Limited gene flow may enhance adaptation to local optima in isolated populations of the Roesel’s bush cricket (Metrioptera roeselii)

Authors


Anna Cassel-Lundhagen, Department of Ecology, Swedish University of Agricultural Sciences, Box 7044, SE-750 07 Uppsala, Sweden. Tel.: +46 (0)18 67 23 70; fax: +46 (0)18 67 28 90; e-mail: Anna.Lundhagen@ekol.slu.se

Abstract

Variation in morphological traits along latitudinal gradients often manifests as size clines. In insects, both positive and negative correlations are seen, and the mechanism behind the response is unclear. We studied variation in seven morphological traits of Roesel’s bush cricket, Metrioptera roeselii, sampled from seven latitude-matched-pair populations that were either geographically isolated from or connected to the species continuous distribution range. The aim was to examine whether morphological traits differed between isolated and continuous populations, and whether latitudinal variation was apparent. The data were used to indicate whether variation in trait means originates from plastic responses to the environment or genetic adaptation to local conditions. To evaluate the influence of gene flow on trait means, we analysed the genetic variation in seven microsatellites. Data showed that individuals from isolated populations display a positive relationship between latitude and body size, whereas individuals from continuous populations show little or no such relationship. The combined morphological and genetic data suggest that the isolated populations have adapted to local optima, while gene flow between continuous populations appears to counteract this process.

Introduction

Variation in individual body size over a species’ distribution range is a well-known phenomenon (Atkinson et al., 1994; Whitman, 2008; Terribile et al., 2009) and often manifests as size clines along latitudinal gradients (James, 1970; Atkinson & Sibly, 1997; Blackburn et al., 1999; Blanckenhorn & Demont, 2004). A positive correlation between size and latitude in endotherms, known as the Bergmann’s rule (Blackburn et al., 1999), is thought to be an adaptation to colder climates through an increased body size to body surface ratio. In ectotherms such as insects, relationships between environmental factors and body size vary between species (Blanckenhorn & Demont, 2004; Whitman, 2008), and it is not clear whether latitudinal variation in morphology is predominantly determined by developmental constraints working on the individual directly or through adaptations to the local environment (see discussions in van Voorhies, 1996; Partridge & Coyne, 1997; Chown & Gaston, 2010). A positive correlation between latitude and body size is, for example, suggested to be: (1) a consequence of developmental processes that cause insect cells to grow larger at lower temperatures (van der Have & de Jong, 1996; van Voorhies, 1996) and/or (2) an adaptation, with larger individuals having higher fitness in colder climates (Cushman et al., 1993; Hassall et al., 2006). Negative correlations also exist between latitude and body size in insects, and these are suggested to result from a reduction in growth period with increased latitude (e.g. Mousseau, 1997; Ciplak et al., 2008; Whitman, 2008; Winterhalter & Mousseau, 2008). Nonlinear or ‘saw-tooth’ relationships between latitude and body size are also possible and can result from complex trade-offs between growth and reproduction (for more details, see Roff, 1980; Nylin & Svärd, 1991).

In addition to variation in local selection pressures, genetic drift and gene flow may vary between regions and populations (Whitlock, 2001; Besold et al., 2008) affecting specific traits (Kirkpatrick & Barton, 1997; Bridle & Vines, 2007; Cassel-Lundhagen et al., 2009). At a species’ range limit, stochasticity may also be pronounced and population densities more varied (Vucetich & Waite, 2003), and all these factors can counteract adaptations to local conditions (Bridle & Vines, 2007). Gene flow from areas of denser populations into regions with lower densities can hinder selection through swamping effects (Garcia-Ramos & Kirkpatrick, 1997; Lenormand, 2002; Alleaume-Benharira et al., 2006; Kawecki, 2008; Rice & Papadopoulos, 2009; Bridle et al., 2010), although such effects may be counteracted by the increase in genetic variance that enables a population to match local conditions (Barton, 2001). Thus, for a trait to evolve to its local optima, strong selection pressures or weak selection in combination with isolation and conserved genetic variance is required (Kirkpatrick & Barton, 1997).

