在自然干扰情况下，物种多样性和遗传多样性协同变化

Molecular Ecology (2009) 18, 1137–1144 doi: 10.1111/j.1365-294X.2009.04102.x © 2009 The Authors Journal compilation © 2009 Blackwell Publishing Ltd Blackwell Publishing LtdParallel changes in genetic diversity and species diversity following a natural disturbance GUILLAUME EVANNO,*†‡ EMMANUEL CASTELLA,§ CÉLINE ANTOINE,§ GABRIELLE PAILLAT§ and JÉRÔME GOUDET* *Department of Ecology and Evolution, Biophore, University of Lausanne, 1015 Lausanne, Switzerland, †INRA, UMR 985 ESE, Ecology and Health of Ecosystems, 35000 Rennes, France, ‡Agrocampus Ouest, UMR 985, 35000 Rennes, France, §Laboratory of Ecology and Aquatic Biology, University of Geneva, 18 chemin des Clochettes, 1206 Genève, Switzerland Abstract We examined the spatial and temporal variation of species diversity and genetic diversity in a metacommunity comprising 16 species of freshwater gastropods. We monitored species abundance at five localities of the Ain river floodplain in southeastern France, over a period of four years. Using 190 AFLP loci, we monitored the genetic diversity of Radix balthica, one of the most abundant gastropod species of the metacommunity, twice during that period. An exceptionally intense drought occurred during the last two years and differentially affected the study sites. This allowed us to test the effect of natural disturbances on changes in both genetic and species diversity. Overall, local (alpha) diversity declined as reflected by lower values of gene diversity HS and evenness. In parallel, the among-sites (beta) diversity increased at both the genetic (FST) and species (FSTC) levels. These results suggest that disturbances can lead to similar changes in genetic and community structure through the combined effects of selective and neutral processes. Keywords: beta diversity, biodiversity, disturbance, floodplain, gastropods, genetic structure Received 21 July 2008; revision revised 5 December 2008; accepted 17 December 2008 Introduction Understanding how variation is generated and maintained is a central question in both ecology and evolutionary biology. Genetic and species diversities are two fundamental measures of biodiversity. These measures have been mainly studied separately, and it is only recently that comparisons between these two levels of biodiversity have been developed (review in Vellend & Geber 2005). Antonovics (1976) already recognized that the patterns of species and genetic diversity share a similar determinism (see also Hu et al. 2006). Random drift, migration and selection/competition influence the evolution of allele frequencies in populations as well as the identity and the abundance of species in communities (Aarssen 1983; Hubbell 2001; Vellend 2005). Mutation influences genetic diversity, and speciation influences species diversity, but both factors affect populations and communities at a larger timescale than the three other forces. If migration, drift and selection act in parallel on genetic and species diversities, a positive relationship between these two levels is expected, henceforth called the species-genetic diversity correlation (SGDC, Vellend 2004). Reviewing studies documenting a SGDC, Vellend & Geber (2005) found that a positive correla- tion was the most frequent pattern, although the sign of the SGDC could also be negative (Karlin et al. 1984), and one study did not find any significant SGDC (Odat et al. 2004). Results from two recent studies suggested that a positive SGDC is likely to arise in habitats undergoing a disturbance, affecting species diversity and genetic diversity (Vellend 2004; Cleary et al. 2006). At a local scale, disturbances (e.g. fire, drought, floods etc.) may decrease population size and thus genetic diversity due to random genetic drift (e.g. Otto & Whitlock 1997). Similarly, a decrease of the total number of individuals in a local community (i.e. community size) could produce a diminution of species diversity via com- munity drift (Orrock & Fletcher 2005; Vellend 2005). Alter- natively, selective pressures imposed by the disturbance may also lead to the diminution of genetic diversity and species diversity through the differential survival of certain genotypes or species, respectively (Vellend 2005). As a Correspondence: Guillaume Evanno, Fax: +33 2 23 48 54 40; E-mail: guillaume.evanno@rennes.inra.fr 1138 G . E VA N N O E T A L . © 2009 The Authors Journal compilation © 2009 Blackwell Publishing Ltd result, local (also called ‘alpha’) species and genetic diver- sities can be expected to decrease following a disturbance (Vellend 2004; Cleary et al. 2006). The among-sites component of diversity (i.e., ‘beta’ diversity) can also be influenced by a disturbance. Environmental filtering (i.e. counter- selection) of species vulnerable to the disturbance in every site might lead to an increased similarity across sites, hence to a decreased beta diversity (Chase 2007). Such an effect may also influence genetic diversity