Widely distributed and regionally isolated! Drivers of genetic structure in Gammarus fossarum in a human-impacted landscape
© The Author(s). 2016
Received: 15 July 2016
Accepted: 18 July 2016
Published: 29 July 2016
The actual connectivity between populations of freshwater organisms is largely determined by species biology, but is also influenced by many area- and site-specific factors, such as water pollution and habitat fragmentation. Therefore, the prediction of effective gene flow, even for well-studied organisms, is difficult. The amphipod crustacean Gammarus fossarum is a key invertebrate in freshwater ecosystems and contains many cryptic species. One of these species is the broadly distributed G. fossarum clade 11 (type B). In this study, we tested for factors driving the genetic structure of G. fossarum clade 11 in a human-impacted landscape at local and regional scales. To determine population structure, we analyzed the mitochondrial cytochrome c oxidase 1 (CO1) gene of 2,086 specimens from 54 sampling sites and microsatellite loci of 420 of these specimens from ten sites.
We detected strong overall genetic differentiation between populations at regional and local scales with both independent marker systems, often even within few kilometers. Interestingly, we observed only a weak correlation of genetic distances with geographic distances or catchment boundaries. Testing for factors explaining the observed population structure revealed, that it was mostly the colonization history, which has influenced the structure rather than any of the chosen environmental factors. Whereas the number of in-stream barriers did not explain population differentiation, the few large water reservoirs in the catchment likely act as dispersal barriers.
We showed that populations of Gammarus fossarum clade 11 are strongly isolated even at local scales in the human-impacted region. The observed genetic structure was best explained by the effects of random genetic drift acting independently on isolated populations after historical colonization events. Genetic drift in isolated populations was probably further enhanced by anthropogenic impacts, as G. fossarum is sensitive to many anthropogenic stressors. These findings highlight the importance of small-scale genetic studies to determine barriers restricting gene flow to prevent further loss of genetic diversity and maintain intact freshwater ecosystems.
Biogeographic studies revealed wide ranges for many freshwater invertebrate species . This holds true in particular for species found in temperate and more northern latitudes, which had to recolonize habitats after glacial periods. Examples range from various insect species with terrestrial adult life stages to purely aquatic species, such as amphipod crustaceans and molluscs. Considering the typically wide ranges and high regional and local abundances, freshwater invertebrates are often regarded as frequent and long distance dispersers (e.g. [2, 3]). Over the past 15 years, however, the paradigm of wide ranges has been increasingly questioned, as molecular studies revealed the presence of morphologically cryptic species in many freshwater invertebrate taxa (e.g. [4–6]). These cryptic species often show rather small and allopatric ranges (e.g. [7, 8]), instead of the presumed broad distribution of the whole cryptic species complex.
One ecologically important freshwater taxon, which was formerly thought to be widely distributed in central and southeastern European streams, is the amphipod crustacean Gammarus fossarum KOCH, 1836. Several recent studies revealed an almost exponentially increasing number of overlooked species within the G. fossarum species complex with enhanced geographic sampling and improved sensitivity of molecular detection methods [9–11]. The highest species diversity, by far, within the G. fossarum species complex was found in the southeastern part of the range, where most of the newly discovered species were local endemics with narrow ranges . However, the four central and western European species, in particular clade 11, still show broad distributions .
Generally, G. fossarum is mainly found in the upper reaches of streams and is sensitive to organic pollution [12, 13], high ammonium concentrations , a lack of oxygen, and acidification . Owing to its high abundances and sensitivity to anthropogenic stressors, G. fossarum is often used in ecotoxicological studies (e.g. [16–18]). However, the precise cryptic species used in these experiments and whether a single or multiple species are used are rarely tested or reported. Validating species assignments prior to experiments is critically important, as studies explicitly investigating G. fossarum type A and B (here referred to as clade 12 and 11, after Weiss et al. ) revealed ecological differences between the species [19–21]. In further studies comparing these two species, clade 11 was found to be more tolerant against tested stressors [14, 22], occurred in areas with higher human impact  and was the better competitor in comparison with clade 12 , but it also showed higher infection rates for various parasites . Additionally, in a direct comparison, populations of clade 11 were less differentiated across hundreds of kilometers than populations of clade 12 , but still significant differentiation within clade 11 was found on a regional scale. These findings agree well with the moderate genetic differentiation found in a broad geographic area for members of clade 11 (e.g. ). Even though these findings may indicate a relatively good dispersal ability for G. fossarum clade 11, it is difficult to predict actual dispersal rates, as they can be influenced by area- and site-specific environmental factors, like water chemistry, stream bed structure, land use and urbanization in the riverine environment, and fragmentation of streams by in-stream barriers, like dams or reservoirs (e.g. [25–27]). However, as understanding the patterns and mechanisms of dispersal and connectivity is crucial for predicting population resilience and long-term adaptability of a species , it is important to determine the actual dispersal rates. An already regularly applied approach for this purpose is the use of genetic markers to estimate effective gene flow between populations, i.e. successful dispersal leading to genetic exchange between populations (e.g. [3, 29, 30]).
Populations of Gammarus fossarum clade 11 are genetically differentiated at the regional scale when considering the whole study area. In contrast differentiation is low at the local scale of few kilometers with evidence of gene flow between populations. These expectations are based on previous genetic analyses of G. fossarum species (e.g. [24, 34–36]).
Genetic variation is significantly partitioned after drainage basins at any spatial scale within the riverine network according to the Stream Hierarchy Model [37, 38], because G. fossarum is confined to aquatic habitats throughout its life cycle. Further, populations show patterns of isolation by distance (IBD) within catchments, especially when considering the waterway distance as this pattern was also found in other population genetic studies of G. fossarum clade 11 (e.g. ).
After testing the two hypotheses, we discuss the findings in order to identify possible drivers of population structure.
Study site and sampling
Sampling sites and number of studied G. fossarum specimens (n) in the Sauerland area
CO1 & msat
2011 & 2013
CO1 & msat
2011 & 2013
CO1 & msat
2011 & 2013
CO1 & msat
2011 & 2013
CO1 & msat
2011 & 2013
CO1 & msat
2011 & 2013
CO1 & msat
CO1 & msat
2011 & 2012
CO1 & msat
2011 & 2013
CO1 & msat
G. fossarum sampling was conducted in 2011, 2012, 2013, and 2014 (Table 1). Samples from six sites in 2011 were provided by Maria Gies and Martin Sondermann (University Duisburg-Essen). To examine changes in the genetic structure of populations over time, seven 2011 sampling sites were resampled in 2013 and one 2011 site was sampled again in 2012 (Table 1). Additional samples from 2013 and 2014 were originally obtained for another study in which a stream was sampled at four sites every 200 m with an in-stream barrier between separating sites two and three. At seven sampling sites in 2014, a similar sampling scheme was used for sewage plants, and, at two sampling sites, similar schemes were used for mining sites. Since no differences in haplotype frequencies were found between those sites (F ST values were not significant, data not shown), the samples were merged into a single sampling site for this study.