Despite the large number of theoretical models describing how different processes will interact and lead to varying phenotypes along a species distribution range, their relative contributions and interaction effects are still not resolved (Kawecki & Ebert, 2004; Bridle & Vines, 2007; Garant et al., 2007). One approach to uncovering the underlying processes responsible for morphology-latitude relationships is to use a study system where the focal species is distributed in a biogeographic region that contains populations connected to the species core distribution area in addition to isolated (ISO) populations. This type of distribution, with known local population histories, is hard to find, which explains the lack of such studies (Chown & Gaston, 2010).

One species ideally suited to such a study is Roesel’s bush cricket, Metrioptera roeselii (Hagenbach, 1822). Morphological traits in species similar to this have moderate to high heritabilities (Begin & Roff, 2004; De Block et al., 2008) that often show latitudinal trends (reviewed by Whitman, 2008). Roesel’s bush cricket is also a strong disperser (Hochkirch & Damerau, 2009), and it has a nonuniform distribution in Europe. From southern to north-eastern Europe, the distribution is continuous, although the majority of Scandinavia and Denmark contains only a few ISO populations. This distribution allows the study of latitudinal gradients and gene flow on morphological variation by comparing ISO and continuous (CON) populations which can be matched for latitude (see Fig. 1). Thus, our aim was to: (1) measure latitudinal trends in morphological size traits for this species and (2) evaluate whether latitudinal trends differed with the level of gene flow by comparing trait sizes of ISO populations against those from populations which are continuous within the core distribution. Our prediction was that if variation in body size is caused purely by developmental processes (e.g. van Voorhies, 1996), then we expect no differences in latitudinal size clines between the ISO and CON populations. In contrast, if body-size variation was owing to adaptation to the local climate (e.g. Hassall et al., 2006), then we expect levels of gene flow to influence latitude-morphology relationships. Thus, we predicted that ISO populations should show stronger morphological responses to latitude because of local genetic adaptation, whereas the CON populations should show lower levels of response because of genetic swamping.

Figure 1.

 Sample sites of Metrioptera roeselii within a study area around the Baltic Sea coast (explanation of abbreviations in Table 1). The grey-shaded area represents a region where the species is widespread and occurs continuously. Within the white areas, the species only occurs in isolated populations.

Materials and methods

The species

Roesel’s bush cricket is 12–18 mm in length and is easily identified by the males’ characteristic stridulation (Bellmann, 1985). Usually, about 1% of the individuals in a population are long-winged, although during rapid colonization events up to 100% of the local population may be macropterous (Vickery, 1965). The preferred habitat of M. roeselii is moist and ungrazed tall grass areas where it feeds on plant matter and small insects (Marshall & Haes, 1988). In continental Europe, the species is common and shows a CON distribution (Harz, 1957; Ingrisch, 1986; de Jong & Kindvall, 1991). At its northern distribution range, in countries surrounding the Baltic Sea, its distribution pattern also includes ISO populations; Sweden – five ISO populations of varying sizes situated from the southern tip of Sweden to the northern coast (http://www.artportalen.se, de Jong & Kindvall, 1991; Ahlén, 1995); Denmark – three known ISO populations (Bavnhoj, 1996); Finland and Baltic states – common and widespread with ISO populations on the Baltic Sea islands (e.g. de Jong & Kindvall, 1991); Germany – common in the central and southern part of the country, whereas absent in the northern-western part (Maas et al., 2002).

