Informing conservation strategies with museum genomics: Long‐term effects of past anthropogenic persecution on the elusive European wildcat

Abstract Like many carnivore species, European wildcats (Felis silvestris) have suffered severe anthropogenic population declines in the past, resulting in a strong population bottleneck at the beginning of the 20th century. In Germany, the species has managed to survive its near extinction in small isolated areas and is currently recolonizing former habitats owing to legal protection and concerted conservation efforts. Here, we SNP‐genotyped and mtDNA‐sequenced 56 historical and 650 contemporary samples to assess the impact of massive persecution on genetic diversity, population structure, and hybridization dynamics of wildcats. Spatiotemporal analyses suggest that the presumed postglacial differentiation between two genetically distinct metapopulations in Germany is in fact the result of the anthropogenic bottleneck followed by re‐expansion from few secluded refugia. We found that, despite the bottleneck, populations experienced no severe genetic erosion, nor suffered from elevated inbreeding or showed signs of increased hybridization with domestic cats. Our findings have significant implications for current wildcat conservation strategies, as the data analyses show that the two presently recognized wildcat population clusters should be treated as a single conservation unit. Although current populations appear under no imminent threat from genetic factors, fostering connectivity through the implementation of forest corridors will facilitate the preservation of genetic diversity and promote long‐term viability. The present study documents how museum collections can be used as essential resource for assessing long‐term anthropogenic effects on natural populations, for example, regarding population structure and the delineation of appropriate conservation units, potentially informing todays' species conservation.


| INTRODUC TI ON
The last decades have been characterized by an increasing and pervasive loss of biodiversity around the globe that has been mainly induced by human activities (Díaz, Settele, Brondízio, Ngo, Guèze, et al., 2019;Pimm et al., 2014). The associated anthropogenic impact on wildlife, however, and specifically the displacement of animals from their natural habitats has existed for a much longer time .
While the extent of the resulting species' loss or reduction is only now being fully acknowledged-for example, in many invertebrate communities-large mammals have been among the first to experience substantial population declines, geographic range contractions, and fragmentation of their habitats (Ceballos & Ehrlich, 2002;Morrison et al., 2007;Ripple et al., 2014). In contrast to large ungulates that were overexploited as game, large carnivores have been particularly affected by human-wildlife conflicts (Treves & Karanth, 2003), such as the extinct thylacine (Paddle, 2000). Perceived as competitors and imminent threats to human livelihoods, large carnivores have been extensively persecuted and consequently extirpated or driven to near extinction in most of Central Europe (Chapron et al., 2014). By the 1850s, several iconic species such as gray wolf, brown bear, and Eurasian lynx had been eradicated from major parts of the continent (Breitenmoser, 1998;Pereira & Navarro, 2015).