Sequencing and genotyping
DNA was extracted from the pereopods of 2,086 specimens using a modified salt-extraction protocol  (see Additional file 2 for details of used protocol). For all specimens, a fragment of the mitochondrial barcoding gene CO1 was amplified with the standard primers HCO2198 and LCO1490  using the following PCR protocol: 1× PCR buffer, 0.2 mM dNTPs, 0.5 μM each primer, 0.5 μl of DNA template, 0.025 U/μl HotMaster Taq (5 PRIME GmbH, Hamburg, Germany), filled to 15 μl with sterile H2O. The PCR settings for CO1 amplification were as follows: initial denaturation at 94 °C for 2 min; 34 cycles of denaturation at 94 °C for 20 s, annealing at 46 °C for 30 s, and extension at 65 °C for 60 s; final extension at 65 °C for 5 min. The PCR products (9 μl) were purified using 0.5 μl of ExoI (20 U/μl) and 1 μl of FastAP (1 U/μl, both enzymes; Thermo Fisher Scientific, Schwerte, Germany). The reaction was incubated for 25 min at 37 °C followed by an inactivation step at 85 °C for 15 min. Purified PCR products were bidirectionally sequenced by GATC-Biotech (Konstanz, Germany).
For 420 specimens from ten sampling sites (Table 1), eight microsatellite loci were additionally amplified: Gam 2, Gam 14 , Gamfos 10, Gamfos 13, Gamfos 18, Gamfos 22, Gamfos 27, and Gamfos 28 . For each primer pair, the originally published sequence was adapted to our analysis system by adding a universal M13 tail (5′-CAC GAC GTT CTA AAA-3′) to the 5′ ends of the forward primers for primers developed by Danancher et al.  and to the reverse primers for those developed by Westram et al. . The amplification of Gam 2, Gam 14, Gamfos 13, Gamfos 27, and Gamfos 28 was performed using the following protocol: 1× PCR buffer, 0.2 mM dNTPs, 0.2 μM sequence-specific untailed primer, 0.05 μM sequence-specific tailed primer, 0.2 μM fluorescently labelled universal M13 primer, 5 % dimethyl sulfoxide (DMSO), 0.5–1 μl of DNA template, 0.025 U/μl HotMaster Taq (5 PRIME GmbH, Hamburg, Germany), filled to 15 μl with sterile H2O. For Gamfos 10, 18, and 22, betaine was used instead of DMSO. PCR settings for the amplification of Gam2, Gam14, Gamfos 13, Gamfos 18, Gamfos 22, and Gamfos 28 were as follows: initial denaturation at 94 °C for 2 min; 36 cycles of denaturation at 94 °C for 20 s, annealing at 60 °C for 30 s, and extension at 65 °C for 30 s; final extension at 65 °C for 45 min. For Gamfos 10 and Gamfos 27, an annealing temperature of 54 °C was used.
Allele sizes were determined using polyacrylamide gel electrophoresis on a LI-COR® 4300 DNA Analyzer with the software SagaGT (LI-COR Biosciences, Bad Homburg, Germany). For Gam 2, the alleles could not be scored because too many stutter bands occurred over a range of 90 base pairs.
Sequence data analysis
Sequences were assembled and edited in Geneious 8.1.2 (http://www.geneious.com, ), and a multiple sequence alignment was computed with MAFFT  as implemented in Geneious (automatic algorithm selection, default settings). Genetic diversity was calculated as haplotype diversity using Arlequin 3.5 . A minimum spanning network was calculated using Arlequin and visualized using HapStar 0.7 . To show the position of the haplotypes in a broader phylogenetic context nine sequences from five different clades (3, 10, 11, 12 and 13), which have been used in the study of Weiss et al.  (GenBank Accession numbers: JN900490, KF521805, KF521817, KF521822, KF521828, KF521829, KF521832, KF521833, KF521835), were added to a dataset of the main haplotypes of this study and a neighbor-joining Tree was calculated with MEGA6 , with evolutionary distances computed using the Kimura 2-parameter method. To test for population differentiation, pairwise F ST values between populations from different sampling sites were calculated using Arlequin. Negative values were set to zero. The significance levels for this and all subsequent F ST calculations were adjusted to account for multiple testing using the false discovery rate control (FDR, ). Additionally, F ST values were calculated using data from multiple years for populations originating from the same stream to test if haplotype frequencies changed significantly over time.
Microsatellite data analysis
Using the program MICRO-CHECKER 2.2.3 , our data set was checked for the occurrence of allelic drop out, stutter bands, and null alleles and null allele frequencies were calculated using the Dempster method . All loci in all populations were tested for deviations from Hardy–Weinberg equilibrium (HWE) and from linkage equilibrium using Arlequin. To estimate genetic diversity, allelic richness was calculated with the program HP-Rare 1.0  using rarefaction to correct for differences in sample sizes.
Genetic distances between populations were calculated as pairwise F ST values in Arlequin. To correct for null alleles, FreeNA  was used, in which the ENA correction was implemented. To test if the allele frequencies in the populations changed over time, F ST values were also calculated between populations originating from the same stream, but sampled in different years.