Sampling

We collected 24 short-winged individuals from each of 14 populations (total 241 males, 95 females) using hand nets during the reproductive season in August and September 2008 (Table 1) and stored them in 99.7% ethanol. In total, nine long-winged individuals were found (2.7%) and only one of those was observed in the ISO populations (0.6%). We sampled two populations from seven different latitudes between N 54.40°–60.43°, with the two populations at each latitude representing an ISO or CON distribution (Fig. 1; Table 1). A population was considered ISO if: (1) it was separated by at least 100 km of land or 50 km of sea from any other population of M. roeselii (well beyond natural dispersal distance of this species) and (2) had been established for more than 10 years to allow local selection pressures to act (Roff, 1997; Carroll et al., 2007). Published records of collected specimens and information from experts regarding its presence and absence indicate that seven such populations occur beyond the species CON distribution limit in northern Europe, and we were able to collect material from all of these. To verify that climatic conditions did not differ between the two sample groups (ISO vs. CON populations at the same latitude), we examined the average monthly temperature and variation in precipitation from May until July during the study year in all sites. This is the critical time when nymphs develop into adults, and these climate measures have been shown to correlate with insect body sizes along latitudinal gradients (i.e. Stillwell et al., 2007; Stillwell & Fox, 2009). There was no difference in either of the two measures between the matched groups along the same latitude (Wilcoxon signed-rank test for temperature –Z = 1.18, P = 0.237 and precipitation –Z = 1.01, P = 0.310; n = 7; Fig. 2).

Table 1.   Basic information about samples of Metrioptera roeselii from sites along the Baltic Sea coast. The populations were either isolated or part of the species continuous distribution, and situated at the species northern range margin.
CodeSite (country)LatitudeNNDate†
  1. = number of sampled individuals size, †Date = collection day.

Continuous populations
 KAAKaarina (FI)60.4313119 August
 TALTallin (EE)59.5223114 August
 VIRVirtsu (EE)58.5717714 August
 LIELiepãja (LV)56.5419318 August
 KLAKlaipeda (LT)55.6923119 August
 KAUKaunas (LT)54.8520219 August
 SLAStawno (PL)54.4015921 August
Isolated populations
 ALAÅland (FI)60.26121231 August
 VASVästerås (SE)59.59121210 Sept.
 SAASaaremaa (EE)58.3722115 August
 THYThyholm (DK)56.6521328 August
SMYSmygehuk (SE)55.34141030 August
 RIBRibe (DK)55.29131027 August
 MARMarielyst (DK)54.6517729 August
Figure 2.

 (a) Average monthly temperatures and (b) the coefficient of variation (CV = SD/mean*100) in amount of precipitations from May to July 2008 at the sampled sites from continuous (open circles) and isolated (full circles) parts of the Metrioptera roeselii distribution.

DNA isolation and genotyping

In order to verify that our geographic definition of isolation also reflected significant genetic distance, we genotyped the 24 individuals from each population using seven microsatellite loci; Metroe05, Metroe07, Metroe08, Metroe19, Metroe20, Metroe24 and Metroe27 (Kaňuch et al., 2010). DNA template (concentration 20–40 ng μL−1) was extracted from muscles of the femur using Chelex 100 (Walsh et al., 1991). Protocols for multiplex PCR reactions were used as described in Kaňuch et al. (2010). Fluorescent-labelled PCR products were separated by capillary electrophoresis in an ABI3730XL Genetic Analyser, and electropherograms were edited in Peak Scanner Software v. 1.0 (Applied Biosystems).

Morphological measurements

To get a good estimate of the individuals’ body size, we measured six morphological traits; body weight (hereafter weight), femur length (femur), head width (head), wing length (wing), pronotum length (pronotum) and male’s cercus/female’s ovipositor length (cercus/ovipositor). Traits were measured using a magnifying lamp (magnification 3×) and a digital slide calliper (accuracy ± 0.03 mm); for legs, the left leg was measured if present, otherwise the right one was used. Body weight was measured on equally wet individuals (stored in ethanol before the weighing) using an electronic balance (± 0.1 mg), and all traits were measured by the same person. Weight was not measured for a small number of individuals because they were missing a leg; thus, these individuals were dropped from analyses where weight was used. All individuals were kept in room temperature and measured within 30 days of collection.

