Volume 100, Issue 2 p. 297-308
Free Access

Phenotypic differentiation in a common garden reflects the phylogeography of a widespread Alpine plant

Eva S. Frei

Corresponding Author

Eva S. Frei

Correspondence author. E-mail: [email protected]Search for more papers by this author
J. F. Scheepens

J. F. Scheepens

Section of Plant Ecology, Institute of Botany, University of Basel, Schönbeinstrasse 6, 4056 Basel, Switzerland

Search for more papers by this author
Georg F. J. Armbruster

Georg F. J. Armbruster

Section of Plant Ecology, Institute of Botany, University of Basel, Schönbeinstrasse 6, 4056 Basel, Switzerland

Search for more papers by this author
Jürg Stöcklin

Jürg Stöcklin

Section of Plant Ecology, Institute of Botany, University of Basel, Schönbeinstrasse 6, 4056 Basel, Switzerland

Search for more papers by this author
First published: 13 October 2011
Citations: 26

Summary

1. Glacial history has affected the phylogeographic structure of numerous Alpine plant species, but its impact on phenotypic differentiation has been little studied. Therefore, we asked whether phenotypic differentiation in a common garden reflects the phylogeographic structure of the widespread Alpine plant Geum reptans L.

2. We combined a molecular investigation with a common garden experiment and investigated genets from 16 populations of G. reptans sampled from the European Alps. Using neutral molecular markers (RAPDs) and Bayesian cluster analysis, we analysed the species’ genetic differentiation and phylogeographic structure. In the common garden, we measured the differentiation of phenotypic traits related to growth, reproduction and leaf morphology.

3. Molecular analysis partitioned the populations into three genetic groups, indicating pronounced phylogeographic structure. Regional molecular variation was correlated with regional phenotypic differentiation.

4. Quantitative trait differentiation (QST) differed from neutral molecular differentiation (GST) in 10 of 11 traits, indicating that selection has contributed to phenotypic differentiation. Significant negative correlations between biomass and precipitation records for site of origin are a further indication of adaptation.

5. Synthesis. The current study compared regional molecular variation and phenotypic differentiation among populations of a widespread species in the context of extreme range changes during glaciations in the Alps. Because the common garden phenotypic differentiation of G. reptans reflects its phylogeographic structure, we conclude that glacial history affected both genotypes and phenotypes. The results suggest that the present-day phenotypic differentiation was caused by genetic drift and limited gene flow between populations in glacial refugia and during post-glacial recolonization, as well as by adaptation to current climatic conditions. Our findings are relevant for understanding the adaptive potential of Alpine plants and predicting potential range shifts in response to future climate change.

Introduction

Our understanding of the historical background of regional differences within plant species has increased substantially since the development of molecular methods (Sunnucks 2000). In the last decade, particular attention has been paid to the way in which the genetic make-up of species has been impacted by the cyclical and extreme range changes that occurred during climatic oscillations and glaciations in the Quaternary (Hewitt 1996, 2000; Taberlet et al. 1998). In the European Alps, phylogeographic studies have demonstrated that the effects of population survival in different glacial refugia outside the Alps are still detectable in regional molecular differentiation in numerous Alpine plant species (Schönswetter et al. 2005; Alvarez et al. 2009). Genetic drift and gene flow through pollen and seed dispersal are considered to be the main opposing evolutionary forces responsible for neutral molecular differentiation in plants (Till-Bottraud & Gaudeul 2002). Therefore, neutral drift and subsequent limited gene flow during glacial survival and recolonization of the Alps may have led to the observed phylogeographic differentiation between populations of Alpine species (Schönswetter et al. 2005). Molecular phylogeographic differentiation is detectable between present-day populations of Alpine plants for two reasons. First, the period of time since the last glaciation (c. 10 000 years) is too short for different phylogeographic lineages to have been obscured by interbreeding. Second, current gene flow is too weak to have completely masked historical effects because it is limited by dispersal barriers, such as the deep valleys and high mountain chains characteristic of the Alps (Körner 2003).

It can be hypothesized that glacial history has also had an impact on the phenotypic differentiation of Alpine species. In contrast to random genetic drift, which leads to neutral differentiation (Nei, Maruyama & Chakraborty 1975), selection leads to adaptive differentiation that maintains or enhances fitness in different environments (Kawecki & Ebert 2004). Adaptive regional differentiation of phenotypic traits has been well documented in widespread plant species (Joshi et al. 2001; Olsson & Ågren 2002; Becker et al. 2006) and can be expected in Alpine species, first because Alpine plants may have experienced historical selection caused by local conditions in glacial refugia outside the Alps (Hewitt 1996), and secondly, because selection caused by current environmental conditions may result from both regional climatic differences over the Alpine belt and local spatial and temporal heterogeneity of distinct Alpine habitats (Till-Bottraud & Gaudeul 2002). Therefore, both neutral processes occurring during glaciations and post-glacial recolonization and historical and current adaptive processes may have affected phenotypic differentiation in widespread Alpine species.

Whereas selection leading to adaptation is a long-term process requiring many generations, phenotypic plasticity allows more rapid adjustment to environmental variation at a more fine-grained scale (Sultan 2000). Phenotypic plasticity complements phenotypic differentiation and is considered a genetic trait in itself (Schlichting & Smith 2002). Regional variation in adaptive plasticity has been observed in lowland species (e.g. Berg, Becker & Matthies 2005) and might be particularly pronounced in Alpine species, because their habitats are subject to wide environmental variation (Gonzalo-Turpin & Hazard 2009).

