Genetic population structure of a top predatory fish (northern pike, Esox lucius ) covaries with anthropogenic alteration of freshwater ecosystems

1. Esox lucius L. (northern pike) inhabits river,


| INTRODUC TI ON
Animal and plant populations are evolving in human-dominated landscapes (Christie & Knowles, 2015;Ortego et al., 2015;Sexton et al., 2013;Smith & Bernatchez, 2008). Studies investigating fish communities and populations have demonstrated that aquatic ecosystems are particularly sensitive to ecological changes (e.g. Whitehead et al., 2017), including habitat fragmentation, which increases isolation of populations through migration barriers (Waples et al., 2017). Alternatively, human activities are involved in connecting previously isolated aquatic ecosystems and fish populations, e.g. through the artificial opening of new routes for migration via the construction of artificial canals that connect previously disconnected catchments (Lin et al., 2020) or intentional transplantation of individuals through stocking and introductions (Laikre et al., 2010).
In particular, stocking as a widespread fisheries-management and conservation practice mediates secondary contacts among fish populations, leading to multiple, difficult-to-predict outcomes, and possibly contributing to either the erosion or reinforcement of genetic differences through hybridisation (Allendorf et al., 2001;Diana et al., 2017;Hansen, 2002;Harbicht et al. 2014;Marie et al., 2012).
Stocking and facilitated natural migration through human-made artificial connections among different river basins have substantially increased the speed of human-assisted secondary contact among subpopulations of fish relative to natural factors alone and thereby can shape contemporary genetic structures through rapid evolution (Lin et al., 2020;van Poorten et al., 2011).
Depending on local ecological conditions, particularly the degree of local natural recruitment, stocked fish either disappear without a trace or become established to varying degrees (Johnston et al., 2018), which eventually leads to admixture of non-native with native populations or even complete replacement of native populations (Englbrecht et al., 2002;van Poorten et al., 2011).
Numerous studies have shown that hybridisation is enhanced at the interspecific level in ecologically perturbed habitats, e.g. in African cichlids Cichlidae (Seehausen et al., 1997), sculpins Cottidae (Nolte et al., 2005), European whitefish Coregonus spp. (Bittner et al., 2010;Vonlanthen et al., 2012), and trout Oncorhynchus spp. (Heath et al., 2010). All of these examples support the idea that the outcome of secondary contact between different populations or lineages is influenced by local ecological conditions, but as yet little is known about how local environmental conditions determine population structure of intensively exploited species at large geographical scales.
Esox lucius L., the northern pike, is a freshwater top predatory fish species. The species may well be affected by contemporary environmental change caused by human activities, including loss of floodplains in rivers and elevated nutrient inputs in lakes that reduce the extent of submerged macrophytes (Craig, 1996;Skov & Nilsson, 2018). To fulfil their life cycle, pike strongly depend on aquatic macrophytes providing shelter for early developmental stages and camouflage to hunt for prey during the juvenile stage (Craig, 1996). After reaching sexual maturation, floodplains, and more generally submerged macrophytes, are equally essential, serving as spawning habitat for the phytophilic pike (Casselman & Lewis, 1996;Raat, 1988). Thus, changes in macrophyte extension through loss of floodplains or eutrophication of lakes affect the reproductive potential of pike, and may in turn also affect the stocking intensity of pike by motivating increased stocking efforts as natural recruitment declines (Cowx, 1994;Guillerault et al., 2018;Hühn et al., 2014). Gandolfi et al. (2017), studying the invasion process of Esox lucius into closely related native Italian Esox flaviae/cisalpinus (Bianco & Delmastro, 2011;Lucentini et al., 2011) populations, observed a mosaic-type distribution of the two species and different degrees of genetic admixture, possibly as a result of the different ecological status of the studied water bodies. Lake Garda provides good ecological conditions for the native E. flaviae/cisalpinus to fulfil its life cycle. Therefore, native pike seem to be able to out-compete the establishment of introduced northern pike. By contrast, other Italian waters with a poorer ecological status have been found to be strongly prone to genetic introgression (Gandolfi et al., 2017).
Similarly, in Danish pike populations at least two regional clusters were identified referring to the hydrogeographic regions of the Baltic and the North Sea (see also Wennerström et al., 2018), which could be further sub-divided at different river catchment scales (Bekkevold et al., 2015). The authors speculated that local-scale deviations from native genetic signatures typical of catchments were the result of human-caused changes in local habitat quality (e.g. availability of submerged macrophytes) or could be due to historical geological alterations (Bekkevold et al., 2015). In addition, stocking may leave a strong legacy in the genetic structure of local pike populations (Möller et al., 2021), which will also be determined by the local environment (Larsen et al., 2005;Nordahl et al., 2019). The reviewed literature suggests that to fully understand the contemporary genetic structures in widely dispersed fish species, such as pike, the interaction of local ecological status in the context of the potential for human-assisted secondary contact is important to consider.
In Europe, in the 2000s the Water Framework Directive (WFD) was introduced as a policy to monitor and improve freshwater ecosystem quality and the ecological status of rivers and large lakes beyond 50 ha in dimension (European Commission, 2000). Therefore, all European rivers and large lakes are today regularly assessed based freshwater ecosystems may influence the genetic structures present in a freshwater top piscivore at both local and regional scales.

