.f.SSVC% var.
Both areas
 Among area182.90.158***6.21102.30.172**5.5
 Among pops within areas18235.70.258***10.210538.41.124***35.7
 Within populations8441795.22.127***83.754410061.850***58.8
 Total8632113.82.543 55516473.147 
Native area
 Among populations7107.00.253***10.25372.51.630***52.4
 Within populations406907.92.236***89.9264390.31.479***47.6
  Total4131014.92.489 269762.83.108 
Introduced area
 Among populations11128.70.264***11.55165.90.650***22.8
 Within populations438887.32.026***88.5280616.12.200***77.2
 Total4491016.02.289 285782.02.850 

All loci of S. canadensis apart from 1B departed significantly (P < 0.05) from HWE. All populations departed from HWE through an excess of homozygotes (positive FIS), although this was only significant for nine populations (see Appendix S1). Because the species is an obligate out-crosser, these deviations are most likely explained by the presence of null alleles. Overall, there was no significant difference in the magnitude of the excess homozygosity between areas (Table 1). Genotypic disequilibria were detected between seven pairs of loci involving 19D and 1B. However, excluding linked loci from the AMOVA analyses had only negligible effects on the partitioning of variance (data not shown). All loci of L. serriola deviated significantly (P < 0.001) from HWE, and all populations had an excess of homozygotes (see Appendices S1 and S3). Indeed, heterozygosity was extremely low in both areas (mean HO = 0.04).

Genetic distance and geographical structure

Mean genetic distance (D) between pairs of populations of S. canadensis in the native (mean, SD; 0.259, < 0.001) and introduced (0.252, < 0.001) areas was not significantly different (Student's t-test = 0.36, d.f. = 74, P > 0.7). Populations were significantly isolated by geographical distance in both areas, although this effect was stronger in the Valais (Z = 27.5, rM = 0.66, P < 0.001) than in the native area (Z = 11.8, rM = 0.38, P < 0.05; Fig. 3). The correlation between altitudinal and genetic distances were not significant in either area (native area: Z = 18.2, rM = 0.07, P > 0.3; introduced area: Z = 42.9, rM = 0.15, P > 0.08).

Figure 3.

Relationship between Nei's (1972) genetic distance and geographical/altitudinal distances between pairs of populations, respectively, for Solidago canadensis (a, b) and Lactuca serriola (c, d) in the native (filled circles) and introduced (open circles) areas.

Native and introduced populations of S. canadensis formed distinct clusters in the UPGMA tree, although bootstrap support for this split was weak (43%; Fig. 4a). Geographically close populations in the Valais tended to group together (e.g. populations Sc3-I and Sc8-I, populations Sc11-I and Sc12-I), although some (e.g. populations Sc2-I and Sc10-I) were genetically more distant from their nearest geographical neighbours. In the native range, there was an approximate north-south divide between clusters of populations. Populations Sc13-N, Sc15-N and Sc19-N formed one cluster, and the four most southerly populations another.

Figure 4.

Unrooted trees from UPGMA cluster analysis based on Nei's (1972) genetic distance between populations of (a) Solidago canadensis and (b) Lactuca serriola. Branch lengths are scaled relative to the distance between populations. Bootstrap values based on consensus trees are displayed when a node is supported by greater than 40% of 1000 trees. Populations from the native range are indicated in bold type.

Mean genetic distances between pairs of populations of L. serriola in the native area (mean, SD; 1.639, 0.079) were significantly greater than mean genetic distances between populations in the introduced area (0.712, 0.003; Student's t = 3.23, d.f. = 15, P < 0.01). Neither isolation by geographical distance (native area: Z = 38.1, rM = 0.11, P > 0.3; introduced area: Z = 16.9, rM = 0.07, P > 0.4) nor by altitudinal distance between populations (native area: Z = 66.2, rM = 0.06, P > 0.4; introduced area: Z = 26.9, rM = 0.19, P > 0.2) was statistically significant.

Native and introduced populations of L. serriola were also separated by the UPGMA cluster analysis, although the native population Ls5-N was genetically closer to the introduced populations (Fig. 4b). The clusters within the native area were generally well supported. The central Valais populations Ls1-N, Ls2-N and Ls4-N were united, although strongest support was found for the clustering of the two most geographically distant populations, Ls3-N and Ls6-N. There was very low bootstrap support (< 40%) for groups in the introduced area.

