Aquatic Toxicology 95 (2009) 17–26
Contents lists available at ScienceDirect
Aquatic Toxicology journal homepage: www.elsevier.com/locate/aquatox
Historical metal pollution in natural gudgeon populations: Inferences from allozyme, microsatellite and condition factor analysis Dries Knapen a,∗ , Hans De Wolf a , Guy Knaepkens b , Lieven Bervoets a , Marcel Eens b , Ronny Blust a , Erik Verheyen c a b c
University of Antwerp, Department of Biology, Ecophysiology, Biochemistry and Toxicology Research Group, Groenenborgerlaan 171, 2020 Antwerpen, Belgium University of Antwerp, Department of Biology, Ethology Research Group, Universiteitsplein 1, 2610 Wilrijk, Belgium Royal Belgian Institute of Natural Sciences, Department of Vertebrates, Molecular Laboratory, Vautierstraat 29, 1000 Brussel, Belgium
a r t i c l e
i n f o
Article history: Received 10 June 2009 Received in revised form 29 July 2009 Accepted 30 July 2009 Keywords: Gudgeon Gobio gobio Genetic adaptation Microsatellites Allozymes Condition factor Metal pollution
a b s t r a c t This study presents the results of a microsatellite and allozyme analysis on natural populations of the gudgeon (Gobio gobio) located in a pollution gradient of cadmium and zinc. Differences among contaminated and reference populations were observed at 2 allozyme loci, as well as a relationship between the fish condition factor and glucose-6-phosphate dehydrogenase genotypes, the locus that showed the largest difference in allele frequencies. The microsatellite data partly confirmed the differentiation pattern that was revealed by the allozyme survey. Our data further suggest that at least 2 microsatellite loci may be affected by natural selection. We thus illustrate that both microsatellite and allozyme loci do not necessarily behave as selectively neutral markers in polluted populations. Estimates of population differentiation can therefore be significantly different depending on which loci are being studied. Finally, these results are discussed in the light of the conservation unit concept, because microsatellites are often used to assess genetic variation in endangered natural populations and to propose measures for conservation or management. © 2009 Elsevier B.V. All rights reserved.
1. Introduction As environmental pollution resulting from anthropogenic activities grows, it poses a serious threat to organisms and populations, and to the ecosystem in which they live, thus adversely affecting the biodiversity on a global scale. The question of how organisms and natural populations react to, and cope with pollution-related environmental changes has already been the focus of numerous studies during the last decades (e.g. Tanguy et al., 2002; Manosa et al., 2001; Bridges and Semlitsch, 2000). Most physiological and toxicological research studying the effects of metal pollution on aquatic organisms has focused on acute, short-term effects of relatively high metal concentrations, which are not necessarily ecologically relevant. Despite the fact that this approach has yielded important insights in the metal regulation and defense mechanisms available to aquatic organisms (Williams and Frausto da Silva, 1996; Boudou and Ribeyre, 1997), much less data exist on effects of long-term exposure to contaminants, and on the ability of organisms inhabiting chronically contaminated habitats to adapt through the process of natural selection to these
∗ Corresponding author. Tel.: +32 3 265 33 49; fax: +32 3 265 34 97. E-mail address:
[email protected] (D. Knapen). 0166-445X/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.aquatox.2009.07.022
changing environmental conditions, which largely determines their chances of survival and reproduction (Bickham et al., 2000). Although research has shown that toxicant stress can modify genetic characteristics of natural populations (e.g. Mulvey et al., 1995), and that adaptive strategies leading to increased physiological tolerance have the potential to evolve at fast rates in populations experiencing environmental stress (Reznick and Ghalambor, 2001), conclusions are often paradoxical. There is a growing concern that several seemingly inconspicuous, long-term effects of pollutants, such as the alteration of genetic variation, genetic adaptation and their potentially associated fitness costs are currently not being sufficiently addressed (Belfiore and Anderson, 2001; Simonsen et al., 2008). In previous studies we examined the impact of historical metal pollution on natural populations of the gudgeon, Gobio gobio (Knapen et al., 2004, 2007). These populations are situated in a river system that is characterized by a pollution gradient of Cd and Zn, making it an appropriate study area that acts as a multi-generation exposure experiment in the field. The experiments reported in Knapen et al. (2004) revealed a significantly higher cadmium tolerance in specimens that were taken from the contaminated sample area compared to those originating from a reference site. This result was mainly attributed to detected differences in the gene and protein expression of the metal binding protein metallothionein in the livers of the studied specimens. Nevertheless, these results
18
D. Knapen et al. / Aquatic Toxicology 95 (2009) 17–26
did not provide direct evidence for a genetic basis of increased cadmium resistance, although a genetic component was likely to be involved since differences in resistance and metallothionein induction remained, and were even reinforced, after a sustained acclimation period to uncontaminated fresh water (Knapen et al., 2004). The role of metallothionein gene and protein expression was further confirmed when two additional populations from the same area were included in Knapen et al. (2007) which focused on the relationship between accumulated metal concentrations and metallothionein induction. Against this background, this study uses both allozymes and microsatellites to investigate the genetic composition of the same four gudgeon populations by comparing the results from these two types of markers. We also measured the condition factor (based on a length–weight relationship of all fish) as an ecologically relevant fitness parameter to assess the adaptive value of the obtained population genetic patterns. Finally, the results of our study will be discussed in the light of the conservation unit concept (Moritz, 1994; Paetkau, 1999) as microsatellites are often used to assess genetic variation in endangered natural populations and to propose measures for conservation or management of these populations. For this purpose, 10 additional gudgeon populations from different unpolluted watercourses were sampled and analysed using microsatellites. 2. Materials and methods 2.1. Study species The gudgeon, G. gobio, is a cyprinid fish species which occurs all over Europe, except in Scandinavia, Italy and Southern Spain (Gerstmeier and Romig, 2000). It is a bottom living, euryoecious species feeding on benthic invertebrates (Maitland, 2000). The gudgeon lives in streams, rivers and lakes, and it is found both in polluted and non-polluted areas. It is rather resistant to low dissolved
oxygen concentrations, organic pollution and high temperatures (Lelek, 1987). Up to 88% of the gudgeons are bound to a home range of about 100 m (Stott, 1967; Lelek, 1987). As a result, measured body-residues of fish were found to reflect the local presence of environmental contaminants (Van Campenhout et al., 2003), which makes it an ideal species for studying local genetic adaptations in the field. Furthermore, the studied gudgeon populations are not affected by stocking. This assures that local adaptations have not been masked, for example as a result of outbreeding depression or translocations. 2.2. Study area and sampling Four sample sites were selected in a downstream pollution gradient of Cd and Zn, located in the Grote Nete drainage basin (Belgium, see Fig. 1). In two of the sample sites in the Molse Nete river (MNA and MNB), Cd and Zn pollution has become a severe problem during the last decades (Clements et al., 1988; De Cooman et al., 1998). Metal concentrations in the surface water of these sites are thoroughly documented. They fluctuate in yearly cycles, reaching a maximum in early spring (source: Surface Water Measuring Network Database of the Flemish Government, www.vmm.be). This yearly maximum coincides with the reproductive period of the gudgeon, which is therefore likely to cause an intense selection pressure on this species during the period of egg development and larval growth, two of the most crucial stages in a fish’s life cycle. Analyses of the surface water revealed dissolved metal concentrations of up to 30 g/l Cd and 2100 g/l Zn in MNA, and up to 20 g/l Cd and 1500 g/l Zn in MNB, whereas the Belgian and European surface water quality objectives are 1 g/l and 200 g/l for Cd and Zn, respectively (Vlaamse Regering, 2000). More information on the pollution history of these sites (including graphs) can be found in Knapen et al. (2007). Two reference sample sites (MNC and WIA) were chosen: one site was selected several kilometers downstream of the
Fig. 1. Maps of the study area and sample sites. (a) Map of Flanders (Belgium) showing the collection sites of gudgeon samples. See Table 1 for population abbreviations. The rectangle indicates the location of Fig. 1b. (b) Map showing the pollution gradient, including the four sample sites in this area. The reference sites MNC and WIA enclose the contaminated sites MNA and MNB. Pie charts depict allele frequencies of the allozyme locus Gpd*.