K E Y W O R D S
admixture, ancestry distribution, ecological modification, secondary contact, stocking on a range of biological indices, from phytoplankton to fish, to assess their ecological status and inform management actions (Hering et al., 2010). Although different taxa respond differently to anthropogenic drivers, the ecological assessment of lakes often tracks eutrophication signals, while hydromorphological changes are a dominant stressor in European rivers (Hering et al., 2013). Both drivers have negative effects on natural recruitment of pike (Craig, 1996;Raat, 1988;Skov & Nilsson, 2018), and hence indicators of ecological status of lakes may relate systematically to the population structure of pike. Genetic population structures within species, however, are currently not considered in the WFD. It is nevertheless likely that the ecological status of water bodies is systematically related to the meta-population structure of individual species. Other assessments of ecosystem status of freshwaters exist, e.g. the degree of wetland loss in rivers (BFN, 2009) or the degree of eutrophication in lakes (Hering et al., 2013), but similar to the WFD assessments, limited data exist to link these water body-specific indicators to genetic structure of individual fish species.
The objective of the present study was to explore whether the ecological status of the inland water bodies in Germany would be associated with the genetic structure of contemporary pike populations, particularly with respect to potential genetic admixture among populations of E. lucius of different hydrogeographic regions (i.e. catchments or basins). We evaluated the presence of population structure that permitted genetic assignment of individuals to their origins, and used this information to identify large-scale and local signs for intra-specific admixture. We then tested whether the observed genetic patterns correlated with the ecological status of the water bodies assessed according to the European WFD and, as a robustness check, according to an index of wetland status of rivers and the trophic state of lakes. We assumed that both indices would correlate with the quality of pike habitat and thus index the degree of natural recruitment of pike in a local ecosystem. We further assumed that the degree of local natural recruitment would affect how local populations would respond to the secondary contact with foreign genotypes (Hühn et al., 2014), and in turn how resilient the genetic structure of a local population would be against genetic invasion. We overall hypothesised that the ecological status of local ecosystems shapes the genetic structure of northern pike across Germany.

| Sampling and DNA extraction
Sampling was performed in 2011 and 2012 covering as many relevant water systems and types as possible over a wide geographical area in Germany. Our standard for inclusion in the study was a minimum of 10 individuals characterised by at least 14 microsatellites. The sample collection comprised specimens from five river catchments draining into the North Sea, six catchments draining into the Baltic Sea, and one catchment draining into the Black Sea (Table 1). Three ecosystem types were covered including 26 lakes, 24 rivers, and three brackish coastal water areas (Table 1). Pike were sampled from water bodies covering th complete range of ecological states according to the WFD, the quality of the wetlands (i.e. wetland loss) following a recent report by the German Agency for Nature Conservation (BFN, 2009) and publicly available information on trophic states in lakes. For three small water bodies <50 ha (Alte Würm, Kleiner Döllnsee, and Schulzensee), ecological status data following the European WFD were not available.
Fin and muscle tissue samples of pike were collected by commercial and recreational fishers, research organisations and state fishery authorities. Samples obtained as frozen tissues were thawed in absolute ethanol (Thomas Geyer) at room temperature and subsequently transferred to fresh ethanol following Eschbach (2012). Samples from research organisations were generally obtained preserved in ethanol, while samples from anglers were obtained air-dried. DNA of all types of samples was extracted with the nexttec TM DNA isolation kit (Biozym Scientific GmbH, Hess. Oldendorf, Germany) according to the manufacturer's instruction.