Determination of drivers of genetic differentiation (CO1 data)
Different landscape genetic approaches were used to identify drivers of genetic differentiation in the study area. First, to test if the distribution of genetic variance can be explained by the partitioning of populations into the six sub-catchments, AMOVA (analysis of molecular variance, ) was implemented in Arlequin. To test if genetic distances were correlated with geographic distances (isolation by distance, IBD), Mantel tests were performed in Arlequin. Pairwise F ST values were used as a measure of genetic distance, and either straight-line or waterway distances were used for geographic distance. Straight-line distances were calculated for all population pairs and waterway distances for populations within the Ruhr catchment (n = 48) using QGIS v. 2.8.2  and the map containing the streams provided by the federal state authority LANUV (Gewässerstationierungskarte des Landes NRW © LANUV NRW (2013)). To test if in-stream barriers influenced the connectivity between populations, another Mantel test was conducted with 21 populations of the Ruhr sub-catchment using pairwise F ST values and the number of in-stream barriers between population pairs per km waterway distance, first for barriers >0.5 m and then for barriers >1 m. The number of in-stream barriers was calculated from the QUIS database . Barriers were mostly weirs and barrages, often classified as only partly or not at all passable for the aquatic fauna in the ELWAS-web portal. To investigate if spatial factors other than geographic distance or catchment assignment shape genetic structure and to infer the number of populations, spatial Bayesian clustering models were implemented in the R package GENELAND, v4.0.5 ([58, 59], R version 3.2.2, ). The number of populations K was allowed to vary between 1 and 20. For six independent runs, 1,000,000 MCMC iterations were calculated, sampling every 100 steps. The maximum number of nuclei in the Poisson-Voronoi tessellation was set to 6,300. For post-processing, a burn-in of 1,500 iterations was used and the pixels along the X-axis were set to 300 and along the Y-axis to 150 according to the ratio of the sampling area. A second AMOVA implemented in Arlequin using the groups detected in GENELAND was used to analyze if this clustering better reflects the population structure than the sub-catchments.
To test which factors determine the association of the individual populations to the GENELAND groups, a discriminant analysis was performed in IBM SPSS v. 23. For this analysis, the environmental variables obtained to characterize sampling sites were independent variables and the main GENELAND groups (consisting of >3 sites) were dependent variables. Additional to the environmental variables, also the geographic position of the sampling sites, represented by their coordinates, was used as a variable to include a proxy for the possible influence of colonization history on genetic structure. To standardize the predictor variables, z-scores were calculated in SPSS prior to the discriminant analysis. Further, multicollinearity between the 25 environmental variables was tested by calculating the variance inflation factors (VIFs) for each variable using the R package “usdm” . Collinearity between variables was assumed for VIFs >10. After calculating the VIFs, the variables, which were highly collinear, were stepwise excluded from the analysis until all values were > 10. Therefore, the following discriminant analysis was conducted with a reduced set of variables. In the discriminant analysis, the stepwise Wilks lambda method was used to select predictors.
We only detected eight haplotypes that were shared between at least two populations (H1, H7, H19–22, H31, and H33) and we observed private haplotypes in 20 populations. At two sampling sites, i.e., KL15 (n = 59) and KL2 (n = 57), populations consisted exclusively of one private haplotype each. Overall, we detected between one and six haplotypes per population, with an average of two haplotypes, and a haplotype diversity of between 0.00 and 0.68, with an average of 0.16 (Additional file 3). The nucleotide diversity ranged from 0.000 to 0.136 with an average of 0.048 (Additional file 3). In seven of the eight populations for which we compared haplotype frequencies among years, we did not observe significant differences. The only exception was population NH, where, in the second sampling year, only the main haplotype of the population (H21) was rediscovered together with a new haplotype (H20), resulting in a small, but significant, F ST value.
Of the seven microsatellites analyzed to complement the CO1 data set, only four were polymorphic in the studied populations (Gamfos 10, 13, 18, and 28). Three of the ten populations were also monomorphic for the same allele at Gamfos 13. We observed between 4 and 23 alleles in all populations at the different loci. We did not detect significant linkage between loci considering all populations. We found evidence for null alleles for three of the loci in different populations. The null allele frequencies, number of alleles, and observed and expected heterozygosity (HO and HE) are given in Additional file 5. Null allele frequencies ranged from 0.00 to 0.29. We observed deviations from HWE in all populations for at least one locus (see Additional file 5). As a measure of genetic diversity, we estimated allelic richness using rarefaction for a minimum sample size of 15 diploid individuals. We observed an average allelic richness over all loci of between 2.5 and 5.3 and we detected private alleles in six of the populations (Additional file 5). In two of the populations sampled twice in different years (NH and E02), the allele frequencies changed over time, resulting in small, but significant, F ST values.
Regional and local differentiation
Drivers of local isolation
Results of AMOVA analysis according to 1) sub-catchments and 2) GENELAND groups
Between GENELAND groups
Source of variation
between populations within groups
Other major barriers located between some of the sampling sites were water reservoirs (see Fig. 2a). One such water reservoir, the Bigge reservoir, separated populations KL2 and KL15, and a second smaller one (Ahauser reservoir) separated both sites from all other sampling sites; both populations consisted only of one private haplotype each (H28 and H29, resp., see Fig. 2a). Another reservoir (Sorpe reservoir) separated QB29 from the other sampling sites. Between QB29 and the next sampled population (RO2), two haplotypes were shared, but the haplotype composition was significantly different. The population GS of the Möhne sub-catchment was separated by the Möhne reservoir from the other populations of this catchment. This reservoir also separated the whole sub-catchment from the Ruhr catchment. In both the GS population and some populations from the Ruhr and Möhne catchment (e.g. KL9, VR7, and LO), we only detected haplotype H1, and we did not detect a barrier effect of the Möhne reservoir based on the CO1 sequence analyses.
To infer the population structure more directly, without manually assigned assumptions regarding affiliations to catchment areas, we performed a clustering analysis with GENELAND. In this analysis, we observed seven distinct groups, termed A–G (letters in circles in Fig. 2a). The biggest group was group A, with 24 populations containing mostly or exclusively haplotype H1. The second biggest group, dominated by H19, was group B with 13 populations. In most of these populations, we also detected H1, although at minor frequencies. This group was geographically split into two subgroups, with two populations located in the western area, and most populations in the east, with no visible connection between the two subgroups. Groups C and D were the easternmost and westernmost groups, both consisting of six populations, dominated by H21 and H33, respectively. While populations of group C shared haplotypes with groups A and B, all haplotypes of group D (H33 to H40) were exclusively found in that group. All but one population of group D contained additional private haplotypes, aside from the main haplotype. Group E consisted of three populations that shared H1 at a minor frequency, but had different dominant haplotypes, which were all derived from H1 (Fig. 1). Groups A to E all contained populations from at least two sub-catchments. The last two groups F and G each consisted only of one population (KL2 and KL15) and contained exclusively one private haplotype (H28 and H29, respectively). Both haplotypes were three mutations apart from H19 and six from each other. Using the GENELAND groups in an AMOVA, we found that this clustering reflects the population structure better than the sub-catchment structuring, as 71.1 % (instead of 19.6 %) of the variation was between groups and only 10 % (instead of 59.8 %) was among populations within groups (Table 2).