D. Knapen et al. / Aquatic Toxicology 95 (2009) 17–26
19
Table 1 Population abbreviations and corresponding streams, geographic coordinates, and drainage basins. N indicates sample size. Abbreviation a,b
MNA MNBa MNCa WIAa , b GN GS GV RD WN ZBB ZWB AB BW M a b
Stream Molse Nete A Molse Nete B Molse Nete C Wimp Grote Nete Groot Schijn Grote Velp Rivierenhof Deurne Witte Nete Zevenborrebeek Zwarte Beek Abeek Berwijn Merkske
Coordinates ◦
◦
51 10 39 N 05 06 18 E 51◦ 09 26 N 05◦ 01 02 E 51◦ 11 22 N 05◦ 07 49 E 51◦ 07 18 N 04◦ 44 24 E 51◦ 08 25 N 05◦ 04 08 E 51◦ 13 23 N 04◦ 34 59 E 50◦ 51 28 N 04◦ 56 54 E 51◦ 12 57 N 04◦ 27 44 E 51◦ 14 01 N 05◦ 05 16 E 50◦ 43 32 N 04◦ 18 53 E 51◦ 05 04 N 05◦ 17 07 E 51◦ 09 36 N 05◦ 42 31 E 50◦ 45 17 N 05◦ 43 24 E 51◦ 25 40 N 04◦ 48 23 E
Basin (Sub-basin)
N
Scheldt (Nete) Scheldt (Nete) Scheldt (Nete) Scheldt (Nete) Scheldt (Nete) Scheldt (Schijn) Scheldt (Demer) Scheldt (Schijn) Scheldt (Nete) Scheldt (Zenne) Scheldt (Demer) Meuse (Meuse) Meuse (Voer) Meuse (Mark)
22 20 20 20 30 30 21 30 30 12 9 29 30 26
Population in which metallothionein gene expression (Knapen et al., 2007) was previously studied. Population in which metal resistance (Knapen et al., 2004) was previously studied.
contaminated sites, another site was located upstream of the source of pollution (Fig. 1b). Both reference sites are part of the same drainage basin, and apart from metal concentrations, physicochemical water quality (pH: 6.8–7.2, BOD: 2.5–3.2 mg O2 /l, hardness: 106–128 mg CaCO3 /l, NO3 − : 0.94–1.7 mg N/l, SO4 2− : 54–62 mg/l) and structural river characteristics are comparable. Furthermore, we sampled a total of 10 additional Flemish (Belgium) gudgeon populations, distributed over Flanders’ two main drainage basins (Fig. 1, Table 1) for microsatellite analysis. Fish were caught by means of electrofishing, using an Electracatch WFC7 generator producing 150 V. At each of the four sites in the pollution gradient, a distance of at least 100 m was sampled until 20 fish were collected. Fish were transported on ice to the laboratory. Liver and muscle tissue were dissected and stored at −80 ◦ C for allozyme analysis. A sample size of 20 specimens was chosen in an effort to reach an acceptable balance between statistical power and adverse effects on natural fish populations that are already under extreme anthropogenic pressure, and of which population size and overall genetic variation were unknown. In all other populations (where tissue sampling was non-invasive), three successive sweeps were performed over a distance of 100 m. Sample size ranged from 9 to 30 (see Table 1 for details on all sampled populations). Tissue samples for DNA extraction were taken from all fish by clipping part of the anal fin. Fin clips were preserved in 100% molecular grade ethanol, and stored at −20 ◦ C prior to DNA extraction. All fish except those used for allozyme analysis were returned in their habitat after sampling. 2.3. Condition factor The study of the condition of fish is based on the analysis of length–weight data and assumes that heavier fish of a given length are in better condition (Bolger and Connolly, 1989; Knaepkens et al., 2002). We used the relative condition factor (K), which was calculated according to Le Cren (1951) and Ricker (1971): K=
w alb
The observed weight of each specimen (w) is compared to its expected weight based on its observed length (l). The expected weight was estimated using a length–weight regression (determined by a and b) of all caught specimens (R2 = 0.96). K indicates whether a specimen is in better (K > 1) or worse (K < 1) condition than an average specimen of the same length. At the population level, the average condition factor allows a quantitative comparison of the condition of two or more populations from different localities. The condition factor of fish is influenced by the develop-
mental stage of the reproductive organs (Lambert and Dutil, 1997). Therefore, all populations in this study were sampled in the same month, and more specifically in November, when the gudgeon is in an inactive reproductive state. 2.4. Allozyme and microsatellite electrophoresis Liver and muscle tissue was homogenized on ice in microcentrifuge tubes with 5 l homogenization buffer (0.2 g/ml sucrose, 0.7 mg/ml dithiothreitol, 0.8 mg/ml ammonium caproic acid, 1.9 mg/ml EDTA) per mg tissue, and subsequently centrifuged for 45 min at 21,000 × g and 4 ◦ C. The supernatant was applied to vertical 6% polyacrylamide gels, which were run for 180 min at 150 V and 4 ◦ C in a continuous tris–citric acid buffer at pH 8.0. Eleven different allozymes were assayed (Table 2) using standard zymographic staining techniques (Harris and Hopkinson, 1976; Richardson et al., 1986). Alleles were numbered according to their electrophoretic mobility from the anodal to the cathodal side. A salt DNA extraction procedure was followed according to Aljanabi and Martinez (1997). We examined microsatellite variation at seven loci: gob3, gob11, gob12, gob15, gob16, gob22 and gob28 (Genbank accession numbers DQ207799, DQ207800, DQ207801, DQ207802, DQ207803, DQ207804, DQ207805, respectively) according to the PCR conditions described in Knapen et al. (2006). Amplification products were scored on a CEQTM 8000 Genetic Analysis System (Beckman Coulter). Allele sizes were determined using a GenomeLabTM DNA Size Standard Kit – 400 (Beckman Coulter) with the CEQTM 8000 Fragment Analysis software (version 8.0.52). Table 2 Enzymes and inferred genetic loci screened electrophoretically in liver and muscle tissue of Gobio gobio, including the number of alleles and genotypes found at each locus. Enzyme
E.C. classification
Locus
Alleles
Genotypes
NADH diaphorase Glucose-6-phosphate isomerase Glucose-6-phosphate dehydrogenase Glutathion reductase Isocitrate dehydrogenase Lactate dehydrogenase Malate dehydrogenase Malic enzyme 6-Phosphogluconate dehydrogenase Phosphoglucomutase Superoxide dismutase
E.C. 1.6.2.2 E.C. 5.3.1.9
Dia* Gpi*
1 1
1 1
E.C. 1.1.1.49
Gpd*
2
3
E.C. 1.6.4.2 E.C. 1.1.1.42 E.C. 1.1.1.27 E.C. 1.1.1.37 E.C. 1.1.1.40 E.C. 1.1.1.44
Gr* Idh* Ldh* Mdh* Me* Pgd*
1 1 1 1 1 3
1 1 1 1 1 6
E.C. 5.4.2.2 E.C. 1.15.1.1
Pgm* Sod*
2 1
3 1
20
D. Knapen et al. / Aquatic Toxicology 95 (2009) 17–26
2.5. Statistical analysis of allozyme and microsatellite data
3. Results