The progressive disappearance of these apex predators was followed by population growth of herbivorous prey species, but also of medium-sized carnivores and mesocarnivores (Prugh et al., 2009;Ripple et al., 2013Ripple et al., , 2014Ritchie & Johnson, 2009). At the same time, forest owners and hunters looking to replace profitable trophies with new prospects turned to hunt smaller carnivores such as the European wildcat (Felis silvestris, Schreber 1777) (Piechocki, 1990). The elusive carnivore was thence (mistakenly) held responsible for livestock damage and presented as a threat to humans, even if the animals' body size, prey spectrum, and habitat needs did not fit this behavior ( Figure 1a) (Müller-Using, 1965). Following the proclamation of a trophy price for hunted wildcats in 1781, populations suffered from massive persecution and experienced intense range contraction ( Figure 1b) (Reinert, 2017). Despite legal protection of the species through national legislations in the early 1900s, the European wildcat experienced a strong population bottleneck between 1920 and 1930 in Central Europe (Piechocki, 1990).
In Germany, wildcats were diminished to few remaining relict populations in low-mountain refugia such as the Harz Mountains, the Palatinate Forest, or the Hesse Highlands (Piechocki, 1990).
Eventually, the continued decline of wildcat populations was crucially-and positively-halted by the ban of steel snap traps in hunting in 1935 (Haltenorth, 1957).
In contrast to alarming global trends (Ripple et al., 2014), ranges of several large vertebrate species are expanding in Central Europe today (Cretois et al., 2020). The European wildcat serves as one successful example of rewilding densely populated European landscapes (Pereira & Navarro, 2015;Venter et al., 2016). Several factors have facilitated the recovery of wildcat populations, among others the reduced hunting pressure, increasing forest cover and progressive rural-to-urban migration of humans following World War II (Pereira & Navarro, 2015). Due to the species' elusive nature and requirement for undisturbed broad-leaved forest habitats, the reexpansion of wildcats originating from secluded low-mountain refugia was initially rather slow and may have even been overlooked in some areas for several decades (Piechocki, 1990).
The ongoing recolonization is currently monitored primarily using hair trapping and subsequent genetic detection (Steyer et al., 2013(Steyer et al., , 2016. Contemporary populations are estimated to comprise 5,000-8,000 individuals in Germany that occur predominantly along the low mountain ranges in the Central and Southwestern parts of the country and can be distinguished as two genetically divergent lineages or metapopulations (Mattucci et al., 2016;Steyer et al., 2016). The origin of this distinct spatial genetic pattern has occupied wildlife managers, stakeholders, and scientists for years as the two populations are located adjacent to each other, without separation by major barriers or an ecological gradient, and individuals from the current contact zone of the two populations hybridize (Mattucci et al., 2016;Steyer et al., 2016).
The recovery of wildcats following their near extermination represents a conservation success story, showcasing the effectiveness F I G U R E 1 Anthropogenic persecution and historical records of European wildcats in Germany. (a) Hunt for disproportionally enlarged wildcats with sticks and dogs (undated copper engraving; von Hohberg, 1695). (b) Number of wildcats reported in historical records (dotted line), of which known cause of death was persecution (red line), or which were found dead (blue line, mostly road kills) in 1800-1980 (modified from Reinert, 2017). The gray rectangle represents the population bottleneck at its strongest magnitude according to Piechocki (1990)   Answering these questions is of fundamental importance for informing applied wildcat conservation strategies that are currently based on minimizing genetic diversity loss through the implementation of dispersal corridors  as well as monitoring regional levels of hybridization with domestic cats (Nussberger et al., 2018).