Most of the phylogeographic studies that have investigated the impact of glacial history on within-species differentiation have used only neutral molecular markers (for a review, see Schönswetter et al. 2005). The present study is among the first to directly compare molecular differentiation to phenotypic differentiation in the phylogeographic context of Alpine glaciations (but see Lagercrantz & Ryman 1990). Here, we investigated whether the phylogeographic structure inferred from putatively neutral molecular markers (RAPDs) is reflected in differentiation of phenotypic traits related to growth, reproduction and leaf morphology. We analysed genets of the widespread Alpine species Geum reptans. Genets were sampled from 16 populations across the species’ range in the Alps (Fig. 1) and used in a common garden experiment, as well as for molecular analysis. Common garden experiments are a powerful tool for revealing genetic differentiation in the phenotypic traits of plants from different regions and populations (e.g. Olsson & Ågren 2002). When a treatment such as competition is included in a common garden experiment, phenotypic plasticity can be measured as variation in the response to this treatment (Pluess & Stöcklin 2005). Therefore, we grew clonal progeny of G. reptans in a common garden with and without the alpine grass Poa alpina L., to test for genetic differentiation in competitiveness. Common garden experiments are not suitable for distinguishing between neutral evolutionary forces (such as drift) and adaptive processes as the causes of phenotypic variation (Kawecki & Ebert 2004). Nevertheless, correlations between traits measured in a common garden and environmental variables at the original sites of the populations may suggest adaptation (Linhart & Grant 1996), and comparisons of QST and FST may indicate whether phenotypic trait differentiation is affected by selection (Merilä & Crnokrak 2001). Neutral molecular differentiation (FST) is a measure of background genetic drift, and any deviation in quantitative trait differentiation (QST) from FST indicates selection (Spitze 1993).

Details are in the caption following the image

Locations (dots) of populations of Geum reptans sampled from the European Alps. Dark grey areas show where the species occurs and white areas where it is absent. Light grey lines represent the borders of the Alpine countries. Map modified from Aeschimann et al. (2004).

We addressed the following questions in our study. (i) Is the regional phylogeographic structure of G. reptans inferred from putatively neutral molecular markers correlated with common garden phenotypic differentiation? (ii) If present, is the regional differentiation in molecular markers and phenotypic traits more pronounced than population differentiation within regions? (iii) Are there indications that, in addition to neutral genetic drift, adaptive processes have affected phenotypic differentiation? (iv) Is there regional variation in competitiveness?

Materials and methods

Study Species

The distribution of G. reptans encompasses the entire European Alps extending eastward to the Carpathians and the mountains of northern Albania and Bulgaria (Conert et al. 1995). The species occurs predominantly on moraines in glacier forelands, moist screes and mountain ridges of siliceous bedrock (Aeschimann et al. 2004). Geum reptans is one of the first pioneers on virgin soils after glacier retreat and persists until competition with other species becomes too strong (Weppler & Stöcklin 2005). The plant reproduces sexually, by producing 1–5 flowering stems with terminal flower heads, and clonally, by forming new rosettes at the tip of stolons (Pluess & Stöcklin 2005).

Common Garden Experiment

In the late summer of 2007, plant material was collected from 16 populations of G. reptans at different sites (Fig. 1; see also Table S1 in Supporting Information). To obtain representative coverage, we collected samples from an area that spanned all biogeographic regions that are assumed to reflect spatial genetic structure within Alpine plant species (Schönswetter et al. 2005). From each population, we sampled a minimum of four stolons with rosettes (ramets) from 20 genets. Each genet was at least 5 m from others to minimize the risk of resampling genotypes. Rosettes were kept in plastic bags in a refrigerator for 5–12 days until they were planted in separate 10 × 10 × 10 cm3 pots filled with a 1 : 1 mixture of river gravel and potting soil. Pots were placed on tables in a greenhouse, and their distribution on the tables was randomized weekly. We applied the organic insecticide Traunem (Andermatt Biocontrol AG, Grossdietwil, Switzerland) to the plants twice to control infestations of Sciaridae. Four weeks before transplantation, the plants were placed outdoors for acclimatization.

On 19 May 2008, the plants were transferred to a common garden in the Central Alps in Davos (altitude 1532 m). From each of the 16 populations, we planted four ramets of 8–14 genets each in the garden (n = 592 plants). We used a randomized block design with each sampled population represented by an equal number of individuals in each of the four blocks. Two of the four ramets of each genet were surrounded by seedlings of Poa alpina (seeds from the Austrian Alps, Otto Hauenstein Samen, Landquart, Switzerland) to provide interspecific competition and to enable us to measure competitiveness. The plants did not require watering, but the garden was weeded regularly, and Poa alpina was clipped four times to prevent it becoming too competitive.

Initial plant diameter was measured immediately after transplantation. Traits related to growth (number of leaves), reproduction (numbers of reproducing individuals, flowers and stolons) and morphology (length, width and number of leaflets of the longest rosette leaf and specific leaf area, SLA) were measured after two growing seasons in June 2009. On 3 July 2009, plants were harvested, and leaf and root biomass were measured separately after drying at 80 °C for 72 h. For biomass partitioning, we calculated root mass as a percentage of total biomass (sum of leaf and root mass) for each plant. To quantify competitiveness, we followed Snaydon (1991). First, for each genet, we calculated the difference in the average log (biomass) between ramets from plants grown with and without competition. Second, to obtain a relative measure of competitive ability, we subtracted the previously calculated difference from one. Higher relative competitive ability of a genet indicated stronger phenotypic plasticity. As a measure of the relative importance of clonal vs. sexual reproduction, we calculated the clonality, that is, the proportion of stolons on all reproductive meristems (flowers and stolons). As an indicator of leaf shape, we calculated the ratio of leaf length to leaf width. To estimate the number of leaflets, we counted all secondary veins branching from the leaf midrib, and as a measure of leaf dissection, the number of leaflets was divided by leaf length. SLA was measured in a subset of plants without competitors (n = 125). Five circular leaf corings with an area of 44 mm2 each were taken from different rosette leaves of an individual plant and dried at 60 °C for 48 h. All leaf corings from an individual plant were weighed together. SLA was then calculated as the fresh leaf area divided by the dry weight of the corings (Cornelissen et al. 2003).