only three were polymorphic (i.e. showed more than 1 allele). At the phosphoglucomutase locus Pgm*, and at the glucose-6-phosphate dehydrogenase locus Gpd*, two alleles were found, while the 6phosphogluconate dehydrogenase locus Pgd* coded for three alleles (Table 2). All invariable, monomorphic loci were excluded for further analysis. Further results are based on a dataset that solely consisted of allele and genotype frequencies of loci Gpd*, Pgd* and Pgm*, unless stated differently. Within each sample site, allozyme genotypes did not show significant deviations from Hardy-Weinberg equilibrium expectations at any of the loci (all p > 0.05). There was no evidence of genotypic linkage disequilibrium at any pair of loci (all p > 0.05). Observed heterozygosities and mean number of alleles did not differ significantly among populations at any of the loci Gpd*, Pgd* and Pgm*. Allele frequencies and Fst-values calculated over all loci differed significantly in every pairwise population comparison which involved a contaminated site and a reference site. However, no genetic differentiation could be demonstrated between the two contaminated sites, nor between the two reference sites (Table 3). This differentiation pattern was further confirmed by a Neighbor Joining analysis based on a Cavalli-Sforza and Edwards distance matrix (Fig. 2a) and a PCA analysis (Fig. 2c). The genetic similarity between both reference sites is remarkable, since they are geographically clearly separated. As a result, there was no significant isolation by distance effect that could explain the observed molecular patterns (r = 0.42, one-tailed p = 0.12). Differences in allele and genotype frequencies were most apparent at locus Gpd*. Allele frequencies of this locus in the four populations are shown in Fig. 1b. Whereas allele 1 is the most abundant allele in the reference populations, it is outnumbered by allele 2 in both contaminated sites (chi-squared test, p = 0.0017). Despite the fact that the upstream reference population MNC is located only 2.3 km from the most contaminated population (MNA), there is a significant genetic differentiation between the populations. No physical migration barriers exist between the sites (personal observation). On the other hand, no differentiation was seen between MNC and WIA, the second reference population located downstream from the pollution gradient. A similar pattern was found at locus Pgd*, but the observed differences were less pronounced. Although 20% more Pgd* heterozygotes were found at MNA, their genotypic distribution did not differ significantly from MNB and MNC. Genotype frequencies of population WIA, on the other hand, did differ significantly from the other three populations due to the presence of a specific allele, which resulted in three unique genotypes. These differences, however, were no longer significant at the Bonferroni-corrected ␣-level. Finally, allele and genotype frequencies of locus Pgm* did not differ among the assayed populations.
3.1. Allozymes in the pollution gradient
3.2. Condition factor
Electrophoretic screening of 11 enzyme systems in 80 G. gobio specimens from four sample sites revealed 11 inferred loci, of which
Fig. 3a shows the average condition factor for each of the four populations in the study area. Populations from the contaminated
Departures from Hardy–Weinberg equilibrium expectations, pairwise linkage disequilibrium among loci and differences in allele and genotype frequencies among populations were investigated using exact probabilities, applying the Markov Chain method using 5000 permutations, 2000 dememorisation steps and 500 batches. These analyses were performed using GENEPOP v3.3 (Raymond and Rousset, 1995). For a direct comparison of the observed and expected heterozygosity and of the mean number of alleles per locus, a correction for differences in sample sizes was calculated by comparing 1000 random cumulative allele counts with Doh (Brzustowski, 2002). Population substructuring was examined by pairwise Fst-values, estimated by theta (), calculated using ARLEQUIN version 2.0 (Schneider et al., 2000). Significance levels of Fst-values were calculated using 5000 permutations. Microsatellite data were bootstrapped 10,000 times using MsatBootstrap 1.1 (http://www.helsinki.fi/science/metapop/Software.htm) before constructing a Neighbor Joining tree using Phylip 3.68 (Felsenstein, 1989, 2004). Population genetic structure was also tested by a hierarchical analysis of molecular variance (AMOVA, Weir and Cockerham, 1984; Excoffier et al., 1992; Weir, 1996), as implemented by ARLEQUIN version 2.0. In this analysis, populations were clustered according to the drainage basin to which they belong (Table 1). PCAGEN (developed by Goudet, http://www.unil.ch/popgen/softwares/) was used to perform a principal component analysis (PCA) on gene frequency data. The inertia of each PCA axis was calculated, and p-values were estimated by 1000 randomizations of genotypes. Mantel tests were performed to test for isolation by distance as implemented by IBDWS 3.02 (Jensen et al., 2005) using 1000 randomizations and Cavalli-Sforza and Edwards genetic distances. Geographic distances were calculated as meters of river length separating the populations, and were obtained using the Mapinfo 8.0 GIS software package. A significance level of 5% was used throughout. Table-wide Bonferroni corrections were used to correct for multiple testing where applicable. Finally, different approaches have been developed to identify loci involved in adaptive population divergence (Storz, 2005). We used BayeScan (v1.0; Foll and Gagiotti, 2008) to identify candidate loci under natural selection. This method uses Bayesian statistics to calculate a posterior probability that each locus is subject to selection. All other regular statistical analyses (ANOVA, Tukey post-hoc tests and Pearson product-moment correlations) were performed in the statistical environment R.
Table 3 p-Values associated with the genic differentiation test (null hypothesis: the allelic distribution is identical across populations) and with the genotypic differentiation test (null hypothesis: the genotypic distribution is identical across populations), and Fst-values for each pair of populations calculated over three variable allozyme loci and six variable microsatellite loci. See Fig. 1 for population abbreviations. Comparison
MNA–MNB MNA–MNC MNB–MNC WIA–MNA WIA–MNB WIA–MNC * **
Genotypic differentiation
Fst
Allozymes
Genic differentiation Microsatellites
Allozymes
Microsatellites
Allozymes
Microsatellites
0.8410 0.0039** 0.0027** 0.0033** 0.0020** 0.2463
0.0608 0.0034** 0.0043** 0.0000** 0.0000** 0.0000**
0.8040 0.0344* 0.0072** 0.0134* 0.0082* (* ) 0.2746
0.1498 0.0085* (* ) 0.0140* 0.0000** 0.0000** 0.0000**
0.00 0.13** 0.16** 0.09** 0.12** 0.00
0.013 0.019* 0.018* 0.071** 0.052** 0.063**
Significant differentiation at ˛ = 0.05. Significant differentiation at ˛ = 0.0083.