| Genetic marker analysis
We employed nuclear as well as mitochondrial markers to infer population structure and to compare our data with published data. Fifteen polymorphic microsatellites ( Table 2, Table S1) for pike were selected for high information content according to Eschbach and Schöning (2013). These were employed to analyse a subset of 1,384 samples of 53 populations with an average sample size of 22.1 ± 9.8 (mean ± SD) individuals per population, and 96.4 ± 94.0 (mean ± SD) individuals per river catchment.
Microsatellites were co-amplified in multiplex polymerase chain reaction (PCR; Table 2) with a Thermocycler T Gradient machine (Biometra) using the Qiagen ® Multiplex PCR Kit (Qiagen). Forward primers were 5′-labelled with fluorescent dyes HEX, NED, or FAM (SMB Services in Molecular Biology GmbH; Table 2). Polymerase chain reaction started with 15 min at 95°C, followed by 35 cycles of 0.5 min at 94°C, 1.5 min at 58°C, 1.5 min at 72°C, and finishing with 10 min at 72°C. Fragments were sized with an Applied Biosystems 3500xL Sequencer equipped with a 24-capillary array.
Haplotype analysis of the mitochondrial cytochrome b gene (cyt b) was carried out to link the present data set with the broad scale phylogeographic analysis by Skog et al. (2014) using the primers described by Grande et al. (2004) Note: High-resolution microsatellites selected according to Eschbach and Schöning (2013) for population genetic analysis of species with low genetic variability. References  Nei & Chesser (1983).

TA B L E 2
Microsatellites used for genotyping sample size was 9.1 ± 2.1 individuals per population and 16.7 ± 8.5 per river catchment. Individual forward and reverse sequences were assembled using Seqman (DNA star package) and the resulting contigs were checked by eye to correct sequencing errors. Sequences of all main and sub haplotypes were deposited at the NCBI database (Acc. no. KY399416-KY399442).  (Felsenstein, 1981) (Bekkevold et al., 2015;Wennerström et al., 2018).

| Analysis of genetic data
However, as a further test we also ran additional models for k = 2 to k = 8 to examine the robustness of our findings on the covariance of genetic structure and environmental status.
We used the STRUCTURE Q value that describes the fraction of the genome inherited from the drainage basin-specific lineage as native ancestry (NA). To express all foreign genetic influences in relation to NA, hybrid indices (HI) for each individual were inferred from the individual NA values. Using the formula HI = 1 -(2 × |0.5 -NA|) results in a value of 1.0, if the native and foreign ancestries contributed equally to an individual's genetic composition (maximal hybrid status, as found in a first-generation hybrid), and a value of 0, if only the native or only the foreign ancestry contributed to an individual's genetic composition. CLUSTAL X Version 2 (Larkin et al., 2007) was used to align all cyt b sequences along with 24 reference sequences of haplotypes described by Skog et al. (2014). The alignment was trimmed to a length of 1,174 bp, which contained the sites that were diagnostic for the groups of haplotypes described by Skog et al. (2014). This alignment was used to confirm the presence or absence of the respective haplotypes in the populations studied here. The relationship among all haplotypes was visualised using a median-joining network as described by Bandelt et al. (1999) that was constructed using the program NETWORK 4.6.1.3 (Fluxus Technology Ltd).