| Sampling
A combination of museum specimens (hereafter referred to as "historical") and samples from extant German wildcat populations ("contemporary") were used in this study ( Figure S1). Historical samples of European wildcats from Germany (1830-2001; n = 175) were collected in 23 zoological museums in Europe (Table S1, sample details in Table SXL1, Appendix S2). Sample material consisted of preserved skins, footpads, fragments of skeletal bones or turbinal bones from the nasal cavity, teeth, and dried remains of tissue found on and in skulls. Contemporary samples originated from wildcat monitoring from 2006-2018 (n = 650) and represent the current wildcat distribution in Germany. These samples encompassed mainly tissues from road kills, as well as blood and noninvasively collected hair samples (Table SXL2, Appendix S2).
For most data analyses, samples were sorted into three temporal groups (prebottleneck, pre-BN, 1830(prebottleneck, pre-BN, -1930postbottleneck, post-BN, 1931postbottleneck, post-BN, -2005extant, 2006 and two geographical groups (Western and Central metapopulation; hereafter referred to as "West" and "Central"). Temporal periods were delimited based on recorded year of sampling. The height of the bottleneck was assumed as 1930 (following Piechocki, 1990

| SNP analyses
All genotyped individuals were screened for possible hybrids and domestic cats based on the 10 diagnostic HYB SNPs contained in the marker panel (von Thaden et al., 2020) using the Bayesian clustering methods implemented in NewHybrids v1.1 beta (Anderson & Thompson, 2002) and STRUCTURE v2.3.4 (Pritchard et al., 2000).
NewHybrids was run under the uniform prior for 200,000 MCMCs, after discarding an initial burn-in of 100,000 sweeps. In STRUCTURE, 200,000 MCMCs were preceded by a burn-in of 30,000 steps, assuming correlated allele frequencies under the admixture model. Ten iterations of K = 2 were combined using CLUMPP (Jakobsson & Rosenberg, 2007). To aid clustering, 22 genotypes of reference domestic cats from von Thaden et al., 2020, were included in these analyses. Subsequently, all domestic cats, potential hybrids, and individuals with an assignment value for wildcats q (wc) ≤ 0.85 (historical, n = 15; contemporary, n = 42) were excluded from further analyses (genetic diversity and structure), leaving 56 historical and 608 contemporary wildcat individuals.
Population structuring was assessed based on the genotyped 84 ID SNPs using three clustering methods and analyzing historical and contemporary samples together. STRUCTURE was run for 10 iterations of 200,000 MCMCs after an initial burn-in of 100,000 for K = 1-15 under the same assumptions as described before. The Evanno method (Evanno et al., 2005) implemented in STRUCTURE harvester (Earl & vonHoldt, 2012) was used to select the most likely K-value. Replicate runs were combined using the LargeKGreedy algorithm of CLUMPP (Jakobsson & Rosenberg, 2007). Principal coordinates analysis (PCoA) was conducted as implemented in GenAlEx v6.502 (Peakall & Smouse, 2012).
Groups for DAPC were assigned according to temporal and geographical classifications.
Genetic differentiation between the populations and sample groups was estimated based on pairwise F ST values calculated with 5,000 permutations (Weir & Cockerham, 1984) in Arlequin v3.5 (Excoffier & Lischer, 2010). To calculate pairwise F ST values between inferred population clusters (K) from STRUCTURE, samples with q (i) < 0.8 to any cluster were excluded. Genetic variability and diversity parameters (observed and unbiased expected heterozygosities, inbreeding coefficient, global pairwise F ST ) and analysis of molecular variance (AMOVA) were calculated using GenAlEx v6.502 (Peakall & Smouse, 2012) and Arlequin v3.5 (Excoffier & Lischer, 2010 (Table SXL3, Appendix S2), as has previously been reported (Dabney et al., 2013;Rohland & Hofreiter, 2007;Yang et al., 1998). Although mtDNA markers generally amplified better than SNP markers (10.3% higher amplification success), no mtDNA haplotypes could be obtained for ten samples that had been successfully SNP genotyped (13.2%). In contrast to the historical samples, all contemporary samples were genotyped successfully for both marker types without exception. The mean allelic dropout rate in SNP genotyping was 5.94% in historical and 0.22% in contemporary samples when comparing the replicates with consensus genotypes (Tables SXL4-5, Appendix S2). False allele rates were 0.36% and 0.03% for historical and contemporary samples, respectively.

| Genetic population structure and differentiation
Clustering analyses revealed changes in spatial genetic structure  (Table 2).
According to AMOVA, the amount of molecular variance between the metapopulations increased from 2% (pre-BN) to 6% in post-BN and extant populations (Table S3) Figure S8 and DAPC, Figure S9).  Figure S10). Values for the population inbreeding coefficient F IS (Table 2) Figure S12).

| Temporal change of haplotype frequencies
Before constructing haplotype networks, 21 mtDNA haplotype sequences were excluded due to ambiguous locality information or domestic cat classification (see Results in Appendix S1 and Table   S1). All haplotypes found in the historical samples corresponded to common extant wildcat haplotypes, with no indication for a loss of maternal lineages (Figure 4). Differences in haplotype frequencies between post-BN and extant populations appeared small (0%-15% per haplotype). However, when comparing historical haplotype frequencies from before the bottleneck (pre-BN, Figure 4; bottom layer) to frequencies of extant populations (Figure 4; second highest layer), changes in geographical haplotype prevalence became apparent (0%-48% difference). Specifically, H05, which is exclusively found in the Western metapopulation in extant samples, appears in two samples from the pre-BN Central populations (22% historical frequency). Also, haplotype H06, which is characteristic  (Table SXL1, Appendix S2).

| Hybridization assessment
Several domestic cats and potentially admixed individuals were detected within both the historical and contemporary sample sets Results from STRUCTURE were congruent.