To test if other environmental factors determine the observed population structure, we gathered data for 25 environmental variables (Additional file 1). As groups F and G only consisted of single sites and populations of group E were highly differentiated from each other, we only conducted the following analyses for the four main groups, A–D. Testing for multicollinearity revealed that some of the variables were highly correlated. We therefore excluded the following variables for subsequent analyses: altitude, MP3, grassland, total nitrogen, and pH, resulting in 20 variables for the discriminant analysis. The stepwise discriminant analyses revealed that the most useful variable for predicting assignments to the GENELAND groups was longitude; all other variables were not included in the model. With the resulting model, 65 % of the populations clustered correctly to the four groups, with differences in prediction performance between groups (see Additional file 8). While all populations of groups C and D clustered correctly, this was the case for only 61.5 % of group B and 50.0 % of group A populations, as these two groups had a broader geographic range with overlapping longitudinal values.
We analyzed a large number of specimens and populations of the freshwater crustacean G. fossarum clade 11 in the anthropogenically heavily impacted Sauerland region to identify factors that influence population structure and limit dispersal in this species.
Our first hypothesis was that populations are genetically differentiated when considering the regional scale of the whole study area, whereas the differentiation is low at the scale of few kilometers. In agreement with the first expectation of this hypothesis we detected strong regional differentiation in the Sauerland area, especially between the eastern and westernmost populations. The westernmost populations (GENELAND group D) contained only haplotypes (H33–H40) separated by at least 14 substitutions from the other haplotypes (H1–H32), but we did not detect this high degree of differentiation using microsatellites. A similar pattern of east-west differentiation was observed in the stonefly Dinocras cephalotes in the same study area . Specifically, two highly divergent haplotype groups were found for the CO1 gene. However, contrary to our findings, Elbrecht et al.  found haplotypes of these groups to be shared across populations of these groups and also detected ongoing gene flow using nuclear genes. As the stonefly has a terrestrial and more mobile life stage, these contrasting patterns may be explained by differences in mobility. The observed divergence between the two G. fossarum clade 11 groups (east-west) is likely the result of independent historic isolation in eastern and western refugia, as suggested by Elbrecht et al. (2014) for the stonefly species. Even though G. fossarum populations belonging to the two divergent groups seem to be isolated, it is likely that they can still interbreed when in contact. This is indicated by the low divergence (between 2.51 and 3.95 %) in comparison to previous estimates of divergence between cryptic species of G. fossarum , also visible in the phylogenetic tree, when comparing intra- and inter-clade divergence. Further, Lagrue et al.  found reproductive isolation between G. fossarum clades only when CO1 sequence divergence was greater than 4 %.
We did, however, not only find strong differentiation between populations at a regional scale, but contrary to our hypothesis also at a local scale. Actually, most of the populations were highly isolated from each other (80 % of pairwise CO1 F ST values were significant), sometimes even within 2 km in the same stream. The strong isolation is further supported by the pattern that many populations contained private haplotypes, which mostly differed by only a single mutation from the main haplotype of the respective population. This indicates long-term isolation because independent mutations were able to accumulate in the populations. In contrast to the overall strong isolation, we found little or no differentiation between some of the populations separated by over 40 km straight-line distances and even higher river distances (especially within GENELAND groups A and D) for the CO1 gene. However, when analyzing a subset of these populations based on the more rapidly evolving microsatellites, we detected strong differentiation between all populations. Therefore, the subtle differentiation found between populations of group A is most likely not caused by ongoing gene flow; instead, low genetic variation is likely due to bottlenecks and the effects of genetic drift [6, 64].
In their study, Westram et al.  concluded that differences in population structure between clade 11 and 12 could hint at interspecific differences in dispersal ability, life history or population size. The authors discussed that differences in differentiation levels reflect more likely species specific rather than being driven by geographic effects. However, the strong local differentiation we found in clade 11 resembled much more the pattern described for clade 12 than that of clade 11 [24, 36]. The discrepancies observed between the data of Westram et al.  and our study suggest that geographic effects can strongly impact on the differentiation of populations and with that probably on the realized dispersal of populations within a species.
Based on the SHM  our second hypothesis was that the populations are structured according to catchment boundaries and show an IBD pattern within catchments. The structuring of populations according to catchments has been shown for other crustaceans [65, 66] and an IBD pattern has been detected in G. fossarum [12, 34, 36], specifically in clade 11 . Studying variance partitioning and using the sub-catchment boundaries in an AMOVA, we detected significant partitioning of genetic variance, but most of the variance (59.8 %) was found between populations within groups, indicating that populations were not primarily structured by sub-catchments, but were locally isolated within sub-catchments. We found a similar pattern when analyzing the correlation between genetic and geographic distances, as the correlation was significant, but very weak for both geographic distance metrics (straight-line and waterway distance), especially for waterway distances. The weak signature of sub-catchment-based structuring and the stronger correlation between genetic and straight-line distances in comparison to waterway distances indicate that processes other than low dispersal within streams are relevant to the genetic structure within this species. This inference is also supported by the GENELAND analysis, where groups A–E contained populations from at least two sub-catchments each and two populations in group B were geographically distant to all other populations of this group. Evidence for overland dispersal in aquatic invertebrates, including G. fossarum, was reported previously (e.g. [36, 67, 68]), and it is assumed to occur by transport via vectors like birds [69–71], large mammals [72, 73], or humans . As some of the springs of the different sub-catchments in our study area are only separated by a few hundred meters and not by mountain ranges, overland dispersal is likely possible. However, as pairwise differentiation was high in general, overland dispersal seems to be more important for rare colonization events on evolutionary time scales rather than for recurrent dispersal at ecological time scales within generations. However, we did not expect the low correlation between waterway and genetic distances. A lack of IBD in other aquatic invertebrates is often attributed to particularly weak (e.g. [6, 74–76]) or strong dispersal abilities (e.g. [27, 76–78]). These results may also be explained by the presence of strong dispersal barriers between populations [79–81]. The slight IBD pattern we found in our study was mainly caused by the strong differentiation between the western- and easternmost populations, as they were separated by the greatest geographic distances and did not share any haplotypes. Otherwise, for many pairwise comparisons, we detected high genetic differentiation at small geographic distances and no differentiation over long distances. As we found strong population differentiation between populations, indicated by high and significant F ST values, the lack of IBD cannot be caused by too strong realized dispersal. However, previous studies found IBD in different areas for G. fossarum (e.g. ); accordingly, the dispersal ability should generally be sufficient to generate this pattern even though the realized dispersal appears to be very limited here. Therefore, it seems likely that other barriers to gene flow exist. These barriers could either be direct, like dams, weirs, or water reservoirs, or indirect if the conditions in connecting areas are unfavorable due to, for example, anthropogenic land use, organic pollution, acidification, large connecting rivers, or strong competition (e.g. [20, 27, 36, 68, 74, 76, 82, 83]). To analyze the influence of direct in-stream barriers, we used a subset of populations to determine if the number of barriers was correlated with the genetic distance between populations. Based on Mantel tests conducted for barriers of >0.5 m and >1 m, we did not observe a correlation with genetic distance, indicating that population isolation did not simply reflect the number of barriers. However, the influence of these barriers on realized dispersal could not be determined using the current marker and sampling scheme. To infer the influence of in-stream barriers more directly, individual barriers should be tested using more rapidly evolving genetic markers. Some of the sampling sites were separated from each other by water reservoirs, and, with one exception, the separated populations were significantly differentiated from all other analyzed populations, indicating a strong barrier effect of these water reservoirs. One exception at a first glance was in the Möhne sub-catchment, where populations were dominated by or even consisted exclusively of H1. However, as discussed previously, the diversity for the CO1 gene in this region was too low to detect barrier effects. Therefore, we conclude that in-stream barriers, especially water reservoirs, can have a substantial impact on dispersal, yet they do not explain the general population structure observed in this study. With the exception of the water reservoirs that separate the clearly differentiated groups F and G from all other populations there were no obvious weir- or dam-related boundaries separating the groups.