D. Knapen et al. / Aquatic Toxicology 95 (2009) 17–26
21
Fig. 2. Comparison of allozyme and microsatellite data. Unrooted Neighbor Joining phylograms based on a Cavalli-Sforza and Edwards distance matrix for the four populations in the pollution gradient based on allozyme (a) and microsatellite (b) data. PCA analysis of the same four populations based on allozyme (c) and microsatellite (d) data.
sites had a higher average condition factor than the reference populations. Surprisingly, MNA, the population from the most contaminated site, exhibited the highest condition factor. Since differences in allozyme allele frequencies were most striking at locus Gpd*, pooled datasets were created including all specimens from populations MNA and MNB (the contaminated sites) on the one hand, and from populations MNC and WIA (the reference sites) on the other hand. Subsequently, the average con-
dition factor was calculated for each Gpd* genotype in each sub-data set. Averaged over all four populations, there was no relationship between Gpd* genotype and condition factor (Fig. 3b). The same was true for the pooled reference populations (Fig. 3d). However, in the pooled populations from the contaminated sites, specimens of the genotype that was most abundant in both contaminated sites (“22”, see Fig. 1b) had a higher condition factor than specimens of the least abundant genotype (“11”, p = 0.02, Fig. 3c).
Fig. 3. Condition factor analysis. Error bars indicate standard errors of the means. Different letters indicate significant differences among sites. (a) Average condition factor of the four populations in the pollution gradient. (b) Average condition factor for each Gpd* genotype over all populations. (c) Average condition factor for each Gpd* genotype in MNA and MNB. Calculations are based on a pooled dataset of both contaminated populations. (d) Average condition factor for each Gpd* genotype in MNC and WIA. Calculations are based on a pooled dataset of both reference populations.
22
D. Knapen et al. / Aquatic Toxicology 95 (2009) 17–26
Observed condition factor patterns were not influenced by age structure as there was no relationship between length and condition factor (r = 0.003). To assess whether osmotic effects in liver resulting from metal exposure contributed to differences in condition factor, metal measurements in liver from Knapen et al. (2004) were re-used and wet weight/dry weight ratios (WDR) were calculated. There was no correlation between metal load in liver tissue and WDR (r = 0.007), and WDR did not differ among populations (p > 0.1). There was no relationship between sex ratio (ranging from 0.5/0.5 to 0.4/0.6 male/female) and the condition factor. 3.3. Microsatellites For locus gob3, we detected a total of 32 alleles in all 14 Flemish populations. Several gob3 amplicons differed by only one nucleotide in length. Simulated random cumulative allele counts showed that on average, one (apparent) new gob3 allele was identified for each additional specimen added to the dataset. Even in populations with a large sample size, the increase in the number of new alleles found as a function of sample size remained linear throughout the entire sample size range. The high variability of this locus, and the relatively small sample size of some populations resulted in Hardy–Weinberg equilibrium deviations in most populations. Locus gob3 was therefore omitted in all further analyses. There were no deviations from Hardy–Weinberg expectations in any of the other six loci in any population, except from two populations (GS and GV), which showed a significant heterozygote deficiency at locus gob12. There was no evidence of genotypic linkage disequilibrium at any pair of loci (all p > 0.1). Genetic variability was high in all populations. Mean observed heterozygosity (Hom ) ranged from 0.481 in population ZWB to 0.661 in population BW. Mean expected heterozygosity (Hem ) ranged from 0.517 (population ZWB) to 0.663 (population MNC). Heterozygosities did not differ among populations (p = 0.9654). The mean number of alleles per locus (Nam ) ranged from 4.0 (population ZWB) to 6.3 (population WN) and did not differ among populations (p = 0.96). Furthermore, Hom , Hem , and Nam did not differ among populations after correcting for sample size (Hom,cor , Hem,cor , Nam,cor ). Allele frequencies over all loci differed significantly between all pairs of populations, except for the population pairs MNA–MNB, MNB–MNC (Table 3) and MNB–GN (data not shown, Table 3 only lists values of the 4 populations in the pollution gradient). Almost all Fst-values, ranging from 0.0052 to 0.1891, were significantly different from 0. As for the above mentioned comparisons, only the MNA–MNB (Table 3) and MNB–GN relationships were confirmed by the analysis of Fst-values. Two loci showed a log(Bayes Factor) between 0.5 and 1 (Fig. 4). According to Jeffreys’ scale of evidence (Jeffreys, 1961) for Bayes factors, a log(BF) between 0 and 0.5 is “barely worth mentioning”, whereas a log(BF) between 0.5 and 1 provides “substantial” evidence, and a log(BF) greater than 1, 1.5 and 2 provides “strong”, “very strong”, and “decisive” evidence respectively for the alternative model (in this case the locus being affected by selection). Locus gob11 showed a log(BF) of 0.73, which indicates that the data favour the selection model over the neutral model at odds of 5.4 to one. For locus gob28, the odds of being affected by selection are 4.3 to one (log(BF) = 0.63). A Neighbor Joining tree (Fig. 5) only partially reflected geographical structuring of the different rivers and drainage basins. Overall however, there was a significant isolation by distance effect (matrix correlation r = 0.97, one-tailed p = 0.038). A hierarchical analysis of molecular variance showed that only 6.4% of the total variation could be attributed to variation among populations within the same drainage basin, and 0% of the total variation could be explained
Fig. 4. BayeScan analysis of six microsatellite loci. Plot of log(Bayes Factor) versus Fst for all six microsatellite loci. According to Jeffreys’ (1961) scale of evidence for Bayes factors, a log(BF) between 0 and 0.5 is “barely worth mentioning”, whereas a log(BF) between 0.5 and 1 provides “substantial” evidence, and a log(BF) greater than 1, 1.5 and 2 provides “strong”, “very strong”, and “decisive” evidence respectively for the alternative model (in this case the locus being affected by selection).
by differences among drainage basins. 93.6% of the total genetic variation was due to variation within populations (Table 4). As for the four populations located in the pollution gradient, the microsatellite analysis showed a similar – but slightly different – pattern compared to the allozyme analysis (Fig. 2b and d). Both contaminated populations (MNA and MNB) were not significantly differentiated, and both populations were significantly differentiated from the populations from both reference sites. However, in contrast to the allozyme data, microsatellite analysis revealed a
Fig. 5. Phylogenetic analysis of all sampled populations. Unrooted Neigbor Joining phylogram based on a Goldstein’s distance matrix for all sampled gudgeon populations based on six variable microsatellite loci. Values on branches indicate bootstrap values that were higher than 50%.