Statistical analyses

We used multilevel/general linear mixed models (GLMM) in MLwiN (version 2.17, Rasbash et al., 2009) to examine the relationship between morphology, sex, latitude and population isolation. The data were structured so that individuals were grouped within the 14 sampled populations, with these sampling sites being included as a random factor in all models. Because male and female Roesel’s bush crickets differ in body size (Bellmann, 1985), sex was included as a fixed factor in all models. Also, because populations were visited on different days, we included the sampling date in all models to account for any size bias which was not already corrected for by the random term. For each of the six morphology measures, we initially compared five models to assess the relative influence of latitude and population isolation on each measure while accounting for sampling site, sex and date of collection; (1) base model (sex and collection date); (2) base + isolation; (3) base + latitude; (4) base + isolation + latitude; (5) base + isolation + latitude + isolation*latitude. To allow for the possibility that males and females respond differently to the effects of latitude and isolation, we then compared these models with models that also included combinations of interactions terms with sex (i.e. sex*latitude, sex*isolation and sex*latitude*isolation). These models were compared and ranked using the Akaike information criterion corrected for sample size (AICc) with parameter estimates for model predictions derived from model averaging based on AIC model weights (Burnham & Anderson, 2002).

Expected and observed heterozygosities as well as allelic richness (mean number of alleles per locus) were calculated using GENETIX v4.05.2 (Belkhir et al., 2001). To test if the sampled populations were in Hardy–Weinberg equilibrium, overall degree of heterozygosity was calculated using score U-tests for each population and a multisample score test for all loci as implemented in GENEPOP v. 4.0.10 (Rousset, 2008). We tested for null alleles, effects of stuttering and large allele dropout using Micro-Checker v. 2.2.3 (van Oosterhout et al., 2004). Genetic differentiation among sampled populations was quantified by calculating FST in GENEPOP v. 4.0.10. Because null alleles are known to overestimate the genetic differentiation among populations, FST estimates were also corrected for possible biases using the so-called ENA method by FreeNA software (Chapuis & Estoup, 2007). We tested for significant correlation between genetic and geographic distance for each FST estimator (i.e. FSTENA) by regressing pairwise estimates of FST/(1 −FST) against ln distance between the populations in a Mantel test (Mantel, 1967; Rousset, 1997). Average differences in genetic estimates of CON and ISO populations were compared by nonparametric Mann–Whitney U-tests. G-statistics were used to test for pairwise FST estimates (Goudet, 1995) adjusting P-values using strict Bonferroni corrections (computed in Fstat 2.9.3.2).

Results

Male and female bush crickets from ISO populations were generally larger than individuals from CON populations at the same latitude: (increase in GLMM trait estimate for individuals from ISO populations: (mean ± SE) femur 0.55 ± 0.24 mm; head 0.18 ± 0.08 mm; thorax 0.24 ± 0.07 mm; cercus/ovipositor 0.19 ± 0.05 (male)/0.26 ± 0.10 (female); weight 28.6 ± 20.1 mg, Fig. 3). However, this pattern was not found for wing size (0.01 ± 0.15 mm, Fig. 4). Also, bush crickets from higher latitudes tended to be larger than bush crickets from lower latitudes (increase in GLMM trait estimate for each degree of elevation in latitude: (mean ± SE) femur 0.07 ± 0.06 mm; head 0.05 ± 0.01 mm; thorax 0.01 ± 0.01 mm; cercus/ovipositor 0.02 ± 0.01 mm (male)/−0.04 ± 0.02 (female); weight 10.8 ± 4.1 mg. Once again wing size showed no distinct relationship (0.04 ± 0.06 mm). The relationship between latitude and morphology was not the same for ISO and CON populations; five of the six morphological measurements showed strong AIC support for models that included an interaction term between latitude and population isolation (Table 2). This interaction shows that individuals from ISO populations generally display a positive relationship between latitude and body size, whereas individuals from CON populations show little or no relationship between morphology and latitude (Fig. 3). These relationships were generally similar for males and females (improvement in AICc < 1 for models that included sex*latitude or sex*isolation interaction terms) with the exception of cercus/ovipositor and wing (see model-averaged predictions in Fig. 3b). For cercus/ovipositor, there was overwhelming support for models to include interactions term for sex*latitude, sex*isolation and sex*latitude*isolation (improvement in AICc = 39.7); for wing, there was support that sex differences existed between the relationship of wing length, latitude and isolation (improvement in AICc for model including sex*latitude and sex*isolation = 2.91).