In view of the tested parameters that cannot explain population structure sufficiently the question arises, which other factors ultimately underlie the apparent population structure. To determine if the structure was more influenced by colonization history (represented by the geographic position of the population) or by environmental factors that may differ between population groups, we conducted a discriminant analysis for the four main groups including variables for both possible drivers. The best predictor for the assignment of populations to the four GENELAND groups was longitude, indicating that the location of each site from east to west determines the affiliation to the CO1 clusters. This clearly suggests that the observed population structure primarily reflects colonization history. If colonization events occur rarely and are initiated by relatively few individuals, this can lead to strong founder effects; genetic drift in small populations can lead to a strong loss of genetic diversity [64, 84]. The maintenance of this structure over time is probably promoted by a high degree of population isolation, leading to small effective population sizes and thereby enhancing the effect of random genetic drift. As G. fossarum is sensitive to many anthropogenic stressors, like organic pollution, a lack of oxygen, acidification , and high ammonium concentrations , isolated populations likely underwent drastic population declines over time, resulting in the loss of genetic diversity in this anthropogenically impacted area. Even though we could not identify a single anthropogenic factor influencing the population structure, we found populations to be highly isolated on local scales, clearly indicating that there are barriers to gene flow, which were not detectable with the methods used here.
In this study we found a considerably higher differentiation between populations of G. fossarum clade 11 in the human-impacted Sauerland area than was expected based on previous genetic analyses. The strong isolation was supported by two independent molecular marker systems, indicating that the realized dispersal was low in the study area. Also contrary to our hypotheses we found only a slight isolation by distance (IBD) pattern and structuring of populations according to river catchments but rather a strong geographic pattern (east-west differentiation). In view of published data, the dispersal ability of G. fossarum clade 11 specimens should be sufficient enough to create an IBD pattern. In the absence of this we conclude that there are barriers preventing gene flow partly, even between neighboring populations. Despite the likely effect of larger reservoirs on connectivity, we could not determine specific anthropogenic factors that directly influence the population structure. In fact, it was best predicted by the independent action of genetic drift at local sites after initial colonization. These effects are likely enhanced by the multitude of anthropogenic stressors, because G. fossarum is sensitive to many anthropogenic stressors. As G. fossarum clade 11 is widely distributed and is frequently found in the Sauerland area we conclude that its colonization ability over long time scales is good. The same holds likely true for its ability to establish new populations based on few colonizers, since long distance dispersal by animal vectors is unlikely to occur frequently . Even though we could not determine specific anthropogenic factors hindering gene flow, it is likely that the dispersal is influenced by these factors because of the exceptionally strong differentiation we found here in contrast to expectations based the findings of an earlier study (Westram et al.  and the much broader distribution of clade 11 in comparison with the other G. fossarum clades. These findings highlight the importance to take regional factors into account when predicting the dispersal ability of species. Further more research is needed to determine the most important barriers restricting gene flow between populations to prevent further losses of genetic diversity and maintain an intact ecosystem.
AMOVA, Analysis of molecular variance; CO1, cytochrome c oxidase 1 gene; FDR, False discovery rate; F ST , Fixation index; HWE, Hardy-Weinberg equilibrium; IBD, isolation by distance; SHM, Stream Hierarchy Model; VIF, variance inflation factor
We thank Hannah Weigand, Jan Macher, Vasco Elbrecht, Markus Patschke, Vivienne Dobrzinski, Lisa Pöttker, Janis Neumann and Tobias Traub for assistance with sampling, Maria Gies and Martin Sondermann for providing some of the samples and Hannah Weigand, Vivienne Dobrzinski, Stefan Gerzmann, Markus Liebert, Anna Eckert and Julia van der Mond for help in the lab. We also gratefully acknowledge the help of the Ruhr-Verband and LANUV in providing the used ecological data. We further thank Ralph Tollrian, Hannah Weigand, Alexander Weigand, Jan Macher, Arne Beermann and Christian K. Feld for helpful discussions and three anonymous Reviewers from Axios Review for valuable comments. We thank the Bioedit copy editing service for correcting language in an earlier version of the manuscript.
This work was funded by a junior research group grant of the Kurt Eberhard Bode Foundation to FL.
Availability of data and material
Sequences of the 40 haplotypes are available in GenBank under the accession numbers KX065366 – KX065405. Individual Sequences and microsatellite data can be made available upon request.
MW and FL conceived and designed the study. MW collected the samples, performed the laboratory work and the data analysis. MW wrote the manuscript together with FL and both authors approved the final version of it.
The authors declare that they have no competing interests.