| Environmental effects on genetic structure
In total, 585 pike individuals from 24 lakes and 392 pike individuals and therefore, we considered the variance attributed to water bodies as a random effect. In addition, to account for a higher probability of natural exchange among individuals sampled in specific water bodies within a basin, water bodies were nested within catchments.
We then modelled HI and NA on a set of predictors using a linear predictor with unknown coefficients and a link function (logit). The predictors considered were: the type of water body (lake or river); its level of modification following the classification of the European WFD (not modified or highly modified); and its ecological status as a numerical covariate from 1 (very good) to 5 (very poor) according to the European WFD or 1 (very good) to 4 (poor) of our index of wetland quality/trophic state (see below). The level of modification was derived from public data on the WFD, which distinguishes heavily modified water bodies from those that are not heavily modified as different policy goals apply to these two conditions. We included degree of modification only as a control variable as we expected the most relevant information in our metric of ecological status and ecosystem type (river vs. lake).
The raw data for the ecological status following the WFD was retrieved from site-specific ecological assessment data supplied by community-based approach for the major taxa groups and uses the poorest assessment results of any of the four taxa to identify the ecosystem status (Hering et al., 2010). Thus, although readily available across Europe, the metric may not perfectly represent the conditions relevant for pike.
As a second index we also used a combined status assessment of the wetland status for rivers (derived from the national wetland report, BFN, 2009)  The key model presented in the paper was fitted to the data following the three hydrogeographical basins (k = 3 in STRUCTURE).
To account for the uncertainties associated with finding, the correct number of genetic clusters in complex data and the possibility that finer-grained population structure was not sufficiently captured by our division into three hydrogeographic regions, we repeated the modelling for results assuming a range of k values from 2 to 8.

| Assessment of genetic markers
All of the 15 microsatellite loci proved to be highly polymorphic with a total number of 9-37 different alleles over all pike populations and a mean number of 3.5-14.5 different alleles per population ( Table 2).
The potential presence of null alleles was detected in 1.6% of alleles over all loci and populations ( Populations showed deviation from Hardy-Weinberg equilibrium in 2.7 ± 1.9 loci (mean ± SD) reflected in significant heterozygote deficiencies in 2.7 ± 1.6 loci ( Figure S1). Linkage disequilibrium was detected in 3.3% of all possible loci combinations after Bonferroni correction ( Figure S2). Because departures were distributed over many loci and populations, all loci were used for population genetic analysis.
A 1,174-bp region of the mitochondrial cyt b gene with 48 variable positions was selected for the network analysis ( Figure S3); 208 sequences (including reference sequences) provided a total of 918 informative sites (excluding sites with gaps and missing data).

| Genetic structure of pike populations in Germany
Analysis of microsatellite genotypes with STRUCTURE suggested Second, PCoA based on F st values was employed (Figure 3).
Although variation was moderate (accumulated variance explained 23.2% by axes 1 and 2) the clear clustering into three groups reflected differences among the drainage basins of the North, Baltic, and Black seas, respectively (Table 3) were generally low, the tree topology proved to be stable and was consistent with the most likely STRUCTURE model predicting three main clades ( Figure S6) reconstruct an individual's ancestry (see supplement for a more detailed analysis).

| Population structure among the major hydrogeographic basins
Admixture analysis based on microsatellites revealed varying pro-   in 4), which was also mirrored in the distance-based consensus tree (Figure 2) and the frequency-based principal coordinate Table S2) of all pike samples. Despite low levels of variation (23.2% accumulated variation of axis 1 and 2) the three main clusters predicted by the most likely model of STRUCTURE ( Figure S6) were clearly resolved and admixed pike populations were positioned within the correct genetic context analysis ( Figure 3). Accordingly, the Rhine population (RHE2) appeared within the Baltic Sea cluster, and pike from GPS were placed within the Black Sea cluster. Proportion of ancestry: Figure 4 Proportion of ancestry: Note: Three genetic clusters of pike populations were identified belonging to the hydrogeographic regions of the North, Baltic, and Black seas, respectively (shaded areas indicate highest proportion of ancestry). Some populations exhibited high shares of non-native ancestry (indicated in italic writing). See Table 1