Figure 3.

 Mean (± 1 SE) of body and trait sizes for males and females along a latitudinal gradient for continuous (open circles) and isolated (full circles) populations of Metrioptera roeselii: (a) head width, femur length and pronotum length and (b) includes the traits wing length, cercus/ovipositor length and body weight. Model prediction lines are derived from AIC-weighted model-averaged parameter estimates in Table 2 for trait means in isolated (solid line) and continuous (dashed line) populations.

Figure 4.

 Correlation between genetic and geographical distance for continuous (open circles) and isolated (full circles) populations of Metrioptera roeselii. The line indicates a significant effect of isolation-by-distance in the continuous populations (Mantel test, r2 = 0.3, P = 0.021).

Table 2.   Relative support for the effect of latitude (lat) and population isolation (iso; i.e. isolated or continuous) for each of the six body measures used in this study. Results show ΔAICc values (where 0.0, which is in bold, is the best ranked model relative to the other models) and AIC model weights (in parentheses) from GLMMs, where collection site was incorporated as a random factor. For each body measure (response variable), a base model was initially run which included the effect of sex and collection date (see text for more details). Subsequent models included different additive (+) and interaction (*) combinations of lat and iso on this base model. Numbers in italics show the number of fixed factor parameters being estimated (k).
Response variableΔAICc values (and AIC weights) for different model structures
Base (k = 3)+ iso (4)+ lat (4)+ iso + lat (5)+ iso + lat + iso*lat (6)
  1. AICc, Akaike information criterion corrected for sample size; GLMM, general linear mixed model.

Head width21.1 (0)19.8 (0)12.3 (0)11.1 (0)0.0 (1)
Femur length1.1 (0.24)2.8 (0.11)2.0 (0.16)3.8 (0.06)0.0 (0.43)
Pronotum length17.3 (0)14.2 (0)18.8 (0)16.0 (0)0.0 (1)
Body weight9.2 (0)4.9 (0.07)7.2 (0.02)3.4 (0.14)0.0 (0.77)
Wing length3.1 (0.09)3.9 (0.06)0.0 (0.45)1.0 (0.27)2.5 (0.13)
Cercus/ovipositor length2.5 (0.18)4.0 (0.09)4.5 (0.07)6.0 (0.03)0.0 (0.63)

The populations from the CON distribution showed a higher degree of expected heterozygosity (Mann–Whitney U-test; Z = 2.49, P = 0.013) and had higher allelic richness (Z = 2.81, P = 0.005) than the ISO populations (Table 3). The global Hardy–Weinberg test also indicated an overall heterozygote deficit in all populations (P < 0.01). Further, we found that null alleles may be present in some loci causing a slight excess of homozygotes in 1–3 loci per population. The estimated total mean frequency of null alleles was 8% (CON) and 7% (ISO), respectively (Table 3). Because of the presence of null alleles, we used a refined FSTENA estimation that takes null alleles into account. Using a Mantel test, we found a significant isolation-by-distance correlation (r2 = 0.3, P = 0.021) in populations from the species CON distribution (Fig. 4). In the ISO populations, there was no correlation and the higher FSTENA values (Mann–Whitney U-test; Z = 5.16, P < 0.001) indicated no dispersal between the populations (Fig. 4). This was also supported by the pairwise FST estimates. For example, the island populations outside Finland and Estonia were more differentiated from the closest mainland site (ALA vs. KAA, SAA vs. VIR, see Table 1) than mainland populations found greater than three times more distant apart (cf. VIR vs. TAL) (Appendix S1).