Consent for publication
Ethics approval and consent to participate
No special permissions were required for sampling G. fossarum. We obtained special permissions to collect specimens at sites in nature reserves.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Illies J, editor. Limnofauna europaea. Stuttgart: Gustav Fischer Verlag; 1978.Google Scholar
- Bunn S, Hughes J. Dispersal and recruitment in streams: evidence from genetic studies. J North Am Benthol Soc. 1997;16:338–46.View ArticleGoogle Scholar
- Bohonak A, Jenkins D. Ecological and evolutionary significance of dispersal by freshwater invertebrates. Ecol Lett. 2003;6:783–96.View ArticleGoogle Scholar
- Witt J, Hebert P. Cryptic species diversity and evolution in the amphipod genus Hyalella within central glaciated North America: a molecular phylogenetic approach. Can J Fish Aquat Sci. 2000;57:687–98.View ArticleGoogle Scholar
- Pfenninger M, Staubach S, Albrecht C, Streit B, Schwenk K. Ecological and morphological differentiation among cryptic evolutionary lineages in freshwater limpets of the nominal form‐group Ancylus fluviatilis (OF Müller, 1774). Mol Ecol. 2003;12:2731–45.View ArticlePubMedGoogle Scholar
- Pauls S, Lumbsch H, Haase P. Phylogeography of the montane caddisfly Drusus discolor: evidence for multiple refugia and periglacial survival. Mol Ecol. 2006;15:2153–69.View ArticlePubMedGoogle Scholar
- Schwenk K, Posada D, Hebert P. Molecular systematics of European Hyalodaphnia: the role of contemporary hybridization in ancient species. Proc Biol Sci. 2000;267:1833–42.View ArticlePubMedPubMed CentralGoogle Scholar
- Katouzian A, Sari A, Macher J, Weiss M, Saboori A, Leese F, et al. Drastic underestimation of amphipod biodiversity in the endangered Irano-Anatolian and Caucasus biodiversity hotspots. Sci Rep. 2016;6:22507.View ArticlePubMedPubMed CentralGoogle Scholar
- Müller J. Mitochondrial DNA, variation and the evolutionary history of cryptic Gammarus fossarum types. Mol Phylogenet Evol. 2000;15:260–8.View ArticlePubMedGoogle Scholar
- Weiss M, Macher J, Seefeldt M, Leese F. Molecular evidence for further overlooked species within the Gammarus fossarum complex (Crustacea: Amphipoda). Hydrobiologia. 2014;721:165–84.View ArticleGoogle Scholar
- Copilaş‐Ciocianu D, Petrusek A. The southwestern Carpathians as an ancient centre of diversity of freshwater gammarid amphipods: insights from the Gammarus fossarum species complex. Mol Ecol. 2015;24:3980–92.View ArticlePubMedGoogle Scholar
- Siegismund H. Genetic differentiation in populations of freshwater amphipods Gammarus roeseli and Gammarus fossarum. Hereditas. 1988;109:269–76.View ArticleGoogle Scholar
- Westram A, Jokela J, Baumgartner C, Keller I. Spatial distribution of cryptic species diversity in European freshwater amphipods (Gammarus fossarum) as revealed by pyrosequencing. PLoS One. 2011;6:e23879.View ArticlePubMedPubMed CentralGoogle Scholar
- Feckler A, Zubrod J, Thielsch A, Schwenk K, Schulz R, Bundschuh M. Cryptic species diversity: an overlooked factor in environmental management? J Appl Ecol. 2014;51:958–967.View ArticleGoogle Scholar
- Meijering M. Lack of oxygen and low pH as limiting factors for Gammarus in Hessian brooks and rivers. Hydrobiologia. 1991;223:159–69.View ArticleGoogle Scholar
- Ladewig V, Jungmann D, Köhler H, Schirling M, Triebskorn R, Nagel R. Population structure and dynamics of Gammarus fossarum (Amphipoda) upstream and downstream from effluents of sewage treatment plants. Arch Environ Contam Toxicol. 2006;50:370–83.View ArticlePubMedGoogle Scholar
- Bundschuh M, Zubrod J, Schulz R. The functional and physiological status of Gammarus fossarum (Crustacea; Amphipoda) exposed to secondary treated wastewater. Environ Pollut. 2011;159:244–9.View ArticlePubMedGoogle Scholar
- Schmidlin L, von Fumetti S, Nagel P. Copper sulphate reduces the metabolic activity of Gammarus fossarum in laboratory and field experiments. Aquat Toxicol. 2015;161:138–45.View ArticlePubMedGoogle Scholar
- Eisenring M, Altermatt F, Westram A, Jokela J. Habitat requirements and ecological niche of two cryptic amphipod species at landscape and local scales. Ecosphere. 2016;7:e01319. doi:10.1002/ecs2.1319.
- Stürzbecher C, Müller J, Seitz A. Coexisting Gammarus fossarum types (Amphipoda) in Central Europe: regular patterns of population dynamics and microdistribution. Crustac Biodivers Crisis Proc Fourth Int Crustac Congr Amst Neth July. 1998;20–24(1998):287–93.Google Scholar
- Müller J, Partsch E, Link A. Differentiation in morphology and habitat partitioning of genetically characterized Gammarus fossarum forms (Amphipoda) across a contact zone. Biol J Linn Soc. 2000;69:41–53.View ArticleGoogle Scholar
- Feckler A, Thielsch A, Schwenk K, Schulz R, Bundschuh M. Differences in the sensitivity among cryptic lineages of the Gammarus fossarum complex. Sci Total Environ. 2012;439:158–64.View ArticlePubMedGoogle Scholar
- Westram A, Baumgartner C, Keller I, Jokela J. Are cryptic host species also cryptic to parasites? Host specificity and geographical distribution of acanthocephalan parasites infecting freshwater Gammarus. Infect Genet Evol. 2011;11:1083–90.View ArticlePubMedGoogle Scholar
- Westram A, Jokela J, Keller I. Hidden Biodiversity in an Ecologically Important Freshwater Amphipod: Differences in Genetic Structure between Two Cryptic Species. PLoS One. 2013;8:e69576.View ArticlePubMedPubMed CentralGoogle Scholar
- Zwick P. Stream habitat fragmentation—a threat to biodiversity. Biodivers Conserv. 1992;1:80–97.View ArticleGoogle Scholar
- Palmer M, Menninger H, Bernhardt E. River restoration, habitat heterogeneity and biodiversity: a failure of theory or practice? Freshw Biol. 2010;55:205–22.View ArticleGoogle Scholar
- Baggiano O, Schmidt D, Sheldon F, Hughes J. The role of altitude and associated habitat stability in determining patterns of population genetic structure in two species of Atalophlebia (Ephemeroptera: Leptophlebiidae). Freshw Biol. 2011;56:230–49.View ArticleGoogle Scholar