| Correlation of hybridisation levels with ecological status and ecosystem type
Employing hierarchical general linear models revealed that the ecological status of the water body as well as the type of ecosystem had a significant effect on the HI of pike populations. Specifically, the decline in the ecological status highly significantly covaried with HI (Table 4). Each unit of decrease of the ecological status led to an increase of HI by a factor (slope) of 0.19 (1.21 in raw scale) ± 0.05 with respect to the intercept (defined as the best ecological status).
Results were consistent when using the ecological status defined according to the European WFD (Table 4) (Table 4). While the hypothesis of a covariance between HI and the type and ecological status of the water body was supported, relationships of HI and the general level of modification of the water body (Table 4) and all two-level interaction effects were not supported (data not shown).
The NA exhibited a similar association with the deterioration of the ecological status of the water bodies ( Figure 6); however, the relationships were not statistically significant (Table 4). This held also true for the predictor type (i.e. lake vs. river) and modification (i.e. general degree of water body modification) of the water bodies F I G U R E 4 Map of Germany illustrating that genetic admixture on population level varied strongly and was not confined to a particular hydro-geographic region or river catchment therein. Black, grey, and white in the pie charts indicate genetic ancestry proportion of Black Sea, North Sea, and Baltic Sea hydrogeographic region, respectively. Different catchments are indicted by the shaded areas. Numbers indicate pooled populations as displayed in Table 3. Large pies indicate river populations, small pies lake populations (Table 4), as well as their two-level interactions (not shown), which were therefore removed from the model with NA.
The results mentioned above for the relationship with the ecological status following the WFD was mirrored by the covariance of the hybridisation index with the indicator for wetland quality/ trophic state (Table 4). As before, increasing losses of the wetlands and increasing eutrophication significantly increased the degree of hybridisation among subpopulations of pike (Table 4, Figure 6). Signs for hybridisation were also more pronounced in rivers relative to lakes (Table 4, Figure 6).
The environmental regression model was also fitted to ancestry coefficients as derived from STRUCTURE analyses assuming a range of k values from 2 to 8 to examine the robustness of the findings. While the same trends for covariation were detected across most values of k, the results were not significant for a k = 2 ( Figures S4 and S5). By contrast, the outcomes of the covariation of ancestry measures with ecological status as detected for a model assuming k = 3 genetic clusters were significant for k values of up to six ( Figures S4 and S5), in all cases showing that the degree of hybridisation increased with declining environmental quality (independent of the index) and was always larger in rivers relative to lakes.

| D ISCUSS I ON
Our analysis of the population genetic structure of pike from 53 sampling sites covered lakes and major rivers from all three hydrogeographic regions in Germany. Admittedly, we were unable to assign the pike genotypes unambiguously to a certain number of genetic clusters, which we attribute to a shallow and complex population structure previously reported for pike populations (Bekkevold et al., 2015;Jacobsen et al., 2005, Pedreschi et  of German pike populations was also supported by our analysis of mitochondrial haplotypes. Pike populations are not known to contain deeply divergent lineages in central Europe (Nicod et al., 2004;Skog et al., 2014). A haplotype NETWORK (Bandelt et al., 1999)  are sufficient to distinguish the lineages of pike studied here. Hence, the strongest support for the existence of three evolutionarily significant units (Moritz, 1994) of pike with different distribution areas was supported by multilocus microsatellite analyses.