| Genetic effects of the anthropogenic bottleneck
Strong population bottlenecks are one of the main drivers of genetic drift, often resulting in significant loss of genetic diversity. The extent of diversity loss and the associated consequences for population fitness are, however, not only dependent on the severity of the bottleneck but also its timescale . Both the duration and magnitude of the anthropogenic bottleneck in German wildcat populations are fairly well documented (Piechocki, 1990, and references therein), and resemble the demographic histories of F I G U R E 4 Temporal haplotype network for historical and extant German wildcats. Each circle represents one haplotype whose name is given in the uppermost layer. The following layers represent frequencies of these haplotypes at different temporal periods: extant (data from Steyer et al., 2016), post-and prebottleneck (historical museum samples from this study). The height of the bottleneck was assumed for the year 1930 (following Piechocki, 1990). Haplotype frequencies are indicated separately for Western (yellow) and Central (purple) geographical groups of German wildcat populations. The size of the circles reflects the percentage frequency of haplotypes. Small white circles indicate missing haplotypes for that time period. The number of dots plus the connecting haplotype equals to nucleotide differences. *, n = 1 sample in Western population (haplotype frequency of 0.15%); **, n = 3 samples in the Western population (haplotype frequency of 0.46%) H23 other medium-sized carnivores in Central Europe (e.g., Eurasian otter or European badger). Interestingly, even though population declines must have been severe until the height of the anthropogenic bottleneck in the early 20th century, we did not find evidence for a profound loss of genetic diversity, but rather a significant alteration in spatial genetic patterns. The most drastic changes were the temporal shifts in genetic population structure (Figure 2 and Figures S3-S7) and the accompanying increase in spatial genetic differentiation between the two metapopulations over time (Table 1, Table S2) Further, the Central population may have suffered stronger population size reductions due to its isolated geographical position or may have been exposed to long-term, recurring fluctuations of their range and population size, which would have likely led to strong drift effects. However, we did not find clear evidence of significant decreases in heterozygosities in either metapopulation (Figure 3a,b, and Table 2), except for the five pre-BN samples from the Central population ( Figure S10). In the latter case, however, sample size is too low to make reliable assertions. Further, our findings may have been influenced by uneven sample sizes, as NewHybrids STRUCTURE we did find a slight decrease of heterozygosity in the Central population (not significant) when analyzing subsets of samples with equal sample sizes (Methods and Results in Appendix S1, Figure   S11). Genetic drift effects may have also acted more rapidly, continuing after the strong population decline due to human persecution. In this case, the effects of drift may be too subtle to identify in common genetic parameters, but pronounced enough to result in the observed spatial genetic patterns, similar to the extant subpopulation structures ( Figure S5). Examples of such fast genetic drift exist for several recolonizing carnivores, such as the Central European wolf population (Szewczyk et al., 2019) or the reintroduced population of Eurasian lynx in the Harz Mountains (Mueller, Reiners, Middelhoff, et al., 2020).
The effects of drift were also noticeable in the mitochondrial sequence data, which suggest that haplotype frequencies have changed notably over time, whereas there was no sign of a complete loss of maternal lineages within the study region ( Figure 4). This is surprising, as some of the historical populations that we analyzed seem to have been completely extirpated and their habitats not recolonized to date (e.g., the Black Forest in Baden-Württemberg; cf. Figure 6). While this finding may suffer from sampling bias, it generally supports the hypothesis of a single, historically panmictic metapopulation.

| Edge effects in an expanding population
Another factor that could have shaped the metapopulation differentiation is distribution edge effects in the (re-)expanding Central population (Williams et al., 2019). Range expansions usually involve a series of founder events which could have acted as a spatial analog of genetic drift in the Central population (Slatkin & Excoffier, 2012).
Consequently, different forms of dispersal may generate distinct spatial patterns (Ibrahim et al., 1996), as could be the case here.
Knowledge about the speed and dimension of wildcat range expansions following the bottleneck has been incomplete preceding the introduction of noninvasive genetic monitoring methods in the last few years (Steyer et al., 2013(Steyer et al., , 2016. Additionally, the presence of the species may have been overlooked for several decades, firstly because wildcats may be easily confused with domestic cats by untrained persons and secondly due to its elusive behavior (Steyer et al., 2013). Poorly documented reintroduction events further complicate the reconstruction, unless accompanied by traceable genetic traits (Mueller, Reiners, Steyer, et al., 2020). For example, the occurrence of haplotype H23 after the bottleneck (Figure 4) is clearly associated with reintroduction efforts in the Spessart low mountain range between 1984 and 2011 (Worel, 2009 (Mattucci et al., 2016;Steyer et al., 2016). Similar differentiation legacies have been found in numerous European species (e.g., brown bear, Davison et al., 2011;roe deer, Sommer et al., 2009;wild boar, Scandura et al., 2008; see also Hewitt, 1999;Schmitt, 2007). This view is questioned, however, by the absence of potential glacial refugia within the current species' range, the lack of evidence for any morphological differentiation or the detection of private historical mitochondrial haplotypes (Sommer & Benecke, 2006). On the contrary, some of the haplotypes appear to be regionally private or distinctly predominant in the extant rather than the historical populations (e.g., West, H05; and Central, H06; Figure 4).
The detection of these currently regional private haplotypes in other areas based on the pre-BN samples demonstrates the considerable impact of the bottleneck on local maternal diversity. One possible explanation could be that the haplotypes were more prevalent once, that is, common in both metapopulations before the bottleneck, and that their occurrence was reduced to local populations through the decline, which is consistent with the nuclear genetic data (SNP results). However, bottlenecks may have very different influences on haplotypic diversity of naturally recovering populations (Sonsthagen et al., 2017), thus our findings can only be classified as indicative.
Further, private haplotypes may have also been present in historical populations, but not been captured in the present study due to the limited number of historical samples.