- Hughes, Schmidt, McLean & W. Population genetic structure in stream insects: What have we learned? In: Lancaster J, Briers RA, editors. Aquatic insects: challenges to populations. Wallingford: CAB International; 2008. p. 268–288. Google Scholar
- Cockerham C, Weir B. Estimation of gene flow from F-statistics. Evolution. 1993;47:855–63.View ArticleGoogle Scholar
- Rousset F. Genetic differentiation and estimation of gene flow from F-statistics under isolation by distance. Genetics. 1997;145:1219–28.PubMedPubMed CentralGoogle Scholar
- Danancher D, Cellot B, Dolédec S, Reynaud D. Isolation and characterization of the first eight microsatellite loci in Gammarus fossarum (Crustacea, Amphipoda) and cross‐amplification in Gammarus pulex and Gammarus orinos. Mol Ecol Resour. 2009;9:1418–21.View ArticlePubMedGoogle Scholar
- Westram A, Jokela J, Keller I. Isolation and characterization of ten polymorphic microsatellite markers for three cryptic Gammarus fossarum (Amphipoda) species. Conserv Genet Resour. 2010;2:401–4.View ArticleGoogle Scholar
- Dumont U, Anderer P, Schwevers U. Handbuch Querbauwerke. Düsseldorf: Ministerium für Umwelt und Naturschutz, Landwirtschaft und Verbraucherschutz des Landes Nordrhein-Westfalen; 2005.Google Scholar
- Siegismund H, Müller J. Genetic structure of Gammarus fossarum populations. Heredity. 1991;66:419–36.View ArticleGoogle Scholar
- Müller J. Invasion history and genetic population structure of riverine macroinvertebrates. Zoology. 2001;104:346–55.View ArticlePubMedGoogle Scholar
- Alp M, Keller I, Westram A, Robinson C. How river structure and biological traits influence gene flow: a population genetic study of two stream invertebrates with differing dispersal abilities. Freshw. Biol. 2012;57:969–981.View ArticleGoogle Scholar
- Meffe G, Vrijenhoek R. Conservation genetics in the management of desert fishes. Conserv Biol. 1988;2:157–69.View ArticleGoogle Scholar
- Hughes J, Schmidt D, Finn D. Genes in Streams: Using DNA to Understand the Movement of Freshwater Fauna and Their Riverine Habitat. Bioscience. 2009;59:573–83.View ArticleGoogle Scholar
- QGIS Development Team. QGIS Geographic Information System. Open Source Geospatial Foundation Project. http://qgis.osgeo.org. 2015
- ATKIS. ATKIS - Objektartenkatalog Basis-DLM. Version 3.2 [Internet]. 2007. Available from: www.adv-online.de/AAA-Modell/ATKIS/
- LAWA. Länderarbeitsgemeinschaft Wasser. Gewässerstrukturgütekartierung in der Bundesrepublik Deutschland: Verfahren für kleine und mittelgroße Fließgewässer. 1st ed. Berlin: Kulturbuch-Verlag GmbH; 2000. Google Scholar
- Raven P, Holmes N, Charrier P, Dawson F, Naura M, Boon P. Towards a harmonized approach for hydromorphological assessment of rivers in Europe: a qualitative comparison of three survey methods. Aquat Conserv Mar Freshw Ecosyst. 2002;12:405–24.View ArticleGoogle Scholar
- European Commission. Directive 2000/60/EEC, Establishing a framework for community action in the field of water policy. Off. J. Eur. Communities. 2000;1–71, Brussels, Belgium.Google Scholar
- Sunnucks P, Hales D. Numerous transposed sequences of mitochondrial cytochrome oxidase I-II in aphids of the genus Sitobion (Hemiptera: Aphididae). Mol Biol Evol. 1996;13:510–24.View ArticlePubMedGoogle Scholar
- Folmer O, Black M, Hoeh W, Lutz R, Vrijenhoek R. DNA primers for amplification of mitochondrial cytochrome c oxidase subunit I from diverse metazoan invertebrates. Mol Mar Biol Biotechnol. 1994;3:294–9.PubMedGoogle Scholar
- Kearse M, Moir R, Wilson A, Stones-Havas S, Cheung M, Sturrock S, et al. Geneious Basic: an integrated and extendable desktop software platform for the organization and analysis of sequence data. Bioinformatics. 2012;28:1647–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Katoh K, Standley D. MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Mol Biol Evol. 2013;30:772–80.View ArticlePubMedPubMed CentralGoogle Scholar
- Excoffier L, Lischer H. Arlequin suite ver 3.5: a new series of programs to perform population genetics analyses under Linux and Windows. Mol Ecol Resour. 2010;10:564–7.View ArticlePubMedGoogle Scholar
- Teacher A, Griffiths D. HapStar: automated haplotype network layout and visualization. Mol Ecol Resour. 2011;11:151–3.View ArticlePubMedGoogle Scholar
- Tamura K, Stecher G, Peterson D, Filipski A, Kumar S. MEGA6: molecular evolutionary genetics analysis version 6.0. Mol Biol Evol. 2013;30:2725–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B. 1995;57:289–300.Google Scholar
- Van Oosterhout C, Hutchinson W, Wills D, Shipley P. MICRO-CHECKER: software for identifying and correcting genotyping errors in microsatellite data. Mol Ecol Notes. 2004;4:535–8.View ArticleGoogle Scholar
- Dempster A, Laird N, Rubin D. Maximum likelihood from incomplete data via the EM algorithm. J. R. Stat. Soc. Ser. B Stat. Methodol. 1977;39:1–38.Google Scholar
- Kalinowski S. Hp‐rare 1.0: a computer program for performing rarefaction on measures of allelic richness. Mol Ecol Notes. 2005;5:187–9.View ArticleGoogle Scholar
- Chapuis M, Estoup A. Microsatellite null alleles and estimation of population differentiation. Mol Biol Evol. 2007;24:621–31.View ArticlePubMedGoogle Scholar
- Excoffier L, Smouse P, Quattro J. Analysis of molecular variance inferred from metric distances among DNA haplotypes: application to human mitochondrial DNA restriction data. Genetics. 1992;131:479–91.PubMedPubMed CentralGoogle Scholar
- Anderer P, Dumont U, Kolf R. Das Querbauwerke-Informationssystem QUIS-NRW. Wasser Abfall. 2007;7-8:10–4.Google Scholar
- Guillot G, Mortier F, Estoup A. Geneland: a computer package for landscape genetics. Mol Ecol Notes. 2005;5:712–5.View ArticleGoogle Scholar
- Guillot G, Santos F, Estoup A. Analysing georeferenced population genetics data with Geneland: a new algorithm to deal with null alleles and a friendly graphical user interface. Bioinformatics. 2008;24:1406–7.View ArticlePubMedGoogle Scholar
- R Core Team. R: A Language and Environment for Statistical Computing. R Found. Stat. Comput. Vienna Austria. 2015. https://www.r-project.org.