| Genetic structure of pike in Germany
Using microsatellite markers, clear signs for genetic substructure were found and manifest as groups of individual sampling sites that containing mostly one genetic ancestry. Genetic clustering also revealed that pike from some sampling sites had genetic ancestry in multiple populations, which could be a sign of genetic admixture or be caused by a lack of resolution of the available genetic data (Lawson et al., 2018). A striking pattern in the data was that this inability to clearly assign a single genetic ancestry appeared to be particularly prevalent in rivers and was also correlated with ecological status. This suggests that a lack of genetic distinctness may not only be caused by the connectivity of ecosystems (which is likely to be more pronounced in rivers compared to isolated lakes, in particular when river catchments are connected via canals and other migration routes) but may also be related to an erosion of population structure that is associated with human activity.
Secondary contacts between divergent pike lineages are very likely to have increased in the last centuries as a result of anthropogenic activities. Humans have facilitated natural migration through human-made artificial connections among different river basins as well as stocking of economically important fish species. The latter represents an important factor that increases the potential for gene flow between populations naturally separated in space (Arlinghaus et al., 2015) and is known to shape the genetic diversity of the congeneric muskellunge (Esox masquinongy) in North America (Roegemont et al., 2019). Both processes are expected to lead to admixed genotypes and may lead to an erosion of population-specific genetic variance. Unfortunately, past stocking is generally not well documented in Germany (Arlinghaus et al., 2015), but certainly has occurred over decades in central Europe in pike and multiple other economically relevant fishes (Cowx, 1994;Guillerault et al., 2018;Kottelat & Freyhof, 2007;Larsen & Berg, 2004).
Our work relies on correlative evidence and thus cannot be conclusive. It is likely that the generally low level of genetic differentiation in pike populations (Wennerström et al., 2018)  Note: The effect of different linear predictors on the hybridisation index and the native ancestry was tested controlling for the random variance attributed to the individuals sampled in specific waterbodies nested within catchments (see Figure 6). The table shows the estimates (in logit scale) and their standard error (SE), the t-value statistics and their p-value (Pr(>|t|) for the model fitted with ecological quality according to EU Water Framework Directive (model 1) and the wetland and trophic status-WTS (model 2). Two-level interactions were non-significant in all cases and removed from the model. The estimates of the categorical variables were shown per one category with respect to the other (intercept). Significance codes: * Significant at the 0.05 probability level; ** Significant at the 0.01 probability level; and *** Significant at the 0.001 probability level. levels were observed in pike populations of the Ems and Weser catchments belonging to the North Sea hydrogeographic region. A possible explanation for the persistence of autochthonous populations is either a low level of local stocking or competitive exclusion of foreign genotypes by better-adapted native populations (Englbrecht et al., 2002;Eschbach et al., 2014;Gandolfi et al., 2017;Larsen et al., 2005;Möller et al., 2021;Roegemont et al., 2019).
In  (Powels et al., 2013). However, we also detected signatures of admixture between rather distant source populations, e.g. between pike of the rivers Oder in the east and Rhine in the west or between the rivers Danube in the south and Eider in the very north of Germany.
Stocking, rather than migration, is a more likely explanation here (Monk et al., 2020), because migration would probably have created a more coherent geographical pattern. Our data are in line with genetic structures of pike populations in Denmark at the intra-specific level (Bekkevold et al., 2015) and Italy at the inter-specific level (Gandolfi et al., 2017), both of which not always reflected natural catchment barriers and were probably caused by successful pike stock enhancement activities in the past. Similar results exist for muskellunge in North America (Roegemont et al., 2019).