RAPD Fingerprinting

Leaf material from eight genets from each population in the common garden experiment (n = 128) was analysed using RAPD fingerprinting (Williams et al. 1990). DNA extraction from dried leaf material and measurement of the DNA concentration were performed as described in the study by Pluess & Stöcklin (2004). After a pilot study to search for suitable primers, we selected the following five oligos for fingerprinting: X5[CGGTCACTGT], M6[GTGGGCTGAC], OPP17[TGACCCGCCT], OPP8[ACATCGCCCA] and OPP9[GTG GTCCGCA]. RAPD-PCR was performed using self-dissolving Illustra™ puReTaq Ready-To-Go PCR Beads (GE Healthcare, Buckinghamshire, UK). The beads contained 10 mm Tris–HCl buffer, 200 μm dNTPs, 1.5 mm MgCl2, 50 mm KCl and 2.5 U polymerase. In addition, 6 ng of DNA, 25 pmol primers and ddH2O to a final volume of 25 μL were added to each PCR bead. PCR amplifications were always run in the same machine (Mastercycler gradient; Eppendorf, Hamburg, Germany) with the following conditions: 120 s at 94 °C for initial denaturing, followed by 34 cycles of 92 °C for 30 s, 36 °C for 30 s and 72 °C for 90 s, with a final extension step of 72 °C for 300 s. PCR products were separated on 2% agarose gels in 1× Tris-borate-EDTA buffer with 100-bp DNA ladders as size standards. Gels were stained with ethidium bromide.

We scored only clear and distinct bands and tested the repeatability of the banding pattern (the absence or presence of bands) in 15 genets with a second complete RAPD analysis (Weising et al. 2005), which revealed an error rate of 4.6%. For data analysis, both monomorphic and polymorphic bands were taken into account (Nei 1973).

Molecular Analyses

To estimate the genetic diversity within populations, we calculated the expected heterozygosity (He; Nei 1973) for each population using popgene version 1.3 (Yeh et al. 1997). GST, a measure of the genetic differentiation between populations (Nei 1973), was estimated using the same program, and 95% confidence intervals were obtained through jackknifing over populations (Miller 1974).

To investigate the genetic structure of the populations, we used a model-based Bayesian cluster analysis to assign genets to genetic clusters. We used the algorithm for dominant markers (Falush, Stephens & Pritchard 2007) and a standard admixture model with independent allele frequencies (Pritchard, Stephens & Donelly 2000) in the program structure version 2.3. After a burn-in period of 100 000 cycles, 100 000 Markov Chain Monte Carlo simulations were performed for values of K (the number of clusters) ranging from 1 to 10. The ad hoc statistic ΔK was used to identify the most likely number of clusters within the dataset (Evanno, Regnaut & Goudet 2005). Molecular data from four genets from each of two additional populations (STAU and TTAU, see Table S1) from the Eastern Alps were included in the cluster analysis to check the continuity of the easternmost phylogenetic group (Fig. 2). These two additional populations were not included in the common garden experiment. Leaf material for these populations was provided by the IntraBioDiv Consortium (Gugerli et al. 2008).

Details are in the caption following the image

Molecular differentiation of genets from populations of Geum reptans sampled from the Alps for (a) K =3 clusters and for (b) K =4 clusters inferred from Bayesian cluster analysis using the program structure. The different clusters (phylogeographic regions) are represented by different shades of grey. Genets are grouped to populations, which are aligned from the Western (left) to the Eastern Alps (right). Bars indicate the assignment probability (Q) of genets to be a member of one of the clusters. The graph shows the simulation run with the highest likelihood for the posterior distribution (Ln P) of data out of the 20 runs for each K.

To test for isolation by distance (Wright 1946), we correlated pairwise genetic distances (Nei 1978) with the geographic distances between populations using Mantel tests. We did this for populations from all phylogeographic regions and for populations from the Central Alpine region separately in genalex version 6.0 (Peakall & Smouse 2006). Assignment of populations to regions is described in detail later. The significance level of the Mantel correlation coefficient R was obtained after performing 1000 permutations. The partitioning of molecular variance between regions, populations within regions and genets within populations was determined using an amova (Excoffier, Smouse & Quattro 1992). Fixation indices were computed and tested using 1000 permutations for each level of the genetic structure: ΦRT for variation between regions, ΦPR for variation between populations within regions and ΦPT for variation within populations. amova and fixation indices were computed using genalex.

Prior to statistical analysis of the common garden experiment, we assigned the populations to the phylogeographic regions inferred from the Bayesian cluster analysis of molecular data. These regions included the Western Alps, Central Alps and Eastern Alps (Fig. 2a). Each population was assigned to a region when its probability of assignment (Q) to one of the three clusters (K = 3) was higher than 70% in the simulation run that had the highest likelihood for the posterior distribution (Ln P) of data out of the 20 runs. We used three regions because for four regions (K = 4; Fig. 2b), the assignment probabilities were too weak to clearly assign the populations from the Central Alps into two groups. We also did not find a split into two well-separated groups when the Bayesian analysis using structure was repeated for only the Central Alpine populations (results not shown).