- Naimi B. usdm: Uncertainty Analysis for Species Distribution Models. R package version 1.1-15. 2015; Available from: http://CRAN.R-project.org/package=usdm
- Elbrecht V, Feld C, Gies M, Hering D, Sondermann M, Tollrian R, et al. Genetic diversity and dispersal potential of the stonefly Dinocras cephalotes in a central European low mountain range. Freshw Sci. 2014;33:181–92.View ArticleGoogle Scholar
- Lagrue C, Wattier R, Galipaud M, Gauthey Z, Rullmann J, Dubreuil C, et al. Confrontation of cryptic diversity and mate discrimination within Gammarus pulex and Gammarus fossarum species complexes. Freshw Biol. 2014;59:2555–70.View ArticleGoogle Scholar
- Zickovich J, Bohonak A. Dispersal ability and genetic structure in aquatic invertebrates: a comparative study in southern California streams and reservoirs. Freshw Biol. 2007;52:1982–96.View ArticleGoogle Scholar
- Foulquier A, Malard F, Lefébure T, Douady C, Gibert J. The imprint of Quaternary glaciers on the present-day distribution of the obligate groundwater amphipod Niphargus virei (Niphargidae). J Biogeogr. 2008;35:552–64.View ArticleGoogle Scholar
- Smith P, Smith B. Small‐scale population‐genetic differentiation in the New Zealand caddisfly Orthopsyche fimbriata and the crayfish Paranephrops planifrons. N Z J Mar Freshw Res. 2009;43:723–34.View ArticleGoogle Scholar
- Bilton D, Freeland J, Okamura B. Dispersal in freshwater invertebrates. Annu Rev Ecol Syst. 2001;32:159–81.View ArticleGoogle Scholar
- Pfenninger M, Salinger M, Haun T, Feldmeyer B. Factors and processes shaping the population structure and distribution of genetic variation across the species range of the freshwater snail Radix balthica (Pulmonata, Basommatophora). BMC Evol Biol. 2011;11:135.View ArticlePubMedPubMed CentralGoogle Scholar
- Figuerola J, Green A. How frequent is external transport of seeds and invertebrate eggs by waterbirds? A study in Donana. SW Spain Arch Für Hydrobiol. 2002;155:557–65.View ArticleGoogle Scholar
- Figuerola J, Green A. Dispersal of aquatic organisms by waterbirds a review of past research and priorities for future studies. Freshw Biol. 2002;47:483–94.View ArticleGoogle Scholar
- Schwentner M, Timms B, Richter S. Flying with the birds? Recent large‐area dispersal of four Australian Limnadopsis species (Crustacea: Branchiopoda: Spinicaudata). Ecol Evol. 2012;2:1605–26.View ArticlePubMedPubMed CentralGoogle Scholar
- Peck S. Amphipod dispersal in the fur of aquatic mammals. Can Field-Nat. 1975;89:181–2.Google Scholar
- Vanschoenwinkel B, Waterkeyn A, Vandecaetsbeek T, Pineau O, Grillas P, Brendonck L. Dispersal of freshwater invertebrates by large terrestrial mammals: a case study with wild boar (Sus scrofa) in Mediterranean wetlands. Freshw Biol. 2008;53:2264–73.Google Scholar
- Cook B, Bunn S, Hughes J. A comparative analysis of population structuring and genetic diversity in sympatric lineages of freshwater shrimp (Atyidae: Paratya): concerted or independent responses to hydrographic factors? Freshw Biol. 2007;52:2156–71.View ArticleGoogle Scholar
- Previsić A, Walton C, Kucinić M, Mitrikeski P, Kerovec M. Pleistocene divergence of Dinaric Drusus endemics (Trichoptera, Limnephilidae) in multiple microrefugia within the Balkan Peninsula. Mol Ecol. 2009;18:634–47.View ArticlePubMedGoogle Scholar
- Watanabe K, Monaghan M, Takemon Y, Omura T. Dispersal ability determines the genetic effects of habitat fragmentation in three species of aquatic insect. Aquat Conserv Mar Freshw Ecosyst. 2010;20:574–9.View ArticleGoogle Scholar
- Schultheis A, Hughes J. Spatial patterns of genetic structure among populations of a stone‐cased caddis (Trichoptera: Tasimiidae) in south‐east Queensland, Australia. Freshw Biol. 2005;50:2002–10.View ArticleGoogle Scholar
- Murphy N, Guzik M, Wilmer J. The influence of landscape on population structure of four invertebrates in groundwater springs. Freshw Biol. 2010;55:2499–509.View ArticleGoogle Scholar
- Slatkin M. Isolation by distance in equilibrium and non-equilibrium populations. Evolution. 1993;47:264–79.View ArticleGoogle Scholar
- Keyghobadi N, Roland J, Strobeck C. Genetic differentiation and gene flow among populations of the alpine butterfly, Parnassius smintheus, vary with landscape connectivity. Mol Ecol. 2005;14:1897–909.View ArticlePubMedGoogle Scholar
- Finn D, Theobald D, Black W, Poff N. Spatial population genetic structure and limited dispersal in a Rocky Mountain alpine stream insect. Mol Ecol. 2006;15:3553–66.View ArticlePubMedGoogle Scholar
- Monaghan M, Spaak P, Robinson C, Ward J. Genetic differentiation of Baetis alpinus Pictet (Ephemeroptera: Baetidae) in fragmented alpine streams. Heredity. 2001;86:395–403.View ArticlePubMedGoogle Scholar
- Dudgeon D, Arthington A, Gessner M, Kawabata Z, Knowler D, Lévêque C, et al. Freshwater biodiversity: importance, threats, status and conservation challenges. Biol Rev. 2006;81:163–82.View ArticlePubMedGoogle Scholar
- Shama L, Kubow K, Jokela J, Robinson C. Bottlenecks drive temporal and spatial genetic changes in alpine caddisfly metapopulations. BMC Evol Biol. 2011;11:278.View ArticlePubMedPubMed CentralGoogle Scholar