Mattucci and colleagues conducted Approximate Bayesian
Computing (ABC) simulations to assess the phylogeographic history of European wildcat populations and estimated a divergence time of ~30,000 years for the Central German metapopulation and other investigated Central European populations (i.e., Belgium, Luxembourg, Switzerland, and Western German metapopulation) (Mattucci et al., 2016). However, the uncertainty of their modal values (0.25-0.75 quantiles) was reportedly high and underlines the challenges associated with ABC inferences. While ABC methods offer a wider application of model-based statistical inference than traditional Bayesian approaches, it has also been the subject of controversial debate in the scientific community (Berger et al., 2010;Robert et al., 2011;Templeton, 2009). Parameter estimation, model selection as well as corresponding assumptions and approximations heavily influence the outcome of the simulations and thus need to be carefully assessed and evaluated (Sunnåker et al., 2013).
In order to take full advantage of the potentials of ABC methods

| Ancient or historical introgression
Introgressive hybridization is a natural phenomenon that may be either beneficial, neutral, or detrimental for the evolutionary trajectory of a species. Hybridization between wild-living taxa and their domestic congeners, however, is usually judged as unfavorable (outside of targeted breeding), as the domestic taxa may genetically swamp their wild relatives and lead to genetic extinction of the latter . Consequently, the assessment of hybridization represents a critical conservation issue. The current habitats of European wildcat populations are situated in landscapes that are densely populated by humans and their domestic cats (EPFI, 2019). Thus, the risks for hybridization between the two species are high and incidences have been reported throughout Europe Lecis et al., 2006;O'Brien et al., 2009;Oliveira et al., 2008;Steyer et al., 2018;Tiesmeyer et al., 2020). In this light, different levels of ancient or historical introgression appropriated before or after the anthropogenic bottleneck could explain the differentiation of the currently observed metapopulation patterns in Germany. This hypothesis is supported by the fact that hybrids have been shown to occur more frequently at the periphery of wildcat ranges (Randi et al., 2001) and a high degree of habitat fragmentation may further enhance these edge effects .
In Switzerland, for instance, recent range expansions have led to increased local hybridization rates (Nussberger et al., 2018;Nussberger, Wandeler, Weber, et al., 2014). The theory of modern introgression is, however, contradicted by the fact that the con-  (Mattucci et al., 2019;Oliveira et al., 2015;Steyer et al., 2018). As previous studies have suggested prehistoric gene flow between the ancestors of European wildcats and domestic cats, future studies focused on ancient introgression will probably be based on whole-genome sequencing (WGS) data and might even take paleogenomic evidence into account in order to unravel these complicated phylogenetic relationships (Driscoll et al., 2007;Howard-McCombe et al., 2021;Ottoni et al., 2017).
While we cannot rule out that any of the four abovementioned factors may have contributed to the origin of the observed spatial differentiation, we argue that it appears to be a direct consequence of the anthropogenically induced bottleneck. Based on our results, the existence of a single, genetically diverse and panmictic metapopulation preceding the anthropogenic persecution appears as the most likely scenario. The local wildcat populations have probably experienced a combination of genetic drift during refugial isolation and range edge effects during the re-expansion following legal protection. Contemporary populations still carry the resulting spatial patterns in their genetic legacy, although Central subpopulations already seem to intermix in the course of their recolonization and consequently lose their distinct genetic substructuring ( Figure S5).
The differentiation between the Western and Central metapopulation, however, will probably take longer to fade out, mainly due to the geographically disjunct location of the current metapopulations.