D. Knapen et al. / Aquatic Toxicology 95 (2009) 17–26 Table 4 Hierarchical Analysis of Molecular Variance (AMOVA) of 14 Flemish gudgeon populations from 2 drainage basins (considered as groups). Source of variation
Percentage of variation
Among groups Among populations within groups Within populations
0.00 6.37 93.63
significant difference between the reference populations MNC and WIA. Table 3 compares differentiation tests and Fst-values. 4. Discussion 4.1. Microsatellites versus allozymes The variation in Fst among loci and types of markers is considered to be a powerful method for examining whether natural selection is playing a major role in the amount of genetic divergence among populations (McDonald, 1994). If some types of markers are more strongly affected by selection than others, we would expect to find differences in the distribution of Fst for different marker types. Under selective neutrality, all loci will be merely affected by the demographic properties of the populations. However, loci under natural selection may have a higher or smaller Fst, depending upon the mode of selection. Specifically, directional selection, which could explain the observed allozyme patterns, decreases genetic variation and increases Fst-estimates (Dhuyvetter et al., 2004; van Straalen and Timmermans, 2002). For example, Lemaire et al. (2000) concluded that six (of 28 allozyme loci tested) were nonneutral as they found higher Fst-values compared to microsatellites in European sea bass populations. In our study, Fst-values calculated based on allozyme data among the three populations in the Molse Nete river (MNA, MNB and MNC) are about ten times higher than those based on microsatellite data (Table 3). Despite the low number of allozyme loci used in our study, this observation could be indicative of selection in the contaminated populations. On the other hand, the finding that microsatellites result in lower levels of differentiation than allozymes is in agreement with the neutral hypothesis as loci with higher heterozygosity are expected to result in lower Fst-values (O’Reilly et al., 2004). Furthermore, the microsatellite data confirm the genetic differentiation between the upstream reference population and the contaminated populations (however, as was mentioned, with a lower Fst-value), despite the fact that population MNA is located only 2.3 km downstream of population MNC without any physical migration barriers between them. Since the microsatellite data, by contrast, do not show a genetic difference between the two contaminated sites, the effect of historical cadmium pollution on microsatellite variability in these populations does not appear to be random. This observation suggests that at least some of the microsatellite loci do not act as neutral markers, which is supported by the evidence provided by the BayeScan analysis showing substantial evidence for two loci being affected by natural selection. However, observed heterozygosities and mean number of alleles indicate that genetic variation was not (yet) affected in these populations. As for the downstream reference population (WIA), the microsatellite data yielded a different pattern. Whereas in the allozyme survey, no genetic differentiation was found between the reference sites, the microsatellite data clearly showed a genetic difference between this population and all other populations. These findings were further confirmed in a comparison of a genic and genotypic differentiation test on both allozyme and microsatellite data (Table 3), and are reflected in a Neighbor Joining phylogram based on a Cavalli-Sforza and Edwards genetic distance matrix (Fig. 2).
23
4.2. Selection or bottleneck? Toxicant-induced changes in the genetic composition of natural populations can be classified into two main categories: changes that occur stochastically, and those that occur selectively (Murdoch and Hebert, 1994). A reduction of the effective population size which occurs in a genotype independent fashion, can alter the degree of inbreeding, may change the gene flow level and/or could induce a bottleneck (Belfiore and Anderson, 2001). Although it is difficult to unambiguously identify a changed population genetic structure as being the effect of a bottleneck event, there are a few indications which can be taken into consideration. First of all, a bottleneck is expected to decrease the overall genetic variation in an affected population as a consequence of the loss of low-frequency and rare alleles and genetic drift (Lande, 1988). It should, however, be mentioned that bottleneck effects also have the potential to increase genetic variation, as low-frequency alleles can become more abundant due to inbreeding (Gillespie and Guttman, 1999). Secondly, population genetic theory predicts that when the mean number of alleles at neutral loci decreases due to a bottleneck, this number will increase faster than heterozygosity when population size is restored as a result of new mutations (Nei, 1975). Finally, if a bottleneck event would occur in different populations, which have a comparable initial gene pool and experience a comparable exposure scenario, allele and genotype frequencies are likely to be affected differently in the respective populations, whereas selection would lead to a comparable genetic population structure (Gillespie and Guttman, 1999). In our study, observed heterozygosities and the mean number of alleles per locus were comparable in all populations. Furthermore, genetic variation did not differ between the two reference populations, which enclose the contaminated sites. It is therefore reasonable to assume that those populations reflect the preexposure genetic variability of all four populations. Since allele and genotype frequencies are also comparable between the two contaminated populations, a bottleneck event seems unlikely since in that case a different outcome is expected for both populations, especially given the strong homing behaviour of the gudgeon which is expected to restrict gene flow. Allozyme patterns in ecotoxicological research however should be interpreted with caution, because allozymes only represent small portions of the genome (Belfiore and Anderson, 2001). Moreover, since they represent a phenotypic rather than a genotypic condition, they can be affected by differential gene expression and post-translational modifications of the protein. Hence, allozymebased molecular patterning may reflect true genetic differentiation, but could equally well represent environmentally induced phenotypic plasticity of the genome (De Wolf et al., 2004). More importantly, all of the effects described above are assumed to occur on neutral loci. Although the neutralist theory, which states that allozyme variation is not selectively adaptive (Kimura, 1991), has long prevailed, and allozymes have been widely used as neutral markers, an increasing number of studies have provided contradictory evidence for this hypothesis (Belfiore and Anderson, 2001). From a selectionist point of view, differences in the function of allozymes can be linked to differences in fitness (Watt, 1994). Selection on allozyme loci which are critical for survival or reproduction in a specific contaminated environment can lead to altered allele frequencies at these loci (Bickham et al., 2000) through selection on the loci themselves, or on the genes that regulate allozyme expression. Furthermore, allozymes can act as non-neutral markers when they are linked to other loci under selection (genetic hitchhiking) and therefore also reflect selection (Baker, 1982). The relationship that was established between allele and genotype frequencies and the presence of metals suggests that the allozymes we assayed do not act as neutral markers, and differ-
24
D. Knapen et al. / Aquatic Toxicology 95 (2009) 17–26
ences among populations are the result of selection, rather than a bottleneck. More specifically, the genetic similarity between the two contaminated populations, and the genetic similarity between the two reference sites (describing pre-exposure variation) point in that direction. In general, observed differentiation patterns based on allozyme data among anthropogenically disturbed populations can therefore be strongly influenced by selection and no longer reflect true population structuring. 4.3. Fitness If selection is at work, one expects differential fitness to occur. However, although it can be expected that resistant genotypes have a higher fitness, evidence also suggests that the evolution of resistance can be associated with a direct fitness cost (Hebert and Luiker, 1996; Coustau et al., 2000) because collateral genetic changes may occur that affect life-history traits directly. Populations MNA and MNB, located in the most contaminated sample sites, show the highest condition factor (Fig. 3a). Although this seems rather unexpected, species with a broad ecological niche, such as the gudgeon, can survive in sub-optimal environmental conditions when there is a reduction in interspecific competition or predation (Postma, 1995). Possibly, other, more sensitive species have disappeared in these sites, creating for example better foraging possibilities and breeding sites for the gudgeon populations. In a previous study, we already demonstrated that fish from the contaminated sites have a higher, genetically based resistance to Cd compared to the reference sites, which was mainly attributed to the induction of the metal binding protein metallothionein in liver (Knapen et al., 2004). This certainly favours these populations, and can partly explain the high condition factor. To establish a more robust conclusion, genotype-dependent fitness differences should be investigated. Our results show that there is a fitness advantage coupled to the presence of a specific allele. Indeed, Gpd* genotypes that carry the most abundant allele (2) in the contaminated populations have a higher condition factor (Fig. 3c). Other studies however have shown that an opposite response can occur as well. For example, Marchand et al. (2004) found that tolerant allozyme genotypes had a reduced condition factor compared to sensitive genotypes, suggesting that a metabolic cost may be associated with stress resistance. The true meaning of the condition factor in stressed populations (i.e. its indicative value for the actual biological fitness) may therefore not always be unambiguous. 4.4. Gpd* and Pgd* If the observed allozyme pattern is, indeed, the result of selection, the question arises whether it is the result of direct selection in which certain allozyme variants yield a direct fitness advantage, or rather the result of genetic hitchhiking. Obviously, this question cannot be conclusively answered without an experimental approach. We can, however, formulate a hypothesis that puts forward the possible mode of action of direct selection by looking at the biochemical role of the two loci which showed significant differences among populations. Glucose-6-phosphate dehydrogenase and phosphoglucomutase are both key enzymes of the pentose phosphate pathway, converting glucose-6-phosphate to ribose-5-phosphate for use in nucleic acid synthesis. However, an at least equally important function of these two enzymes is the concomitant production of NADPH, the major cytoplasmic reducing compound (Biagiotti et al., 2000). Evidence shows that because of this characteristic, glucose-6phosphate dehydrogenase is an important regulatory enzyme in the NADPH-dependent defence against oxidative stress (Winzer et al., 2002). Metals such as Cd and Zn cause oxidative stress by disturbing
the pro-oxidant/anti-oxidant equilibrium of the cell. For example, because of the high affinity of metals for the sulphydryl groups of reduced glutathione (Nigam et al., 1999; Shaikh et al., 1999), they cause an increased oxidative state of the cell, which amongst other effects damages cell membranes. Furthermore, a link has been established between the regulation of metallothionein metal load and the cellular redox status (Maret, 1994, 2000). As mentioned earlier, previous work on G. gobio showed that liver metallothionein plays an important role in metal resistance in the contaminated populations (Knapen et al., 2004). Although it appears that the expression of an alternative Gpd* allele is advantageous in conditions that induce oxidative stress, a direct fitness cost is possibly associated with the expression of that allele. This hypothesis creates new possibilities for future research, in which the relative importance of selection acting directly on the allozyme loci, or selection acting on loci regulating allozyme expression, should be addressed.