at selected loci. In contrast, genetic and community drift will tend to lower the similarity among populations and communities, potentially leading to a higher beta diversity following the disturbance (Vellend 2004). These theoretical predictions suggest that drift and selection may have different effects on beta diversity. Studies based on experimental communities found a lower species beta diversity among disturbed communities consistent with an environmental filtering of species vulnerable to the distur- bance (Chase 2007; Jiang & Patel 2008). In contrast, for both genetic and species diversity, Vellend (2004) found a higher structure in secondary forests than in primary forests of central New York State (USA). This latter result is consistent with a major role for community drift. As a result, it seems that a pattern of increased beta diversity is more likely to be revealed under natural conditions than under experimental settings. An additional explanation to Vellend’s finding may be that in natural communities, the intensity of a disturbance can vary among sites. Drift and selection may alter species and genetic diversities in heavily disturbed sites, while the patterns of diversity could be more stable in localities undergoing a lower degree of disturbance. Such a spatial heterogeneity in the intensity of disturbances could then contribute to the overall dissimilarity among sites. So far, this hypothesis has not been investigated in studies measuring a SGDC. In addition, no study has yet undertaken a temporal approach to measure genetic and species beta diversities both before and after a disturbance. Here, we investigate the outcome of a natural disturbance on species and genetic diversities in a metacommunity of freshwater gastropods living in a floodplain habitat located in southeastern France. We used species abundance data collected in 1999, 2002 and 2003 to describe the changes in species diversity in five water bodies. Genetic data collected in 2001 and 2003 on Radix balthica, an ubiquitous species of the metacommunity, were used to document the recent changes in genetic diversity. An extended period of drought occurred from 2002 to 2003 (Luterbacher et al. 2004; Mouthon & Daufresne 2006) and differentially affected the five study sites. We took advantage of this drought to test whether genetic diversity and species diversity respond similarly to an exceptional natural disturbance. Based on previous empirical and theoretical studies of the effects of disturbances on genetic and species diversity (Vellend 2004; Orrock & Fletcher 2005; Cleary et al. 2006; Chase 2007), we made three predictions: (i) both genetic and species alpha diversity should be lower after the disturbance; (ii) genetic beta diversity inferred from neutral markers should increase following the diminution of local population size-enhancing random genetic drift; and (iii) species beta diversity should increase due to community drift and/or heterogeneous effects of the disturbance among sites. Materials and methods Study sites and community surveys The five study sites are floodplain pools located in cut-off meanders of the Ain river, southeastern France (see Figure S1, Supporting information). PL is located in the former meander called ‘Le Planet’, while BX1 and BX2 and PN1 and PN2 are situated in the meanders called ‘Les Brotteaux’ and ‘Puits-Novet’, respectively (Figure S1 and Table S1, Supporting information). In order to quantify the intensity of the disturbance, we estimated the number of days during which each site dried out (see Appendix S1, Supporting information, for a description of the methods used). Drought-frequency data revealed that between 1998 and 2003 BX1, PL and PN2 almost never completely dried out whereas BX2 and PN1 dried out for 30% of that period (Table S2, Supporting information). However, drought frequency was not equal across the years, and from April 2002 to March 2003, BX2 and PN1 dried out 72% longer than during the period October 2000–September 2001. These results clearly show that from 2002 to 2003, an exceptional drought period affected our study area (see also Mouthon & Daufresne 2006). This exceptional drought period was observed all over Western Europe (see for instance Chuine et al. 2004 or Luterbacher et al. 2004). Four dates were kept, at which the species diversity of (almost) all five sites was surveyed: September 1999, Sep- tember 2002, and April and September 2003. Three to 10 quadrats (50 × 50 cm) were taken in each pool (Antoine 2002), and the aquatic vegetation and the upper sediment were thoroughly sampled with a hand net (500 μm mesh). The material collected was kept in 70° alcohol and sorted under binocular in the laboratory. Gastropods were identified and counted, empty shells were not included. We used the nomenclature of Falkner et al. (2001). Molecular analyses For the genetic analyses, we used Radix balthica (L. 1758), a lymnaeid gastropod that was