| Limitations and methodological issues
Surprisingly, we did not find any clear indication of genetic drift or other phenomena that accompany sharp reductions in population size. This may be explained by the type of bottleneck, as substantial genetic diversity loss may only become detectable if population sizes are below a certain threshold (e.g., N < 200 in Hoban et al., 2014). Hoban et al. (2014) conclude in their simulation study that the detection and monitoring of genetic erosion may be unfeasible -even considering many genetic markers -if the effective population size exceeds several hundreds. Although the latter seems unlikely for the wildcat, it cannot be excluded that the anthropogenic bottleneck event may not have been as severe as previously stated by Piechocki (1990) and/or that local wildcat populations might have been overlooked until recently. This is because standardized genetic monitoring, which represents the most effective detection method for this elusive species, only started in the early 2000s (Steyer et al., 2013). As a matter of fact, we did not find any signs of inbreeding (F IS values; Table 2) or strong loss of genetic diversity.
Genetic data are commonly used to assess the timing of bottleneck events in conservation genetic studies, but the ability to reliably determine these is subject to several factors (Peery et al., 2012). While the timescale and severity of the bottleneck markedly influence the resulting genetic effects for a population or species (Hundertmark & Daele, 2010;Sonsthagen et al., 2017), they are also depending on the level of pre-and postdecline diversity, as ancient reductions in genetic diversity may mask recent declines (Dussex et al., 2015). More specifically, if genetic diversity is already low, for example, due to an ancient decline, before experiencing a recent bottleneck, genetic losses may be impossible to detect (Cornuet & Luikart, 1996;Dussex et al., 2015;Sonsthagen et al., 2017). Further, long-lived species may be able to preserve genetic diversity over shorter periods of decline (Hailer et al., 2006;Johnson et al., 2008;White et al., 2014). Various methods exist for the detection of bottlenecks, ranging from classical tests for loss of heterozygosity and changes in allele distribution frequencies to more recent approaches based on genomic data (Allendorf, 2017;Cammen et al., 2018;Luikart et al., 1998). All of these methods rely on a series of assumptions and models, such as a model and rate for mutations, absence of gene flow, mutation-drift equilibrium before decline, or uniformity of reproductive success (Gattepaille et al., 2013;Hoban, Mezzavilla, et al., 2013). Further, demographic reconstructions suffer from bias and confounding factors such as the presence of population structure (Peter et al., 2010;Sousa et al., 2012), insufficient sampling or the choice of marker systems including their associated ascertainment bias Williamson-Natesan, 2006). As we could not exclude the violation of several of the above-mentioned assumptions, we did not conduct bottleneck tests in this study but focused instead on the evaluation of spatial patterns. Based on our findings -especially the lack of evidence for common genetic bottleneck effects -subsequent studies are advised to incorporate a more equally distributed sampling scheme and probably a much higher number of markers (if not WGS data) to further elucidate the demographic history of European wildcats in Germany.
Reconstructing demographic histories based on genetic analyses of collection material involves both opportunities and challenges (Newbold, 2010). General limitations of data from museum resources may result from errors (e.g., ambiguous locality information or wrong species identity) and/or biases (e.g., spatial, environmental, temporal, and taxonomic) (Graham et al., 2004;Newbold, 2010;Soberón et al., 2000). Some of these limitations may be overcome and require careful scrutiny, which led us to exclude 27 of the collected historical specimens from the analyses. Further, regional and temporal biases often make representative, standardized sampling of a specific geographical or temporal group impossible, resulting in low sample numbers that hamper sound statistical inferences, for instance for the Central pre-BN individuals (n = 5) in this study. One option to test for potential sampling effects lies in subsampling contemporary populations for more balanced sample sizes, as we did here (Methods and Results in Appendix S1, Figures S4-S7 and S11-S12, Tables S5-S10).
Concomitantly, time series do not exist for many species and populations, and are often heavily biased (see above). Considering the resulting gaps in our historical data, we may have missed parts of genetic diversity in the present assessments. Although heavily persecuted, killed wildcats have probably never reached the status of more prominent historical game species, such as the wolf, lynx, and red deer, and may have been considered less as a trophy worth of taxidermic preservation. This lesser interest in collecting wildcats for natural history museums may have influenced our findings and should be considered. For example, it seems possible that wildcats were already scarce before the mid-19th century, that is, before the specimens in this study were collected, so that the bottleneck event may actually extent much farther into the past and consequently might have affected our results.
Another significant aspect is the choice of marker system and the number of loci used to detect genetic erosion Peery et al., 2013). Here, we used a panel of 96 SNPs, of which 84 were selected for maximized heterozygosity in contemporary wildcat populations (von Thaden et al., 2020). The selection of highly polymorphic markers is typically associated with some degree of ascertainment bias, which may have affected the results in this study. Firstly, an optimized panel of polymorphic markers does not provide unbiased estimates of genetic indices, may potentially overestimate current diversity, and, owing to the restricted number of SNPs, does not represent genome-wide diversity (Geibel et al., 2021). Secondly, the absence of signs of genetic erosion may stem from the SNP panel being designed using solely contemporary samples. Accordingly, the SNPs may not completely reflect historical diversity, potentially leading to underestimation of genetic losses, lower estimates of pairwise F ST and lack of genetic structure in historical samples. Indication for masked historical diversity is, for instance, observable in the DAPC ( Figure S9), where pre-BN samples form an adjacent, but separate cluster as compared to extant individuals. Further, it is important to note that the typed SNPs are bi-allelic (Albrechtsen et al., 2010;Malomane et al., 2018). Other than multiallelic microsatellites, bi-allelic SNP markers are not as prone to exhibit loss of alleles. Consequently, the detected diversity changes in this study were limited to changes in allele frequencies rather than number of alleles, whereas the latter have been assessed as the best indicator for monitoring genetic erosion following declines with high statistical power (Hoban et al., 2014). We chose to use SNP markers because they have proven to yield higher amplification successes and lower error rates when genotyping degraded sample materials, also due to their short amplicon lengths (<120 bp; von Thaden et