4.5. The management unit concept In order to describe, and more importantly, to manage the full array of genetic variation within a species, several types of conservation units have been proposed. The two most important types of conservation units that are currently being used are evolutionarily significant units (ESUs) on the one hand, focusing on unique and independent evolutionary trajectories of sets of populations, and management units (MUs) on the other hand (Moritz, 1994; Paetkau, 1999; Parker et al., 1999). MUs are functionally independent populations, or groups of populations (Moritz, 1994), in which local population dynamics are determined rather by birth and death than by immigration and emigration (Moritz et al., 1995). Within a MU, population genetic structure can therefore be dominated by processes such as local genetic adaptation. MUs can be considered as logical components for local short-term conservation management, used as the building blocks for managing genetic variation within ESUs, in order to conserve ecological and evolutionary processes in the long term (Moritz, 1999). However, the most important premise of MUs (as well as of ESUs) is that they only apply when the populations under study are naturally divided into groups (Moritz et al., 1995), making them not straightforward to use effectively in most contemporary Western European watercourses. Since, for example, pollution of the environment has been demonstrated to exert stronger coefficients of selection than natural processes (Reznick and Ghalambor, 2001), anthropogenically induced natural selection is likely to be able to change population genetic structure in a way that is not instantly detectable by genetic markers traditionally used to identify conservation units, such as microsatellites and mitochondrial DNA sequences. The real world applicability of ESUs has been thoroughly discussed (e.g. Paetkau, 1999; Crandall et al., 2000; Goldstein et al., 2000; Fraser and Bernatchez, 2001). However, only few attempts have been made to assess the impact of human disturbance of natural populations on the recognition and usability of MUs (e.g. Hedrick et al., 2001). If we were to delineate MUs based on this dataset, our data suggest that each population meets the criterion for MU recognition (Moritz, 1994), except for the population pair MNA–MNB, and possibly the cluster MNA–MNB–GN, or even MNA–MNB–MNC–GN. The degree of divergence within this group is not entirely clear, as the Fst-values and the results from the genetic differentiation test are contradictory in some cases (see Table 3). Allele frequencies between all other pairwise comparisons were significantly different. Non-significant Fst-values between other population pairs than the four populations mentioned above should be investigated further within the context of possible artificial mixing, before final conclusions are drawn because of their geographic separation.
D. Knapen et al. / Aquatic Toxicology 95 (2009) 17–26
However, the relationship among populations MNA–MNB– MNC–GN is more complex than it may seem at first sight. We demonstrated that the MNA and MNB population, located only 2.3 km downstream of population MNC, is clearly differentiated from population MNC due to metal contamination, both at the organismal level as well as at allozyme and microsatellite loci. Populations MNA and MNB, both contaminated populations, were almost identical at these parameters. Therefore, MNA–MNB should be considered as a functionally separate unit from population MNC. However, both units should not be treated as MUs in the traditional meaning of the MU concept. They cannot be seen as sources of pristine, geographically restricted genetic variation, as changes in genetic composition have been anthropogenically induced. Using these populations as a source to complement threatened populations, or complementing these populations with specimens from related MUs, could result in unpredictable effects due to the potentially complex nature of local genetic adaptation in these populations. 4.6. Concluding remarks In conclusion, we have shown that long-term exposure to metals can induce changes at the population genetic level in natural fish populations, which can be detected both at microsatellite as well as at allozyme loci. These findings are relevant for the mere interpretation of microsatellite and allozyme data observed in anthropogenically impacted populations, but also pose some ethical and scientific questions as to how such populations should be approached from a nature conservation perspective. For example, should such forms of anthropogenically modified sources of genetic variation be conserved, or should they be treated as nonendemic strains? Should these populations be protected, or even supplemented with specimens from other populations if they are on the verge of extinction? What will be the effect of environmental cleanup operations, removing chemical migration barriers and allowing locally adapted populations to interbreed with neighbouring populations? Furthermore, we have illustrated that both microsatellite and allozyme loci do not necessarily behave as selectively neutral markers in polluted populations. Estimates of population differentiation can therefore be significantly different depending on which loci are being studied, especially when the number of loci being used is small. On the one hand, this could mean that observed differentiation patterns no longer reflect the overall, natural population substructuring. On the other hand, if those loci that are affected by natural selection are appropriately identified, they can serve as useful markers to study (either indirectly or directly, as put forward in the case of the Gpd* and Pgd* allozyme loci) the underlying micro-evolutionary processes, for example leading to local genetic adaptation to toxicants. Finally, it should be clear that exposure to toxicants does not always, or necessarily lead to tolerance changes in future generations. In our previous publications on the same gudgeon populations, we have effectively shown a significantly higher cadmium tolerance in specimens that were taken from the contaminated site (Knapen et al., 2004, 2007). However, others report no changes in physiological performance in offspring from genetically tolerant parents. For example, Brown et al. (2009) showed no heritable tolerance differences in rainbow trout offspring after 17␣-ethynylestradiol (EE2) exposure despite the fact that the offspring was produced only by fish that survived EE2 exposure. A comparable result was found in Kolok and L’Etoile-Lopes (2005) for copper in fathead minnows. It should be noted however that these studies were limited to one generation only, whereas the gudgeon populations in our study have been exposed to metals for many decades. This underlines the importance of multi-generation