Linear Modelling

We used mixed-effects modelling to investigate genetic effects (effects of phylogeographic region, population and genet) and environmental effects (effects of competition) on traits measured in the common garden. To analyse the frequency of reproduction with a binomial error distribution, we fitted generalized linear mixed models (GLMMs) with a logit link function. For all continuous variables with normal error distributions, we fitted linear mixed models (LMMs). In both models, we used restricted maximum likelihood (REML). These models perform better with unbalanced datasets (in this case, an unequal number of populations per region) than classical anovas (McCulloch & Searle 2001). Mixed-effects models were calculated using the function lmer in the r package lme4 (Bates & Maechler 2009). The most complex model included the initial plant diameter as a covariate and the factors Competition and Region, as well as their interaction, as fixed effects. The factors Block, Population (nested in Region) and Genet (nested in Population) were treated as random effects in the model. The covariate was included to account for initial size differences. Block was included as a random effect to account for possible spatial heterogeneity in the common garden. To test for the significance of the fixed effects, conditional F-tests were performed as recommended for mixed-effects models (Faraway 2006). We estimated random effects by calculating their variances and tested the significance of the random effects using likelihood ratio tests following Pinheiro & Bates (2000). We checked all model assumptions using diagnostic plots constructed in the r packages lattice (Sarkar 2009) and asur (Fabbro 2007). Biomass, number of leaves, number of flowers and numbers of stolons were natural logarithm-transformed to conform with model assumptions. Tukey’s HSD post hoc tests were used to test differences between means of trait values for each pair of regions.

To determine how much variation in experimental plants’ traits could be attributed to genetic effects, we used linear models with the factors Region, Population and Genet nested in each other and fitted as random effects. The variances were extracted from the models with the function VarCorr in the r package lme4 (Bates & Maechler 2009), and the corresponding variance components (V) were calculated based on the statistics by Crawley (2007).

To analyse regional variation in phenotypic plasticity, we fitted linear models with the relative competitive ability in terms of growth (i.e. of leaf and root mass). The effects of Region, Population and Genet were nested in each other and tested with anovas. Another anova was run to test for regional differences in genetic diversity (He). Contrast tests using the function mancontr in the r package asur (Fabbro 2007) were performed to examine differences between means of the competitive abilities between regions and between means of the He values between regions.

All statistical analyses described earlier were performed using the statistical language r version 2.10.0 (R Development Core Team 2009).

QSTFST Analysis

To evaluate whether any of the observed phenotypic differentiation was due to selection, we compared the quantitative trait differentiation (QST) of all phenotypic traits measured in the common garden with a neutral molecular differentiation index (GST). In theory, when a trait is differentiated in a neutral manner, QST should equal GST. In contrast, a trait is assumed to have been under selection when QST differs from GST; that is, unifying selection has occurred when QST < GST, and diversifying selection has occurred when QST > GST (Merilä & Crnokrak 2001). We calculated QST according to the formula used by Spitze (1993). Instead of extracting variance components from classical anovas (Spitze 1993), we used a REML approach and calculated mean QST values and 95% confidence intervals through jackknifing over populations (O’Hara & Merilä 2005). To investigate whether QST differed significantly from GST, we checked whether the 95% confidence intervals of the mean QST values overlapped with the GST value. All calculations were performed using r version 2.10.0 (R Development Core Team 2009).

Correlation Analysis

To determine whether observed trait differentiation was related to climate, we performed a Pearson’s correlation analysis in r version 2.10.0 (R Development Core Team 2009) between all phenotypic traits and climatic data from the site of population origin. Climatic data were obtained from the WorldClim database (http://www.worldclim.org), a set of global climate grids with a spatial resolution of 150 arcsec containing monthly climatic data from 1950 to 2000 (Hijmans et al. 2005). From the WorldClim data points surrounding each site, we selected the one that differed least in altitude from the population location. Temperature data were corrected for the difference in altitude by adding or subtracting 0.0055 °C m−1 (Ozenda 1988). We calculated total annual precipitation, annual summer temperature (mean for the months June–August) and mean, minimum and maximum annual temperatures based on the monthly climatic data. We averaged the annual data for each parameter over the last 50 years. Because variables other than temperature can change with elevation, we analysed correlations between the altitude of the site of origin and all phenotypic traits. By correlating climatic data and the altitude of the site of origin with the residuals obtained using anovas on phenotypic traits with Region as fixed effect, we could make a stronger case for adaptation. This is because the region effect could be due to both neutral differentiation and regional adaptation, whereas correlations after the removal of the region effect would indicate Alpine-wide adaptation to local conditions.

Results

Molecular Differentiation and Phylogeographic Structure

A total of 53 different RAPD markers were scored in all investigated genets. Only two of these 53 markers were monomorphic. The genetic diversity in all the studied populations of G. reptans was He = 0.14 ± 0.04 (mean ± SD), with a range of 0.07–0.21. The values of He differed significantly between phylogeographic regions [anova (Region): F2,15 = 11.89, P < 0.001], with the lowest genetic diversity He = 0.08 ± 0.01 (mean ± SD) occurring in the West Alpine populations (Fig. 3a). The average genetic differentiation between populations was GST = 0.395 (95% CI 0.388–0.399).

Details are in the caption following the image

Regional differences in (a) genetic diversity and (b) competitive ability for root mass of genets from 16 populations of Geum reptans from three different phylogeographic regions in the Alps. Genets were grown with and without competition from the grass Poa alpina in a common garden to measure competitiveness (sensu Snaydon 1991). Bars show means + SE. F- and P-values are from anovas. Significance of differences between regions was obtained with contrast tests at the α = 0.05 level and indicated by different letters.

Bayesian cluster analysis of the molecular data resulted in a distinct phylogeographic structure, with three genetic clusters (K = 3) having the best ad hoc statistical fit (ΔK; Figure S1). For K = 3, the populations were grouped into West, Central and East Alpine groups (Fig. 2a). For K = 4, the Central Alpine group was split into two groups, dividing the Central Alpine populations into western and eastern groups, but with a large admixture zone and no clear geographic boundary (Fig. 2b). Pairwise genetic distances ranged from 0.02 to 0.27 and were significantly correlated with geographic distances, both when all populations were included in the Mantel test (R = 0.80, P < 0.01) and when the Central Alpine populations were analysed separately (R = 0.58, P < 0.01; Fig. 4). amova revealed that the molecular differentiation between the three phylogeographic regions explained 36% of the total molecular variation, whereas differentiation between populations within regions accounted for only 10% (Table 1).