| Conclusions and implications for conservation
European wildcats in Germany have survived centuries of population decline and massive anthropogenic persecution. Today, populations are expanding their ranges, appear to recolonize most of their former habitats, and even seem to advance to newly occupied areas ( Figure 6, with details in Appendix S1 Methods and Results ;Steyer et al., 2016;Reinert, 2017;Balzer et al., 2018). Based on the presence data alone, the extant wildcat populations appear viable and thriving and offer reason to hope for a successful re-establishment of the species within the next decades into the presently remaining gaps of its historical distribution ( Figure 6; Balzer et al., 2018).
The long-term viability of a species, however, is not only dependent on the sheer number of individuals, but also of the genetic makeup of its populations (Hoban et al., 2020). While the species' recovery in a cultivated landscape can certainly be evaluated as a conservation success, genetic monitoring of wildcats is still required to assess postdecline development of wildcat populations (Steyer et al., 2016;Tiesmeyer et al., 2018 Reinert (2017) to official contemporary distribution data (shades of green) from the European Environment Agency (EEA, 2020, Copenhagen, Denmark) and national reports (Birlenbach & Klar, 2009). Wildcat presence is depicted on a 10 × 10 km grid 2018) and will likely be revisited by future studies based on WGS. For now, the endangered wildcat populations in Germany appear under no imminent threat from genetic factors and consequently viable in long term. While an active reconnection of the disjunct populations is not absolutely essential from a genetic perspective, it will certainly facilitate the ongoing range expansion of the species. The resulting convergence of West and Central populations may promote the restoration of genetic diversity in German wildcat populations to levels seen before the onset of massive persecution.
In conclusion, our study demonstrates how the inclusion of historical genetic data, for example, from museum records, serves as an important tool to understand a species' demographic history and take appropriate and effective conservation actions (Barnosky et al., 2017;Fenderson et al., 2020;Meineke et al., 2018).

ACK N OWLED G M ENTS
This study was partially funded by the Landes-Offensive zur

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
SNP genotyping data are available in the Dryad repository: https:// doi.org/10.5061/dryad.31zcr jdmr. Sample information for this study is available in the Supplementary excel files (Tables SXL1-2 in Appendix S2).