Table 3.   Expected (HE) and observed (HO) heterozygosities, allelic richness (AR) and the mean frequencies of null alleles present (NA) in seven microsatellite loci in continuous and isolated populations of Metrioptera roeselii. Significant heterozygote deficiencies are indicated in bold (P < 0.05). Explanations to population abbreviations are found in Table 1.
SiteHEHOARNA
Continuous populations
 SLA0.760.6412.30.08
 KAU0.710.599.10.08
 KLA0.710.569.40.10
 LIE0.720.568.60.09
 VIR0.730.7011.10.04
 TAL0.730.5910.40.09
 KAA0.680.557.00.09
 Mean ± SE0.72 ± 0.010.60 ± 0.029.7 ± 0.70.08 ± 0.01
Isolated populations
 MAR0.720.627.10.07
 RIB0.670.616.00.05
 SMY0.530.354.40.10
 THY0.660.605.40.05
 SAA0.660.527.10.10
 VAS0.630.586.30.04
 ALA0.690.626.90.06
 Mean ± SE0.65 ± 0.020.56 ± 0.046.2 ± 0.40.07 ± 0.01

Discussion

Individuals from ISO populations were generally larger at higher latitudes. In contrast, there was no such correlation for individuals from the CON distribution despite similar climatic and seasonal variation for the two groups. Because a correlation between insect morphology and latitude was only found for the ISO populations, this suggests that the morphology-latitude pattern is linked to the genetic composition of the populations rather than simply being a physiological response to the climate. Thus, the hypothesis that any latitudinal correlation in morphology results from general developmental processes (van Voorhies, 1996) is less likely, as this would be expected to produce similar morphology-latitude patterns for the ISO and CON populations. Such conclusion would not be possible if we had only included one replicate per latitudinal level, as is usually the case when studying climatic gradients (Gienapp et al., 2008). However, because our data are based on wild-caught individuals, the results must still be seen as indicative and need to be followed up by common garden experiments to validate that environmental conditions are not confounding the results (Stillwell, 2010). Such experiments will also allow us to specifically conclude about the mechanisms behind the observed variation. In cases when a species is difficult to rear under controlled conditions, our approach may be the only way to discriminate between genetic and nongenetic influences.

If we assume that there are local optima for trait sizes at certain latitudes (Bolnick & Nosil, 2007), we would expect morphology to vary with latitude in those populations that could respond to this selection pressure. The morphological variation in the ISO populations of our study suggests such an ability (Atkinson et al., 1994; Angilletta & Dunham, 2003). The reason why this did not happen in the populations from the CON distribution is possibly because of recurrent immigration and gene flow from surrounding regions (with different local optima), which genetically swamp these range margin populations (Garcia-Ramos & Kirkpatrick, 1997; Lenormand, 2002; Alleaume-Benharira et al., 2006; Rice & Papadopoulos, 2009; Bridle et al., 2010). This is supported by the genetic data that indicate considerable gene flow between the CON populations, with high genetic diversity over all populations and a weak but apparent isolation-by-distance pattern over a large geographic region. The populations that are characterized as ISO, on the other hand, apparently lack significant gene flow as indicated by random scatter of pairwise FST values and reduced genetic diversity. Thus, one interpretation of our findings is that range margin populations that lack significant gene flow can develop unique adaptations which differ from populations in other parts of a species distribution (Cassel-Lundhagen, 2010).

Our results stand in contrast to other studies that have found strong genetic divergence in adaptive traits despite high levels of gene flow (reviewed in Leinonen et al., 2008). However, the studies in the meta-analysis by Leinonen et al. (2008) may not be completely representative of natural patterns as many study systems are likely to have been chosen because they are located in contrasting environments (Demont et al., 2008) and/or are already known to be phenotypically divergent. On the contrary, theoretical simulations have shown that migration load will significantly affect trait means when population sizes are reduced (Bridle et al., 2010) and fitness gradients are steep (Alleaume-Benharira et al., 2006), as may be expected at species range margins (Vucetich & Waite, 2003).