DISCUSSION

Genetic diversity and population structure in the mountain areas

A reduction in allelic diversity when populations pass through a genetic bottleneck, either due to sudden reductions in population size or due to colonization events, is predicted to be a common phenomenon for alien species introduced to a new range (Novak & Mack, 2005). We therefore expected genetic diversity within (Hypothesis 1) and differentiation among (Hypothesis 2) introduced populations of these species to be lower than in the native areas. The lower genetic diversity of populations of S. canadensis in the Valais compared to the native area, and in particular the loss of rare alleles, is in line with Hypothesis 1 and is symptomatic of a founder effect. Because S. canadensis was deliberately introduced on a few occasions as an ornamental species (Weber, 1994), it is likely that the bottleneck we observed in the Valais is typical of the rest of Europe. However, the effect might be exacerbated in the Valais, where gene flow is primarily restricted by the high flanking mountain chains to east-west movements along the Rhône valley. Accordingly, and contradicting Hypothesis 2, the stronger geographical structuring of populations in the Valais (expressed as a slightly higher between-population component of genetic variance, greater bootstrap support for clusters of populations and stronger isolation by distance) might be explained by repeated founder events during colonization of temporary and rather isolated ruderal habitats (Chauvet et al., 2004).

Bottlenecks due to founder effects are more common for inbreeding species, for which most genetic variation is partitioned among rather than within populations (Chaboudez, 1994; Bossdorf et al., 2005; and references therein; Edwards et al., 2006). Our discovery of higher levels of genetic variation in L. serriola in the introduced than in the native area, contradicting Hypothesis 1, is therefore unusual. Since the total number of alleles detected in each area were rather similar, the greater richness of distinct multilocus genotypes in the introduced area must stem largely from the higher recombination observed between individuals (greater heterozygosity and lower selfing rate). Increased out-crossing would result from the conversion of among- to within-population genetic variation by the admixture of multiple source populations with different genetic constitutions, whether from a single or several areas of the native range (e.g. Novak & Mack, 1993; Squirrell et al., 2001; DeWalt & Hamrick, 2004; Genton et al., 2005; Dlugosch & Parker, 2008). Support for admixture is found in the much weaker population structure and greater genetic variation within populations of L. serriola in the introduced area, in line with Hypothesis 2. Similar trends have also been observed for other plant invaders of the American West (Novak & Mack, 1993; Neuffer & Hurka, 1999; Garnatje et al., 2002). This pattern also suggests that the Wallowa Mountain populations stem from a common, richer gene pool of admixed populations that invaded the area once (or several times from areas with similar genetic compositions) and subsequently spread, as has been implicated for other alien species (e.g. Kolbe et al., 2004; Genton et al., 2005). However, this assumes that the strong population structure observed in the Valais is typical for L. serriola across its native range, and so our results should be extrapolated with caution to the wider range of the species.

The presence of null alleles might have affected our results by leading to an underestimation of allelic richness and heterozygosity. For S. canadensis, the mean predicted frequency of null alleles was not significantly different between areas, and so does not affect our interpretation of a bottleneck for this species. For L. serriola, the putative proportion of null alleles was greater in the introduced area, which rather strengthens our finding of multiple introductions and population admixture in this area.

Genetic diversity along altitudinal gradients

Analogous with bottlenecks on introduction to a new range, populations spreading along environmental gradients (e.g. during post-glacial recolonization) also undergo reductions in genetic variation and increased differentiation due to drift (e.g. Chauvet et al., 2004). The same applies to small and isolated populations at the periphery of the core distribution of a species, such as those towards the altitudinal limit (Lesica & Allendorf, 1995; Frankham, 1996). We therefore expected to observe a decrease in genetic diversity with altitude in both the native and the introduced areas (Hypothesis 3), as has been observed for other species expanding along altitudinal gradients (Ohsawa & Ide, 2008 and references therein). Both these phenomena might explain the decline in genetic variation we observed approaching the altitudinal limit of S. canadensis.