Details are in the caption following the image

Correlation of pairwise genetic with geographic distances between 16 populations of Geum reptans from three different phylogeographic regions in the Alps. The graph shows correlation coefficients and P-values from Mantel tests for all populations (filled and open dots) and for Central Alpine populations separately (open dots).

Table 1. Analysis of molecular variance (amova) with data of genets from 16 populations of Geum reptans from three different phylogeographic regions in the Alps
Source of variation d.f. MS Estimated variance Variation (%) Fixation indices
Between regions 2 86.5 2.6 36 ΦRT = 0.36**
Between populations within regions 13 9.9 0.8 10 ΦPR = 0.16**
Between genets within populations 112 3.9 3.9 54 ΦPT = 0.46**
Total 127 100.3 7.3
  • d.f., degrees of freedom; MS, mean squares. Significance of 1000 permutations: **P < 0.01.

Phenotypic Differentiation

Significant regional differentiation was present in all the measured traits related to growth, reproduction and leaf morphology (Table 2). Plants originating from the West Alpine and East Alpine regions had a greater total biomass, leaf and root mass and more rosette leaves than Central Alpine plants (Table 3; Fig. 5a). The biomass allocation to roots was highest in the West Alpine plants and decreased towards the east (Fig. 5b). In contrast, the Central Alpine plants had more flowers and stolons than plants from the other two regions (Fig. 5c). The frequency of plants reproducing either by flowers, stolons or with both reproductive modes increased (Table 3) as the proportion of stolons on all reproductive meristems (Fig. 5d) decreased from the western to the eastern region. The ratio of leaf length to leaf width and the degree of leaf dissection (number of leaflets divided by leaf length) both increased from the western to the eastern region (Fig. 5e, f). Mean SLA was highest in the West Alpine plants (Table 3). Region explained 8.5–27.2% of the variation in growth, 5.7–25.5% of the variation in reproductive traits and 21.6–46.9% of the variation in leaf morphology (Table 4).

Table 2. Summary of mixed-effects model analysis of genetic effects (region, population and genet) and environmental effects (competition) on growth, reproduction and leaf morphology of genets from 16 populations of Geum reptans from three different phylogeographic regions in the Alps
Covariate Competition Region Competition × Region Population Genet
MS F 1 MS F 1 MS F 2 MS F 2 s 2 χ21 s 2 χ21
Growth
 Total biomass 20.0 58.1*** 112 323.8*** 1.4 3.9* 0.2 0.5 0.04 10.0** 0.06 6.0*
 Leaf mass 21.5 56.5*** 152 399.1*** 1.4 3.6* 0.2 0.4 0.05 12.8*** 0.05 4.9*
 Root mass 17.1 51.5*** 58.2 175.6*** 1.9 5.7** 0.3 0.9 0.03 8.0** 0.08 10.5**
 Root mass/total biomass 86.3 2.8 11 007 356.2*** 279 9.0*** 21.0 0.7 8.96 42.0*** 8.73 15.7***
 Number of leaves 8.2 34.9*** 31.5 134.1*** 2.1 8.7*** 0.1 0.4 0.02 4.8* 0.06 12.8***
Reproduction
 Number of flowers + stolons 0.7 2.6 2.3 8.4** 1.7 6.3** 0.4 1.6 0.02 2.8 0.11 14.5***
 Clonality 1353 2.0 32.5 0.1 6696 9.9*** 229 0.3 0.01 0.9 6.66 33.9***
 Frequency of reproduction 630 24.8*** 619 7.0** 591 5.2* 8.0 3.0 0.31 8.7** 2.29 33.1***
Leaf morphology
 Leaflets/length 4.1 16.6*** 7.7 31.3*** 5.7 23.3*** 0.2 0.9 0.03 1.0 0.03 5.0*
 Leaf length/width 10.0 2.4 175 72.2*** 29.0 12.0*** 16.7 6.9** 0.05 1.3 0.32 4.8*
 Specific leaf area 2.2 1.4 14.4 9.1*** 0.07 0.5 0.00 0.0
  • Fixed effects: Mean squares (MS) and F-values (F) are from conditional F-tests. Random effects: Variances (s2) and Chi-square values (χ2) are from likelihood ratio tests. Population is nested in Region, Genet is nested in Population. n = 462 for all traits, with exception of n = 125 for specific leaf area. The covariate is initial plant diameter. Block (random effect) was never significant and is not shown. Significance levels are represented by asterisks: *P < 0.05; **P < 0.01; ***P < 0.001.
Table 3. Means (SE) for traits related to growth, reproduction and leaf morphology of genets from 16 populations of Geum reptans from three different phylogeographic regions in the Alps. Genets were grown with and without competition from the grass Poa alpina in a common garden
Alpine region Competition
West Central East Without With
Growth
 Total biomass (g) 9.29a (0.6) 6.42b (0.3) 11.13a (2.2) 10.96 (0.5) 3.77 (0.2)
 Leaf mass (g) 5.80a (0.4) 4.38b (0.3) 8.10a (1.6) 7.59 (0.3) 2.22 (0.1)
 Root mass (g) 3.49a (0.2) 2.04b (0.1) 3.04ab (0.6) 3.37 (0.1) 1.55 (0.1)
 Root mass/total biomass (%) 41.1a (0.9) 35.7b (0.5) 27.5c (1.6) 31.2 (0.5) 42.3 (0.6)
 Number of leaves 21.4a (0.9) 15.8b (0.6) 32.8c (3.9) 22.9 (0.8) 13.1 (0.6)
Reproduction
 Number of flowers + stolons 1.8a (0.2) 3.0b (0.2) 1.9ab (0.2) 3.0 (0.2) 2.3 (0.1)
 Clonality (%) 47.3a (7.0) 21.1b (2.7) 2.6b (0.6) 25.8 (3.4) 23.5 (3.9)
 Frequency of reproduction (%) 37.0a (4.7) 48.4b (2.7) 54.2ab (10.4) 50.6 (3.2) 41.4 (3.2)
Leaf morphology
 Leaflets/length (cm−1) 1.68a (0.04) 1.90b (0.03) 2.73c (0.13) 1.78 (0.03) 2.02 (0.04)
 Leaf length/width 5.04a (0.16) 5.76b (0.11) 7.45c (0.68) 4.99 (0.09) 6.34 (0.16)
 Specific leaf area (mm2 mg−1) 10.48a (0.3) 9.14b (0.1) 9.64ab (0.5) 9.49 (0.1)
  • Mean values identified by the same letter did not differ significantly from one another at the α = 0.05 level, using Tukey’s HSD post hoc tests.
Details are in the caption following the image