In addition, traits affected by many unlinked loci of small effects may suffer from gene flow to a larger extent (Kirkpatrick & Barton, 1997; Ronce & Kirkpatrick, 2001) than single locus traits (Kawecki, 2000) and traits controlled by a few dominant loci. This may be the reason why we find a different pattern in wing lengths compared with the other size traits (Table 3; Fig. 4), where wing lengths expressed variation among latitudes but little variation between ISO and CON populations. Recent studies suggest that wing size is mainly controlled by a few genes (McKechnie et al., 2010), evolves independently from other body-size traits (David et al., 2006; Hoffmann et al., 2007; McKechnie et al., 2010) and can evolve rapidly when conditions are altered (Huey et al., 2000).

It is also known that orthopteran development is affected by a multitude of environmental factors (Whitman, 2008). For example, the abundance, quality and diversity of food resources (Arnett & Gotelli, 1999; Unsicker et al., 2008; Ho et al., 2010), predation and parasite pressure (Danner & Joern, 2003, 2004; Danyk et al., 2005) as well as level of competition (Ciplak et al., 2008) affect body size. A higher degree of competition and a higher pressure from predators and parasites could potentially make it difficult for individuals in the CON populations to produce local optimal trait sizes if they are present in more diverse communities. Although current data do not allow us to test the influences of such variation, we argue that none of these can be the sole mechanism as we expect variation between sites to result in variation in local averages rather than variation in latitudinal trends. The ISO populations, for example, are found both on the European continent, on islands in the Baltic Sea as well as on the Scandinavian peninsula (Fig. 1) which could potentially lead to variation in species community composition (Gillespie & Roderick, 2002). However, there is no systematic grouping of sites on islands or mainland along the sampled gradient, and we therefore see no reason to suspect that habitat conditions change in such a systematic way that would lead to gradually changing trait means.

The presence of a latitudinal trend in the ISO populations also points against random genetic drift as the driving force. If, for example, random alterations in trait averages would appear at the colonization or establishment phase, no particular pattern related to latitude is expected. Future experimental studies in common garden settings and monitoring of local species communities will enable us to evaluate alternatives to the adaptation hypothesis that we currently favour and exclude any confounding factors.

When vulnerability of populations is discussed, population isolation is often viewed negatively and a threat to the viability of the population (Berggren, 2001; Farwig et al., 2009). The reasoning for this includes reduced immigration that lowers the possibility of rescue effects and increases the risk of detrimental genetic effects (Macdonald & Johnson, 2001; Frankham, 2005). These risks may also be present for the populations examined in this study; however, our findings suggest that there may be other aspects of population isolation that might mitigate the negative isolation effects (Garant et al., 2007). In case that ISO populations can compensate for the lack of gene flow by better adapting to local optima, populations with different degrees of isolation may show a variation in resilience to changing climatic conditions.

Irrespective of the relative influences of the underlying mechanisms, the results from this study highlight the paradox between the observed detrimental effects of genetic bottlenecks and isolation in threatened species (Vrijenhoek, 1994; Willi et al., 2006) and the lack of such problems in invasive species (Frankham, 2004; Perez et al., 2006; Hufbauer, 2008). If ISO populations can evolve faster than those that are affected by continuous gene flow, this could explain why species can become invasive when they are introduced into new and previously unoccupied areas. The system of CON and ISO populations along the same latitudinal gradient that was used in this study is ideal to further explore the influence of climate change and interacting processes on evolutionary responses, an area that lacks empirical data (Gienapp et al., 2008).

Acknowledgments

We thank Andrea Kaňuchová and Frida Holma for assistance in the field and two anonymous reviewers for valuable comments on an earlier version of the manuscript. The study was supported by the Swedish University of Agricultural Sciences.

Ancillary