| Impacts of ecosystem status on hybridisation
We detected various degrees of mixed ancestries in a range of sampling sites, either documenting admixture between genetically distinct populations or a lack of clear population differentiation. This effect was significantly associated with rivers and increased with the degradation of the ecological status of ecosystems. While the ecological status indicator of the WFD does not directly indicate which ecological factor was involved, our wetland/trophic state index is strongly suggestive that the loss of key pike habitat, particularly loss of access to submerged macrophytes, correlated with the increasing admixture rates. The quality of this inference depends on the sample sizes that were available for each population as well as the degree of differentiation between the presumed source populations.
Given the low resolution of microsatellites, it is certainly useful to revisit specific populations with a more powerful study design and genome wide marker coverage. Before such research becomes available, the current analysis suggests that the pike lineages hybridised upon secondary contact. This result bears general questions on why hybridisation proceeded with different intensity in different pike populations and whether pike of different origins are indeed isolated to some extent when they are brought into secondary contact.
Note again that the secondary contact might have been caused by F I G U R E 6 Relationships of native ancestries and hybridisation indices with habitat type, strength of modification, ecological quality (1 = very good to 5 = poor according to EU Water Framework Directive and the combined wetland/trophic status for pike ecology-WTSfrom 1 = very good to 5 = very poor) as obtained with HGLM analysis (see Table 4 for details). Baltic coastal waters (BAL2, BAL3 and BAL4) and freshwaters without ecological information (AWU, SUS, KDO) were excluded from analysis. IDs of water bodies are explained in Table 1. Significance codes: n.s. non-significant, * Significant at the 0.05 probability level, ** Significant at the 0.01 probability level and *** Significant at the 0.001 probability level stocking or any other factors, e.g. increased connectivity of water bodies among catchments.
We found that the individual admixture levels in pike, expressed as an HI, were not confined to a specific hydrogeographic region or any particular river catchment therein. Instead, it turned out that the HI increased significantly with decreasing ecological quality of a water body, and this result was robust for k values from three to six and for two different indices of ecological status. Albeit not statistically significant, we observed a congruent decrease of native ancestry with habitat modification. Thus, our data suggest environmental change drove genetic changes in pike populations and individuals by affecting the frequency of hybridisation among populations brought into secondary contact. The fact that the HI was only slightly lower in water bodies with low modifications as compared to the HI of highly modified waters implies that the admixture as such occurs in all populations and is not restricted to highly modified habitats.
Our analysis yielded a significantly higher HI in pike populations in rivers as compared to lake-dwelling pike, which is probably due to fundamental ecological differences between the two habitat types such as the increased natural connectivity in rivers, resource availability, productivity, habitat structure, and community composition (Hof et al., 2008;Irz et al., 2006). Most importantly, however, rivers and lakes vary in stability and disturbance frequency, including exposure to floods or minimum water flows, which occur more frequently in lotic than in lentic systems. Rivers of central Europe also have been more strongly modified, e.g. by removal of connectivity to floodplains and habitat simplification (BFN, 2009), which represent a central component of their disturbance regime and at the same time constitute essential spawning habitat for pike. In a meta-analysis comparing resistance of limnic, marine and terrestrial ecosystems towards invasive species, Alofs and Jackson (2014) demonstrated that lentic systems displayed a higher biotic resistance than lotic systems, which is in accordance with our findings of different susceptibilities towards hybridisation in river and lake pike populations.
Our observation that a lack of population genetic differentiation and signs for admixture in pike appears to be favoured in ecologically perturbed water bodies raises important questions about the mechanisms. The effect could first be caused by an increase of foreign genotypes that managed to invade a weakened native population (Englbrecht et al., 2002;Gandolfi et al., 2017) and it could be affected by demographic processes (Mathieu-Bégné et al., 2019). Alternatively, genetically admixed fish could be more competitive in the face of anthropogenic changes to the ecosystem. This would resemble the first step of a hybrid speciation scenario, where intraspecific hybrids are expected to be most successful when parental populations are not at their optimum (Abbott et al., 2013;Nolte & Tautz, 2010). Stelkens et al. (2014) showed that particularly the interactions of genetic variants between distant Saccharomyces strains can lead to a better survival in environments of decreasing quality. Thus, hybridisation can create biodiversity, resulting in novel phenotypes and adaptive change in response to environmental change (Charlesworth & Willis, 2009;Edmands, 2007;Sefc et al., 2017). Examples of these processes can be found among invaders conquering new environments that were not occupied by populations of the respective species before, as it was found for Cottus hybrids in the river Rhine (Nolte et al., 2005;Stemshorn et al., 2011), but also for spiders (Krehenwinkel & Tautz, 2013) and plants (Keller & Taylor, 2010). Likewise, in a previous study we observed increased intraspecific genetic diversity of zander (Sander lucioperca) in water bodies, where this fish species had been introduced in the late 19th century, a pattern that would be in line with an advantage of admixed individuals in the course of an invasion (Eschbach et al., 2014). Thus, new combinations of genes from different evolutionary backgrounds might enable fast adaptation, and thus increase the chance to survive under declining environmental conditions. However, careful future studies are needed to distinguish the adaptive scenario outlined here from neutral explanations that are related to abrupt changes in propagule pressure in fluctuating environments.

| Conclusions and implications
Our study revealed a novel relationship between ecosystem status, assessed under the European WFD and via a wetland/trophic state index, and the genetic structure of northern pike across Germany.
Due to the limitations in sample size, our wide geographical scope and the lack of time series data, our findings of the relationship of genetic structure and environmental status in pike necessitates more careful analysis using genetic markers of higher resolution. However, our work supports the novel hypothesis that habitat degradation can affect the genetic integrity within pike. Efforts to improve the ecological quality of lakes and rivers could therefore promote the maintenance of genetic structure. In the case of pike, this would mean reconnecting floodplains with rivers, or wetlands to lakes or brackish lagoons (Nordahl et al., 2019), and reducing nutrient inputs into lakes to increase macrophyte abundance. These actions would probably increase the pike population size and help to maintain local genetic biodiversity.

ACK N OWLED G EM ENTS
Particular thanks go to Yvonne Klaar, Asja Vogt, Jasmin Spamer, and Elke Bustorf for technical assistance in the lab, as well as Petra Kersten for invaluable advice on microsatellite analysis. We are also indebted to Ute Mischke and Jakob Sölter for kindly providing additional ecological data of the different water bodies. We thank Jochem Kail for creating the map in Figure 4 and help with sourcing environmental data. We are very much obliged to the many fishers, anglers, fisheries authorities and research organisations for providing tissue samples as well as a lot of helpful information. We are grateful to the whole Besatzfisch und

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.