Quantitative trait differentiation in (a, b) growth, (c, d) reproduction and (e, f) leaf morphology of genets from 16 populations of Geum reptans from three different phylogeographic regions in the Alps. Genets were grown with and without competition from the grass Poa alpina in a common garden. Bars show means + SE based on the pooled error variance from anovas. For significance of differences between regions see letters in Table 3.

Table 4. Variance components V (%) of genetic effects (region, population and genet) on growth, reproduction and leaf morphology of genets from 16 populations of Geum reptans from three different phylogeographic regions in the Alps
Region Population Genet
Growth
 Total biomass 10.6 6.5 0.7
 Leaf mass 8.5 7.8 0.1
 Root mass 15.3 4.6 7.9
 Root mass/total biomass 27.2 11.8 0.1
 Number of leaves 24.3 3.2 6.9
Reproduction
 Number of flowers + stolons 7.7 4.0 20.9
 Clonality 25.5 0.0 34.8
 Frequency of reproduction 5.7 2.0 29.4
Leaf morphology
 Leaflets/length 46.9 1.0 3.6
 Leaf length/width 26.5 1.3 2.1
 Specific leaf area 21.6 2.9 0.0

Populations within regions were significantly differentiated for six of the 11 traits investigated (Table 2). The initial plant diameter covariate had a significant influence on several traits related primarily to growth (Table 2). Block had no significant influence on the investigated traits (results not shown). Population and Genet generally explained less variation than did Region, except in the case of reproductive traits (Table 4).

Differentiation in Competitiveness

The competition treatment affected most plant traits significantly. Only clonality was not affected by competition (Table 2). The significant interaction between Competition and Region on the ratio of leaf length to leaf width (Table 2) indicated that the effect of competition on this trait was different in plants from different regions. The effect of competition was found to be negative for most traits but positive for the biomass allocation to roots, the ratio of leaf length to leaf width and the degree of leaf dissection (Table 3; Fig. 5). Strong regional differentiation in competitive ability was observed for root mass (anova (Region): F2,15 = 3.52, P < 0.05). Plants originating from the Central Alps exhibited the highest relative competitive ability for root mass (0.73 ± 0.04, mean ± SE) compared with plants from the other Alpine regions (Fig. 3b). Similar but non-significant trends in competitiveness were observed for leaf mass (results not shown).

Neutral Drift vs. Adaptation

Comparisons of QST and GST indicated that trait differentiation was affected by past selection. QST values differed significantly from GST values, with no overlap between the 95% confidence intervals of QST and GST in any traits, with the exception of the ratio of leaf length to leaf width (Table 5).

Table 5. Estimates of quantitative trait differentiation (QST) in growth, reproduction and leaf morphology, and comparisons of QST with neutral molecular differentiation (GST) of genets from 16 populations of Geum reptans sampled from the Alps and grown in a common garden
Q ST (95% CI) Q ST vs. GST
Growth
 Total biomass 0.471 (0.450–0.492) >
 Leaf mass 0.610 (0.525–0.695) >
 Root mass 0.421 (0.403–0.440) >
 Root mass/total biomass 0.525 (0.498–0.553) >
 Number of leaves 0.299 (0.278–0.321) <
Reproduction
 Number of flowers + stolons 0.141 (0.116–0.166) <
 Clonality 0.091 (0.079–0.104) <
 Frequency of reproduction 0.086 (0.077–0.095) <
Leaf morphology
 Leaflets/length 0.464 (0.421–0.507) >
 Leaf length/width 0.348 (0.311–0.396) =
 Specific leaf area 0.150 (0.140–0.160) <
  • CI, confidence interval. Symbols show whether mean QST values differed significantly from GST (mean 0.395; 95% CI, 0.388–0.399), indicating diversifying selection (>) or unifying selection (<). Neutral drift (=) is indicated when QST was equal to GST.

Significant correlations of traits with climatic data and the altitude of site of origin were found, suggesting adaptation. Total biomass, leaf mass, root mass and the number of leaves were positively correlated with maximum annual temperature, while all of these traits except root mass were negatively correlated with total annual precipitation (Table 6; Fig. 6a). None of the assessed reproductive traits were correlated with any of the environmental variables (Table 6). Regarding leaf morphology, SLA was positively correlated with maximum annual temperature (Fig. 6b). Finally, the biomass allocation to roots was positively correlated with the altitude of site of origin (Fig. 6c). When the effect of region was removed statistically, the positive correlations of traits with temperature and altitude became insignificant, but the negative correlations of total biomass, leaf mass and the number of leaves with total annual precipitation remained significant (see Table S2).