the only species present in all five water bodies over the study period. Snails were collected in September 2001 and April 2003. Between 17 and 30 individuals were sampled in each site at each time, and a foot sample was used for DNA extraction (Evanno et al. 2006). Genetic diversity analyses were performed using C H A N G E S I N GE N E T I C A N D S P E C I E S D I V E R S I T Y 1139 © 2009 The Authors Journal compilation © 2009 Blackwell Publishing Ltd the amplified fragment length polymorphism (AFLP) method. The AFLP procedure was carried out on a total of 247 individuals using two selective primer combinations: E-ACT × M-CTG and E-AGG × M-CAT, which produced 109 and 81 polymorphic bands, respectively (see Evanno et al. 2006 for a detailed description of the protocol). Electrophoreses were run on an automated sequencer ABI 377, fragments being visualized using genescan 3.1.2. (Applied Biosystems) and scored with binthere (Garnhart and Kocher, University of New Hampshire) according to Evanno et al. (2006). The AFLP data set is available online (Appendix S2, Supporting information). Data analyses To estimate genetic alpha diversity, we computed the gene diversity HS (Nei 1987) for each population and sampling date using the software hickory 1.0 (option ‘full model’, Holsinger et al. 2002). Species alpha diversity was estimated by the evenness (E, similar to HS), and by the species richness (SR, equivalent to allelic richness for codominant markers). HS is the probability that two randomly chosen alleles at a locus in a population are different (Nei 1987). Likewise, the evenness [equal to 1 − λ, λ being the Simpson (1949) concentration] here refers to the probability that two randomly chosen individuals in a locality are from different species. By considering the community as a single locus and species as alleles at this locus, evenness was calculated using fstat 2.9.3 (Goudet 1995). fstat is primarily designed to analyze diploid data but is also suitable for haploid data (as explained in the help file) and, thus, species-abundance data. To compare species diversity across the years, we estimated rarefied species richness by re-sampling 10 000 times 10 individuals (the lowest sample size) in each site using ecosim (Gotelli & Entsminger 2004). Importantly, the estimators we used (E, HS and rarefied SR) are all robust to sampling artefact and thus to the unequal number of samples collected across sites and years. Genetic structure across sites (or genetic beta diversity) was estimated by FST, the fraction of total genetic variance attributable to differences among populations. It is usually computed in an analysis-of-variance framework following Weir & Cockerham (1984) as: where , and are the components of variance of allele frequencies among populations, between individual- sand within individuals, respectively. To account for the dominant nature of the markers we used, FST was estimated by θB, a Bayesian equivalent of FST implemented in hickory (option ‘full model’: simultaneous estimation of f and θB) and by ΦST using arlequin 2.0 (Schneider et al. 2000). Credible intervals of 95% were computed for Bayesian estimates (Holsinger et al. 2002), and ΦST values were tested by 10 000 permutations in an analysis-of-molecular-variance (amova) framework (Stewart & Excoffier 1996). Spatial pairwise ΦST were also computed between sites for the two sampling dates. Similarly, we estimated the community structure (or species beta diversity) by computing FSTC, the exact equivalent of FST. FSTC refers to the proportion of total species diversity due to differences among communities and is computed as: where and are the components of variance of species frequencies among localities and among individuals within localities, respectively ( since the community is considered as a haploid locus). Compared to genetic data, species data are often characterized by large differences in samples size across localities. For instance in 2003, although using the same sampling scheme, we found 41 individuals in BX2 and 1902 in PL (Table S3, Supporting information). For the genetic data, sample sizes were much more homo- geneous since their range varied between 17 and 30 individuals. FST is based on a ratio of variance components estimated from a hierarchical analysis of variance of allele frequencies, and thus it weights samples according to their size (Weir & Cockerham 1984). For species data, the ratio of the largest to the smallest sample is 79. Variance components estimated from such unbalanced samples would give an enormous weight to the largest compared to the smallest sample. In order to give a similar weight to the different samples we used a rarefaction procedure: FSTC was repeatedly estimated from data sets consisting of 24 individuals (the lowest sample size from September 1999 and April 2003) sampled without replacement from each sample. The procedure was repeated 1000 times. For this rarefaction analysis, FSTC