25
studies, both in the field as well as under laboratory conditions. References Aljanabi, S.M., Martinez, I., 1997. Universal and rapid salt-extraction of high quality genomic DNA for PCR-based techniques. Nucleic Acids Res. 25, 4692–4693. Baker, J., 1982. Selective effects of insecticides on within-species variation: the lessons to be learnt when considering the environmental effects of pollutants. Agric. Ecosyst. Environ. 7, 187–198. Belfiore, N.M., Anderson, S.L., 2001. Effects of contaminants on genetic patterns in aquatic organisms: a review. Mutat. Res. 489, 97–122. Biagiotti, E., Bosch, K.S., Ninfali, P., Frederiks, W.M., Van Noorden, C.J.F., 2000. Posttranslational regulation of glucose-6-phosphate dehydrogenase activity in tongue epithelium. J. Histochem. Cytochem. 48, 971–977. Bickham, J.W., Sandhu, S., Hebert, P.D.N., Chikhi, L., Athwal, R., 2000. Effects of chemical contaminants an genetic diversity in natural populations: implications for biomonitoring and ecotoxicology. Mutat. Res. 463, 33–51. Bolger, T., Connolly, P.L., 1989. The selection of suitable indices for the measurement and analysis of fish condition. J. Fish Biol. 34, 171–182. Boudou, A., Ribeyre, F., 1997. Aquatic ecotoxicology: from the ecosystem to the cellular and molecular levels. Environ. Health Perspect. 105 (Suppl. 1), 21–35. Bridges, C.M., Semlitsch, R.D., 2000. Variation in pesticide tolerance of tadpoles among and within species of ranidae and patterns of amphibian decline. Conserv. Biol. 14, 1490–1499. Brown, K.H., Schultz, I.R., Nagler, J.J., 2009. Lack of a heritable reproductive defect in the offspring of male rainbow trout exposed to the environmental estrogen 17␣-ethynylestradiol. Aquat. Toxicol. 91, 71–74. Brzustowski, J., http://www2.biology.ualberta.ca/jbrzusto/Doh.php, 2002. Clements, E.D., Neal, E.G., Yalden, D.W., 1988. The National Badger Sett Survey. Mammal Rev. 18, 1–9. Coustau, C., Chevillon, C., ffrench-Constant, R., 2000. Resistance to xenobiotics and parasites: can we count the cost? Trends Ecol. Evol. 15, 378–383. Crandall, K.A., Bininda-Emonds, O.R.P., Mace, G.M., Wayne, R.K., 2000. Considering evolutionary processes in conservation biology. Trends Ecol. Evol. 15, 290–295. De Cooman, W., Florus, M., Devroede-Vander Linden, M.P., 1998. Karakterisatie van de bodems van de Vlaamse onbevaarbare waterlopen. AMINAL, VMM, Brussels, Belgium. De Wolf, H., Blust, R., Backeljau, T., 2004. The use of RAPD in ecotoxicology. Mutat. Res. 566, 249–262. Dhuyvetter, H., Gaublomme, E., Desender, K., 2004. Genetic differentiation and local adaptation in the salt-marsh beetle Pogonus chalceus: a comparison between allozyme and microsatellite loci. Mol. Ecol. 13, 1065–1074. Excoffier, L., Smouse, P., Quottro, J., 1992. Analysis of molecular variance inferred from metric distances among DNA haplotypes: application to human mitochondrial DNA restriction data. Genetics 131, 479–491. Felsenstein, J., 1989. PHYLIP – Phylogeny Inference Package (version 3.2). Cladistics 5, 164–166. Felsenstein, J., 2004. PHYLIP (Phylogeny Inference Package) version 3.6. Distributed by the author. Department of Genome Sciences, University of Washington, Seattle. Foll, M., Gagiotti, O., 2008. A genome-scan method to identify selected loci appropriate for both dominant and codominant markers: a Bayesian perspective. Genetics 180, 977–993. Fraser, D.J., Bernatchez, L., 2001. Adaptive evolutionary conservation: towards a unified concept for defining conservation units. Mol. Ecol. 10, 2741–2752. Gerstmeier, R., Romig, T., 2000. Zoetwatervissen van Europa. Tirion Uitgevers BV, Baarn, the Netherlands. Gillespie, R.B., Guttman, S.I., 1999. Chemical-induced changes in the genetic structure of populations: effects on allozymes. In: Forbes, V.E. (Ed.), Genetics and Ecotoxicology. Taylor & Francis, Philadelphia, USA. Goldstein, P.Z., DeSalle, R., Amato, G., Vogler, A.P., 2000. Conservation genetics at the species boundary. Conserv. Biol. 14, 120–131. Harris, H., Hopkinson, D.A., 1976. Handbook of Enzyme Electrophoresis in Human Genetics. North-Holland Publishing Company, Amsterdam, the Netherlands. Hebert, P.D.N., Luiker, M.M., 1996. Genetic effects of contaminant exposure—towards an assessment of impacts on animal populations. Sci. Total Environ. 191, 23–58. Hedrick, P.W., Parker, K.M., Lee, R.N., 2001. Using microsatellite and MHC variation to identify species, ESUs, and MUs in the endangered Sonoran topminnow. Mol. Ecol. 10, 1399–1412. Jeffreys, H., 1961. Theory of Probability, third ed. Clarendon Press, Oxford, UK, p. 432. Jensen, J.L., Bohonak, A.J., Kelley, S.T., 2005. Isolation by distance, web service. BMC Genet. 6, 13. Kimura, M., 1991. The neutral theory of molecular evolution: a review of recent evidence. Jpn. J. Genet. 66, 367–386. Knaepkens, G., Knapen, D., Bervoets, L., Hänfling, B., Verheyen, E., Eens, M., 2002. Genetic diversity and condition factor: a significant relationship in Flemish but not in German populations of the European bullhead (Cottus gobio L.). Heredity 89, 280–287. Knapen, D., Bervoets, L., Verheyen, E., Blust, R., 2004. Resistance to water pollution in natural gudgeon (Gobio gobio L.) populations may be due to genetic adaptation. Aquat. Toxicol. 67, 155–165. Knapen, D., Taylor, M.I., Blust, R., Verheyen, E., 2006. Isolation and characterization of polymorphic microsatellite loci in the gudgeon, Gobio gobio (Cyprinidae). Mol. Ecol. Notes 6, 387–389.