Populations of L. serriola did not decrease significantly in genetic variation with altitude in either area. However, levels of variation within individual populations of this highly inbreeding species might be more contingent on the size and diversity of their founding gene pool than on gene flow among neighbouring populations along environmental gradients. This would be especially so given the short-lived ruderal nature of L. serriola populations (Chauvet et al., 2004; Hooftman et al., 2006). The absence of an altitudinal population structure is therefore unsurprising in the case of L. serriola, for which gene flow between populations is very low and relatively independent also of the geographical distance between populations.

Implications for plant invasions of mountains

The close genetic affinities of sometimes quite remote populations indicate the importance of long-distance anthropogenic dispersal, for example by commercial activities, for the distribution of both species in the Valais. As an example, the Täsch population (Sc1-I) of S. canadensis in the Mattertal valley is more closely related to a potential source in the central Valais around Sierre (Sc8-I), than to Visp (Sc9-I) which is at the mouth of the valley. This can account for the rapidity of some invasions that do not rely purely on natural dispersal. It also suggests that propagule pressure is unlikely to be a major limitation to the spread of alien species within mountain regions, at least in anthropogenically disturbed areas.

The lack of differentiation between populations of S. canadensis along the altitudinal gradients in both areas indicates that these populations are not isolated by barriers to gene flow – due for example to altitudinal variation in phenology or pollinator activities – as has been hypothesized for other species (Reisch et al., 2005; Ohsawa & Ide, 2008). The connectivity between high/low altitude populations in the native range suggests that any potentially detrimental effect of gene flow on local adaptation will not prevent S. canadensis from invading its full altitudinal range in areas where it has been introduced (Garant et al., 2007).

Our results clearly demonstrate that plant invasions in mountain areas are not necessarily associated with genetic bottlenecks. Furthermore, the loss of allelic diversity of S. canadensis in the Valais has apparently not affected its ecological amplitude along altitudinal gradients. Indeed, the highest population sampled in the Valais (1414 m a.s.l.) corresponds to the highest elevations at which the species can be found in its native range in eastern North America (see Methods). Neutral and quantitative measures of genetic variation are weakly correlated for several reasons (see Reed & Frankham, 2001), and this demonstrates that variation in neutral markers is not necessarily a good measure for judging the ecological behaviour of a species. Thus, assuming variation in quantitative traits is related to the ecological range of a species (Hoffmann & Blows, 1994), we would not expect reductions in allelic diversity during a bottleneck to necessarily reduce ecological amplitude. Consequently, plant invaders might be expected to colonize the full ecological range found in their native area, even if they experience population bottlenecks (Dlugosch & Parker, 2008).

However, a contrasting scenario is presented by L. serriola. For species such as this with a strong population structure in the native range, the multiple introduction and admixture of previously isolated lineages can increase the genetic variation of introduced populations by the conversion of between- to within-population genetic variation (Taylor & Keller, 2007). Greater recombination within introduced populations could lead to a larger ecological amplitude through hybrid vigour (increased heterozygosity) and by exposing novel genetic combinations to selection (Holt et al., 2005; Heliyanto et al., 2006). The likelihood of observing such an effect is therefore higher for an inbreeding species than an out-crossing species for which most genetic combinations have already been ‘tried and tested’. It is thus intriguing that the upper altitudinal limit of L. serriola in the Wallowa Mountains is approximately 500 m higher than in the Valais (Alexander et al. in press b), despite the close climatic similarity of these areas, and that this is associated with greater genetic variation and reduced differentiation of populations in this area. This suggests that its ecological amplitude might indeed be enhanced as a result of population admixture in the introduced range. To test this hypothesis, detailed studies are required to quantify the link between mating system, genetic variation and the ecological amplitude of alien species along environmental gradients.

ACKNOWLEDGEMENTS

We are much indebted to Claudia Michel and Daniel Schlaepfer for help with laboratory work, and to Chris Kettle, Walter Durka, Alex Widmer and Sophie Karrenberg for discussions during preparation of the manuscript. John Semple kindly determined the S. canadensis specimens. Special thanks are due to Clemens van de Wiel of Plant Research International, Wageningen, for providing the microsatellite primers for L. serriola, and to Catherine Parks and the La Grande Forestry and Range Laboratory for providing support in the Wallowa Mountains. Four anonymous reviewers provided many helpful comments on the manuscript. This work was funded by a grant of the Swiss National Science Foundation to HD.

Ancillary