estimates were calculated using the hierfstat package (Goudet 2005) for R (R Development Core Team 2006). We did not apply the rarefaction procedure to genetic data since FST is unlikely to be biased by small differences in sample size (Weir & Cockerham 1984). Overall and pairwise FSTC were only computed for Sep- tember 1999 and April 2003 in order to compare the results with those obtained from genetic data (collected in September 2001 and April 2003). Importantly, we used the statistic FSTC instead of standard measures of species beta diversity like Sorensen’s or Jaccard’s indices because we needed identical statistics for genetic and species data. To test whether genetic (respectively species) alpha diversity decreased after the disturbance we used pairwise t-tests to compare site-specific HS (respectively E) computed in September 2001 (or September 1999) and April 2003. To test for an increase in beta diversity, we compared: (i) 95% credible intervals computed in hickory for pre- and post-disturbance global FST; and (ii) 95% confidence intervals calculated for F a a b w ST σ σ σ σ 2 2 2 2+ + σa 2 σb 2 σw 2 F a a b STC = + σ σ σ 2 2 2 σa 2 σb 2 σw 2 0= 1140 G . E VA N N O E T A L . © 2009 The Authors Journal compilation © 2009 Blackwell Publishing Ltd the pre- and post-disturbance global FSTC from the rarefac- tion analysis. To test specifically for an increase in species beta diversity, we used the distributions of FSTC generated by randomizations above. We generated the distribution of the differences in FSTC after and before the disturbance and reasoned that under the null hypothesis of no difference in FSTC, this difference should not differ from zero. Under the alternative hypothesis of an increased beta diversity after the disturbance (one-sided test), this difference in FSTC should be positive, and very few values from the distribution should be negative. We thus estimated the probability that the two FSTC are equal as the proportion of differences less than or equal to zero. Results Species and genetic alpha diversity We identified a total of 5228 individuals from 16 species, among which three were only present in 1999 and three others only in 2003 (see Table S3, Supporting information). Species richness over all sites was 13 in 1999 and 2003 (Table S3, Supporting information). Evenness averaged over all sites was 0.75 in 1999 and 0.44 in 2003. This decline is significant (t = 2.2, d.f. = 4, P < 0.05) and stronger in BX2 and PN1 than in BX1, PL and PN2 (Fig. 1a). Rarefied SR decreased from 1999 to 2003 in BX1, BX2 and PN1, whereas it remained stable in PN2 and PL (Fig. 2). A similar decline was observed in gene diversity: computed over 190 polymorphic AFLP loci, HS averaged over popu- lations was 0.315 (Bayesian CI: 0.309–0.321) in 2001 and 0.202 (Bayesian CI: 0.198–0.206) in 2003. This significant reduction of gene diversity (t = 5.8, d.f. = 4, P < 0.01) was stronger in BX1, BX2 and PN1 than in PN2 and PL (Fig. 1b). Species and genetic beta diversity The decrease in species alpha-diversity was accompanied by a parallel increase in species beta diversity. Overall FSTC was 0.13 (95% CI: 0.07–0.18) in September 1999 and 0.23 Fig. 1 Evenness (a) and gene diversity ±95% Bayesian CI (b) calculated for each site before (white) and after the disturbance (grey). FSTC calculated as the mean of pairwise FSTC for the focal site vs. all other localities (c), FST calculated similarly for each site (d). C H A N G E S I N GE N E T I C A N D S P E C I E S D I V E R S I T Y 1141 © 2009 The Authors Journal compilation © 2009 Blackwell Publishing Ltd in April 2003 (95% CI: 0.17–0.30), the increase being significant between the two years (P = 0.007, one sided test based on 1000 randomizations). FSTC for each site (calculated as the mean of all pairwise FSTC for the focal population) are given in Fig. 1(c) and show the increase of global community structure from 1999 to 2003 with FSTC only decreasing slightly in PN2. Similarly, genetic beta diversity significantly increased after the disturbance with overall θB and ΦST being 0.065 (Bayesian CI: 0.055–0.076) and 0.069 (P < 10–5) in September 2001, and 0.169 (Bayesian CI: 0.148–0.192) and 0.160 (P < 10–5) in April 2003, respectively (see also Table S4, Sup- porting information). FST for each site (calculated as the mean of pairwise ΦST of the focal population) revealed a strong increase in population structure from 2001 to 2003 (Fig. 1d). Discussion The striking result of this study is the parallel change in species and genetic alpha and beta diversities over a short time period during which a natural disturbance occurred. In 1999, species alpha diversity was high and evenly distributed across sites, hence beta diversity was low. In 2001, genetic alpha diversity was also relatively high and evenly distributed among Radix balthica populations. In 2003, the situation had drastically changed: following an extended period of drought that started in