26
D. Knapen et al. / Aquatic Toxicology 95 (2009) 17–26
Knapen, D., Reynders, H., Bervoets, L., Verheyen, E., Blust, R., 2007. Metallothionein gene and protein expression as a biomarker for metal pollution in natural gudgeon populations. Aquat. Toxicol. 82, 163–172. Kolok, A.S., L’Etoile-Lopes, D., 2005. Do copper tolerant fathead minnows produce copper tolerant adult offspring? Aquat. Toxicol. 72, 231–238. Lambert, Y., Dutil, J.D., 1997. Condition and energy reserves of Atlantic cod (Gadus morhua) during the collapse of the northern Gulf of St. Lawrence stock. Can. J. Fish. Aquat. Sci 54, 2388–2400. Lande, R., 1988. Genetics and demography in biological conservation. Science 241, 1455–1460. Le Cren, E.D., 1951. The length–weight relationship and seasonal cycle in gonad weight and condition in the perch (Perca fluviatilis). J. Anim. Ecol. 20, 201–219. Lelek, A., 1987. Threatened fishes of Europe. In: Germany. The Freshwater Fishes of Europe, vol. 9. Aula Verlag, Wiesbaden. Lemaire, C., Allegrucci, G., Naciri, M., Bahri-Sfar, L., Kara, H., Bonhomme, F., 2000. Do discrepancies between microsatellite and allozyme variation reveal differential selection between sea and lagoon in the sea bass (Dicentrarchus labrax)? Mol. Ecol. 9, 457–467. Maitland, P.S., 2000. The Hamlyn Guide to Freshwater Fish of Britain and Europe. Octopus Publishing Group Ltd, London, UK. Manosa, S., Mateo, R., Guitart, R., 2001. A review of the effects of agricultural and industrial contamination on the Ebro delta biota and wildlife. Environ. Monit. Assess. 71, 187–205. Marchand, J., Quiniou, L., Riso, R., Thebaut, M.T., Laroche, J., 2004. Physiological cost of tolerance to toxicants in the European flounder Platichthys flesus, along the French Atlantic Coast. Aquat. Toxicol. 70, 327–343. Maret, W., 1994. Oxidative metal release from metallothionein via zincthiol/disulfide interchange. Proc. Natl. Acad. Sci. U.S.A. 91, 237–241. Maret, W., 2000. The function of zinc metallothionein: a link between cellular zinc and redox state. J. Nutr. 130, 1455S–1458S. McDonald, J.H., 1994. Detecting natural selection by comparing geographical variation in protein and DNA polymorphisms. In: Golding, B. (Ed.), Non-Neutral Evolution: Theories and Molecular Data. Chapman & Hall, New York, USA, pp. 88–100. Moritz, C., 1994. Defining ‘Evolutionarily Significant Units’ for conservation. Trends Ecol. Evol. 9, 373–375. Moritz, C., Lavery, S., Slade, R., 1995. Using information from allele frequency and phylogeny to define units for conservation and management. In: Nielsen, J.L. (Ed.), Evolution and the Aquatic Ecosystem: Defining unique units in population conservation. American Fisheries Society, Maryland, USA. Moritz, C., 1999. Conservation units and translocations: strategies for conserving evolutionary processes. Hereditas 130, 217–228. Mulvey, M., Newman, M.C., Chazal, A., Keklak, M.M., Heagler, M.G., Hales, L.S., 1995. Genetic and demographic responses of mosquitofish (Gambusia holbrooki, Girard 1859) populations stressed by mercury. Environ. Toxicol. Chem. 14, 1411–1418. Murdoch, M.H., Hebert, P.D.N., 1994. Mitochondrial DNA diversity of. brown bullhead from contaminated and relatively pristine sites in the Great Lakes. Environ. Toxicol. Chem. 13, 1281–1289. Nei, M., 1975. Molecular Population Genetics and Evolution. North-Holland, Amsterdam, Netherlands. Nigam, D., Shukla, G.S., Agarwal, A.K., 1999. Glutathione depletion and oxidative damage in mitochondria following exposure to cadmium in rat liver and kidney. Toxicol. Lett. 106, 151–157. O’Reilly, P.T., Canino, M.F., Bailey, K.M., Bentzen, P., 2004. Inverse relationship between Fst and microsatellite polymorphism in the marine fish, walleye pollock
(Theragra chalcogramma): implications for resolving weak population structure. Mol. Ecol 13, 1799–1814. Paetkau, D., 1999. Using genetics to identify intraspecific conservation units: a critique of current methods. Conserv. Biol. 13, 1507–1509. Parker, K.M., Sheffer, R.J., Hedrick, P.W., 1999. Molecular variation and evolutionarily significant units in the endangered Gila topminnow. Conserv. Biol. 13, 108–116. Postma, J.M., 1995. Adaptation to Metals in the Midge Chironomus riparius. PhD thesis, University of Amsterdam, the Netherlands. Raymond, M., Rousset, F., 1995. GENEPOP (version 1.2): population genetics software for exact tests and ecumenicism. J. Hered. 86, 248–249. Reznick, D.N., Ghalambor, C.K., 2001. The population ecology of contemporary adaptations: what empirical studies reveal about the conditions that promote adaptive evolution. Genetica 112, 183–198. Richardson, B.J., Bavenstock, P.R., Adams, M., 1986. Allozyme Electrophoresis: A Handbook for Animal Systematics and Population Studies. Academic Press, New York, USA. Ricker, W.E., 1971. Methods for Assessment of Fish Production in Fresh Waters. Blackwell Scientific Publications, Oxford, UK. Schneider, S., Roessli, D., Excoffier, L., 2000. Arlequin, Version 2.000: A Software for Population Genetic Data Analysis. Genetics and Biometry Laboratory, University of Geneva, Switzerland. Shaikh, Z.A., Vu, T.T., Zaman, K., 1999. Oxidative stress as a mechanism of chronic cadmium-induced hepatotoxicity and renal toxicity and protection by antioxidants. Toxicol. Appl. Pharmacol. 154, 256–263. Simonsen, V., Laskowski, R., Bayley, M., Holmstrupa, M., 2008. Low impact of metal pollution on genetic variation in the earthworm Dendrobaena octaedra measured by allozymes. Pedobiologia 52, 51–60. Storz, J.F., 2005. Using genome scans of DNA polymorphism to infer adaptive population divergence. Mol. Ecol. 14, 671–688. Stott, B., 1967. The movements and population densities of roach (Rutilus rutilus (L.)) and gudgeon (Gobio gobio (L.)) in the river Mole. J. Anim. Ecol. 36, 407–423. Tanguy, A., Boutet, I., Bonhomme, F.Y., Boudry, P., Moraga, D., 2002. Polymorphism of metallothionein genes in the Pacific oyster Crassostrea gigas as a biomarker of response to metal exposure. Biomarkers 7, 439–450. Van Campenhout, K., Bervoets, L., Blust, R., 2003. Metallothionein concentrations in natural populations of gudgeon (Gobio gobio): relationship with metal concentrations in tissues and environment. Environ. Toxicol. Chem. 22, 1548–1555. van Straalen, N.M., Timmermans, M.J.T.N., 2002. Genetic variation in toxicantstressed populations: an evaluation of the “genetic erosion” hypothesis. Hum. Ecol. Risk Assess. 8, 983–1002. Vlaamse Regering, 2000. Vlarem II: Vlaams Reglement Milieuvergunning - Editie 2000. Kwaliteitsdoelstellingen voor oppervlaktewater in het Vlaamse Gewest. Vlaamse Regering, Brussels, Belgium. Watt, W.B., 1994. Allozymes in evolutionary genetics: self-imposed burden or extraordinary tool? Genetics 136, 11–16. Weir, B.S., Cockerham, C.C., 1984. Estimating F-statistics for the analysis of population structure. Evolution 28, 1358–1370. Weir, B.S., 1996. Genetic Data Analysis II: Methods for Discrete Population Genetic Data. Sinauer Associates Inc., Sunderland, Massachusetts, USA. Williams, R.J.P., Frausto da Silva, J.J.R., 1996. The Natural Selection of the Chemical Elements. Clarendon Press, Oxford, UK. Winzer, K., Van Noorden, C.J.F., Kohler, A., 2002. Glucose-6-phosphate dehydrogenase: the key to sex-related xenobiotic toxicity in hepatocytes of European flounder (Platichthys flesus L.)? Aquat. Toxicol. 56, 275–288.