Table 6. Correlations of climatic data and the altitude of site of origin with traits related to growth, reproduction and leaf morphology of genets from 16 populations of Geum reptans sampled from the Alps and grown in a common garden
Prec T mean T min T max T summer Alt
Growth
 Total biomass −0.50* 0.46 0.25 0.60* 0.48 −0.21
 Leaf mass −0.55* 0.46 0.26 0.57* 0.47 −0.31
 Root mass −0.33 0.40 0.24 0.57* 0.42 0.03
 Root mass/total biomass 0.42 −0.14 0.05 −0.10 −0.15 0.50*
 Number of leaves −0.51* 0.46 0.07 0.69** 0.53* −0.37
Reproduction
 Number of flowers + stolons −0.10 0.02 0.14 0.35 −0.22 −0.27
 Clonality −0.09 0.26 0.32 0.29 −0.10 0.24
 Frequency of reproduction −0.43 −0.15 −0.25 −0.11 −0.15 −0.08
Leaf morphology
 Leaflets/length −0.38 0.14 −0.02 0.12 0.16 −0.49
 Leaf length/width −0.01 −0.14 −0.28 −0.10 −0.13 −0.25
 Specific leaf area 0.07 0.44 0.23 0.63** 0.48 −0.05
  • Values show Pearson’s correlation coefficient r. Prec, total annual precipitation; Tmean, Tmin, Tmax, mean, minimum and maximum annual temperatures; Tsummer, annual summer temperature (mean June–August); Alt, altitude. Climatic data are obtained from the WorldClim database (Hijmans et al. 2005) and averaged over the years 1950–2000. Significance levels are represented by asterisks: *P < 0.05; **P < 0.01.
Details are in the caption following the image

Correlations of phenotypic traits of Geum reptans with (a) total annual precipitation, (b) maximum annual temperature and (c) the altitude of site of origin. Genets from 16 populations of G. reptans were sampled from the Alps and grown in a common garden. Climatic data are obtained from the WorldClim database (Hijmans et al. 2005) and averaged over the years 1950–2000. The graphs show correlation coefficients and P-values from a Pearson’s correlation analysis.

Discussion

Molecular Differentiation and Phylogeographic Structure

The average genetic diversity (He = 0.14) within populations of G. reptans was low in comparison with the genetic diversity reported in 20 other widespread plant species (He = 0.22; Nybom 2004). Genetic bottlenecks resulting from small population sizes (Nei, Maruyama & Chakraborty 1975) during glacial survival or during post-glacial recolonization might explain this low overall level of genetic diversity. Moreover, genetic diversity was significantly lower in the West Alpine populations than in the populations from the other two regions (Fig. 3a). The lower genetic diversity of West Alpine populations may be caused by their predominantly clonal mode of reproduction (Fig. 5d). The high genetic diversity of the Central Alpine populations (Fig. 3a) may be explained by the ongoing admixture of two previously separated gene pools.

We detected a high level of genetic differentiation between populations of the perennial G. reptans (GST = 0.40) compared with the average value inferred for short-lived (GST = 0.32) and long-lived (GST = 0.19) perennials from other RAPD studies (Nybom 2004). The high genetic differentiation observed in G. reptans is probably related to its particularly low seed dispersal in the Alpine landscape (Tackenberg & Stöcklin 2008). In addition, regional differentiation (36%, amova; Table 1) was much higher than population differentiation within regions (10%), which concurs with the distinct phylogeographic structure inferred from Bayesian cluster analysis.

The pronounced molecular phylogeographic structure in G. reptans (Fig. 2) fits well with results from previous biogeographic studies of Alpine plants. The two main boundaries that split the Alps into three regions were previously described from floristic data (Merxmüller 1952; Ozenda 1988). A third boundary partitioning the Central Alps into two regions, as previously suggested, for example by Ozenda (1988), was not strongly supported by our findings (Fig. 2b). We thus propose that the Central Alpine group may have originated from an admixture of two originally separated gene pools in the central area of the Alps. Two Central Alpine groups were found in a phylogeographic study addressing G. reptans that included more populations and used AFLP markers (Thiel-Egenter et al. 2009). Our molecular data and the weakly supported previous evidence of a boundary in the Central Alpine region (see Fig. 4b in Thiel-Egenter et al. 2009) suggest that gene flow between the two Central Alpine groups is quite substantial, probably because of an absence of pronounced dispersal barriers, despite the isolation by distance pattern (Fig. 4) within the entire Alpine belt. The observed isolation by distance indicates that gene flow is more common between closer populations.

The distinct west–east structure observed in the siliceous species G. reptans concurs with glacial refugia on siliceous bedrock being longitudinally oriented at the southern and eastern border of the Alps (Alvarez et al. 2009). Therefore, we suggest that the phylogeographic structure and the strong regional differentiation indicated from our molecular analysis is largely a result of genetic drift and limited gene flow during the survival of G. reptans in different glacial refugia. Subsequent weak gene flow between Alpine regions due to dispersal barriers, such as the deep valleys in the Western Alps (Aosta valley; Fig. 1) or the high mountain chains in the Eastern Alps (Grossglockner mountains; Fig. 1), may have contributed to the regional differentiation of this species at a later stage.

Phenotypic Differentiation

The present study is among the first to compare phenotypic population differentiation to neutral molecular differentiation in the context of glacial history in the Alps. The regional structure of G. reptans derived from molecular markers closely parallels the regional phenotypic differentiation observed in this species in the common garden. We detected strong regional differentiation in all the assessed phenotypic traits and for competitiveness (Tables 2 and 3; Fig. 3b). Moreover, the regional phenotypic differentiation was clearly higher than the differentiation between populations within phylogeographic regions (Table 4).

Size differences in plants at the beginning of the experiment influenced several traits significantly (Table 2). The initial size differences might have been a consequence of maternal effects to some degree (Weiner et al. 1997). Because we used initial plant diameter as a covariate, these size differences should not have obscured the effects of the tested genetic factors (region, population and genet) and environmental factors (competition).

Several main patterns of regional phenotypic differentiation were found, including (i) significantly lower vegetative biomass and more flowers and stolons in Central Alpine plants than in plants originating from the Western or Eastern Alps (Fig. 5a,c), (ii) a decrease in the biomass allocation to roots and in clonality from west to east (Fig. 5b,d) and (iii) an increase in the ratio of leaf length to leaf width and in the degree of leaf dissection from west to east (Fig. 5e,f). Regional differentiation in competitiveness was also indicated, as the Central Alpine plants suffered less from competition and displayed a higher competitive ability than did plants from the other two regions (Fig. 3b).

We suggest that a home advantage effect could explain the enhanced reproductive output and the high competitive ability of Central Alpine plants because the common garden was located in the Central Alps. Central Alpine plants experienced conditions similar to those existing in their sites of origin, which may have enabled them to achieve higher fitness compared to plants originating elsewhere. However, a home advantage effect cannot adequately explain the decrease of trait values from east to west or vice versa. Therefore, we conclude that past evolutionary processes, either neutral or adaptive, may have played a role in the regional differentiation of biomass, reproduction and leaf morphology in G. reptans.

Neutral Drift and Glacial History

Our results suggest that strong historical effects, including genetic drift and subsequent limited gene flow during glacial survival and post-glacial recolonization, have affected phenotypic differentiation in G. reptans. Differentiation in the ratio of leaf length to leaf width and in the number of leaflets divided by leaf length was especially strong in East Alpine plants compared to those from other regions (Table 3), and this differentiation in leaf morphology might have been caused either by neutral drift in a separate eastern gene pool or by selection. Neutral evolutionary processes leading to differentiation in quantitative traits have frequently been neglected in previous studies of widespread plant species (e.g. Joshi et al. 2001; Olsson & Ågren 2002). We emphasize the relevance of such neutral processes for present-day phenotypic differentiation and suggest that neutral phenotypic differentiation might be a more general phenomenon in widespread Alpine plants than previously assumed.

Indication of Adaptation

Some of the observed phenotypic differentiation in G. reptans may be explained by historical selection imposed either by environmental conditions during survival in glacial refugia outside the Alps or by current environmental conditions during recolonization of the Alps. Past selection (QST ≠ GST) is indicated for almost all traits (Table 5). Thus, selection has played an important role in shaping phenotypic differentiation in this Alpine species, suggesting a relatively high adaptive potential with respect to growth, reproduction and leaf morphology. Strong phenotypic differentiation resulting from adaptation has also been documented in common lowland species (Joshi et al. 2001; Becker et al. 2006).

To prove that adaptation to particular local conditions has occurred, reciprocal transplantation experiments would be necessary (Kawecki & Ebert 2004). However, adaptation is suggested in our study by the significant correlations of several traits measured in the common garden with climatic data and with the altitude of site of origin (Table 6). For example, biomass and SLA were correlated with climatic variables at the original sites of populations. Reduction in biomass associated with decreasing temperature might be an adaptive strategy of G. reptans to reduce freezing damage at locations with low temperatures, as observed in other Alpine plant species (Körner 2003). The reduced biomass in plants from locations with high total annual precipitation (Fig. 6a), which also includes snowfall, could be an adaptation to extended periods of snow cover and a shortened growing season. The negative correlation between biomass and precipitation remained when the regional effect was removed statistically (see Table S2), emphasizing the importance of precipitation for the local adaptation of traits related to growth. Reduced SLA values (i.e. greater leaf thickness) in plants originating from locations with low temperatures (Fig. 6b) are probably also an adaptation to climatic variation (Scheepens, Frei & Stöcklin 2010). Adaptation to environmental conditions related to altitude is indicated by a significant positive correlation between biomass allocation to roots and the altitude of site of origin (Fig. 6c). An increase in fine root mass with altitude might be related to reduced mycorrhizal infection at high altitudes (Nespiak 1953; Körner & Renhardt 1987). Although comparisons of QST and GST indicated selection for reproduction, none of the reproductive traits were correlated with climatic data or the altitude of site of origin. Therefore, differentiation in reproduction might be partially explained by adaptation to historical conditions or to current environmental conditions that were not measured.

Conclusions

We used a phylogeographic approach to investigate regional phenotypic differentiation in the widespread Alpine plant G. reptans at the scale of the European Alps. We demonstrated that the molecular phylogeographic structure paralleled similar, strong common garden phenotypic differentiation. Our results suggest that historical evolutionary forces (such as neutral drift and limited gene flow during survival in glacial refugia and post-glacial recolonization) have affected differentiation of both the genotypes and phenotypes of G. reptans. Based on comparisons of QST and GST and correlations of trait values from a common garden with precipitation data from site of origin, we conclude that adaptation to climatic differences may at least partially explain the observed phenotypic differentiation. We suggest that the extreme historical range changes that occurred during climatic oscillations and Alpine glaciations in the Quaternary have left their mark in the phenotypic differentiation patterns of widespread plant species. Therefore, the results of this study might be of relevance for estimating the adaptive potential of Alpine species and the consequences of their potential range shifts in response to future climate change.

Acknowledgements

We thank Reinhard Frei, Jean-Nicolas Haas and Marianne Heberlein for help in the field, Serge Aubert for support in collecting plant material in the French Alps, Thomas Hahn for help with GIS and Veronica Preite and Stefanie von Felten for statistical advice. We are grateful to Felix Gugerli and the IntraBioDiv Consortium for DNA samples of East Alpine populations, to Christine and Kai Huovinen for enabling the common garden experiment in Davos and to Martin Heggli, Melissa Dawes, Kristina Giano and two anonymous referees for critical comments on the manuscript. The study was funded by the Swiss National Science Foundation (project no. 3100AO-116785 to J. Stöcklin).